Modelling the reinjection of deep-water

after methane extraction in Lake Kivu

Alfred Wüest

Lukas Jarc

Martin Schmid

Eawag
Surface Waters - Research and Management
Seestrasse 79, CH-6047 Kastanienbaum, Switzerland

December 2009

[Link]
2
Abstract
The deep-water of the East African Lake Kivu contains large amounts of
dissolved methane (~60 km3 STP) and carbon dioxide (~300 km3 STP). To utilize
the methane as an energy resource valued at many billion dollars and to reduce the
risk of an uncontrolled gas eruption, the governments of the Democratic Republic of
the Congo and Rwanda have decided to begin extracting the methane. The
Government of Rwanda has issued first concessions for pilot plants to start
extraction in the fall 2008.
There are two major risks involved with extracting methane from Lake Kivu: (i)
disrupting the lake stratification, which could increase the probability of a gas erup-
tion, and (ii) enhancing the transport of nutrients from the deep-water to the surface
water, which could negatively affect the lake ecosystem. These two risks should be
kept in mind while trying to (iii) minimize the methane loss and maximize the extrac-
tion of the naturally produced methane in the lake. In this study, a one-dimensional
model was developed and applied to predict the stratification, in addition to the gas
and nutrient concentration profiles, for a 100 year period of various methane
extraction Scenarios in Lake Kivu. The goal of the modelling is to find an optimal
strategy that acceptably fulfils all three concerns (safety, lake ecology and socio-
economic benefit).
The resulting simulations let us conclude that (a) the deep-water MUST NOT
be diluted with surface water to adjust its density before reinjection into the lake,
and that (b) the degassed deep-water can not be reinjected above 200 m depth,
where this would lead to both strongly increased nutrient fluxes into the surface layer
and unacceptable algae growth. Despite these two restrictions it is possible to
harvest ~90% of the maximal possible methane without risking a gas eruption and
without sacrificing the ecological integrity of the lake. The key findings are listed in
the executive summary.

How to read this report?

This report consists of two parts: (PART I) a summary (Sections 1 and 2),
which contains the conclusions (executive summary) and an outline of the decision-
making approach; and (PART II) a technical part, where the lake data (Section 3),
the modelling approach (Sections 4), the evaluation procedure (Sections 5 and 6)
and the model simulations (Section 7) are given in detail. The Appendix contains
useful technical details (such as formulas). Readers interested only in the outcome
and implications can refer to PART I of this report. PART II serves as a detailed
documentation of (i) the model input data, (ii) the interpretation of the simulation
results and (iii) for further planning and future revisions of model calculations.

3
Acknowledgments
We would like to like to thank Natacha Pasche for sharing her knowledge
and data on the nutrient cycling in Lake Kivu, and we are grateful to many people
from the National University of Rwanda in Butare (Claudien Kabera, Jean
Népomuscène Namugize, Sylvain Nzaramba, Antoine Ntamavukiro), from the
Institut Supérieur Pédagogique de Bukavu (Boniface Kaningini, Pascal Isumbisho,
Fabrice Muvundja Amisi, Georges Alunga), and from Eawag (Michael Schurter,
Christian Dinkel, Ruth Stierli, Bernhard Wehrli) who supported the field work on
Lake Kivu. In addition we would like to thank Jean-Pierre Descy and Michel
Halbwachs as well as the crews of the different research vessels used during the
sampling on Lake Kivu. Jeff Carpenter and KellyAnn Ross are acknowledged for
linguistic improvements.
The following individuals provided input and feedback for the Scenario
construction as well as the preparation and reviewing of this document: John Boyle,
François Darchambeau, Michel Halbwachs, Finn Hirslund, Philip Morkel, Charles
Nyirahuku, Klaus Tietze, and Augusta Umutoni.
Last, but not least, we thank the Belgian Technical Cooperation for the
financial support and discussions, in particular Dirk Brems, Pierre Dulieu and
Ahmad Parsa as well as the Swiss SNF/DEZA for supporting the scientific project
207021-109710 (Nutrient cycling in Lake Kivu), which massively contributed to the
understanding of the fascinating lake.

4
Table of contents
Abstract 3
How to read this report? 3
Acknowledgments 4
Table of contents 5

PART I: Summary of Results 7

1. Executive summary 7
2. Background to executive summary 15
2.1 Lake Kivu: stratification of methane and nutrients 15
2.2 Guiding Principles for sustainable methane extraction 19
2.3 Modelling and Scenario evaluation procedure 23
2.4 Evaluation of the extraction Scenario simulations 31
2.5 Evaluation of the robustness of selected Scenarios 33
2.6 Conclusions and optimal extraction Scenario 35

PART II: Technical Documentation 39

3. Lake data used for model simulations 39
3.1 CTD Profiles 39
3.2 Water chemistry 39
4. Model description and model assumptions 43
4.1 Introduction 43
4.2 Surface inflow 44
4.3 Subaquatic inflow 44
4.4 Surface outflow 45
4.5 Turbulent Diffusion 48
4.6 Temperature 48
4.7 Salinity 50
4.8 Density 50
4.9 Methane 52
4.10 Carbon dioxide 54
4.11 Phosphate and ammonium 55
4.12 Lake development without extraction - Scenario O 56
5. Definition of the extraction Scenarios 59
5.1 Design Parameters 59
5.2 Intermediate Zone reinjection - Scenarios IZ1 to IZ5 62
5.3 Potential Resource Zone reinjection - Scenarios PR1 63
5.4 Resource Zone reinjection - Scenarios RZ1 to RZ6 63
5.5 Summary of the simulated extraction Scenarios 63
6. Scenario evaluation criteria 65
6.1 Extraction Objectives and quantification by Attributes 65
6.2 Definitions of the Attributes 66
7. Simulation results of all Scenarios 71
7.1 Scenario IZ1: Reinjection into Intermediate Zone with dilution 71

5
 7.2 Scenario IZ2: Reinjection into Intermediate Zone with dilution 76
7.3 Scenario IZ3: Reinjection into Intermediate Zone with dilution 80
7.4 Scenario IZ4: Reinjection into Intermediate Zone - no dilution 84
7.5 Scenario IZ5: Reinjection into Intermediate Zone - no dilution 88
7.6 Scenario PR1: Reinjection into Potential Resource Zone 92
7.7 Scenario RZ1: Single reinjection into Resource Zone 96
7.8 Scenario RZ2: Double reinjection into Resource Zone 100
7.9 Scenario RZ3: Long-term effects of gas removal efficiency 104
7.10 Scenario RZ4: Resource Zone Reinjection 108
7.11 Scenario RZ5: Resource Zone Reinjection 112
7.12 Scenario RZ6: Resource Zone Reinjection 116
7.13 Summary of Attributes for all Scenarios 120
8. Appendix 123
8.1 AQUASIM 123
8.2 Model parameters 124
8.3 Formulas used in the model 126
8.4 Dynamic processes and stoichiometry 129
8.5 Exchange at the lake surface 130
8.6 Standard units used 130
8.7 Sensitivity analysis 131
8.8 Data profiles used in the model 138

6
PART I: Summary of Results
1. Executive summary
The deep-water of the East African Rift Lake Kivu contains ~60 km3 (STP) of
dissolved methane and ~300 km3 (STP) of dissolved carbon dioxide. While the
gases pose a serious risk to all oxygen-dependent life in the vicinity of the lake, the
methane gas also provides an accessible (and partly renewable) energy resource,
valued at many billions of dollars, for the Kivu region.
The Governments of Rwanda and DR of the Congo decided to award conces-
sions for methane extraction to private investors under specific management pre-
scriptions (MP 2009). These requirements will (i) ensure the safety of the population,
and (ii) conserve the integrity of the lake ecosystem while (iii) maximizing the
methane harvest by minimizing the methane loss to the atmosphere and to the oxic
surface water. This report summarizes the results of model calculations that aim at
supporting the development of an optimal strategy of methane extraction from Lake
Kivu.
Most of the methane is stored in the deep part of the lake, indicated as the
Resource Zone in Figure 1.1. The principle for extracting the methane from this
Zone is straightforward (Figure 1.2): initially a tube is placed vertically into the
Resource Zone. As the deep-water moves upwards in the tube it experiences de-
creasing pressure and bubbles are formed. The bubbling water is lighter (buoyant)
and drives an upward flow through the tube. Deep-water is pulled upward while

Figure 1.1 Definition of the terms for the identification of the water bodies, as listed in Table
1.1. The boundaries are drawn according to the vertical structure of the methane profile. For
volumes of water and methane, see Table 1.1. The triangles and circles depict observed
methane concentrations; the black line represents a continuous methane profile, assuming
that methane concentrations are proportional to CO2 concentrations.

7
continuously forming bubbles, which results in a self-sustaining (self-siphoning) flow.
The gases are then stripped from the withdrawn deep-water using a separator near
the lake surface. Following the washing of the extracted raw gases, the refined
gases are transferred in underwater pipelines to the shore, where the methane can
be used for distribution, bottling or electricity.
The gas-depleted deep-water then needs to be reinjected into the Lake. This
reinjection could potentially lead to long-term negative changes including: (1)
excessive algae growth causing nuisance to the population and damage to the lake
(such as (i) zones deprived of oxygen and related fish kills, (ii) changes in the
composition of the plankton, which is the food base for fisheries, or (iii) toxic algae
blooms) as a result of enhanced nutrient fluxes from the deep-water to the surface,
and (2) weakening the extremely stable density stratification (i.e. salty/heavier water
at depth, underneath of fresher/lighter water at the surface), which is presently
preventing the gases from diffusing (and mixing) upwards and is therefore crucial for
safety as well as for the methane accumulation in the lake.
The concept of the gas extraction described above is schematically identical
for the different proposed projects that may vary in design and operation parameters.
Therefore, the aim of this study is to estimate the effects of different extraction
Scenarios on the development of Lake Kivu’s density stratification, nutrient fluxes
and gas concentrations. The vertical distribution of the gases is of particular concern
as this affects the safety. The extraction Scenarios (Table 2.3.2) were developed in
co-operation with engineers working on Lake Kivu methane extraction projects.
These Scenarios, as well as the “no extraction” Scenario (O) without methane

Figure 1.2 Methane-rich deep-water is extracted from the RZ. After stripping the gases the
methane-depleted deep-water is reinjected into the lake. This report studies the conse-
quences for different reinjection Scenarios releasing the methane-depleted water into the IZ
(Scenarios IZ1 to IZ5), into the PRZ (Scenario PR1) and the RZ (Scenarios RZ1 to RZ6).

8
exploitation, were simulated with a one-dimensional model specifically designed for
this purpose and implemented with the software AQUASIM. Density stratification,
gas concentrations and nutrient fluxes were simulated for extracting operations over
a 100 year period.
In order to understand the simulation results, it is important to take into
account a few particularities of Lake Kivu’s stratification. These specialties comprise:
(1) As a consequence of the density stratification, the deep-water of Lake Kivu is
decoupled from the surface waters, i.e. there is almost no exchange of water mass
between the deep waters below 260 m depth and the surface water. (2) Due to this
confinement gases as well as nutrients and other dissolved substances accumulate
to very high concentrations in the deep waters. (3) Despite the strong stratification
and the suppressed vertical exchange of water, there is still upward transport of the
nutrients and gases from the deep-water towards the surface. This upward flux is
due to the slow movement, so-called upwelling of the water layers (i.e. uplift) of up to
~1 m per year. This upwelling is caused by the inflow of subaquatic sources at
various depths within the deep-water. In this report it is assumed that ~36% of the
lake water outflow to the Ruzizi River stems from such subaquatic inflows.

The results of the simulations can be summarized as follows:

A) Guiding Principle “Safety”

(1) Maintaining stable lake stratification - The best approach to eliminate any risk
for a gas eruption would be to completely remove all the gases from Lake
Kivu immediately. However, this would not permit the utilization of methane
as a resource during the next few decades and is therefore not considered an
option. The most ideal approach is to the keep the lake stratification as stable
as possible in order to minimize the risk of a gas eruption and furthermore to
minimize the loss of the methane resource. Most of the Scenarios (Table
2.3.2) simulated here will be able to maintain 50% of the natural stability,
which we consider more than acceptable in view of the decreasing gas
contents due to the extraction operation.

(2) Disadvantage of shallow reinjection - The lake stratification is drastically

weakened when water is reinjected into the Biozone (not acceptable) or into
the Intermediate Zone. This disadvantage is reduced with deeper reinjection
(Potential Resource Zone).

(3) Advantage of deep reinjection - Reinjecting water into the Resource Zone is
the Scenario with the least disturbance to the density stratification.

9
B) Guiding Principle “Lake Ecological Integrity”

(4) Minimizing nutrient fluxes into the Biozone - The best approach for conser-
ving the lake ecological integrity would be to extract methane from the
Resource Zone water and then to reinject this water back into the same zone.
The deep-water would then be depleted in methane but the stratification of all
other water properties would remain almost unchanged. Particularly the
nutrient fluxes into the Biozone, which affect algae growth, would remain at
their natural level.

(5) Release into the Biozone - Methane-depleted reinject-water, which consists of

extremely nutrient-rich deep-water, must STRICTLY NOT be released into
the Biozone. Its release would increase the nutrient flux into the Biozone (and
subsequently cause algae growth) to an unacceptable level even for a small
rate of methane extraction (equivalent to a few MW of electrical energy
production).

(6) The role of upwelling by subaquatic springs - Subaquatic springs in Lake

Kivu induce a continuous upward flow of water (~1.3 km3 per year) and cause
an upwelling (uplift) of the lake water layers towards the surface (~0.7 m per
year into the Biozone). If methane-depleted reinject-water is released directly
below the Biozone into the Intermediate Zone, the nutrient content below the
Biozone would increase. The upwelling water would consequently flush an
increased amount of nutrients into the Biozone.

Despite that upwelling is slow, the nutrients would reach the Biozone already
after a few decades. Therefore, the reinjection of nutrient-rich water into the
Intermediate Zone (defined as layer from 60 to 200 m depth) would only be
acceptable for small extraction rates of less than a few 10 MW of equiva-
lent electricity production, which is much less then the anticipated extraction
of several 100 MW of equivalent electrical energy. For those high methane
extraction rates, reinjecting deep-water into the middle-to-upper
Intermediate Zone is not acceptable.

(7) Dilution of reinject-water - If reinject-water is to be re-stratified in the Interme-

diate Zone (small amounts only, such as during pilot plant operations or for
short periods of time), then the density of the methane-depleted deep-water
has to be decreased to meet the density of the lake water in the Intermediate
Zone. This density adjustment can be achieved by diluting the reinject-water
with near-surface (lighter) water. Removing water for dilution from the surface
and releasing it at a greater depth would immediately increase the upward
flow of deep-water into the Biozone. Therefore, dilution-water (if used at all)
has to be taken from STRICTLY BELOW 80 m depth in order not to increase
the upward flux of nutrients.

10
 However, any dilution, including with water from below 80 m depth, will
increase the flux of nutrients to the surface in the long term. Extraction
operations without dilution-water have a clear advantage for the
ecological integrity of the lake. Furthermore, using dilution-water only
negligibly benefits the amount of harvestable methane, and therefore we
conclude that there is no justification of using dilution-water.

C) Guiding Principle “Economic benefit”

(8) Maximum methane harvest - The maximum methane harvesting would be

achieved if the reinject-water were released at the surface, which would
prevent any dilution of methane-containing deep-water. Additionally, this
approach would minimize the loss of methane to the Biozone. However, the
deep-water is enormously nutrient-rich and therefore this Scenario is unac-
ceptable. From the tested Scenarios, we found that the maximum methane
harvest would be achieved for the Intermediate Zone reinjection Scenario.
This Scenario serves, therefore, as a reference for the maximum methane
harvest. Depending on the formation rate of new methane, ~50 to ~70 km3
can be harvested during the next 100 years.

D) Finding a compromise among the three Guiding Principles

(9) Structured decision-making - A so-called structured decision-making approach

(Reckhow 1994; Reitsma 1996) has been chosen to compare the advantages
and disadvantages of the different possible extraction Scenarios (Table 2.3.2).
This allows a comprehensible and transparent way of comparing the
Scenarios along well-defined Objectives (Table 2.2.2). Our approach is to
stipulate that the overall lake stability shall never be lowered by more than
50% relative to natural (undisturbed; no extraction) and that the upward flux of
nutrients are never enhanced by more than 25% relative to natural conditions.
Several Scenarios fulfil the requirements of these Objectives (Table 2.4.1).

(10) Optimal Scenario – There are some limited differences among the Scenarios
that fulfil the Objectives to maintain “Safety“ and “Lake ecological integrity”.
We concluded that reinjection without dilution into the Resource Zone
carries the most advantages. This Scenario allows almost a maximal amount
of methane to be extracted. In addition, this Scenario is not sensitive towards
the “completeness” of methane extraction at the extraction facility. Methane
that is not effectively extracted, and which is released back into the lake (with
the reinject-water), would thereby return back into the Resource Zone. There-
fore, such incompletely extracted methane could be harvested in the (far)
future and would not be lost as a resource. Therefore, it is not acceptable
diluting the Resource Zone with water which is from outside the Resource
Zone (such as by using dilution-water).

11
 It is realistic to harvest the methane from Lake Kivu while following the three
Guiding Principles concerning safety, maintaining the lake ecological integrity
and achieving high methane output. The ideal Scenario, which fulfils these
restrictions on internal nutrient cycling and maintaining the lake stratification,
will most likely access reduced amounts of methane. The available methane
in the lake will be reduced by ~2% whereas the maximal methane harvest
can be reduced by up to ~12%, relative to the economically most favourable
Scenario. The differences of methane harvest among the Scenarios, which
fulfil the Principles concerning safety and lake ecological integrity, are only a
few %.

Due to this optimization potential of a few % in methane harvest, we recom-

mend to analyze the best strategy in more details. Such a study has to inclu-
de: (i) the optimization of the methane extraction efficiency and of the power
conversion efficiency, (ii) the flexible removal of CO2 and (iii) the optimized
vertical stacking of the reinjection water.

(11) Robust strategy - Some of the assumptions included in the model simulations,
are not well-known (or not yet defined) and need to be reconsidered in future
modelling, after an understanding of the lake processes and the extraction
processes has improved and more planning information has become avai-
lable. The two most relevant uncertainties with respect to the model output
are (i) the rate of new methane formation, and (ii) the discharge-rate of the
subaquatic sources. Furthermore, (iii) the rate of methane extraction, which is
dependent on political and economic developments, is not yet well-defined.
Fortunately, the comparison of the different Scenarios (evaluation procedure
in Section 6) show that the “ranking” of the different Scenarios does not
depend sensitively on these three assumptions.

Probably the most important uncertainty is the concentration level at which

methane can still be efficiently and economically harvested. In this report, we
assumed this level at 5 mol m-3. If practical experience should turn out that
this level is significantly different, then the details (especially the timing) of
the extraction strategy (as described in RZ6) need to be adapted.

(12) Future strategy adjustment - The results of the simulations presented in this
report are sufficiently robust to be used as a guide for projecting the methane
extraction from Lake Kivu into the upcoming decades. Nevertheless, adjust-
ments to the strategy or to the targets will be necessary in the future. The
model contains unknowns and the simulation output carries uncertainties (see
item 11) including the two most important internal lake processes: the water
inflows from the deep-water sources, and the new methane formation in the
lake. Furthermore, their constancy is not guaranteed in the future (natural

12
 variability). It is therefore essential to continuously monitor the evolution of
the stratification as well as the gas and nutrient concentrations in the lake.
Such monitoring will allow for a reaction to unforeseen developments by adju-
sting the methane harvesting. We recommend re-evaluating the optimal
strategy whenever major concessions are granted and the simulations
should be repeated at least every 10 to 20 years by using the newest profiles
and adjusting the model parameters according to the latest knowledge on the
functioning of the lake. The harvesting strategy should then be optimized
based on these new observations and simulation results, as well as the
experience gained from the ongoing gas extractions and the technology
improvements in methane harvesting methods.

Table 1.1 Water bodies of Lake Kivu (terms as used in report)

Lake-internal Area Water volume CH4 volume Estimated CH4
water body [km2] [km3] [km3] new formation
[km3 yr-1] (*)
Biozone 2’370 to 2’020 134
(0 to 60 m depth)
Intermediate Zone 2’020 to 1’280 229 11.7 0.17 (*)
(60 to 200 depth)
Potential Resource 1’280 to 1’050 70 8.5 0.05 (*)
Zone (200 to 260 m)
Resource Zone 1’050 to 0 118 44.7 0.24 (*)
(260 to 485 m depth)
Total 2’370 550 63.9 0.45 (*)
(“) This new CH4 formation rate (120 g-C m-2 yr-1) was used to reproduce concentrations
measured by Tietze (1978) and Schmid et al (2005a,b). The most recent data (Pasche et al
submit), however, indicate that this new CH4 formation rate is an overestimate. For planning
purposes, we recommend using 50% of the above listed values as the most probable new
CH4 formation rate.

13
14
2. Background to executive summary
This Section provides background on the Lake Kivu functioning (internal lake
processes), the definition of the extraction Scenarios, the model simulations, and the
evaluation of the extraction Scenarios. In addition the procedure for determining the
optimal Scenarios is explained.
Information on the modelling approach (Sections 3, 4, 5), the decision-making
process (Section 6) and the model simulations (Section 7), as well as the Appendix
(Section 8) is presented in more detail in the second part of the report.

2.1 Lake Kivu: stratification of methane and nutrients

Vertical structure of lake water - Lake Kivu is one of the most fascinating
lakes on earth. Surrounded by mountains, this Rift Lake is located at an altitude of
1462 m between Rwanda and the Democratic Republic of the Congo (DRC). The
lake water body has a maximum depth of 485 m, contains a volume of 550 km3 and
covers a surface area of 2370 km2. During the dry season the lake surface cools off,
causing convective mixing (cooler water is heavier than warmer water) from the
surface down to ~50 to ~70 m depth. This surface layer - called the Biozone in this
report (Table 2.1.1) - is oxygenated every year during convective mixing in the dry
season. The Biozone ecosystem hosts all the “higher” life in Lake Kivu, such as
fishes, zooplankton, etc. that cannot survive without oxygen. The oxygen originates
from the atmosphere and the algae, which grow in the Biozone. Below the Biozone
(i.e. below ~60 m) the lake water body is permanently density-stratified and anoxic
(free of oxygen) at all times. Enormous amounts of gases have accumulated there
during hundreds of years; about ~300 km3 (STP) of carbon dioxide (CO2) and ~60
km3 (STP) of methane (CH4), are trapped in the deep waters of the lake (STP: gas
volume at 0 °C and 1 atm). The gasses are dissolved in the water similar to the
carbon dioxide dissolved in bottled mineral water.

Methane in Lake Kivu - The CH4 in Lake Kivu is a renewable energy re-
source, as new CH4 is continuously formed. Due to the unusually stable stratification
and the large depth of Lake Kivu, the CH4 remains stored in the deep water for a
term of several hundred years. CH4 is transported to the Biozone via upwelling water,
where it is either consumed by microorganisms or diffused into the atmosphere. The
microorganisms need either oxygen, sulphate, or nitrate to degrade CH4. Because
these three substances are only available in the upper layers, CH4 is well preserved
in the deep-water and can accumulate undisturbed.
The CH4 stems from (i) bacterial fermentation of settling organic matter (main-
ly in the IZ and PRZ) whereas in the RZ in addition also (ii) CO2 reduction and/or (iii)
inflow of geogenic CH4 contribute to the build-up of the CH4 resources. The exact
composition of these three different sources is still the subject of scientific research.

15
Table 2.1.1 Definition of the terms used for the lake water bodies

In this report, we use consistently the following terms to distinguish the

different vertical layers of the Lake Kivu water body:

Biozone (BZ) - This is the upper, oxygenated water body of Lake Kivu (~50 to ~70 m deep;
Figure 1.2), where algae provide food for zooplankton and fish. The Biozone be-
comes homogenised during the dry (cooler) season but is strongly stratified during
the rainy (warmer) season. The Biozone water contains oxygen, reaching to ~60 m
during the dry (cooler) season. During the rainy (warmer) season the oxygen is partly
depleted in the lower Biozone. The CH4 content in the Biozone is negligible. If not
indicated otherwise, the Biozone refers to the top 60 m.

Intermediate Zone (IZ) - This zone is located between the Biozone and the Potential
Resource Zone and ranges from 60 to 200 m depth. The Intermediate Zone contains
11.7 km3 of CH4 (Table 1.1), which is not expected to be harvestable within the 100 yr
of simulation.

Potential Resource Zone (PRZ) - In this zone, which is confined by the Intermediate Zone
above 200 m and by the Resource Zone below 260 m depth, a significant part of the
CH4 is stored (8.5 km3, Table 1.1). Some of this CH4 may become harvestable within
the next few decades, if CH4 accumulation is continuing or if future technologies
advance.

Resource Zone (RZ) - The Resource Zone contains the entire water body below the
Potential Resource Zone from 260 m to the maximum depth of 485 m. This water
body not only contains most of the CH4 at harvestable level (~44.7 km3) but also
enormous amounts of nutrients. This deep-water is CO2-rich and salty and therefore
significantly denser (heavier) than the upper water layers. In this report we occa-
sionally distinguish between Upper RZ (260 to 320 m depth) and Lower RZ (320 to
485 m depth).

Main Density Gradient - The steep gradient from 255 to 262 m depth forms the highest den-
sity difference within the vertical structure of Lake Kivu. Although there are other
density interfaces in the lake, this gradient at 260 m is special as it distinctly sepa-
rates the Resource Zone (extractable CH4) from the rest of the lake above.

Reinject-water - The nutrient-rich deep-water, after CH4 extraction, plus the added dilution-
water (if used) needs to be reinjected into the lake (Fig. 2.3.1). Although the deep-
water is gas-depleted, it still contains large amounts of CO2 and some traces of CH4.
The core of this report addresses the question on how to best manage the
“undesirable” reinject-water. Reinject-water is unwanted in the upper lake layers as it
contains too much nutrients and reinject-water is unwanted in the deep-water layers
due to dilution of the CH4 resource.

Dilution-water, dilution factor - see Table 2.3.1.

16
However, we know, that above the RZ, the formation of CH4 can be explained by the
settling organic matter only, whereas in the RZ ~40% of the CH4 is produced from
fresh sediment material and ~60% contains “old” carbon. The scientific background
addressing the sources and sinks of CH4 in Lake Kivu are available in: Deuser et al
(1973), Tietze (1978), Tietze et al (1980), Schoell et al (1988), Jannasch (1975) and
Pasche et al (2009, 2011, submit).

The current rates of CH4 new formation and of CH4 accumulation are not
accurately known (Pasche et al submit). Therefore, the harvesting strategy has to be
designed robustly such that the economic viability (such as the maximum CH4
harvest) or important Design Parameters (Table 2.2.1) should not sensitively depend
on these not well-defined Planning Parameters (Table 2.2.1). The present
technically extractable amount (of the CH4 stored below 200 m depth) is ~42.3 km3
(STP). For a typical conversion factor of 2.6 TWh km-3, this amount of CH4 equals an
electrical energy of ~110 TWh. If harvested over 50 yr, this energy corresponds to
an electricity power of ~250 MW. For the model simulations we used an annual CH4
new formation rate of 0.45 km3 yr-1 (= 120 g-C m-2 yr-1) for the volume below the
Biozone (Table 1.1). Thereof ~0.24 km3 yr-1 is produced within the Resource Zone
below 260 m depth. This CH4 new formation is equivalent to a steady-state energy
generation of ~68 MW. This assumed new CH4 formation rate is based on the analy-
sis in Schmid et al (2005a), available at the time of the simulations. However, from
recent investigations of the lake-internal carbon fluxes by Pasche et al (submit), we
conclude that the true rate of new CH4 formation is probably approximately half the
value given above (~0.23 km3 yr-1 for the entire lake; ~0.12 km3 yr-1 for the RZ).

Risk of gas supersaturation - The downside of the gases in Lake Kivu is the
danger which they expose to all oxygen-dependent life in the lake region. A gas
eruption in the lake could lead to an unimaginable disaster of an apocalyptic
dimension, far beyond the events at the comparable “killer lakes” Monoun in 1984
(Sigurdsson et al 1987) and Nyos in 1986 (Kling et al 1987; Sigvaldason 1989)
which claimed 37 and more than 1700 lives, respectively. The gases present in Lake
Kivu could replace more than a 100 m thick air layer if suddenly released from the
lake water (this height is calculated by dividing the total amount of lake gas by the
lake surface area). Although the probability of such a catastrophic chain reaction
leading to a complete degassing of the lake is presently low, the gases need to be
removed from the lake in the future decades. Without CH4 extraction, the gas accu-
mulation - at the current estimated rate of new formation - could reach dangerous
(saturation) levels towards the end of this century (Schmid et al 2005a).
Although CH4 concentrations are only ~21% of CO2, the probability of an
eruption is mainly determined by CH4. This is because of the low solubility of CH4; its
contribution to the total gas pressure exceeds that of CO2 (Tietze 1978; Schmid et al
2004). Consequently a gas release from Lake Kivu would be triggered by super-
saturation of CH4. However, CO2 would thereby be stripped into the newly-generated
gas bubbles and would account for most of the released gas volume. The origin of
the CO2 is mainly from the volcanically active underground, and to a smaller part,
from the degradation of organic matter (carbon in algae).

17
 Vertical fluxes of methane - The CH4 and CO2 remain trapped in Lake Kivu
to an extent which is unique compared to other lakes and ocean basins on Earth.
Why Lake Kivu’s CH4 concentrations reach such high levels is still a poorly ans-
wered question (Pasche et al submit). Besides the high rate of CH4 new formation,
the strong density stratification of the lake is another important reason. As a result of
this stratification, water masses are only very gently (almost not) stirred in the verti-
cal direction. Therefore the upward flux of CH4 is small and it can reside for long in
the deep layers. In addition, the CH4 trapped in the deep anoxic zone, is preserved
and microorganisms can not degrade or remove it from those deep layers. Under
such confined anoxic conditions, CH4 could remain for hundreds or thousands of yr.
However once CH4 has been transported upwards into the oxic (oxygen-containing)
Biozone, then CH4 is oxidized to CO2. This transformation can only take place in the
top ~80 m, as the microorganisms need oxygen (or sulphate or nitrate; here a minor
part) for the oxidation. There are two transport mechanisms by which CH4 moves
upwards:

(1) Deep subaquatic inflows into the lake push water, and all the substances
dissolved therein, upwards. This so called “upwelling” is a slow vertical
upward movement with a velocity determined by the subaquatic water
discharge (up to ~42 m3 s-1 or ~1.3 km3 yr-1, details in Section 4.3 and
Appendix 8.2) divided by the cross sectional area of the lake (< 2000 km2,
decreasing with depth, details in Appendix). The upwelling velocity is varying
with depth, but is in the range of up to 1.0 m yr-1 (Figure 4.3.4). The upward
transport of CH4 (or any other substance) can then be calculated by the
upwelling velocity times the concentration. As an example: the CH4 upward
flux at 260 m depth (= loss from the Resource Zone) is 23 g-C m-2 yr-1 * 1050
km2 = 24’000 t-C yr-1 = 0.045 km3 yr-1 = ~0.1% yr-1 of the CH4 contained in the
Resource Zone (Table 1.1). Therefore we can conclude that CH4 would
remain almost 1000 yr in the deep-water despite the continuous upwelling by
the subaquatic sources.

(2) As water masses in lakes are never at complete rest, there are always some
(that are especially slow in Lake Kivu) currents caused by storms,
atmospheric pressure changes or other external forces. These (slow)
horizontal currents lead to friction between water layers that lie on top of each
other and flow at slightly different velocities. This friction causes some
turbulent vertical mixing, which leads to an upward diffusion from higher to
lower concentration (CH4 is increasing with depth). As an example, this
turbulent upward transport of CH4 in 260 m depth (i.e. the loss from the
Resource Zone) is ~11 g-C m-2 yr-1 * 1050 km2 = 11’000 t-C yr-1 = 0.021 km3
yr-1 = ~0.05% yr-1 of the CH4 contained in the Resource Zone (Table 1.1).
This flux implies that CH4 would reside for ~2000 yr, if turbulent diffusion were
the only process to transport CH4 out of the deep-water layers.

18
 When comparing these two types of upward fluxes, which occur simulta-
neously and independently of each other, it becomes evident that upwelling
(subaquatic sources) is generally more important than the turbulent upward diffusion
below ~100 m depth (in the previous example, the diffusion was about the largest in
the entire lake, as there the gradient is the strongest). The vertical fluxes in Lake
Kivu are unusual compared to other lakes for two reasons: (i) The turbulent mixing in
the deep-water is unusually weak, as highlighted by the presence of about 340 so-
called “double-diffusive” homogeneous layers (forming an unprecedented double-
diffusive staircase; Schmid et al 2010) below 120 m depth, which would quickly be
destroyed if the deep waters were to contain much turbulence. (ii) The subaquatic
springs comprise a significant flow both in absolute terms (~42 m3 s-1 = 1.3 km3 yr-1)
as well as in relative terms of the lake throughflow (~⅓). These subaquatic springs
cause the unusual upwelling of water, CH4 and nutrients.

Vertical fluxes of nutrients - During the decomposition of settling organic

matter (predominantly algae in Lake Kivu), phosphate (PO4) and ammonium (NH4),
assimilated during algae growth, are released back into the deep-water. Over the
decades, the accumulation of nutrients leads to unusually high concentrations,
similar to CH4. The upwelling (subaquatic springs), described above, transports the
nutrients into the Biozone and supports algae growth at a medium intensity (so-
called mesotrophic). In absolute terms, the current estimate of the upward transport
is about 1900 t-P yr-1 of PO4 and about 15’900 t-N yr-1 of NH4. Such a mesotrophic
level of algae growth is usually considered as an indicator for a healthy ecosystem
and a desirable water quality.
If the extraction operation would release nutrient-rich deep-water at or near
the lake surface, the algae growth would increase. The amounts of released deep-
water of a full-scale CH4 extraction would enhance the nutrient fluxes into the
Biozone by factors (!) and these nutrients would stimulate a tremendous algae
growth over the entire lake surface (and not only near the release site). Subsequent-
ly, the oxygen-containing Biozone would shrink in depth, while the oxygen-free
deep-water masses below would expand by the same volume. It is not possible to
predict the impact on such a complex aquatic system in detail, but adverse effects,
such as occasional fish-kills, toxic algae blooms, changes in species composition,
and degradation of the already poor fish biodiversity would be expected. Releasing
large amounts of deep-water (i.e., releasing large amounts of nutrients) into the
Biozone is ultimately against the Guiding Principle of “conserving the lake ecological
integrity” (next Section) and is therefore unacceptable as a planning option.

2.2 Guiding Principles for sustainable methane extraction

Motivation - After the March 2007 stakeholder meeting in Gisenyi (NCEA

2007, [Link] focussing on the monitoring during the upcoming
CH4 extraction, it was realized, that principles, technical requirements and prescrip-

19
tions are needed to responsibly guide the extraction planning and operations of the
different investors. The ministries in charge of the Governments of Rwanda and DR
of the Congo asked a group of experts to develop Management Prescriptions for
the Development of Lake Kivu Gas Resources (MP 2009) in order to achieve safe,
environment-friendly and economic ways of CH4 extraction. These discussions
revealed a set of agreed-on basic principles, which need to be maintained during the
entire CH4 extraction. Some of these principles concern the extraction technology
per se, which has to comply with “safe, environment-friendly and efficient procedures
and operation of the CH4 extraction”. These technology-related principles and
guidelines are part of MP (2009) and are not subject of this report.

Guiding Principles - We adopt the same agreed-on basic principles for our
analysis. These principles, which we call “Guiding Principles” in the following, are
used for the evaluation and inter-comparison of the different extraction Scenarios,
which represents the core of this report. These Guiding Principles are (in priority (i)
to (iii)):

(i) Reduce the probability of a catastrophic gas release from the lake;
(ii) Conserve the ecological integrity of the lake;
(iii) Maximize the benefit from the CH4 resources in the lake.

There are evidently industry-specific professional standards (best practice,

codes of conduct) aimed at preventing harm to people and the environment, which
could be caused by the operation. Foremost, the operation has to avoid any risk of
causing a catastrophic gas release (such as by breaking pipes, etc). This report is
not addressing these issues, which are part of the Impact Assessment and have to
be observed independent of the adopted extraction approach. The aim of this report
is to identify an optimal extraction Scenario, by estimating the implications and the
consequences that those different Scenarios (Table 2.3.2) might have.

Objectives and their measures - In order to perform such an assessment in

a comprehensible and transparent way (Table 2.2.1), we need to define Objectives,
against which the outcome of the different Scenarios can be numerically measured.
For this particular problem, we defined seven Objectives ((a) to (g) in Table 2.2.2):

(a) Keep lake stratification as stable as possible, at least at 50% of natural

(b) Reduce content of the gases in the lake, at least by 25% over the next
century
(c) Increase required activation energy for (and thereby reduce chance of)
initiation of bubble formation
(d) Minimize additional nutrient input to Biozone, at maximum 25% increase
(e) Minimize loss of CH4 (i) to Biozone and (ii) by the extraction operation
(f) Allow new formation of CH4 in deep-water
(g) Avoid dilution of CH4 resources.

20
 In order to measure how well these seven Objectives are fulfilled, we assign
eight quantifiable Attributes to these Objectives (Table 2.2.2). These Attributes are
well-defined, measurable, physical quantities, which can be calculated at any mo-
ment for all Scenario simulations. The definitions of the eight Attributes are summa-
rized in Table 2.2.2 and detailed explanations are provided in Section 6.2. The
quantitative evaluation (comparison, etc) of these Attributes for the various
Scenarios follows in the next Section.

Table 2.2.1 Terms for structured decision-making framework

We use the following terms for the structured decision-making process:

Guiding Principles - the three basic principles to follow while extracting the CH4 from the
lake. These principles concern specifically the “safety of the population”, the “lake
ecological integrity” and the “economic benefit” from the CH4 resource.

Objectives - the goals of the extraction project are expressed in a practical / operational
way, such that the Objectives become easily comprehensible, quantifiable and finally
measurable.

Attribute - the physical term, with absolute units, which is used to measure to which degree
each Scenario fulfils an Objective.

Scenario - there are many alternative options to construct and operate extraction plants.
The Scenarios are those extraction approaches, which we analyzed in this report. A
Scenario is specified by a well-defined set of Design Parameters.

Consequence Table - for each Scenario model simulations are performed. The
Consequence Table is the inter-comparison of the Attributes, which resulted from
the simulations of these Scenarios.

Trade-offs - discussion of the advantages and disadvantages for those Scenarios that fulfil
the Objectives (and consequently also the Guiding Principles).

Design Parameter - the key parameters of an extraction plant defining the extraction
operation, such as withdrawal and reinjection depths, the extraction water flow, etc
(Table 2.3.1).

Planning Parameter - the managerial boundary conditions, such as: location of the plant on
the water surface, number of extraction plants, licensed CH4 harvesting rate, assumed
rate of CH4 new formation.

21
Table 2.2.2 Guiding Principles, Objectives and measurable Attributes

Guiding Objectives Attributes to Objective Physical measure

Principles and unit (1)
(i) Safety (a) Keep lake Energy required for Schmidt stability
stratification as complete mixing (vertically)
[J]
Reduce probability of stable as of the entire stratified lake
catastrophic gas erup- possible water body (higher value =
tion from lake more stable)
(b) Reduce gas CO2 content in lake = Content of CO2
content in lake Integral of CO2 over entire
[km3]
lake volume
CH4 content in lake = Content of CH4
Integral of CH4 over entire
[km3]
lake volume
(c) Increase Minimal energy required for Safety margin
activation initiating bubbles (at the
[J m3]
energy for (and most critical depth) in water
reduce chance column below 200 m depth
of) initiation of (higher value = more stable
bubble formation and safer)

(ii) Lake ecological (d) Minimize Vertical upward flux of P at Vertical upward flux
integrity additional 80 m depth (= input into of P
nutrient input to Biozone) (lower value =
[t-P yr-1]
Conserve ecological Biozone more sustainable, less
integrity of the lake disturbance)
Vertical upward flux of N at Vertical upward flux
80 m depth (= input into of N
Biozone) (lower value =
[t-N yr-1]
more sustainable, less
disturbance)

(iii) Economic (e) Minimize loss Available CH4 = Harve- Sum of already
benefit of CH4 (i) to sted(2) CH4 (= amount of harvested CH4 plus
Biozone and (ii) CH4 extracted from deep- remaining CH4
Maximize benefit from by operation water) plus remaining CH4 content in lake
lake CH4 resources content in lake [km3]
(f) allow accu-
mulation of
newly formed
deep-water CH4
(g) no dilution of CH4 harvest = sum of Sum of harvested
CH4 resources already harvested(2) and and harvestable CH4
still harvestable(3) CH4 [km3]
(1)
Definitions and calculation formula of the Attributes are detailed in Section 6.2.
(2)
Harvested CH4 = integral of CH4 already extracted (to power production) for entire simulated
operation period
(3)
Harvestable CH4 = integral of harvestable CH4 over entire lake volume (harvestable: CH4 > thres-
hold concentration; here arbitrarily assumed as 5 mol m-3 (Figure 1.2).

22
2.3 Modelling and Scenario evaluation procedure

(i) Goal and procedure - In this Section we explain the procedure to identify
the optimal approach for extracting the CH4 from Lake Kivu, while observing the
Guiding Principles formulated in Section 2.2. To achieve this goal, we proceed by
the following steps (order 1 to 6):

(1) definition of the Scenarios

(2) definition of the modelling approach
(3) definition of the Attributes for evaluating the Scenarios
(4) evaluation of the simulations
(5) evaluation of the robustness of selected Scenarios
(6) choice of the optimal Scenario.

(ii) Basic elements of the extraction operation - A large variety of

approaches for CH4 extraction are conceivable. Many of them may not be of interest,
as they would not meet the three Guiding Principles (safe, sustainable and efficient).
Here we focus only on plant designs, which are qualitatively similar to the currently
projected plants, i.e., water is taken to the surface, degassed, and reinjected into the
lake. Other options are conceivable but currently not thought to be economical. We
also do not discuss the technology applied (independent of the option chosen),
which is not relevant to the analyses of the lake system response, but may be
relevant to the economic benefit.
There are five “plant-defining” parameters, in the following called “Design
Parameters”, which are basic elements of all considered Scenarios, and which
determine the outcome of the extraction, both regarding the achievable CH4 harvest
as well as the change of the lake stratification. These five Design Parameters,
defined in Table 2.3.1, comprise:

(1) withdrawal depth and depth range

(2) reinjection depth (of the depleted deep-water) and depth range
(3) dilution-water flow (% of extraction flow) and dilution-water intake depth
(4) removal (%) of CH4
(5) removal (%) of CO2.

Further “plant-defining” parameters are conceivable, but we conclude that these five
Design Parameters are the most relevant because (i) they comprise the key ele-
ments of an extraction operation and because (ii) the Attributes, used for the
Scenarios evaluation, are sensitive towards these parameters.

(iii) Scenario definitions - In the past much analyzing was performed to

establish an extraction concept that would fulfil the core ideas of the Guiding
Principles and the MP (2009). There are three extreme options, which would each
fulfil one of the respective three Guiding Principles (i) to (iii) in Section 2.2:

23
Table 2.3.1 Definitions of the five Design Parameters

The Design Parameters, used for the Scenario analysis, are defined as follows:
Withdrawal depth - CH4-containing water is withdrawn from the Resource Zone below
260 m depth. If water is taken in over a vertical range, then the withdrawal depth
refers to the centre of the intake. The withdrawal range refers to the vertical extent
of the intake. Some Scenarios have two intakes indicated by withdrawal depth 1,
2 and withdrawal range 1, 2 (Table 2.3.2).
Reinjection depth - After stripping CH4 (and other gases) from the extracted deep-water,
the depleted deep-water is returned to the lake. The reinjection depth indicates
the vertical centre, if the release occurs over a vertical reinjection range. The
reinject-water contains also dilution-water (if used for density adjustment; see
below). If the density of the reinjection water is different from the lake water density
at the reinjection depth, then the released water restratifies above (less dense) or
below (denser) the reinjection depth. Some Scenarios have two return outlets,
indicated by reinjection depth 1, 2 and reinjection range 1, 2 (Table 2.3.2).
Dilution-water - If the reinject-water needs to be restratified within a narrow or precise
vertical range, it was usually considered as necessary to add less dense surface
water to the CH4-depleted deep-water in order to adjust the density before release
into the lake. As the density of the dilution-water depends on the withdrawal depth,
the dilution intake depth needs to be specified as well. The dilution factor is the
ratio of dilution-water flow divided by extraction-water flow. The reinject-water is the
sum of the extracted deep-water plus the dilution-water.
Removal of CH4 - During the initial gas stripping and the scrubbing (gas washing) pro-
cesses not all the gases are completely separated. The removal of CH4 gives the
percentage of the CH4 that is removed from the CH4 contained in the deep-water
(typically 70 to 96%). The rest of the CH4 is returned with the reinject-water into
the lake. Depending on the re-stratification depth, the returned CH4 is lost (if rein-
jection is shallow) or can be harvested later (if reinjection is deep).

Removal of CO2 - After the initial gas stripping, most of the CO2 is washed back into the
scrubbing-water (Figure 1.2). However the gas-washing process is not perfect and
some deep-water CO2 will be removed together with the harvested CH4 (via power
production) or released to the atmosphere. The removal of CO2 gives the percen-
tage of the CO2 that has been removed from the CO2 contained in the deep-water
(typically 40 to 90%). The rest of the CO2 is returned with the reinject-water and the
scrubbing-water back into the lake (Figure 2.3.1).

24
Figure 2.3.1 Using self-siphoning, CH4-rich deep-water is extracted from the RZ. After strip-
ping the gases, CH4-depleted deep-water is reinjected into the lake. The gas mixture is
washed using scrubbing-water to redissolve H2S and CO2 back into lake water (see scrub-
bing-water outlet at 40 to 60 m depth). The remaining CH4 is piped to power production on-
[Link] this study, we analyse the effect of releasing the reinject-water into the Intermediate
Zone (Scenarios IZ1 to IZ5), into the Potential Resource Zone (Scenario PR1) and into the
Resource Zone (Scenarios RZ1 to RZ6). The Scenarios are defined in Table 2.3.2 and
detailed in Section 5.

• The approach for maximum risk reduction (i) would be to rapidly extract all
CH4 and to remove all CO2
• The best approach for conserving the lake ecological integrity (ii) would be
using the CH4 from the Resource Zone and reinjecting the extracted deep-
water back into the same Resource Zone, without removing the CO2 in order
not to change the lake stratification. The deep-water would then be depleted
in CH4 but the stratification of all other water properties would remain
unchanged.
• The most economical (iii) approach would be to draw the deep-water entirely
down to the deepest zones and to return the reinject-water into the Biozone in
order to prevent any dilution of the CH4-containing deep-water and to allow
complete removal of all CH4 contained in the deep-water.

It appears obvious that none of these three options fulfils a-priori all three
Guiding Principles and therefore an optimal, feasible and acceptable concept may
consist of a compromise among those three positions. In addition, a viable extraction
Scenario has to also meet management aspects - here called Planning Parameters,
such as extraction location on the lake (accessibility) or rate of CH4-harvesting.

25
 Prior and during the establishment of this report we consulted with interested
engineers and MININFRA and established a list of extraction approaches. We
analysed those options for similarities and differences and condensed them into a
shortlist of extraction “Scenarios”. We expect these Scenarios to be distinct enough
that the sensitivity of the Attributes (of the Objectives) towards those Scenarios can
be estimated. The analysis in this report was performed for the following Scenarios
(summarized in Table 2.3.2 and detailed in Section 6):

O no extraction
IZ1 intake: Lower RZ; reinjection: middle part of IZ; with dilution-water
IZ2 intake: Upper RZ; reinjection: upper part of IZ; less CH4 removal,
with dilution-water
IZ3 intake: Lower RZ; reinjection: upper part of IZ; with dilution-water
IZ4 intake: Lower RZ; reinjection: middle part of IZ: no dilution-water
IZ5 intake: Upper RZ; reinjection: lower part of IZ; less CH4 removal,
no dilution-water
PR1 intake: RZ; reinjection into PRZ
RZ1 intake: Lower RZ; reinjection: Upper RZ
RZ2 two intakes from RZ (Upper and Lower); two reinjections into RZ (Upper
and Lower); withdrawal and reinjection over entire RZ
RZ3 same as RZ2, except for higher gas removal
RZ4 two intakes from RZ (Upper and Lower); two reinjections into RZ (Upper
and Lower); preparation step for RZ5
RZ5 intake: entire Lower RZ; reinjection: Upper RZ
RZ6 first RZ4 until Upper RZ exploited; then RZ5 until completion.

(abbreviations: RZ = Resource Zone; PRZ = Potential Resource Zone; IZ = Intermediate

Zone; BZ = Biozone; see Figure 1.1).

(iv) Modelling approach - The stratification of Lake Kivu is modelled as a

function of (extraction) time by using the software AQUASIM ([Link]
[Link]/). The following natural processes are included (details in Section 4 and
Appendix 8.1):
• inflow, outflow and upwelling (caused by the subaquatic discharges)
• turbulent diffusion (vertical mixing of water masses; slow process)
• CH4 new formation at the lake sediment surface
• CH4 oxidation in the oxygen-containing Biozone
• nutrients (PO4, NH4) assimilation (algae growth) in Biozone
• nutrients redissolution from settling matter at the lake sediment surface
• exchange of CH4 and CO2 with the atmosphere
• lake-internal water dislocation induced by the extraction operation

26
Table 2.3.2 Methane extraction Scenarios O, IZ1 to IZ5 and PR1
Sce- Aim Realization Advantages Disadvantages
nario
no CH4 extraction no vertical water move- natural gases would
O ment in lake, except accumulate up to
upwelling by subaquatic eruption
springs
stacking reinject-water CH4 withdrawal at 475 clear separation of enormous (up to 5-
IZ1 above the PRZ; maxi- m and release at 150 m harvestable water fold) nutrient fluxes
mizing CH4 harvest by (IZ); reinject-water (RZ) and reinject- and enhanced salt
drawdown of entire density adjusted by water (IZ), well flux into BZ; mas-
CH4-containing deep- dilution-water taken at below BZ sive reduction of
water to deepest zone 40 m depth stratification
Extraction close to CH4 withdrawal at 320 clear separation of 60% increased nu-
IZ2 shore; else as m (Upper RZ) and harvestable water trient fluxes into BZ;
Scenario IZ1 reinjection at 90 m (IZ); (RZ) and reinject- loss of ~8 km3 of
reinject-water density water (IZ), well CH4 (relative to IZ4);
partly adjusted by dilu- below BZ poor CH4 extraction;
tion-water from 10 m reduction of strati-
fication (1)
Extraction close to CH4 withdrawal at 350 clear separation of Up to 2.6-fold nu-
IZ3 shore; else as m (Lower RZ) and harvestable water trient fluxes into BZ;
Scenario IZ1 reinjection at 90 m (IZ); (RZ) and loss of ~4 km3 (rela-
reinject water density reinjecting water tive to IZ4) of CH4;
adjusted by dilution- into IZ, well below large reduction of
water from 50 m depth BZ stratification (2)
maximizing CH4 har- CH4 withdrawal at 475 clear separation of doubling of the
IZ4 vest by drawdown of m (RZ) and release at harvestable water nutrient fluxes and
entire CH4-containing 150 m (IZ); reinject- (RZ) and reinject- enhanced salt flux
deep-water to deepest water will sink further, water (IZ) well into the BZ; lake
zone; no dilution-water as density is not adjus- below BZ; high stratification redu-
ted (no dilution-water) CH4 harvest ced to half of natural
Extraction close to CH4 withdrawal at 320 clear separation of 20% increased nu-
IZ5 shore; stacking the re- m (Upper RZ) and harvestable water trient fluxes into BZ;
inject-water directly reinject at 190 m (lower (RZ) and reinject- loss of ~7 km3 (rela-
above PRZ (no IZ); reinject-water sinks, water (IZ), well tive to IZ4) of CH4
dilution-water) as density not adjusted below BZ due to poor extrac-
tion (1)

high CH4 harvest by CH4 withdrawal at 475 RZ remains modification of stra-

PR1 drawdown of entire m (RZ) and release at undiluted; nutrient tification; loss of
CH4-containing deep- 240 m (PRZ); reinject- fluxes to BZ not ~2.5 km3 CH4 (relati-
water to deepest water will sink but not affected; stability ve to most economic
location; avoiding high beyond the Main reduced by only IZ4); PRZ diluted
nutrient flux into BZ Chemocline; expansion ~20%
of the PRZ
(1)
CH4 below 320 m depth not harvested; (2) CH4 below 350 m depth not harvested. We consider this
not as an important disadvantage, as CH4 can still be harvested at a later term.
Abbreviations used: RZ = Resource Zone; PRZ = Potential Resource Zone; IZ = Intermediate
Zone; BZ = Biozone.

27
Table 2.3.2 Methane extraction Scenarios RZ1 to RZ6 (continuation)

Sce- Aim of Scenario Realization Advantages Disadvantages

nario
protecting the PRZ CH4 withdrawal at CH4 can build up loss of ~2.5 km3 CH4
RZ1 and retaining the 475 m (Lower RZ) within PRZ (relative to most
stratification by and reinjection at during 100 yr to economic IZ4); parts of
stacking reinject- 290 m (Upper RZ); harvestable level RZ diluted
water to the Upper reinject-water will
RZ remain in Upper RZ
extraction of CH4 homogeneous CH4 no changes to at end of extraction,
RZ2 from entire RZ withdrawal below the lake eco- lake left with large
RZ3 without changing 275 m (RZ) depth logical integrity; amount of CH4; CH4
stratification of the over entire RZ high flexibility in concentration and
lake harvest rate CH4/CO2 ratio decrea-
sing over time; reduc-
tion in power production
efficiency (Halbwachs
2009)
extraction of CH4 CH4 withdrawal at no changes to accurate depth com-
RZ4 from the entire RZ 310 m (Upper RZ) the lake eco- pliance for reinject-
by stacking and reinjection at logical integrity; water required;
reinject-water into 270 m (Upper RZ); flexibility in accurate adjustment
Upper and Lower CH4withdrawal at harvest rate (with CO2) of reinject-
RZ separately 410 m (Lower RZ) water density required;
and reinjection at parts of RZ diluted
325 m (Lower RZ)
drawdown of CH4 CH4 withdrawal at no changes to accurate depth com-
RZ5 in the entire RZ by 400 m (Lower RZ) lake ecological pliance for reinject-
stacking the and reinjection at integrity; water required;
reinject-water into 270 m (Upper RZ) flexibility in accurate adjustment
Upper RZ harvest rate (with CO2) of reinject-
water density required;
parts of RZ diluted
harvesting the First Scenario RZ4 no changes to accurate depth com-
RZ6 Upper RZ before until CH4 used up in lake ecological pliance for reinject-
diluting it with Upper RZ; thereafter integrity; water required; needs
reinject-water Scenario RZ5 flexibility in operational flexibility:
originating from harvest rate change of withdrawal
the Lower RZ and reinjection depths
(RZ4 to RZ5) after ~25
yr; parts of RZ diluted

Abbreviations: RZ = Resource Zone; PRZ = Potential Resource Zone; IZ = Intermediate Zone;

BZ = Biozone.

28
In order to simulate (i) the stratification, (ii) the gas contents and (iii) the nutrient
fluxes, all relevant quantities need to be included as dynamic model variables. The
simulated quantities comprise the following output (all as a function of depth):
• all density variables: temperature, salinity, CO2 and CH4
• concentrations of dissolved gases: CO2 and CH4
• concentrations of nutrients: PO4 and NH4
• turbulent diffusion and upwelling.

We use the term CO2 for dissolved gaseous CO2, which is less than the total
dissolved inorganic carbon (CO2,total, including carbonic acid, bicarbonate and carbo-
nate; details in Section 4.10 and Appendix 8.3). The latter ions are included in the
salinity. Both contributions (CO2 and ions) are included in the density calculations.
Also lake-internal processes induced by the extraction operation are included
in the model: The two most relevant are the water withdrawal (leading to down-
welling between the withdrawal and the reinjection depths) and the reinjection
(leading to upwelling above the reinjection depth). Beside the water masses, the
constituents in the withdrawal and reinject-water are also accounted for. The Design
Parameters directly influence the properties of the reinject-water and thereby affect
the lake stratification in the long-term, especially the vertical distribution of the gases.
The one-dimensional advection-diffusion-reaction model is run for 100 yr from
2004 to 2104 for all Scenarios (IZ1 to RZ6) listed in Table 2.3.2. Scenario O (no
extraction, no intervention) is used for model calibration as well as a reference for
comparing the Scenarios (see evaluation below). The simulation time frame of 100
yr is adequate for the entire extraction of Lake Kivu. A choice of 50 yr for the same
volume of extracted water would double all fluxes and rates of change, but would not
alter the lake system response or the structure of the results. Therefore, the chosen
time frame has no influence on the evaluation of the outcome.
The lake stratification data (six dynamic model variables and related quanti-
ties) have been evaluated (Table 7.1 to 7.12 and 7.13) after 0 (2004), 50 (2054) and
100 yr (2104). Details of the model construction are given in Section 4.

(v) Definition of Attributes measuring the Objectives - The planning of the

large-scale CH4 extraction unfolds a list of questions concerning the effects that the
operation may have on Lake Kivu and the risks related to its gases. Obvious
questions on any of the different extraction Scenarios are:

o How does the long-term stratification of Lake Kivu develop?

o How does the long-term risk for gas outburst develop?
o Are there critical layers with a high probability for gas outburst?
o How large are the changes of the lake-internal nutrient fluxes in comparison
to the undisturbed (natural) fluxes?
o What is the maximal total amount of CH4 to be possibly harvested?
o What amount of CH4 is “lost” (by oxidation) as a result of the extraction in
addition to the naturally occurring oxidation?
o How does the vertical distribution of the CH4 change during extraction?

29
The model simulations provide answers to such questions. To perform the trade-offs
between the different Scenarios quantitatively and comprehensively, we specified
eight Attributes, which are well-defined physical quantities (Table 2.2.2; Section 6.2).
These Attributes basically measure to which degree the Objectives are fulfilled. The
Attributes comprise:

• Schmidt stability = the mechanical energy required for complete (vertical)

mixing of the entire stratified lake water body. Unit: [J]
• Safety margin = the minimal mechanical energy required to initiate bubbles (at
the most critical depth) in the water column below 200 m depth. Unit: [J m-3]
• Phosphorus upward flux = vertical upward flux of P into the Biozone (at
80 m depth). Unit: [t-P yr-1]
• Nitrogen upward flux = vertical upward flux of N into the Biozone (at 80 m
depth). Unit: [t-N yr-1]
• CO2 content = CO2 content in entire lake. Unit: [km3]
• CH4 content = CH4 content in entire lake. Unit: [km3]
• CH4 harvested = CH4 led to production since start of extraction. Unit: [km3]
• CH4 harvestable = total amount of CH4 left in the lake at concentrations
exceeding the economically viable (harvestable) concentration, presently
assumed as 5 mol m-3. Unit: [km3].

All eight Attributes (Table 2.2.2; Section 6.2) have a value at any given point
in time for any of the Scenario simulations. As indicated above, the evaluation was
carried out for 2004 (0 yr), 2054 (50 yr) and 2104 (100 yr).
To potentially limit the complexity of the optimisation by reducing the number
of Attributes, we checked for interdependencies among the Attributes and found that
the Schmidt stability, the safety margin and the CO2 (or total gas) content are corre-
lated. This implies that one of these three Attributes alone is enough to characterize
the Guiding Principle “Safety”. Also the PO4 and NH4 upward fluxes are strongly cor-
related, and consequently one of those two Attributes already contains the full infor-
mation related to the Guiding Principle “Lake ecological integrity”. Therefore we base
the final evaluation on a reduced set of the following four independent Attributes:

• Schmidt Stability
• P flux into Biozone
• Sum of “CH4 harvested” (production) plus “CH4 content” (remaining in lake)
• CH4 harvest = sum of “CH4 harvested” plus “CH4 harvestable”.

In the next Section, we apply these four Attributes to all model simulations and list
the results in the Consequence Table (2.4.1). Subsequently, this table is used as a
base for determining the optimal Scenario.

30
2.4 Evaluation of the extraction Scenario simulations

The eight Attributes for all Scenarios simulations (Section 7) are summarized
for 0, 50 and 100 yr of extraction in Table 7.13. An excerpt for the essential four
Attributes is provided In Table 2.4.1, which allows a first triage to be made, based on
how well the Scenarios meet the Objectives. The comparison of the Scenarios O to
RZ6 in Table 2.4.1 reveals the following findings:

(i) Safety (Guiding Principle (i)): The largest modifications of the lake stratifi-
cation occur when reinject-water is released into the Intermediate Zone (Scenarios
IZ1, IZ2, IZ3, IZ4; Table 2.4.1). Reinjection into the Potential Resource Zone (or the
lower reaches of the Intermediate Zone) reduces the stratification much less (Scena-
rios IZ5, PR1). In contrast, the changes to the stratification are small, as long as wa-
ter is reinjected into the Resource Zone (Scenarios RZ1 to RZ6). In these Scenarios,
CH4 is allowed to further accumulate in the PRZ and the IZ, which implies that a
solution for degassing these zones may be required at a later stage.

(ii) Lake ecological integrity (Guiding Principle (ii)): For this Objective, the
distinction between Intermediate Zone reinjection and deep reinjection is even more
pronounced than for the Guiding Principle “Safety”. Whereas the Potential Resource
Zone and Resource Zone reinjection (PR1 and RZ1 to RZ6) cause almost no
change in the (natural) nutrient upward fluxes, the Intermediate Zone reinjections
(IZ1 to IZ5; Table 2.4.1) cause a significant unwanted enhancement of the fluxes.
The addition of dilution-water - if dilution-water is taken from the Biozone –
enhances immediately (with the onset of the extraction) the nutrient upward fluxes
into the Biozone (Scenarios IZ1, IZ2, IZ3; Table 7.3.1). The dilution-water withdrawal
from the Intermediate Zone enhances the nutrient fluxes eventually as well, but only
after several decades and much more gradually.
In several hundred years the nutrient upward fluxes will gradually increase
independently of the extraction and of the Scenario chosen. This increase is
unavoidable given the enormous reservoir of nutrients in the deep-water.

(iii) Economic benefit (Guiding Principle (iii)): The maximum CH4 harvest
would be achieved, if the deep-water would be reinjected at the surface. This option
is however not acceptable. Of the analyzed Scenarios, the maximum CH4 harvest
would occur for the Intermediate Zone reinjection (Scenario IZ4, Table 2.4.1).
Although this Scenario IZ4 fails on grounds of the other two Guiding Principles
(Table 2.4.1), it serves as an ideal reference for the maximum CH4 harvest (set to
100% for the comparison in Table 2.4.1).
Fortunately, for the concession and planning process, the available CH4 (sum
of CH4 let to power production and CH4 still in the lake) varies only within 1% among
the Scenarios (PR1 and RZ1 to RZ6), as there is almost no additional loss of CH4 to
the Biozone. There are however significant differences concerning the CH4 harvest
among the Scenarios (Tables 2.4.1 and 2.4.2). The reason is related to the different

31
Table 2.4.1 Consequence Table relating Attributes of all Scenarios (1).
Sce- Objective Objective Objective Objective Assessment
nario “Safety” “Lake “Economic “Economic
ecology” benefit” benefit”
Schmidt Nutrient Available Methane
stability fluxes methane harvest (2)
O stability intact, natural no benefit no benefit no option
but gas accumu-
lating to dange-
rous level
IZ1 only 19% of natu- 4.9-fold of 90 % of 97 % of failed on Obj.
ral after 100 yr, natural after maximum maximum “Safety” and on
lake homogeni- 50 yr Obj “Lake
zed, salty surface ecology”
IZ2 61% of natural 1.6-fold of 92 % of 84 % of failed on Obj.
after 100 yr natural after maximum maximum “Lake ecology”
50 yr and poor on
Obj. ”Economic
benefit”
IZ3 48% of max. after 2.6-fold of 96 % of 89 % of failed on Obj.
100 yr natural after maximum maximum “Lake ecology”
50 yr and poor on
Obj. “Safety”
IZ4 48% of max. after 2.0-fold of 100 % of 100 % of failed on Obj.
100 yr, natural after maximum maximum “Lake ecology”
50 yr and poor on
Obj. “Safety”
IZ5 82% of natural 1.2-fold of natu- 93 % of 83 % of weak on all
after 100 yr ral after 100 yr maximum maximum three Obj.

PR1 81% of natural natural 97 % of 92 % of option

after 100 yr maximum maximum

RZ1 100% of natural natural 97 % of 90 % of option

after 100 yr maximum maximum
RZ2 100% of natural natural 98 % of 96 % of option
after 100 yr maximum maximum
RZ3 87% of natural natural 98 % of 96 % of option
after 100 yr maximum maximum
RZ4 91% of natural natural 98 % of 88 % of option
after 100 yr maximum maximum
RZ5 85% of natural natural 98 % of 89 % of option
after 100 yr maximum maximum
RZ6 87% of natural natural 98 % of 88 % of option
after 100 yr maximum maximum
(1)
Dark shading indicates failure fulfilling the Objective, intermediate shading indicates weak
performance compared to best scenarios.
(2)
Methane harvest = CH4 harvested plus CH4 harvestable (Table 2.2.2) relative to maximum of
Scenario IZ4 (100%).

32
vertical distributions of CH4 (Figures 7.1.5 to 7.12.5) that result for the different
Scenarios and the definition of the level at which the remaining CH4 is still
economically harvestable. The advantages and disadvantages of those Scenarios
are evaluated in Table 2.4.2 and in Section 2.6.
In addition to the Attributes of the “economic benefit” (available methane,
methane harvest, Table 2.4.1) other factors will be equally important for the overall
economic benefit of the CH4 harvesting process. Such factors will include the
temporal development of the CH4 concentration and the CH4/CO2 ratio at the
extraction depth, the percentage of energy consumption for the plant operation, the
conversion efficiency of CH4 to electrical energy, or the required distance from the
shore to reach a certain extraction depth.

2.5 Evaluation of the robustness of selected Scenarios

(i) Role of model assumptions - It is important that the optimal (chosen)

Scenario is not only fulfilling the Guiding Principles and the associated Objectives. In
addition the optimal Scenario should also be robust towards the critical uncertainties
in the model assumptions. The model simulations and the Consequence Table
(2.4.1) are most sensitive to three model assumptions:

• upwelling flow (rate of uplift, caused by the subaquatic springs),

• CH4 new formation rate, and
• CH4 harvesting rate.

The first two assumptions are not well-known and the third has not been
decided yet. For the evaluation of potential influences, robustness tests have been
performed. Therefore, we varied the assumptions for the simulations as follows:
- The present in-situ CH4 new formation rate (assumed: 120 g-C m-2 yr-1 is
equivalent to 0.29 km3 yr-1 within the Resource Zone + Potential Resource Zone;
Table 1.1), was varied from 32 to 120 g-C m-2 yr-1.
- The subaquatic water inflows (sources at several depths with estimated total
flow of 42 m3 s-1 (Table 8.2.2), corresponding to a vertical upwelling advection of 0.7
m yr-1 in 80 m depth) were varied from 10.5 m3 s-1 to 42 m3 s-1.
- A constant water flow rate of 70 m3 s-1 (corresponding to an approximate
energy equivalent of ~100 to ~200 MW over 100 yr) was used in all Scenarios. For
the robustness test, we further assumed a faster extraction rate of ~600 MW
equivalence, which is assumed to be reached after 15 yr of gradual increase (and
constant extraction flow afterwards).

(ii) Results of the robustness tests - The simulation runs, with the varying
assumptions as listed above, show that the structure of the results (Table 2.4.1 and
2.4.2) are not affected by these assumptions. Therefore, we saw no need to gene-
rate further Consequence Tables (2.4.1) for other combinations of assumptions.

33
Table 2.4.2 Advantages/disadvantages for Scenarios fulfilling the Objectives
Sce- Methane Advantages Disadvantages Prefe-
nario harvest (1) rence

PR1 92 % of o simple concept of extraction o stability decreased to 81% of medium

maximum and release natural after 100 yr (still very
o high CH4 harvesting and stable)
energy production rate o most plants close together (at
end of extraction process) near
deepest location
o reinjected CH4 is lost
RZ1 90 % of o simple concept of extraction o most plants close together (at medium
maximum and release end of extraction process) near
deepest location
o CO2 in lake reduced by only
~15% of today
o reinjected CH4 is partly lost
RZ2 96 % of o no loss of CH4 by incomplete o more complex extraction and re- low
maximum extraction (reinjected CH4 is lease; potential short-circuiting
reused) between extraction and
reinjection
o CH4 conc. and CH4/CO2 ratio
decrease continuously with time
o CO2 in lake reduced by only 10
% of today
RZ3 96 % of o no loss of CH4 by incomplete o more complex extraction and re- low
maximum extraction (reinjected CH4 is lease; potential short-circuiting
reused) between extraction and
o CO2 in lake reduced by 30 % reinjection
of today o CH4 conc. and CH4/CO2 ratio at
extraction depth decrease
continuously with time
RZ4 (2) 88 % of o no loss of CH4 by incomplete o more complex extraction and re- not a
maximum extraction (reinjected CH4 is lease Scenario
partly reused) o CH4 conc. and CH4/CO2 ratio at (with RZ5
extraction depth decreasing only)
drastically with time
RZ5 89 % of o simple concept of extraction o most plants close together (at medium
maximum and release end of extraction process) near
o reinjected CH4 is only partly deepest location
lost o CH4 conc. and CH4/CO2 ratio
o high energy production rate decrease abruptly with time
o CO2 in lake reduced by 40%
of today
RZ6 88 % of o no loss of CH4 by incomplete o more complex extraction and re- high
maximum extraction (reinjected CH4 is lease
reused) o most plants close together (at
o high energy production rate end of ext process) near deepest
o CO2 in lake reduced by 35% location
of today o CH4 conc. and CH4/CO2 ratio
decrease continuously with time
(1)
Methane harvest = CH4 harvested plus CH4 harvestable (Table 2.2.2) relative to maximum of
Scenario IZ4 (100%, Table 2.4.1).

34
2.6 Conclusions and optimal extraction Scenario

(i) Trade-offs - Table 2.4.2 summarizes the advantages and disadvantages

of the Scenarios which fulfil the three Guiding Principles. We argue that the optimal
Scenario should not only fulfil the Attributes towards the Objectives, and robustly
resist against (i.e. not sensitively depend on) uncertainties in the model assumptions,
but the optimal Scenario should also have technological and managerial advantages.

PR1 and RZ1 - The advantages of PR1 and RZ1 are the simple concept and
the high CH4 harvest. The main disadvantages for PR1 are (i) the high density at the
reinjection depth which leads to forming a thick layer above the main gradient with a
relatively low CH4 concentration (Figure 7.6.5) and (ii) the loss of CH4 in the re-
inject-water (re-inject-water will not be drawn-down in the next century). This loss is
somewhat reduced for RZ1, especially if after a few years (decades) the CH4 in the
re-inject-water will hopefully be completely eliminated. The performance of RZ1 can
be improved relative to the modelled Scenario, if the density adjustment of the re-
inject-water is optimized by the fraction of CO2 removed. The control of CO2 removal
in an extraction process depends on being able to control or vary the pressure of
separation. This can in turn vary the CO2 extraction from below 20% to above 80%.
Hence the re-inject water density can be varied enough to control its density in order
to match lake water density at its preferred reinjection depth. This separation pres-
sure also affects the CH4 extraction efficiency and the energy production rates.

RZ2 and RZ3 - The difference between RZ2 and RZ3 is only the fraction of
the gases removed, which may be of interest for an optimized extraction operation.
The main advantage of these Scenarios is the entrainment of the re-injected CH4
into the Resource Zone, implying that this recycled CH4 is not lost for future
extraction. These Scenarios have two major disadvantages: (i) The extraction and
reinjection is distributed over the entire RZ, which could lead to short-circuiting
between release and intake flow. In principle this risk could be avoided, by a more
elaborate vertical structuring of the intake and reinjection. (ii) The CH4 concentration
in the extracted water is continuously decreasing (if CH4 extraction is higher than the
formation of new CH4) and will eventually reach a level of non-viable low CH4
concentrations. To avoid the large loss of the remaining CH4, the extraction needs to
switch to a Scenario of the RZ1 type.

The implications are twofold: (1) These RZ2/RZ3 Scenarios would be an

opportunity to create a long-term steady-state CH4 extraction, in balance with the
new formation rate, and retaining a long-term gas content at all levels in the lake.
However, such a goal conflicts with the objective to harvest most of the CH4 in the
near future and to reduce the overall gas concentrations for improved safety. (2) If
CH4 will substantially be used in the near future, then RZ2/RZ3 would would only be
a short-term (a few years) option in order not to decrease the concentration below a
critical level. For these two reasons, we can not recommend these two Scenarios as
an extraction option.

35
 RZ6 - The goal of RZ6 is to minimize the disadvantages of all the RZ injection
Scenarios. By extracting the Upper and Lower RZ separately (RZ4), the density
difference within the RZ is maintained and potential CH4 loss by the reinject-water in
the first few years (decades) is avoided. The density stratification will reduce the risk
of short-circuiting. During the second part of the extraction (RZ5) it is necessary to
adjust the density of the reinject-water carefully (by removing increasingly more CO2)
such that the reinjected water restratifies at the top of the RZ and does not cause a
large mixed layer with low CH4-content (as in RZ1), which would be lost for further
extraction. As for RZ1, the performance of RZ6 can be improved relative to the
modelled Scenario, if the density adjustment of the reinject-water is optimized during
the second part of the extraction operation. However, retaining reasonably high CO2
levels is also a goal (in order to facilitate siphoning for future extraction). Therefore it
is imperative that the extraction plants are designed in such a way to allow for the
flexible removal of CO2 for full control of the reinject-water density.

(ii) Conclusions - Several important conclusions from the overall comparison

(summary in Table 7.13 and Consequence Table 2.4.1) as well as from Table 2.4.2
can be drawn:

(a) It is possible to extract the CH4 in a responsible manner complying with

safety precautions and maintaining lake ecological integrity. Concerning safety and
lake ecological integrity, the impacts of deep-reinjection Scenarios (PR1, RZ1 to
RZ6) are very similar.

(b) Using dilution-water shows no advantages over Scenarios which operate

without dilution-water. As dilution-water increases nutrient upward fluxes and as
dilution-water can - under unfavourable conditions - dilute CH4 resources, there is no
compelling or acceptable reason for using dilution-water.

(c) If the CH4 extraction efficiency is lower than planned (i.e the CH4 content in
the reinject-water is too high), it becomes a significant advantage to release the
reinject-water into the Resource Zone. Returned CH4 remains then a (potential)
resource and can be harvested later. However, CH4 released into the Intermediate
Zone can not be extracted in the future and is therefore lost (Scenarios IZ1 to IZ5).

(d) Compared to the most economic Scenario (IZ4), the RZ reinjection

Scenarios reduce the available CH4 by only ~2%. The differences of available CH4
among the Scenarios are particularly small (~1% only).

(e) However, there is substantial difference concerning the CH4 harvest,

which reflects the different vertical distribution of CH4 left behind in the lake after 100
yr of extraction. For the definition of CH4 harvest we have chosen a realistic, but
arbitrary CH4 threshold concentration of 5 mol m-3. Although, we can assume that
improved technology and experience will lower the threshold for economic CH4
harvesting below 5 mol m-3, the lowest economically extractable CH4 concentration

36
expected in the future remains uncertain. Therefore it is sensible to remain conser-
vative and opt for a Scenario that maintains a favourable vertical distribution of CH4
to the very end of the extraction process.

Compared to the most economic Scenario (IZ4), the CH4 harvest is reduced
in the worst case by up to ~12% (PR1, RZ1 and RZ6). However, with an optimized
density management (by flexible removal of CO2) in the second half of the extraction,
the CH4 harvest can be increased. If the density gradient above the Resource Zone
(at 260 m depth) is maintained during the extraction, the Potential Resource Zone
will most probably become harvestable in a few decades. Under those two optimal
conditions, the loss of CH4 harvest (relative to IZ4) will be significantly less than
10%.

From these considerations and the comparison of the advantages and

disadvantages in Table 2.4.2, we concluded the following ranking of the Scenarios:

- Best - and recommended - Scenario: RZ6

- Second - but still acceptable - Scenarios: PR1 or RZ1
- the other Scenarios are not recommended.

The best and recommended Scenario corresponds to Plan A1/A2 and Plan B in MP
(2009). We, however, recommend studying the details of a flexible removal of CO2
and optimized vertical stacking of the reinjection water in more detail.

37
38
PART II: Technical Documentation

3. Lake data used for model simulations

3.1 CTD Profiles

In 2004 a set of 13 vertical profiles of temperature and conductivity (CTD)
were acquired in the northern half of Lake Kivu with a Sea-Bird SBE-19 which was
equipped with a Sea-Bird SBE-22B combined pH and oxygen sensor. With the
exception of some cool intrusions observed at 180 and 250 m depth (see below;
Schmid 2010), temperature and salinity were horizontally homogeneous in the
deep-water. The four deepest reaching profiles were averaged to produce the
temperature and salinity profiles used in the model. These temperature and salinity
profiles are listed in Table 8.8 of the Appendix.

3.2 Water chemistry

Water samples were analyzed with standard methods for alkalinity (Alk),
nutrient contents and main ionic composition. Salinity (S) was calculated from
conductivity and ionic composition, according to Wüest et al (1996), and agreed
within 2% with the values calculated by adding the measured ionic content. Historic
conductivity and salinity values are inconsistent; the levels measured by Tietze
(1978) in the deep-water are about 10% higher than our values, whereas those of
Degens et al (1973) are lower by almost a factor of two. However, recalculating
conductivity from the ionic composition given by Degens et al (1973) yields values
comparable to ours which indicates that their salinity calculations were incorrect.
CO2 was calculated from pH and alkalinity including the effects of temperature and
salinity and the first dissociation constant of H2CO3 (Cai and Wang 1998). The
uncertainty in this calculation is mainly due to pH, since an error in pH of ±0.05
units produces an error in the CO2 concentrations of approximately ±10%. The
equations used in the model are listed in Appendix 8.3.
CH4 was measured with a Capsum Mets CH4 sensor, which was calibrated
for the high concentrations in Lake Kivu. The standard deviation observed by the
manufacturer in the calibration was 0.4 mol m-3. In the deep-water of Lake Kivu
this corresponds to an error of 2 to 5%. The sensor recorded the CH4 concentra-
tion every 0.5 s. In the field, it exhibited an unexpectedly slow response time of
46 min. The measured concentrations were extrapolated by fitting time series of
200 to 600 consecutive samples to an exponential curve. The error due to the
fitting procedure was 0.5% (below the main density gradient at 260 m depth), 2 to
3% (within the main density gradient), and up to 10% above 230 m depth. It was
larger at lower concentrations due to the higher relative noise of the instrument.
However, we lack a proper error analysis for the sensor and applied method.

39
 -1
Salinity (g L )
0 1 2 3 4 5 6
0
50 temperature
salinity
100
CO 2 concentration
150 CH 4 concentration
200
Depth (m)

250
300
350
400
450
500
22 23 24 25 26 27
Temperature (ºC)
0.00 0.02 0.04 0.06 0.08 0.10
-1
Gas concentrations (mol L )

Figure 3.2 Vertical profiles of temperature (T), salinity (S), CH4 and CO2, as observed in
February 2004 (Schmid et al 2005a,b). The CH4 profile used in this study (dashed, set
proportional to the CO2 concentration) slightly deviates from the observed data (open
circles and triangles). Data in Table 8.8 (Appendix).

In November 2003, CH4 and CO2 concentrations were measured by Michel

Halbwachs by transferring water through a polyethylene tube to the lake surface,
separating the gas and the water phases, measuring their flow rates, and determi-
ning the CH4 and CO2 concentrations in the gas phase with a portable GA2000
infrared gas analyzer. The gas analyzer was verified by measuring lake gas
samples with another identical instrument before and after the expedition with a
standard of 40% CH4 and 60% CO2. Errors were in the range of ±0.2% (volume),
i.e., less than 2% of CH4 concentrations observed in the lake. The measurement
error for the Schlumberger gas flowmeter is about ±2%, while the error of the water
flow can be neglected. Consequently, a total error of ±4% is assumed. Measured
ratios of CH4 to CO2 concentrations ranged between 0.196 and 0.224. A conti-
nuous vertical CH4 profile (Figure 1.1) was constructed from the data by assuming
that CH4 concentrations were 20.9% of CO2 concentrations everywhere in the
water column. M. Halbwachs (pers comm) remeasured the concentrations in 2009
with this “tube-method” and found values consistent within a few %.

40
 Compared to previous measurements in the 1970s (Tietze 1978), the
measured CH4 concentrations in the deep-water have increased by 15 to 20% and
CO2 concentrations by about 10%. The increase in the CH4 concentrations is
clearly significant if we exclude a systematic error of much more than 5% relative
to the measurements of Tietze (1978), while the observed 10% increase in CO2 is
hardly significant. Profiles of CO2 and CH4 are listed in Table 8.8 (Appendix).
The concentrations of the two most abundant nutrients, phosphate (PO4)
and ammonium (NH4), were measured by means of a photometric analysis of
water samples taken at 20 m vertical intervals with a Niskin bottle (Pasche 2009).
Profiles of phosphate and ammonium are listed in Table 8.8 (Appendix).

41
42
4. Model description and model assumptions

4.1 Introduction
The goal of the modelling is to describe the development of the Lake Kivu
stratification under the effects of the extraction operation for a period of 100 yr. The
model provides the development of all variables relevant for the density stratifica-
tion (T,S, gases), the gas content (CO2 and CH4) and the nutrient fluxes (PO4 and
NH4). For any point in time, the eight Attributes (Table 2.2.2) can be calculated and
these Attributes allow determining to which degree the Guiding Principles (Table
2.2.2) are fulfilled. The outcome of all 12 Scenarios (Table 2.3.2) can then be com-
pared (Table 2.4.1) and assessed relative to the formulated Objectives of the
extraction project.
A one-dimensional reactive-diffusive-advective model was set up with the
lake module of the software AQUASIM 2.1 (Reichert 1994, [Link]
[Link]/). Simulations were performed with a 1-m vertical grid and with variable
time stepping chosen by the integration algorithm. In this Section 4 the processes
included in the model are explained and formulated. All equations used in the
model are listed in Sections 8.3 to 8.5 (Appendix). The model parameters are
given in Appendix 8.2. The model contains the following natural processes:

• surface inflow, upwelling (caused by the subaquatic sources inflow), and

outflow
• turbulent and molecular diffusion (vertical mixing of water masses)
• gas exchange for CH4 and CO2 between lake surface and atmosphere
• CH4 new formation and CO2 release at the lake sediment
• CH4 oxidation in the oxygen-containing Biozone
• algae growth in the Biozone: assimilation of nutrients (PO4, NH4)
• Sediment area-proportional release of nutrients by mineralisation of
settling organic matter.

In addition, the model includes the processes induced by the extraction

operation, such as lake-internal water dislocation (upwelling, dowelling and chan-
ges of concentrations at the reinjection depth). This modelling approach includes
all the processes relevant for the evaluation of the Scenario simulations (Section 5)
and we are convinced that no relevant process is neglected. Despite this
comprehensive approach, the model has obviously uncertainties and limitations,
comprising:
• uncertain CH4 new formation
• uncertain discharges and varying concentrations of deep-water sources
• uncertain feedback between changing stratification and input depth of
deep-water sources
• assumption of long-term steady state, although lake is not at steady-state
• one-dimensional model (not considering local effects at reinjection points)

43
 • no feedback (model coupling) between nutrient input, primary production
and CH4 new formation
• dilution-factors are not continuously adjusted for the changing density
• uncertain feedback between changed stratification and diffusivity
• potential feedback between decreasing CH4 concentrations due to
extraction and CH4 release from the sediments is neglected
• nitrogen fixation by cyanobacteria from the atmosphere is ignored
• removal of NH4 by nitrification and denitrification is not included.

4.2 Surface inflow

The surface inflow to Lake Kivu consists of river flow from 127 small rivers,
contributing 2.4 km3 yr-1 (75 m3 s-1), and precipitation (1400 mm yr-1) onto the
2370 km2 large surface, yielding 3.3 km3 yr-1 (Table 4.4; Muvundja et al 2009). In
the model, the net surface inflow is considered, which reduces the above water
flow by the evaporation from the lake surface, which is about 3.4 km3 yr-1.
For the model simulations, the input of phosphate (PO4), ammonium (NH4)
and salinity (S) by the rivers and the atmospheric deposition is considered (Table
4.4). The inclusion of these dynamic variables in the model is described in the
following Sections below.

4.3 Subaquatic inflow

The net surface inflow to Lake Kivu, estimated to ~2.3 km3 yr-1 in the pre-
vious Section, is smaller than the Ruzizi River outflow of ~3.6 km3 yr-1 (Degens et
al 1973; Muvundja et al 2009), calling for an additional source of inflow into the
lake of ~1.3 km3 yr-1. Also the fact that the salt loss by the Ruzizi outflow (Table 4.4)
is larger than the salt inflow by the 127 rivers indicates a more salty additional input.
From the previous model calculations (Schmid et al 2005a) we can conclude that
several subaquatic sources, at several depth levels, contribute to this inflow. The
water input by this subaquatic source displaces the lake water layers causing a
slow upwelling flow.
Also the vertical structure of various water constituents calls for subaquatic
source at several depth ranges. For example, the persistent strong temperature
gradient at 260 m depth (Figure 4.7.2) with a step of 0.7 °C within 7 m, calls for an
active process to sustain this sharp interface. Otherwise it would be destroyed by
molecular heat diffusion within a few years. Also horizontal structures in that depth
range provide evidence of such sources. A distinct negative temperature signal
(and a weak salinity structure) was observed at 250 m depth which was strongest
in the profiles measured in the North-Eastern part of the lake. Similar weaker
signals were also observed at 180 m depth. Such signals can be explained by

44
inputs of cooler and less saline water. These inflows are most probably sustaining
the steep density gradients observed at several depths in the profiles. Especially,
the main gradient at 255 to 262 m depth is maintained by the inflow at ~250 m
depth. Furthermore, the observed decrease in CH4 is due to dilution by subaquatic
sources, although some minor oxidation (by oxygen, sulphate or nitrate contained
in the subaquatic inflows) can not be excluded.
The observation that the gradient at 260 m depth got stronger since the
work of Tietze (1978) and Newman (1976), indicates that some subaquatic flows
may have become stronger since the 1970s (Schmid et al 2010).
The flow rates, temperatures and salt concentrations of these inputs as well
as the geothermal heat flux were fitted to reproduce the observed profiles in a time
frame of 990 yr prior to 2004. Figure 4.3.3 shows the upwelling flow resulting from
deep-water inputs and Figure 4.3.4 shows the corresponding upwelling velocity.
The depth ranges and the discharges of the subaquatic sources, which fit best the
model results, are listed in Table 8.2.2 (Appendix).
In the next step, concentrations of CH4, CO2, phosphate (PO4), salinity (S)
and ammonium (NH4) in the deep inflows were estimated (Table 8.2.2, Appendix)
such that their deep-water concentrations in the lake remained constant within 30
yr of simulation with the historic CH4 new formation rate (see below) and initial
conditions from 2004. The model assumptions are listed in Table 8.2.1 (Appendix).
The assumption is that Lake Kivu was close to steady-state until 1944 and
gas concentrations increased since then. Setting the measured concentrations
from 2004 as initial values (instead of unavailable data from 1944 with lower values)
probably leads to a slight overestimation of the concentrations in the sources. This
effect is of minor importance for the comparison of different Scenarios but for the
Scenario O, without gas extraction, the total amount of CH4, CO2, PO4 and NH4
may be overestimated in the long term. It is presently impossible to measure the
estimated properties of the subaquatic water inputs mentioned above.
It is well possible that the intrusion depths of the subaquatic sources will
alter if the vertical density stratification changes in the future. In the model those
depths are kept constant, since the AQUASIM code does not allow programming
the inflows as a function of the vertical density profile. Furthermore, we do not
really know whether and how these depths would change. Therefore all Scenarios
which lead to a major change of the density stratification contain an uncertainty of
the effective intrusion depth of the subaquatic sources.

4.4 Surface outflow

The Ruzizi River is the only surface outflow from Lake Kivu with an average
flow of ~3.6 km3 yr-1 (114 m3 s-1; Muvundja et al 2009). Together with the evapo-
ration of ~3.4 km3 yr-1, the total water input and output balances at ±7.0 km3 yr-1.
For water constituents affected by a surface exchange process with the
atmosphere (such as temperature and gases) the Ruzizi outflow is not relevant.
Therefore the outflow of heat, CH4 and CO2 can be neglected.

45
 Although nutrients leave the lake via Ruzizi, their outflow is small compared
to sedimentation (Pasche et al 2009; Muvundja et al 2009) and therefore nutrient
outflow - although included - is not relevant for the model output. As salinity has
only a small sedimentation component the Ruzizi River outflow is the main sink for
salinity.
The outflow of salt (salinity), nutrients (P and N components) and CO2 are
listed in Table 4.4.

Table 4.4 Mass inflows (surface and subaquatic) and outflows (Ruzizi River)
Depth Water Salt flux (2) Phosphate Nitrogen CO2,total (5)
flow flux (3) flux (4)
[m3 s-1] (1) [106 t yr-1] [t-P yr-1] [t-N yr-1] [kt-C yr-1]
Inputs
at surface 75 0.24 230 (1) 5400 (1) ~60
at 180 m depth 22 1.46 370 5100 15.3
at 250 m depth 15 1.28 450 4300 22.0
at 310 m depth 1 0.11 0 620 2.2
at 365 m depth 1.5 0.26 20 930 12.5
at 425 m depth 0.8 0.15 0 0 7.6
at 465 m depth 1.25 0.24 150 1660 15.0
Degradation sed 284 (7)
Output
Ruzizi River -114 -3.89 -70 (6) -300 (6) -505
Net sediment - ? -280 1900 ?
To atmosphere -3 - - not included
(1)
Water balance according Muvundja et al (2009): rivers = 2.4 km3 yr-1; Ruzizi outflow = 3.6 km3 yr-1
(2)
Salinity values of inflows are: 0.1, 2.1, 2.7, 3.4, 5.5, 5.8 and 6, respectively (see Table 8.2.2 Appendix)
(3)
Phosphate values of inflows are: 0.003, 0.017, 0.032, 0, 0.012, 0, 0.12 mol m-3
(4)
Dissolved nitrogen values of inflows are: 0.043, 0.52, 0.65, 1.4, 1.4, 0, 3.0 mol m-3
(5)
CO2,total (definition in Section 4.10) enters as any of the carbonate components (Table 8.2.2, Appendix). In
the lake water, this CO2,total forms carbonate species according to the pH and alkalinity in the lake.
(5)
Assumption on the CO2,total release by organic matter degradation: 705 mol s-1 .

46
 Area per depth Area gradient
0 0

50 50

100 100

150 150

200 200
Depth (m)

Depth (m)
250 250

300 300

350 350

400 400

450 450

500 500
0 500 1000 1500 2000 2500 -8.E+06 -6.E+06 -4.E+06 -2.E+06 0.E+00
2 2 -1
A(z) (km ) dA / dz (m m = m)

Figure 4.3.1 Cross sectional area A(z) of Figure 4.3.2 Area change per depth dA/dz(z)
the lake as a function of depth z. (negative scale, as area decreases with z).

Upwelling flow Advection

0 0

50 50

100 100

150 150

200 200
Depth (m)

Depth (m)

250 250

300 300

350 350

400 400

450 450

500 500
0 20 40 60 80 100 120 0.0 0.5 1.0 1.5
3 -1 -1
Upwelling flow Q (m s ) Upwelling advection w (m yr )

Figure 4.3.3 The upwelling flow Q (z) at Figure 4.3.4 The upwelling velocity w(z) is
depth z is the sum of all subaquatic water the upwelling flow Q(z) divided by the area
inflows below depth z. A(z) at depth z; w(z) = Q(z)/A(z).

47
4.5 Turbulent Diffusion
Below 120 m depth (down to maximum depth) double-diffusive staircases (a
sequence of sharp interfaces and homogeneous layers) can be observed over
almost the entire water column (Schmid et al 2010). Above 120 m depth none have
been found so far. Therefore, below 120 m depth, turbulent vertical diffusivity was
set equal to 0.15*ε/N2 (Osborn 1980; Wüest et al 2000) for heat and - based on our
analysis of the fluxes through the double diffusive layers - a factor 10 lower for salt
and gases (Schmid et al 2010). Here, N2 = (-g/ρ)*(dρ/dz) is the stability of the
density stratification (Figures 4.5.3/4), which is continuously calculated during the
simulations according to Chen and Millero (1986) and Schmid et al (2004)
(equations in Appendix 8.3). The energy dissipation ε below 120 m depth is on the
order of 1.0x10-10 W kg-1, which was confirmed by three independent methods
(Schmid et al 2005a).
The continuous calculation of density stratification generates a numerical
problem due to the following positive feedback: Turbulent diffusion causes a
weakening of the density stratification which leads to increased diffusion, weake-
ning the stratification further. Therefore vertical heat diffusivity was limited to a
maximum value of 10-4 m2 s-1 (Figures 4.5.1/2). The effect of this damping is that
convective mixing caused by unstable stratification, which would require a value of
about 10-3 m2 s-1 or even more for turbulent vertical diffusion, can not be simulated
correctly. This deficiency is not critical, as the mass balances (most important)
remains correct.
The natural diffusive transport in the deep-water is weak compared to the
dominant upward flux by upwelling and a maximum value of 10-4 m2 s-1 for turbu-
lent vertical heat diffusion is rather too high for the basic simulation Scenario O (no
extraction). Since vertical transport is dominated by upwelling in this depth range, it
is not critical how the diffusivity is parameterized, as long as the results of the
simulations are not unstable.
Above 120 m depth no double-diffusive layers were observed (Schmid et al
2010) and the temperature minimum at 68 m indicates occasional seasonal con-
vective mixing of the Biozone to this depth. As there is no seasonality implemented
in the model, turbulent vertical diffusion was kept constant above 120 m depth and
the values were adapted to reproduce the observed salinity profile. The top 50 m of
the lake are mixed annually and the top 18 m of the Biozone are almost completely
mixed (scale of about a week).

4.6 Temperature

Heat input (geothermal heat flux) - The temperature gradient in Lake Kivu
is simulated with a homogeneous geothermal heat flux of 0.02 W m-2 at the sedi-
ment surface. This value agrees well with the independent estimates of small-scale
double-diffusive heat fluxes (Schmid et al 2010). The temperatures of the sub-
aquatic sources were fitted to reproduce the observed temperature profile in a time
frame of 990 yr. A comparison of the temperature measurements from 1973

48
 Turbulent vertical diffusion Turbulent vertical diffusion
0 0
2004
50 2014 50
2024 2004
100 2054 100 2014
2104 2024
150 150 2054
2104
200 200
Depth (m)

Depth (m)
250 250

300 300

350 350

400 400

450 450

500 500
0.0E+00 2.0E-06 4.0E-06 6.0E-06 8.0E-06 1.0E-05 0.0E+00 5.0E-05 1.0E-04 1.5E-04 2.0E-04 2.5E-04
2 -1 2 -1
K(z) (m s ) K(z) (m s )

Figure 4.5.1 Vertical heat diffusivity K(z) Figure 4.5.2 Vertical heat diffusivity K(z)
below 120 m depth (Scenario O, no above 120 m depth (Scenario O, no
extraction). extraction).

Stability Stability
0 0
2004
50 50 2014
2004
2024
2014
100 100 2054
2024
2104
2054
150 150
2104
200 200
Depth (m)

Depth (m)

250 250

300 300

350 350

400 400

450 450

500 500
-0.001 0 0.001 0.002 0.003 0.004 1.E-09 1.E-08 1.E-07 1.E-06 1.E-05 1.E-04 1.E-03 1.E-02
2 -2 2 -2
Stability N (s ) Stability N (s )

Figure 4.5.3 Observed (2004) and simu- Figure 4.5.4 Observed (2004) and simu-
lated stability N2 profiles (Scenario O, no lated stability N2 profiles (Scenario O, no
extraction; linear scale). extraction; log scale).

and 2004 shows a slow warming of the upper part of the lake. The temperature of
the subaquatic source at 180 m depth was fitted again to reproduce this observed
warming in a time frame of 30 yr.

49
 Heat output at the lake surface - The model does not include the seasonal
variation of temperature due to atmospheric heat exchange. The temperature at
the lake surface was set to 23.06 °C, which was the minimum temperature found in
2004 at 60 m depth. This temperature minimum is assumed to be a relict from the
last cooling-induced convective mixing.

4.7 Salinity

The vertical structure of the salinity can be understood solely by the inter-
play of the input via subaquatic sources (defining the high deep-water concentra-
tions), the upwelling of the salts to the lake surface and the surface flushing by the
low-salinity river inflow and precipitation. Basically the subaquatic salt input balan-
ces the sink via the Ruzizi River plus some negligible sedimentation within the lake.

Salinity input (surface and subaquatic inflow) - In the model, six sub-
aquatic sources sustain and shape the strong vertical salinity structure in the water
column (Table 4.4). The flow rates and the salinities of these water inputs have
been fitted to reproduce the observed profiles over a time scale of 990 yr (Table
4.4). The discharges of all the subaquatic inflows as well as their different
concentrations of the water constituents are listed in Table 8.2.2 (Appendix).

Salinity output (surface outflow) - The Ruzizi outflow (~114 m3 s-1;

Muvundja et al 2009) transports about 3.9x106 t yr-1 of salt from the surface of Lake
Kivu to Lake Tanganyika (Table 4.4).

4.8 Density

In this report and in the model simulations, the density of Lake Kivu water has
been calculated by the following approximation:

ρ(S, T, CH4, CO2) = ρ(T) * (1 + 0.75x10-3 m3 kg-1 * S + 0.284x10-6 m3 g-1 * 44.0099

g mol-1 * CO2 - 1.25x10-6 m3 g-1 * 16.04 g mol-1 * CH4) [kg L-1]

ρ(T) = 0.999839 + 6.7914x10-5 * T - 9.0894x10-6 * T2 + 1.0171x10-7 * T3 -

1.2846x10-9 * T4 + 1.1592x10-11 * T5 - 5.0125x10-14 * T6 [kg L-1]

CH4 and CO2 are the gas concentrations [mol m-3]. T = temperature [°C], and S =
salinity [kg m-3]. The equations for calculating CO2 (the gaseous CO2 concen-
trations) from CO2,total , are given in Appendix 8.3.

50
 Salinity Temperature
0 0

50 2004 50 2004
2014 2014
100 100
2024 2024
2054 2054
150 150
2104 2104
200 200
Depth (m)

Depth (m)
250 250

300 300

350 350

400 400

450 450

500 500
0 1 2 3 4 5 6 22.5 23.5 24.5 25.5 26.5
-3
S (kg m ) T (°C)

Figure 4.7.1 Observed (2004) and simula- Figure 4.7.2 Observed (2004) and simula-
ted salinity S profiles (Scenario O, no ted temperature T profiles (Scenario O, no
extraction). extraction).

Density Density
0 0

50 50 ρ(T)
2004
ρ(T,S)
2014
100 100 ρ(T,S,CH4)
2024
ρ(T,S,CO2)
2054
150 150 ρ(T,S,CH4,CO2)
2104
200 200
Depth (m)

Depth (m)

250 250

300 300

350 350

400 400

450 450

500 500
0.998 0.999 1 1.001 1.002 1.003 0.996 0.997 0.998 0.999 1.000 1.001 1.002 1.003
-3 -3
Density ρ (kg m ) Density ρ (kg m )

Figure 4.8.1 Observed (2004) and simu- Figure 4.8.2 Effects of temperature,
lated density profiles (Scenario O, no salinity, CH4 and CO2 on the lake water
extraction). density for profiles in 2004.

51
4.9 Methane

Methane new formation in the lake - CH4 formation was assumed to be

homogeneous at the sediment surface; i.e., the CH4 input into the lake is proportio-
nal to the sediment area at each depth of the lake. Formation within the water
column (open lake water) was considered to be negligible. It was further assumed
that the historic value of CH4 new formation, 30 g-C m-2 yr-1, had been constant for
a sufficiently long time to approach steady-state (in the model we arbitrarily used
990 yr). We further assumed that the CH4 formation increased from 1944 to 1974
to 70 g-C m-2 yr-1 and then to 120 g-C m-2 yr-1 between 1974 and 2004 (Figure
4.9.1). The three values of CH4 formation (historic, 1974 and 2004) were determi-
ned by fitting to the CH4 profiles observed in 1974 (Tietze 1978) and 2004 (Schmid
et al 2005a,b). Of course, the chosen temporal development is one of several
realistic possibilities, but using another time frame would not alter the average CH4
new formation needed to produce the observed increase in concentrations bet-
ween 1974 and 2004 (Schmid et al 2005a). The future development of the CH4
new formation is not well predicted and therefore the fitted value for 2004 was
used for the simulation of gas extraction. The CH4 new formation is one of the
major uncertainties in the model. Potential effects of different new formation rates
have been tested (Section 2.5). However no changes, relevant for the Conse-
quence Table 2.4.1 have been observed.
The most probable CH4 new formation rate of today is somewhere between
30 and 120 g-C m-2 yr-1. According to Pasche et al (submit) the most plausible esti-
mate is ~60 g-C m-2 yr-1, but includes a large error of ~30 g-C m-2 yr-1.

140
CH4 new formation (gC m yr )
-1

120
-2

100

80

60

40

20

0
1914 1944 1974 2004 2034 2064 2094
Year

Figure 4.9.1 CH4 formation rate (g-C m-2 yr-1) as implemented in the model, based on best
fit to the measured profiles in 1974 and 2004.

52
 Methane Carbon dioxide
0 0

50 2004 50 2004
2014 2014
100 100
2024 2024
2054 2054
150 150
2104 2104
200 200
Depth (m)

Depth (m)
250 250

300 300

350 350

400 400

450 450

500 500
0 10 20 30 40 0 50 100 150 200
-3 -3
CH4 (mol m ) CO2 (mol m )

Figure 4.9.1 Observed (2004) and simula- Figure 4.9.2 Observed (2004) and simula-
ted CH4 profiles (Scenario O, no ted CO2 profiles (Scenario O, no extraction).
extraction).

Total gas pressure

0

50 2004
2014
100
2024
2054
150
2104
200
Depth (m)

250

300

350

400

450

500
0 10 20 30 40
p (atm)

Figure 4.9.3 Total gas pressure and hydro-

static pressure (Scenario O, no extraction).

53
 Methane loss (sink) - The CH4 in the lake is transported towards the
Biozone by the upwelling flow and by turbulent diffusion. CH4 is consumed by
oxidation once it reaches the shallower water layers, which contain oxygen (most
important) as well as sulphate and nitrate (less important, Section 2.1). Therefore a
self-limiting process transforming CH4 to CO2 by oxidation above 60 m depth with
a rate of 10-6 s-1 was implemented in the model. The sum of oxidized (and thus lost)
CH4 for Scenario O (no extraction) is 6.4 km3 in the first 50 yr and 15.4 km3 in the
total 100 yr (calculated from the CH4-flux in 70 m depth). With regard to the estima-
ted CH4 new formation below 60 m depth, ~45 km3 during 100 yr (Table 1.1), the
CH4 concentrations in Lake Kivu are not in a steady-state and will continue to grow,
if not extracted.

4.10 Carbon dioxide

Modelling of CO2 - Modelling carbon dioxide (CO2) in the lake water bears
some complexity, as the CO2 dissociates into different dissolved carbonate
components in water: H2CO3 (dissolved, covalently bound), HCO3- (bicarbonate),
and CO32- (carbonate). We have chosen to simulate in the model the “total CO2“,
defined by the sum:

CO2,total = ΣCO2 = H2CO3 + HCO3- + CO32-

and to estimate the various carbonate components from this CO2,total and from pH.
The three carbonate species equilibria are calculated from the dissociation con-
stants pK1 and pK2 and pH, which depend on salinity and temperature by:

pK1 = f(T, S)
pK2 = f(T, S).

The three carbonate components are then calculated by

H2CO3 = f(CO2,total, pH, pK1, pK2)

HCO3- = f(H2CO3, pH, pK1)
CO32- = f(HCO3-, pH, pK2).

All five equations are detailed in Appendix 8.3. In this report, the term CO2 relates
only to the dissolved gas component (H2CO3), as this component is responsible for
the partial gas pressure and the potentially dangerous release of CO2 gas. The
ionic forms of CO2,total (i.e. HCO3- and CO32-) are included in the salinity.

Sources of CO2 - CO2 is assumed to be produced proportional to the sedi-

ment area at a rate of 120 g-C m-2 yr-1 (1:1 with CH4, 750 mol s-1) and additional
CO2,total, enters by the subaquatic sources (200 mol s-1, Table 8.2.2). It is assumed

54
that the decomposition of organic matter leads to the production of the same
amount of CO2 as of CH4. The total CO2 input is then 950 mol s-1 (Table 4.4).
For a layer at depth z, the accumulation of CO2 is therefore practically pro-
portional to the sediment area per unit depth, -dA/dz(z), which increases as a
function of depth (Figures 4.3.1/2). The lake water contains approximately 10-times
more CO2,total than CH4. Based on the estimated concentrations in the subaquatic
springs and the CH4 new formation rate, we expect that ~90% of the CO2 in the
lake is of magmatic origin, whereas the rest stems from the decomposition of
organic material.
The oxidation of CH4 above 60 m depth represents only a minor source of
CO2 (Section 4.9).

Loss (sinks) of CO2 - CO2 is leaving the lake water by the Ruzizi River out-
flow (minor contribution) and by degassing into the atmosphere (Table 4.4). These
two CO2 fluxes are parameterized in the model as follows:
(i) Ruzizi River outflow = water discharge (m3 s-1) * lake surface
concentration (mol m-3),
(ii) Gas exchange with the atmosphere: vgas exchange * (CO2 surface eq. - CO2)*
surface area. The parameterizations used in this equation, are detailed
in Appendix 8.3.

4.11 Phosphate and ammonium

Sources of nutrients - The external inputs of PO4 and NH4 have been esti-
mated based on field measurements in 2006 to 2008 (Muvundja et al 2009). The
comparison with the lake internal nutrient cycling shows (Pasche et al 2009) that
the input of phosphate (PO4) and ammonium (NH4) into the Biozone is largely
dominated by the lake internal loading. Therefore, in this model, the major and only
important source for nutrients for the Biozone is the upwelling of nutrient-rich deep-
water. The reason for the high deep-water nutrient concentrations is the minerali-
zation (degradation) of the lake internal organic material (plankton) at the sediment
surface in combination with the long residence time of the deep water.
Mineralization is assumed proportional to CH4 with the stoichiometric coeffi-
cients of the organic material found in the sediment traps in 2007 (molar ratios C:P
= 216 and N:P = 13.9 (Pasche et al submit; Appendix 8.4). Degradation of organic
matter in anaerobic water transforms carbons in equal parts to CH4 and CO2.
Although, direct nitrogen fixation by algae or bacteria from the atmosphere
may be an additional nutrient source, this process is not included in the model.

Sinks of nutrients - To account for the primary production (algae growth at

the lake surface), a fast self-limiting process (limiting in PO4 and NH4), removing
PO4 with a rate of 10-6 s-1 in the Biozone, was implemented. It was observed that
PO4 is the limiting factor for most of the time, but occasionally NH4 appears to be
limiting (Pasche et al 2009; Muvundja et al 2009). This sink of PO4 implies that any

55
PO4 entering the Biozone is consumed on average within 12 days. NH4 is removed
in parallel with the ratio of the corresponding stoichiometric coefficients (Appendix
8.4). Removal of NH4 by nitrification and denitrification is not included in the model,
implying that NH4 surface concentrations are overestimated.
The Ruzizi outflow transports ~50 t-P yr-1 as PO4 and an additional ~20 t-P
-1
yr as plankton from Lake Kivu to Lake Tanganyika (Table 4.4; Muvundja et al
2009). This small phosphorus export has no influence on this model as the nutrient
cycle is not closed. Simulated PO4 release by algae degradation (mineralization) is
constant and does not depend on PO4 consumption (i.e. primary production). Con-
sequently, a potential positive feedback between enhanced nutrient upwelling,
higher primary production and therefore increased mineralization cannot be
reproduced by the model.

Phosphate Ammonium
0 0

50 2004 50 2004
2014 2014
100 100
2024 2024
2054 2054
150 150
2104 2104
200 200
Depth (m)

Depth (m)

250 250

300 300

350 350

400 400

450 450

500 500
0 0.1 0.2 0.3 0.4 0 1 2 3 4 5 6 7
-3 -3
PO4 (mol m ) NH4 (mol m )

Figure 4.11.2 Observed (2004) and simu-

Figure 4.11.1 Observed (2004) and simu-
lated NH4 profiles (Scenario O, no
lated PO4 profiles (Scenario O, no
extraction).
extraction).

4.12 Lake development without extraction - Scenario O

With these assumptions, detailed in the above Sections 4.1 to 4.11, we

performed model simulations for 100 yr duration without CH4 extraction (Scenario
O, natural development). We used these results from Scenario O in the following

56
Scenario simulations (Section 7) as a reference. The “no extraction” simulations
reveal the following (see various figures in this Section 4):
Salinity (Figure 4.7.1) and temperature (Figure 4.7.2) are, according to the
assumptions, in a quasi-steady state. For this reason, salinity (Figure 4.7.1) and
temperature (Figure 4.7.2) change only slightly over the 100 yr of simulation.
Although we know from various other evidence that the lake is not exactly in
steady state, we consider this assumption as reasonable, as changes to salinity
and temperature may occur in “both directions”.
As salinity and temperature are the main components defining the density
(Figure 4.8.1), also the density profile is not much affected over 100 yr of simula-
tion. Therefore the simulated stability N2 of the density stratification (Figures 4.5.3,
4.5.4) and the Attribute “Schmidt Stability” (Table 7.1) also remain almost constant.
The major changes over 100 yr of simulation are related to the gases and
the nutrients. CH4 (Figure 4.9.1) would increase to almost double the current
concentration in the Resource Zone, whereas CO2 (Figure 4.9.2) would increase
only slightly. As an effect of both, the total gas pressure (Figure 4.9.3) is expected
to become very critical below 250 m depth (Resource Zone) within the next century.
Also the nutrient concentrations of PO4 (Figure 4.11.1) and NH4 (Figure
4.11.2) would rise by up to 50% in the deep-water (below Biozone) and therefore
also the nutrient fluxes to the Biozone would increase by about ~55 % (Table 7.1).
The simulations without extraction also show the effect of the interaction of
the assumptions. If for example the water discharge of the subaquatic sources
would weaken, then the accumulation of CH4 in the deep-water would intensify
(and the risk would increase even faster), because the “dilution” effect of the
inflowing source water would be reduced. Therefore it is important to monitor the
development of the vertical stratification of all relevant lake water properties (gases,
temperature and salinity, nutrients) to detect potential changes in due time.

57
58
5. Definition of the extraction Scenarios

5.1 Design Parameters

The extraction Scenarios, tested in the following simulations, are characteri-
zed by five plant-defining “Design Parameters”. Here we focus only on plants
which are qualitatively similar to the currently projected plants, i.e., water is taken
to the surface in a vertical pipe, degassed, and reinjected again, by a vertical pipe
into the lake (Figure 1.2). These five Design Parameters (Table 2.3.1), comprise:

(1) withdrawal depth and withdrawal depth range

(2) reinjection depth (of the depleted deep-water) and reinjection depth range
(3) dilution-water flow (% of extraction flow) and dilution-water intake depth
(4) removal (%) of CH4, and
(5) removal (%) of CO2.

Withdrawal depth and withdrawal depth range - The withdrawal of the to-
be-extracted water is located in the Resource Zone (Figure 1.2), as the concen-
tration in the PRZ and the IZ are currently too low for viable extraction.
The flow towards the holes of the intake pipe is not horizontal close to the
intake (Figure 5.1). The vertical width of the “selective withdrawal” depends on the
flow Q [m3 s-1] as well as the local stability N2(z) [s-2] of the stratification at depth z
(Kataoka et al 2000; Woods 2001). The withdrawal width can be approximated by

D ≈ 1.8 * (Q/N)1/3 [m]

where the factor of 1.8 can vary from 1.6 to 2.0 according to Table 2 in Fan (2008).
For example, an extraction flow Q of 1 m3 s-1 taken from the deepest reaches (460
to 480 m depth; N2 in Figures 4.5.3, 4.5.4), would cause a vertical withdrawal
range of ~16 m. In the Upper RZ (stronger stratification) the corresponding with-
drawal range would be narrower.

Although, the selective withdrawal appears not to cause concerns, it may be

important to maintain a smooth and gentle flow with only small vertical distortion of
the horizontal layers by spreading the flow over a certain vertical range and by ad-
ding holes to the intake pipe. The values listed in Table 5.5 indicate the withdrawal
depth (centre of such a withdrawal range) and the (full) vertical range. For the
model simulations, a minimum withdrawal range of 10 m is used (Table 5.5). If the
withdrawal pipe is not reaching the deepest location, the model assumptions imply
that the lake volume below the intake is only affected by the turbulent diffusion and
the natural upwelling. Therefore the simulation results occasionally show much
higher concentrations in the very lowest reach of the lake
Figure 5.1 Schematic diagram defining the thickness D of the withdrawal layer (depth
range from where water is entering the pipe, point sink) for the flow Q in the ambient lake
stratification N2. Details in Fan (2008).

Reinjection depth and reinjection depth range - The depth of the reinjec-
tion is a key parameter for this study, as its level affects strongly the future deve-
lopment of the stratification and the vertical distribution of the gases and density in
particular. The pros and cons of the different reinjection depths - the key outcome
of this study - are discussed throughout but specifically in the Sections 2.4 to 2.6.
Two different concepts have intensively been discussed: the “Density
Gradient Lowering Method” and the “Density Structure Maintaining Method“ (MP
2009). The idea of the first approach is to stack the (lighter) CH4-depleted reinject-
water on top of the RZ (or on top of the PRZ) while “Lowering” the RZ to the
deepest point and completely emptying the RZ from CH4. The second approach
conserves the vertical structure of the lake density stratification, as the CH4-
depleted deep-water is reinjected into the RZ. However, in order to avoid dilution of
the RZ, which would decrease the efficiency of the extraction as well as the energy
production (at least below the “non-economic” threshold concentration of CH4), it is

60
still necessary to stack the CH4-depleted reinject water on top of the high-concen-
tration RZ water. This can only be achieved by carefully controlling the density of
the reinject water via adjusting the CO2 removal (MP 2009).
Same as the withdrawal, also the reinject-water should be spread over a
certain vertical range (Figure 5.1). The advantages are lower release velocity, less
turbulence and more efficient mixing of the reinject-water with ambient lake water,
as the reinjection entrains into a larger layer while enhancing the mixing. Density
differences between reinject-water and ambient lake water are quickly reduced and
as a result the reinject-water plumes sink or rise only short vertical distances. For
example, the reinjection plume from the KP1 pilot plant has been observed to
stratify only about 10 m below the reinjection depth at 90 m, while the natural
restratification depth would have been located around 180 m (Nzayisenga et al
2009).

Dilution - The CH4-depleted deep-water is usually denser than the ambient

lake water at the chosen reinjection depth (the opposite is theoretically possible if
almost all CO2 is removed). Reasons for the higher density are the higher salinity
(Figure 4.7.1) and CO2 content (Figure 4.9.2) at greater depth. In addition the ex-
trusion of gas cools the deep-water by up to ~0.5 oC. Usually the reinject-water is
also lighter than the originally extracted deep-water (again the contrary is possible
under special circumstances when almost all CO2 is retained).
To ensure that the released water re-stratifies within the chosen reinjection
depth range (no sinking or rising of the plumes) two measures can be taken: (i)
The reinject-water can be diluted with near-surface water to adjust its density to the
density of lake water near the reinjection depth, or (ii) the reinject-water can be
spread (horizontally or vertically) with multiple outlets to efficiently mix it with
ambient lake water.
The simulations (Section 7) demonstrate two major disadvantages by using
dilution-water and therefore we recommend refraining from its use (Section 2.4). If
no dilution-water is used (what we encourage), enough mixing of the reinject-water
(usually heavier) with the ambient lake water is needed to avoid sinking or rising
plume within the lake.

Removal efficiency for CH4 - An ideal plant would have 100% removal
efficiency for CH4. Not extracted CH4 is returned to the lake with the reinject-water.
If the reinject-water is released at shallow depth, the upwelling will remove this
remaining CH4 from the lake (details below) and hence the non-extracted CH4 is
lost. However, when reinjecting into the RZ, non-extracted CH4 remains within the
RZ and is not lost for future extraction. If a plant operation, for whatever reasons,
performs inefficient removal of CH4, it does not cause a loss of CH4.

Removal efficiency for CO2 - The fraction of CO2 removed is relevant for
two reasons: (i) the remaining CO2 concentration affects the density of the reinject-
water and (ii) the CO2 content in the extracted gas mixture affects the efficiency of
the power production. As CO2 has a much higher solubility compared to CH4, the
CO2 removal is not critical for the safety of the lake (see below), but its removal is

61
still a welcomed safety advantage. Plants with the possibility to adjust the CO2
removal, can adjust the density of the reinject-water within a certain range.

Plant size - This study does not consider the pros and cons of the size of
individual plants, as for the development of the lake stratification in a one-dimen-
sional model only the water flows of the overall extraction operation are relevant. If
large plant modules would be located in close proximity to each other, flow inter-
actions could become relevant and therefore 3-dimensional modelling could be-
come necessary. However, as the lake currents are not well known, a conservative
planning is required. To make the different extraction simulations comparable the
water flow was set to 70 m3 s-1 for all Scenarios. This flow is probably in the lower
range of the expected future CH4 extraction rate, as this flow corresponds to a
typical electrical energy production of 100 to 200 MW.

Energy production - The plant efficiency defines the electric power output
per flow of CH4 led to the power production unit. Investigating these conversions is
not the subject of this study, and we used a constant value of 2.6 kWh m-3 (Morkel
pers comm) to convert CH4 to electrical energy. Note that the estimated energy
output may decrease with time, as the conversion may decrease due to lower CH4
concentrations as a function of time. In our simulations, the comparison of the
extraction is done based on CH4 only, but we list the electrical energy for practical
reasons.

5.2 Intermediate Zone reinjection - Scenarios IZ1 to IZ5

Because the early reasoning on the most economic way of CH4 harvesting
asked for reinjection in the upper part of the Intermediate Zone, reinjections at 90,
150 and 190 m depth have been tested (Table 5.5). The idea of the Intermediate
Zone reinjection is to keep the CH4-containing Resource Zone and the layer of
reinjection as far as possible apart, so that no dilution of the Resource Zone can
occur. The main disadvantage of this option is the large additional nutrient input
into the Intermediate Zone which ends up - due to upwelling - in the Biozone after
a few years / decades (Table 2.3.2). This disadvantage is especially important for
the Scenarios IZ1, IZ2 and IZ3, where dilution-water is used to adjust the density of
the reinject-water. Together with the nutrients, also the upwards flux of dissolved
CH4 to the surface layer is increased, where it is oxidized and therefore lost for a
potential future extraction. No dilution-water is used for IZ4 and IZ5 (Table 5.5) and
therefore the induced upward fluxes of nutrients and CH4 are reduced. In terms of
CH4 harvest, Scenario IZ4 provides most output.
The details of the Scenarios IZ1 to IZ5 are provided in Table 5.5. The
different extraction Scenarios are described in Table 2.3.2.

62
5.3 Potential Resource Zone reinjection - Scenarios PR1

The idea of the Potential Resource Zone reinjection (Scenario PR1) is (i) to
profit from the advantage of the vertical separation of the CH4 resources (deep)
and the reinjection (above the Resource Zone) and (ii) to reduce the disadvantage
of the nutrient upflux compared to the IZ scenarios by reinjection into the medium-
depth of 240 m. The details of Scenario PR1 are provided in Tables 5.5 and 2.3.2.

5.4 Resource Zone reinjection - Scenarios RZ1 to RZ6

The Intermediate Zone reinjection has the disadvantage of increasing
nutrient fluxes to the lake surface and of weakening the density stratification
(Section 2.4). To avoid these negative effects, it has been suggested to reinject the
depleted deep-water back into the Resource Zone. This approach has the great
advantage that the lake stratification is almost undisturbed, except for the changes
within the RZ, which is, however, far away form the Biozone. The main disadvan-
tage of these options is the danger that, due to the weak stratification within the
Resource Zone, the CH4-depleted reinject water may mix with the high-CH4 con-
centration water and thereby the CH4 resource may become diluted. Besides the
CH4 concentration also the CH4/CO2 ratio is decreasing, because the reinject-water
contains more of the original CO2 than CH4.
Six different Resource Zone reinjection Scenarios (RZ1 to RZ6) have been
simulated to test the subtle differences in search for an optimal Scenario. The idea
of RZ1 is to keep the withdrawal and reinjection depth vertically separated as far
as possible within the RZ. Some of the Scenarios (RZ2 to RZ4) have two withdra-
wal and two reinjection layers with the aim to manage the Upper RZ and the Lower
RZ separately. Scenario RZ6 is a combination of two (first RZ4 and later RZ5).
The Design Parameters of Scenarios RZ1 to RZ6 are detailed in Table 5.5
and the Scenarios are described relative to each other in Table 2.3.2. The so-
called “Plan A1/A2 and Plan B” in the Management Prescriptions (MP 2009)
corresponds in this report appoximately to Scenario RZ6. The scenario RZ4 (cor-
responding to A1/A2) can only be run until the Upper RZ is used up (a few years/
decades). It is not a stand-alone Scenario and not examined in the evaluation
procedure.

5.5 Summary of the simulated extraction Scenarios

In Section 7 the results of the simulated extraction Scenarios are presented
and the evaluation criteria (Section 6) are quantified. The following Table 5.5
summarizes the Design Parameters for all extraction Scenarios as used in the
simulations.

63
Table 5.5 Summary of the Design Parameters used for Scenario definitions

64
6. Scenario evaluation criteria

6.1 Extraction Objectives and quantification by Attributes

As laid out in Section 2, the CH4 extraction concept and implementation has
to live up to the following Guiding Principles (order of priority):

(i) Reduce the probability of a catastrophic gas release from the lake
(ii) Conserve the ecological integrity of the lake
(iii) Maximize the benefit from the CH4 resources in the lake.

In order to make these principles operational, we define seven “concrete

Objectives, which can be used to evaluate different extraction Scenarios, defined
in Section 5. These seven Objectives are:

(a) Keep lake stratification as stable as possible, at least at 50% of natural

(b) Reduce gases content in lake, at least by 25% over the next century
(c) Increase required activation energy for (and thereby reduce chance of)
initiation of bubble formation
(d) Minimize additional nutrient input to Biozone, at maximum 25% increase
(e) Minimize losses of CH4 (i) to Biozone and (ii) by the extraction operation
(f) Allow new formation of deep-water CH4
(g) Avoid dilution of CH4 resources.

With the model simulations we want to quantify how well these seven Objectives are
fulfilled. Therefore, we assign quantifiable Attributes to those Objectives (Table
2.2.1). The simulations allow for any given time and for any Scenario to calculate
those Attributes. This procedure allows quantifying how well the Objectives are
fulfilled for any moment, and also - more important - it allows the different Scenarios
to be compared over several decades of extraction. We have used the following
eight Attributes, which are well-defined physical quantities to measure the Objectives:

• Schmidt stability = required energy for complete mixing of the lake

• Safety margin = minimal energy required to initiate bubbles
• Phosphorus upward flux = upward flux of P into the Biozone
• Nitrogen upward flux = upward flux of N into the Biozone
• CO2 content = CO2 content in the lake
• CH4 content = CH4 content in the lake
• CH4 harvested = amount of CH4 led to power production
• CH4 harvestable = amount of economically harvestable CH4 in the lake.

Although the eight Attributes have a value at any point in time, we listed them
only for 2004 (0 yr), 2054 (50 yr) and 2104 (100 yr). Definitions are provided in the
following Section 6.2 and values of the Attributes are summarized in Table 7.13.

65
6.2 Definitions of the Attributes

Schmidt stability - The Schmidt stability is an overall measure of how

strong a water body is density-stratified. Numerically, the Schmidt stability indi-
cates the amount of energy [J] needed to completely homogenize a particular
water body. An alternative unit of the Schmidt stability SS [J m-2] provides the
energy needed per lake surface area Ao. It is calculated as the vertical integral
over the lake volume of the density difference between the actual stratification and
the homogenised density times the dislocation from the centre of mass (Schmidt
1928):
max depth

SS = g ∫ (z − z )(ρ
0
* z − ρ )Az dz [J]

(g = gravitational acceleration, Ao = lake surface area). The Schmidt stability allows

assessing the changes in the strength of the stratification (large SS = higher
stability).
The effects of the extraction operation are mainly (1) to remove CO2 and
thereby to lower the density of the deep-water and (2) to move denser (more saline)
water to higher levels in the lake. Both effects lower the Schmidt stability of the
stratification. As shown in the Tables 7.1 to 7.12, all Scenarios lower the stability.
However the differences among the Scenarios are very large: whereas the IZ
reinjections change the Schmidt stability massively, the RZ reinjections have
almost no effect on the Schmidt stability (comparison in Table 7.13).
It may be interesting to realize that Lake Kivu is among the lakes with the
highest Schmidt stability (> 300,000 J m-2). Compared to this potential energy, the
kinetic energy in the lake (currents and waves due to wind energy input) is only a
few 100 J m-2 (Schmid et al 2005a) and therefore the interior turbulence is very
weak. Without extraction, the Schmidt stability would remain almost constant, even
while the gas concentrations rise to dangerous levels, since the density in the lake
is mainly determined by salinity and temperature.

Safety margin - As shown above, the Schmidt stability is an excellent

measure to quantify the overall density stratification of a lake. However, it gives no
information on how close a gas-saturated water body is from eruption. For this
reason, we also introduced the Attribute “safety margin”, which expresses how
much energy is needed [J m-3] to bring the most critical water layer (in the lake) to
spontaneous degassing which could potentially trigger a catastrophic eruption.
The safety margin is calculated in two steps. First we evaluated for each
depth z1 to which level z2 a (virtual) water parcel would need to be lifted until its
total gas pressure reaches the hydrostatic pressure. At the level z2 this water
parcel would become 100% saturated (and bubbles could be created without any
further energy input). In a second step, the mechanical energy SM [J m-3] to lift this
water parcel (from z1) to this estimated level z2 was calculated by:

z2
SM = ∫ g (ρ ( z1) − ρ ( z ) )dz [J m-3]
z1

66
In the Figure 7.2.1 to 7.2.12 SM is plotted as function of depth.
This definition of the safety margin does not include the volume or amount
of gases available in the critical depth range. In practice, the amount of gases
available to initiate and to fuel a chain reaction is highly relevant for eruptive de-
gassing. Close to the surface, the safety margin becomes undefined, as there the
total gas pressure is too low and also the required energy SM decreases towards
zero. However, within the Biozone and the Intermediate Zone, the safety margin is
irrelevant, since gas concentrations are not sufficient to cause any risk (no self-
sustaining degassing). For this reason, we consider in this particular case of Lake
Kivu the safety margin only below 200 m depth (Potential Resource Zone and
deeper), where the gas concentrations are relevant for the safety.
The Scenario O (no extraction) shows that after 100 yr the most dangerous
layer for degassing is not found at 260 m depth (Figure 6.2.1) but near the deepest
reaches. Although, the total gas pressure is closest to saturation near 260 m depth,
the strong density gradient above 260 m depth suppresses vertical movement of
water parcels. The highest risk will be located below 400 m depth (Figure 6.2.1).
Although water parcels in this deepest layer would have to be lifted further to reach
saturation compared to 260 m depth, the density gradient in the deepest reach is
weaker and therefore less energy is required to move water vertically. The values
discussed in Section 7 represent the lowest energy required for the most critical
layer below a depth of 200 m.

Phosphorus and nitrogen upward fluxes - The nutrient fluxes at z = 80 m

depth are calculated from the local nutrient concentration gradients dC/dz (z) times
the local vertical diffusivity K(z), as well as from the local concentration C(z) times
the local upwelling velocity w(z). Taking into account the lake area A(z) at the
given depth leads to upward transports of:
Safety margin
0

50 2004
2104 no extracion
100

150

200
Depth (m)

250

300

350

400

450

500
0 500 1000 1500 2000 2500 3000
-3
Energy needed to lift until ptotal = phydrostatic (J m )

Figure 6.2.1 Safety margin for Scenario O (no extraction). In the year 2104 the gas con-
centrations are estimated to approach saturation in the deepest part of the lake, and the
safety margin is therefore very low (~30 J m-3).

67
 Fadvective = A(z) C(z) w(z) [mol s-1]
Fdiffusive = - A(z) K dC/dz [mol s-1]

Ftotal = Fadvective + Fdiffusive [mol s-1]

The evaluation of the model simulations showed that the upwelling (advective
transport) is more important than the diffusive transport for most of the deeper
water body, but in 80 m depth, where turbulent diffusion is already higher, both
fluxes contribute to similar parts. Any changes to the lake which enhance the water
upward flow or the entrainment into the Biozone, also enhance the nutrient trans-
port into the Biozone and are therefore considered as critical for the Guiding
Principle of the “lake ecological integrity”. The simulated future nutrient upward
transport is therefore compared to the current nutrient fluxes (2004) and to the
fluxes for the “no extraction” Scenario O (Tables 7.1 to 7.12).
According to the Management Prescriptions (MP 2009), the upwards
nutrient fluxes should not be increased by the extraction operation by more than
25% compared to the Scenario O (no extraction).

Gas content in the lake - The amount (MLake) of CH4 or of CO2 present in
the lake at any point in time is calculated by vertically integrating the gas concen-
trations C(z) times the lake area A(z):

max depth

M Lake = ∫ A( z ) C ( z ) dz .
0
[km3]

The CO2 in the lake is relevant in a threefold way: (i) as the gas adds to the partial
pressure and therefore to the risk, parts of the CO2 should be removed during
extraction; (ii) the CO2 supports the self-siphoning in the pipe and is therefore a
welcomed source of buoyancy and finally (iii) the CO2 reduces the efficiency of
power production (such as by turbines). The ideal management of the CO2 is
therefore an optimization issue.

CH4 harvested (to power production) - The total amounts of “effectively”

harvested CH4 is the time-integrated instantaneous flow of CH4 supplied to the
power production (turbines). The instantaneous CH4 flow is calculated as the
product of the water withdrawal rate Q(t), the gas concentration C(t) at withdrawal
depth and the fraction fturb of CH4 led to power production relative to the CH4 in the
withdrawal water:

t end

M Turbine = ∫ Q(t ) C (t )
0
f turb dt [km3]

68
 CH4 harvestable (still in the lake) - The vertical distribution of the remai-
ning CH4 in the lake varies for the different Scenarios and with time. As CH4 at
higher concentrations is easier extractable, and since below some threshold con-
centration (depending on the technology used) CH4 extraction will not be econo-
mically viable, the value of the CH4 remaining in the lake will depend on the vertical
distribution of CH4. We express this “dilution” phenomenon by the so-called “CH4
harvestable”, defined by

zm

M harvestable = ∫ A( z )
0
CH 4 harvestable ( z ) dz [km3]

which accounts only for CH4 exceeding a defined threshold CH4,limit. In the
definition above, CH4,harvestable is counted only if CH4 > CH4,limit and set equal to 0
otherwise. The threshold value CH4,limit was set to 5 mol m-3 for the evaluation of
model simulations.

69
70
7. Simulation results of all Scenarios
In this Section the simulations of all 12 Scenarios (Table 5.5) are docu-
mented. They comprise five Scenarios with reinjection into the Intermediate Zone
(IZ1 to IZ5), one with reinjection into the Potential Resource Zone (PR1) and six
with reinjection into the Resource Zone (RZ1 to RZ6).
The simulation results serve as input for the evaluation of the effects on
the lake (Section 6) and for comparison between the Scenarios. The major model
output are shown graphically for the years 2004 (simulation start), 2014, 2024,
2054, and 2104 (simulation end). These figures (7.1.1 to 7.12.8), containing
much technical details, can be consulted for planning purposes. The outcome of
the simulations is briefly discussed concerning the three Guiding Principles (here
listed as “stability and safety”, “nutrients”, and “methane and energy”). Numerical
values of the Attributes (evaluation criteria, Tables 7.1 to 7.12) are briefly discus-
sed for all 12 simulations and summarized numerically in Table 7.13.

7.1 Scenario IZ1: Reinjection into Intermediate Zone with dilution

Design parameters - For Scenario IZ1 Scenario IZ1

the withdrawal depth is 475 ± 5 m and the rein-
jection is from 100 to 200 m. With this setting Extraction flow 70 m3 s-1
the entire RZ and part of the PRZ would be CH4 reinjected 5.9 %
drawn down and could be emptied. The draw- CO2 reinjected 60 %
down volume in the lake (downwelling below Withdrawal depth 475 m
183 m depth) is 212 km3. This volume is with- Withdrawal range 10 m
drawn 1.04 times during the 100 yr of simu-
Dilution factor 2
lation (221 km3). In the past, this Scenario has Dilution withdrawal depth 40 m
been considered as the obvious option for the Dilution withdrawal range 10 m
CH4 extraction.
Reinjection depth 150 m
For this Scenario IZ1 dilution-water is Reinjection range 100 m
used to adjust the density to approximately 150
m depth of the ambient lake water (dilution factor of ~2). The depth of dilution-water
intake was set to 40 m. As dilution-water, taken from the Biozone, is released below
the Biozone (in 150 m depth) an additional upwelling flow (equal to dilution-water
flow) is caused in the lake (Figure 7.1.1: upwelling between 150 and 40 m). As a
consequence the nutrient fluxes to the surface layer are strongly enhanced com-
pared to Scenario O (no extraction).

Stability and safety - The deep-water volume between the withdrawal depth
and the reinjection depth is moving downward (Figure 7.1.1), towards the deepest
reaches of the lake (= withdrawal level). The layer at the reinjection depth widens

71
Table 7.1 Evaluation of Scenario IZ1

No Extraction Scenario IZ1

2004 2054 2104 2004 2054 2104
PO4 upflux 80 m depth 1.9 2.4 2.9 4.0 11.7 8.1 103 tyr-1
NH4 upflux 80 m depth 15.9 21.0 26.0 34.3 113.9 82.1 103 tyr-1
Schmidt stability 8.8 8.7 8.7 8.8 3.2 1.7 1014 J
CH4 in lake (i) 64.6 81.7 96.3 64.6 28.7 18.2 km3
CH4 harvestable (ii) 42.3 58.9 69.0 42.3 14.2 3.3 km3
CH4 harvested (iii) 47.3 71.4 km3
Average energy output 281 212 MW
CO2 in lake 290 315 336 290 164 86 km3
CO2 removed from lake 86 119 km3

downward (Figure 7.1.3), with the downwelling flow below ~170 m (Figure 7.1.1). As
a result of this water volume replacement, the density stratification “flattens” and the
density difference between IZ and RZ shrinks drastically (Figure 7.1.3). Therefore,
the Schmidt stability decreases by more than 80% (compared to Scenario O,
undisturbed) after 100 yr of simulation (Table 7.1). After 100 yr, the surface water is
twice as salty as today.
Due to increasing density at the dilution-water intake (as a result of the rein-
jection into the IZ) the required dilution factor needs to be increased with time which
would increase the upflow (upwelling more salts and nutrients into the Biozone). This
positive feedback is not included in the model, but in reality this would further reduce
the density stratification and lower the stability.
The safety margin shows that the minimal energy needed to lift any water
parcel until the gas pressure reaches the hydrostatic pressure would be found in the
Intermediate Zone above 200 m depth. The top 200 m we consider as not critical for
the risk of an eruption (Section 6.2) and therefore we evaluate the safety margin only
below 200 m depth. Even after 100 yr, the safety margin at 200 m depth is still com-
fortingly high (184 J m-3) and the risk of an uncontrolled degassing remains low.
Compared with Scenario O (no extraction), the safety margin improves massively
below 330 m depth (Figure 7.1.2), where dissolved gas concentrations are highest
(Figures 7.1.5 to 7.1.8). However above 330 m depth, IZ1 reduces the safety margin,
as the stability is significantly weakened.

Nutrients - The nutrient transport to the surface layer strongly increases for
IZ1. Right after the start of the extraction the PO4 flux in 80 m depth would double
and after 50 yr of simulation the flux is even 4.6 times as high as for Scenario O
(Figure 7.1.4). This enhanced nutrient fluxes would lead to excessive algae growth.
Scenario IZ 1, therefore, obviously fails on the Objective of maintaining the lake
ecological integrity.

Methane and energy - The increased upwelling between the dilution-

water intake and the reinjection (Figure 7.1.1) depth leads to an additional
upward transport of CH4 and therefore to an additional loss of CH4 due to oxi-

72
dation of about ~7 km3 during the first 50 yr and ~10 km3 CH4 (relative to IZ4) for
the entire 100 yr (oxidation and gas exchange to the atmosphere are the only
ways to loose CH4). The CH4 concentration in the intake water (Figure 7.1.7)
would remain high for ~50 yr and then drop by a factor of ~2 within one decade
from ~20 to ~10
mol m-3 (Figure 7.1.7).
The harvested amount of CH4 during 100 yr is 71.4 km3 and the correspon-
ding average electrical energy production is ~212 MW (Table 7.1).

73
 Advection Safety margin
0 0
dilution water depth dilution water depth
50 50

100 100
reinjection depth reinjection depth
150 150

200 200
Depth (m)

Depth (m)
250 250 2004

300 300 2104 no extraction

2104 Scenario IZ1
350 350

400 400

450 extraction depth 450 extraction depth

500 500
-50 -40 -30 -20 -10 0 10 0 1000 2000 3000 4000 5000 6000
-1 -3
Upwelling advection (m yr ) Energy needed to lift until ptotal = phydrostatic (J m )

Figure 7.1.1 Simulated profiles of vertical Figure 7.1.2 Safety margin for Scenario IZ1
velocity (+: upwelling; -: downwelling) for compared to Scenario O.
Scenario IZ1 (orange) compared to natural
(black).

Salinity Phosphate
0 0
dilution water depth dilution water depth
50 2004 50 2004
2014 2014
100 100
2024 2024
reinjection depth 2054 reinjection depth 2054
150 150
2104 2104
200 200
Depth (m)

Depth (m)

250 250

300 300

350 350

400 400

450 extraction depth 450 extraction depth

500 500
0 1 2 3 4 5 6 0 0.1 0.2 0.3 0.4
-3
S (kg m ) PO4 (mol m-3)

Figure 7.1.3 Simulated salinity profiles for Figure 7.1.4 Simulated PO4 profiles for
Scenario IZ1. Scenario IZ1.

74
 Methane Carbon dioxide
0 0
dilution water depth dilution water depth
50 2004 50
2014
100 100
2024
reinjection depth 2054 reinjection depth
150 150
2104
200 200
Depth (m)

Depth (m)
250 250
2004
300 300
2014
2024
350 350
2054
2104
400 400

450 extraction depth 450 extraction depth

500 500
0 5 10 15 20 25 0 50 100 150 200
-3 -3
CH4 (mol m ) CO2 (mol m )

Figure 7.1.5 Simulated CH4 profiles for Figure 7.1.6 Simulated CO2 profiles for
Scenario IZ1. Scenario IZ1.

Methane extraction concentration Total gas pressure

25 0
dilution water depth
50

20 100
reinjection depth
150
2004
15 200 2014
CH4 (mol m-3)

Depth (m)

2024
250 2054
2104
10 300 hydrostatic

350

5 400

450 extraction depth

0 500
2004 2024 2044 2064 2084 2104 0 10 20 30 40
Time (year) Pressure (atm)

Figure 7.1.7 Simulated CH4 extraction Figure 7.1.8 Simulated total gas pressure
concentration as a function of time for profiles for Scenario IZ1.
Scenario IZ1.

75
7.2 Scenario IZ2: Reinjection into Intermediate Zone with dilution

Design parameters - Scenario IZ2 is

similar to IZ1, as CH4-containing deep-water is Scenario IZ2
withdrawn from the RZ (320 ± 5 m) and rein- Extraction flow 70 m3 s-1
jected at the upper end of the IZ (90 ± 20 m
CH4 reinjected 30 %
depth). Dilution-water is also used, however,
CO2 reinjected 60 %
withdrawn in 10 m depth in order to minimize
its flow (although seasonal variations of the Withdrawal depth 320 m
density remains a challenge) and the dilution Withdrawal range 10 m
factor is therefore only 1.1. The water volume Dilution factor 1.1
3 Dilution withdrawal depth 10 m
between 90 and 320 m depth (298 km ) is
drawn down (Figure 7.2.1) by the extraction Reinjection depth 90 m
volume of 221 km3 during the simulated 100 yr Reinjection range 40 m
(0.74 times recycling).
Similarly as in IZ1, the dilution-water causes an additional upwelling flow
between the reinjection depth (90 m) and the Biozone, which also increases the
nutrient upward flux as in Scenario IZ1. In this Scenario IZ2, this aspect is even
more critical (Table 7.2), as the nutrient-rich deep-water is injected directly below the
Biozone (90 ± 20 m; not entirely accounted for by the upflux in 80 m depth).

Stability and safety - The water of the IZ and the PRZ between the withdra-
wal depth (320 m) and the reinjection depth (90 m) is moving downward, towards the
withdrawal level. The layer at the reinjection depth widens vertically and descends
with the downwelling flow in the lake. As a result, the density stratification “flattens”
and the density difference between the surface and 320 m depth is drastically redu-
ced (Figure 7.2.3), and the surface water becomes significantly saltier. The RZ
below is not affected. The vertical density gradient shrinks substantially and the
Schmidt stability decreases by up to 40% after 100 yr of simulation compared to
(undisturbed) natural (Table 7.2).

Table 7.2 Evaluation of Scenario IZ2

No Extraction Scenario IZ2
2004 2054 2104 2004 2054 2104
PO4 upflux 80 m depth (1) 1.9 2.4 2.9 2.5 3.9 3.0 103 tyr-1
NH4 upflux 80 m depth (1) 15.9 21.0 26.0 20.9 37.3 29.4 103 tyr-1
Schmidt stability 8.8 8.7 8.7 8.8 5.6 5.3 1014 J
CH4 in lake (i) 64.6 81.7 96.3 64.6 56.2 60.8 km3
CH4 harvestable (ii) 42.3 58.9 69.0 42.3 31.8 33.8 km3
CH4 harvested (iii) 20.8 30.5 km3
Average energy output 123 90 MW
CO2 in lake 290 315 336 290 213 193 km3
CO2 removed from lake 51 69 km3
(1)
Nutrient fluxes shown are slight underestimates, as some of the reinject water is released above 80 m depth.

76
The safety margin (Figure 7.2.2) shows that the minimal energy needed to lift any
water parcel until the gas pressure reaches the hydrostatic pressure would be found
in the deep RZ (~450 m depth: ~30 J m-3) where the gas would also accumulate
during the 100 yr period. Therefore, the safety margin in the deepest layers would
shrink (almost as for Scenario O) to dangerous levels and the Guiding Principle of
safety would not be fulfilled. This simulation makes obvious that CH4-containing
water from the deepest reaches need to be removed not later than in a few decades.

Nutrients - The transport of nutrients to the surface (PO4, Figure 7.2.4)

strongly increases. Right after the start of the extraction the upward flux in 80 m
depth would be enhanced by ~40% and after 50 yr by ~60% compared to natural
(Table 7.2). These numbers even don’t show the full impact, since part of the
nutrient-rich water is reinjected above 80 m depth. This Scenario IZ2 obviously fails
on the Guiding Principle of maintaining the lake ecological integrity.

Methane and energy - The increased upwelling flow between the dilution-
water intake and the reinjection depth leads to an additional loss (relative to IZ4) of
CH4 via oxidation of about ~8 km3 during the 100 yr of simulation. The main loss of
CH4 for this Scenario IZ2 is however due to the inefficient extraction of CH4 (defi-
nition above). The 30% of the CH4 reinjected into the IZ is diluted and lost for future
harvesting. Therefore, this Scenario also fails on the Guiding Principle of maximizing
the economic benefit. This simulation shows drastically, that the fraction of reinjected
CH4 needs either to be minimized (if released into the IZ or the PRZ), or - if techni-
cally not possible - needs to be reinjected into the Resource Zone, such that the
reinjected CH4 remains “harvestable” for later.
The harvested amount of CH4 during 100 yr is 30.5 km3 and the correspon-
ding average electrical energy production is ~90 MW (Table 7.2).

77
 Advection Safety margin
0 0
dilution water depth dilution water depth
50 50
reinjection depth
100 100
reinjection depth
150 150

200 200 2004

Depth (m)

Depth (m)
250 250 2104 no extraction

300 extraction depth 300

350 350 extraction depth

400 400

450 450

500 500
-50 -40 -30 -20 -10 0 10 0 1000 2000 3000 4000 5000 6000
-1 -3
Upwelling advection (m yr ) Energy needed to lift until ptotal = phydrostatic (J m )

Figure 7.2.1 Simulated profiles of vertical Figure 7.2.2 Safety margin for Scenario IZ2
velocity (+: upwelling; -: downwelling) for in 2104 (green) compared to Scenario O (red).
Scenario IZ2 (orange) compared to
Scenario O (black).

Salinity Phosphate
0 0
dilution water depth dilution water depth
50 2004 50
2014 reinjection depth
100 100
2024
reinjection depth
2054 2004
150 150
2104 2014
200 200 2024
Depth (m)

Depth (m)

2054
250 250 2104

300 extraction depth 300 extraction depth

350 350

400 400

450 450

500 500
0 1 2 3 4 5 6 0 0.1 0.2 0.3 0.4
-3
S (kg m ) PO4 (mol m-3)

Figure 7.2.3 Simulated salinity profiles for Figure 7.2.4 Simulated PO4 profiles for
Scenario IZ2. Scenario IZ2.

78
 Methane Carbon dioxide
0 0
dilution water depth dilution water depth
50 50
2004
reinjection depth 2014 reinjection depth
100 100
2024
2054
150
2104
150
2004
200 200 2014

Depth (m)
Depth (m)

2024
250 250 2054
2104
300 extraction depth 300

extraction depth
350 350

400 400

450 450

500 500
0 10 20 30 40 0 50 100 150 200
-3
CH4 (mol m-3) CO2 (mol m )

Figure 7.2.5 Simulated CH4 profiles for Figure 7.2.6 Simulated CO2 profiles for
Scenario IZ2. Scenario IZ2.

Methane extraction concentration Total gas pressure

20 0
dilution water depth
18 50
reinjection depth
16 100

14 150

12 200
CH4 (mol m )
-3

Depth (m)

10 250

8 300 extraction depth

6 350 2004
2014
4 400 2024
2054
2 450 2104
hydrostatic
0 500
2004 2024 2044 2064 2084 2104 0 10 20 30 40
Time (year) Pressure (atm)

Figure 7.2.7 Simulated CH4 extraction Figure 7.2.8 Simulated total gas pressure
concentration as a function of time for profiles for Scenario IZ2.
Scenario IZ2.

79
7.3 Scenario IZ3: Reinjection into Intermediate Zone with dilution

Design parameters - The description of

Scenario IZ3 is similar to IZ2, as the CH4-con- Scenario IZ3
taining deep-water is also extracted (slightly
Extraction flow 70 m3 s-1
deeper) in 350 ± 20 m depth (RZ), dilution-
water from 50 m depth (dilution factor = 1.0) is CH4 reinjected 15 %
used, and the reinject-water is released in 90 ± CO 2 reinjected 60 %
20 m depth. Withdrawal depth 350 m
3 Withdrawal range 20 m
The draw-down lake volume of 320 km ,
below 90 m depth, is withdrawn 0.69 times Dilution factor 1.0
during the 100 yr of simulation (221 km3). Dilution withdrawal depth 50 m
Besides the vertical setting, the diffe- Reinjection depth 90 m
rence to IZ2 is that only half the amount of CH4 Reinjection range 20 m
is reinjected into the lake for IZ3. Therefore, we
expect a better efficiency (less loss) concerning the CH4 harvested.
Smaller differences are due to the slightly deeper withdrawal depth, causing
slightly higher nutrient fluxes and causing slightly higher density. Therefore we
expect slightly higher nutrient fluxes as well as slightly lower stability after 100 yr of
simulation.

Stability and safety - The water volume between reinjection (90 m) and
withdrawal (350 m) depth is entirely replaced after 100 yr (Figures 7.3.3 to 7.3.6)
and as a result this layer is almost homogenised. The density structure remains
strong only below 350 m (to 485 m depth; Figure 7.3.3). Therefore the Schmidt
stability drops after 100 yr to only 48% of the natural stability. Exploiting the
remaining deepest layers will additionally reduce the stability. Therefore this
Scenario does not fulfil the Guiding Principle concerning safety.

Table 7.3 Evaluation of Scenario IZ3

No Extraction Scenario IZ3

2004 2054 2104 2004 2054 2104
PO4 upflux 80 m depth 1.9 2.4 2.9 2.9 6.3 4.6 103 tyr-1
NH4 upflux 80 m depth 15.9 21.0 26.0 24.6 61.8 44.7 103 tyr-1
Schmidt stability 8.8 8.7 8.7 8.8 4.5 4.1 1014 J
CH4 in lake (i) 64.6 81.7 96.3 64.6 47.4 49.3 km3
CH4 harvestable (ii) 42.3 58.9 69.0 42.3 24.4 22.4 km3
CH4 harvested (iii) 32.6 46.2 km3
Average energy output 193 137 MW
CO2 in lake 290 315 336 290 207 177 km3
CO2 removed from lake 16 22 km3

80
 The safety margin (Figure 7.3.2) shows that the minimal energy needed to lift
any water parcel until the gas pressure reaches the hydrostatic pressure would be
found in the deep RZ (~450 m depth: ~30 J m-3) where the gas would accumulate
during the 100 yr period. Therefore, the safety margin in the deepest layers would
shrink (almost as for Scenario O) to dangerous levels and the Guiding Principle of
safety would not be fulfilled. It is evident that CH4-containing water from the deepest
reaches needs to be removed not later than in a few decades.

Nutrients - This Scenario IZ3 would strongly increase the nutrient transport to
the surface. After 50 yr, the flux would be enhanced by 160% (Table 7.3) relative to
natural (Scenario O). Such an intensification of the nutrient cycling is in contradiction
to the Guiding Principle of maintaining the lake ecological integrity.

Methane and energy - The increased upwelling flow between the dilution-
water intake depth and the reinjection depth leads to an additional loss of CH4 (rela-
tive to IZ4) due to the oxidation of ~4 km3 during the 100 yr of simulation. Compared
to Scenario IZ2 the loss to the Biozone is smaller, as for IZ3 the CH4 content of the
reinjected water (15%) is smaller than for IZ2 (30%). But still, as CH4 in the reinject-
water is high, the reinject-water forms a large column above 350 m depth with a not
harvestable CH4 concentration of ~3 mol m3 (Figure 7.3.5). Relative to the most
economic Scenario (IZ4, below), this Scenario IZ3 reveals that ~8.5 km3 of CH4
harvest would be lost (Table 7.13). Therefore, IZ3 also fails on the Guiding Principle
of maximizing the economic benefit.
The harvested amount of CH4 during 100 yr is 46.2 km3 and the correspon-
ding electrical energy production is ~137 MW (Table 7.3).

81
 Advection Safety margin
0 0
dilution water depth dilution water depth
50 50

100 100
reinjection depth
reinjection depth
150 150

200 200
Depth (m)

2004

Depth (m)
250 250
2104 no extraction
300 300
extraction depth extraction depth
350 350

400 400

450 450

500 500
-50 -40 -30 -20 -10 0 10 0 1000 2000 3000 4000 5000 6000
-1
Upwelling advection (m yr ) Energy needed to lift until ptotal = phydrostatic (J m-3)

Figure 7.3.1 Simulated profiles of vertical Figure 7.3.2 Safety margin for Scenario IZ3
velocity (+: upwelling; -: downwelling) for in 2104 (green) compared to Scenario O (red).
Scenario IZ3 (orange), compared with
Scenario O (black).

Salinity Phosphate
0 0
dilution water depth dilution water depth
50 50

100 100
reinjection depth reinjection depth
2004 150
150
2014 2004
2024 200 2014
200
Depth (m)
Depth (m)

2054 2024
250 2104 250 2054
2104
300 300
extraction depth extraction depth
350 350

400 400

450 450

500 500
0 1 2 3 4 5 6 0 0.1 0.2 0.3 0.4
-3 -3
S (kg m ) PO4 (mol m )

Figure 7.3.3 Simulated salinity profiles for Figure 7.3.4 Simulated PO4 profiles for
Scenario IZ3. Scenario IZ3.

82
 Methane Carbon dioxide
0 0
dilution water depth dilution water depth
50 50

100 100
reinjection depth reinjection depth
150 150
2004 2004
2014 2014
200 200
2024

Depth (m)
Depth (m)

2024
2054
250 250 2054
2104
2104
300 300
extraction depth extraction depth
350 350

400 400

450 450

500 500
0 10 20 30 40 0 50 100 150 200
-3
CH4 (mol m-3) CO2 (mol m )

Figure 7.3.5 Simulated CH4 profiles for Figure 7.3.6 Simulated CO2 profiles for
Scenario IZ3. Scenario IZ3.

Methane extraction concentration Total gas pressure

25 0
dilution water depth
50

20 100
reinjection depth
150

15
CH4 (mol m-3)

200
Depth (m)

250

10 300
extraction depth
350 2004
2014
5 400 2024
2054
450 2104
hydrostatic
0 500
2004 2024 2044 2064 2084 2104 0 10 20 30 40
Time (year) Pressure (atm)

Figure 7.3.7 Simulated CH4 extraction Figure 7.3.8 Simulated total gas pressure
concentration as a function of time for profiles for Scenario IZ3.
Scenario IZ3.

83
7.4 Scenario IZ4: Reinjection into Intermediate Zone - no dilution

Design parameters - This Scenario IZ4 is

identical to IZ1 except that no dilution-water is used Scenario IZ4
for density adjustment. It is assumed that the Extraction flow 70 m3 s-1
reinject-water is mixed efficiently into the ambient
CH4 reinjected 5.9 %
lake water by technological measures (multiple and
CO2 reinjected 60 %
spatial spreading of the outlets). In the model, this is
achieved by homogenous distribution of the re- Withdrawal depth 475 m
inject-water into the ±50 m thick reinjection layer. Withdrawal range 10 m
3
The draw-down lake volume is 239 km Dilution factor 0
(downwelling below 164 m depth). This volume is (no dilution)
withdrawn 0.92 times (221 km3) during the 100 yr of Reinjection depth 150 m
simulation. Reinjection range 100 m
Note that the density of the lake water layer
receiving the reinject-water (100 to 200 m depth) rises rapidly and reaches the
density of the Potential Resource Zone in less than 20 yr (Figure 7.4.3). Thereafter,
further density increase causes an expansion of the receiving layer which reaches
down to ~360 m depth after 50 yr (Figures 7.4.3 to 7.4.6). The deep layers below
(RZ) remain stratified and with higher CH4 concentrations.

Stability and safety - The lake volume between the reinjection depth and the
withdrawal depth is moving downwards to the deepest reaches (= withdrawal level).
The IZ, particularly the layer at the reinjection depth, widens rapidly in the vertical
direction (see above) and is descending with the draw-down. Similar as in Scenario
IZ1, the density stratification flattens, and the density difference between surface
and the down-moving front shrinks (Figure 7.4.3), however, at a much lower rate
than for IZ1. The Schmidt stability decreases by about 50% after 100 yr of simulation,
compared to natural (Table 7.4). The density gradient between the surface and the
reinjection depth however remains much stronger than in Scenario IZ1.

Table 7.4 Evaluation of Scenario IZ4

No Extraction Scenario IZ4

2004 2054 2104 2004 2054 2104
PO4 upflux 80 m depth 1.9 2.4 2.9 1.9 4.8 5.2 103 tyr-1
NH4 upflux 80 m depth 15.9 21.0 26.0 15.9 43.8 46.1 103 tyr-1
Schmidt stability 8.8 8.7 8.7 8.8 5.1 4.2 1014 J
CH4 in lake (i) 64.6 81.7 96.3 64.6 35.9 31.5 km3
CH4 harvestable (ii) 42.3 58.9 69.0 42.3 7.6 9.2 km3
CH4 harvested (iii) 47.2 67.8 km3
Average energy output 280 201 MW
CO2 in lake 290 315 336 290 220 182 km3
CO2 removed from lake 86 120 km3

84
 The analysis of the safety margin does not reveal much difference to IZ1. The
safety margin increases with depth and therefore the lowest relevant value is taken
in 200 m depth (170 J m-3). As no dilution water is used, the overall stratification is
less degraded than for IZ1. But still, the safety margin is not improved relative to
Scenario O (natural, no extraction; Table 7.13). However the risk of degassing re-
mains low due to the extraction of gases, which reduces the gas content massively
(Table 7.4).

Nutrients -The nutrient transport to the surface layer doubles after ~50 yr of
simulation compared to the natural development of the lake. This increase is due to
the nutrient-rich deep-water, released below the Biozone and naturally lifted into the
Biozone by the upwelling flow of the subaquatic sources (see IZ1). This Scenario IZ4
therefore fails on the Guiding Principle of maintaining the lake ecological integrity.

Methane and energy - It is not well-known, how much the turbulent vertical
mixing would increase due to the weakening of the PRZ stratification. From Figure
7.4.4 it is obvious that the CH4 loss to the Biozone increases with extraction time.
However, compared to the natural development, the extraction and reinjection of the
CH4-depleetd water causes in fact a reduction of the CH4-loss of ~3 km3, mainly
because the concentrations below the Biozone become lower.
The CH4 concentration at the deep-water intake remains high and relatively
constant for about 50 yr before it drops drastically by almost a factor of 4 during a
~20 yr period (Figure 7.4.7) from ~20 to ~5 mol m-3.
The harvested amount of CH4 during 100 yr is 68 km3 and the corresponding
electrical energy production is ~200 MW (Table 7.4). This Scenario is the one with
the highest CH4 harvest.

85
 Advection Safety margin
0 0

50 50

100 100
reinjection depth reinjection depth
150 150

200 200
Depth (m)

Depth (m)
250 250
2004
300 300
2104 no extraction
350 350

400 400

450 extraction depth 450 extraction depth

500 500
-50 -40 -30 -20 -10 0 10 0 1000 2000 3000 4000 5000 6000
-1 -3
Upwelling advection (m yr ) Energy needed to lift until ptotal = phydrostatic (J m )

Figure 7.4.1 Simulated profiles of vertical Figure 7.4.2 Safety margin for Scenario IZ4
velocity (+: upwelling; -: downwelling) for in 2104 (green) compared to Scenario O (red).
Scenario IZ4 (orange), compared with
Scenario O (black).

Salinity Phosphate
0 0

50 2004 50 2004
2014 2014
100 100
2024 2024
reinjection depth 2054 reinjection depth 2054
150 150
2104 2104
200 200
Depth (m)

Depth (m)

250 250

300 300

350 350

400 400

450 extraction depth 450 extraction depth

500 500
0 1 2 3 4 5 6 0 0.1 0.2 0.3 0.4
-3 -3
S (kg m ) PO4 (mol m )

Figure 7.4.3 Simulated salinity profiles for Figure 7.4.4 Simulated PO4 profiles for
Scenario IZ4. Scenario IZ4.

86
 Methane Carbon dioxide
0 0

50 2004 50 2004
2014 2014
100 100
2024 2024
reinjection depth 2054 reinjection depth 2054
150 150
2104 2104
200 200
Depth (m)

Depth (m)
250 250

300 300

350 350

400 400

450 extraction depth 450 extraction depth

500 500
0 5 10 15 20 25 0 50 100 150 200
-3 -3
CH4 (mol m ) CO2 (mol m )

Figure 7.4.5 Simulated CH4 profiles for Figure 7.4.6 Simulated CO2 profiles for
Scenario IZ4. Scenario IZ4.

Methane extraction concentration Total gas pressure

25 0

50

20 100
reinjection depth
150
2004
15
CH4 (mol m-3)

200 2014
Depth (m)

2024
250 2054
2104
10 300 hydrostatic

350

5 400

450 extraction depth

0 500
2004 2024 2044 2064 2084 2104 0 10 20 30 40
Time (year) Pressure (atm)

Figure 7.4.7 Simulated CH4 extraction Figure 7.4.8 Simulated total gas pressure
concentration as a function of time for profiles for Scenario IZ4.
Scenario IZ4.

87
7.5 Scenario IZ5: Reinjection into Intermediate Zone - no dilution

Design parameters - Scenario IZ5 is

similar to IZ2, as CH4-containing deep-water is Scenario IZ5
withdrawn from the RZ (320 ± 5 m) and reinjected
Extraction flow 70 m3 s-1
at the lower end of the IZ (190 ± 5 m depth). How-
ever, dilution-water is avoided (compared to IZ2). CH4 reinjected 30 %
It is again assumed that the reinjected water is CO2 reinjected 60 %
mixed efficiently into the ambient lake water by Withdrawal depth 320 m
technological measures (see IZ4). Withdrawal range 10 m
The water volume between 190 and 320 m Dilution factor no dilution
depth (158 km3) is drawn down by the extraction Reinjection depth 190 m
volume (221 km3) during the simulated 100 yr Reinjection range 10 m
(1.4-times replaced).

Stability and safety - The water of the PRZ, which is between the withdrawal
(320 m) and the reinjection depth (190 m) is moving downward to the withdrawal
level. The layer at the reinjection depth widens vertically and descends following the
downwelling in the lake (Figure 7.5.1). The layers above the reinjection depth (190 m)
are only indirectly affected. Due to the natural upwelling (subaquatic sources) the
receiving water layer is also moving upwards. Subsequently, a quite homogenous
layer develops between 320 m and ~120 m depth with increasing concentrations of
salinity, gases and nutrients at the upper end (~120 m) of this new layer. The lake
waters above 100 m and below 320 m depth are only slightly modified.
As an effect, the overall stratification weakens much less (than for the pre-
vious Scenarios), and the Schmidt stability decreases by only ~18% after 100 yr of
simulation compared to Scenario O (Table 7.5). For this Scenario IZ5, the reduction
of the water column stability is not a concern.
The safety margin, remains on a natural level in the upper half of the lake. In
the RZ the safety margin drops, due to the accumulation of gases, with time and

Table 7.5 Evaluation of Scenario IZ5

No Extraction Scenario IZ5

2004 2054 2104 2004 2054 2104
PO4 upflux 80 m depth 1.9 2.4 2.9 1.9 2.4 3.4 103 tyr-1
NH4 upflux 80 m depth 15.9 21.0 26.0 15.9 20.9 30.4 103 tyr-1
Schmidt stability 8.8 8.7 8.7 8.8 7.7 7.2 1014 J
CH4 in lake (i) 64.6 81.7 96.3 64.6 60.1 61.7 km3
CH4 harvestable (ii) 42.3 58.9 69.0 42.3 33.8 33.7 km3
CH4 harvested (iii) 20.2 30.5 km3
Average energy output 120 90 MW
CO2 in lake 290 315 336 290 257 241 km3
CO2 removed from lake 52 79 km3
after 100 yr it reaches a level of only ~33 J m-3 in ~450 m depth. As there is no
extraction of gases in the deepest part, the safety would be almost as critical as
without extraction (Scenario O; Table 7.13). Therefore, also extraction of CH4 from
the deepest parts needs to start not later than in a few decades.

Nutrients - Due to the lower reinjection depth, and due to avoiding dilution-
water, the upwelling as well as the nutrient concentration below the Biozone are
lower. As a result, the upward transport of nutrients is significantly lower for this
Scenario, then for IZ1 to IZ3 (with dilution-water), and is enhanced by only ~20%
after 100 yr relative to Scenario O. The Guiding Principle of “lake ecological integrity”
would just marginally be fulfilled.

Methane and energy - Despite the deep (190 m) reinjection and despite no
dilution-water is used, the additional loss to the Biozone is about ~7 km3 (relative to
IZ4) during the 100 yr of extraction (Figure 7.5.5, Table 7.5). The main loss of CH4
for this Scenario IZ5 is, however, due to the large volume of not harvestable CH4 (~3
to 4 mol m-3) in the lake water above 320 m depth (Figure 7.5.5). This “wasted” CH4
of high concentration is due to the inefficient extraction of only 70% of CH4 (see
above). The 30% of the reinjected CH4 is diluted into the IZ and PRZ and lost for
future harvesting. Therefore, this Scenario fails the Guiding Principle of maximizing
the economic benefit.
There is a drastic drop in the CH4 concentration at the intake (320 m depth)
between 2025 und 2035 from ~16 to ~6 mol m-3 (although this could be avoided by
continuously moving the intake deeper). The harvested amount of CH4 during 100 yr
is 30.5 km3 and the corresponding electrical energy production is ~90 MW (Table
7.5).

89
 Advection Safety margin
0 0

50 50 2004

100 100 2104 no extraction

150 150
reinjection depth reinjection depth
200 200
Depth (m)

Depth (m)
250 250

300 extraction depth 300 extraction depth

350 350

400 400

450 450

500 500
-50 -40 -30 -20 -10 0 10 0 1000 2000 3000 4000 5000 6000
-1 -3
Upwelling advection (m yr ) Energy needed to lift until ptotal = phydrostatic (J m )

Figure 7.5.1 Simulated profiles of vertical Figure 7.5.2 Safety margin for Scenario IZ5
velocity (+: upwelling; -: downwelling) for in 2104 (green) compared to Scenario O (red).
Scenario IZ5 (orange), compared with
Scenario O (black).

Salinity Phosphate
0 0
2004 2004
50 2014 50 2014
2024 2024
100 2054 100 2054
2104 2104
150 150
reinjection depth reinjection depth
200 200
Depth (m)

Depth (m)

250 250

300 extraction depth 300 extraction depth

350 350

400 400

450 450

500 500
0 1 2 3 4 5 6 0 0.1 0.2 0.3 0.4
-3 -3
S (kg m ) PO4 (mol m )

Figure 7.5.3 Simulated salinity profiles for Figure 7.5.4 Simulated PO4 profiles for
Scenario IZ5. Scenario IZ5.

90
 Methane Carbon dioxide
0 0
2004
50 50 2014
2004
2014 2024
100 100 2054
2024
2054 2104
150
2104 150
reinjection depth
200 200
reinjection depth

Depth (m)
Depth (m)

250 250

300 extraction depth 300

extraction depth
350 350

400 400

450 450

500 500
0 10 20 30 40 0 50 100 150 200
-3
CH4 (mol m-3) CO2 (mol m )

Figure 7.5.5 Simulated CH4 profiles for Figure 7.5.6 Simulated CO2 profiles for
Scenario IZ5. Scenario IZ5.

Methane extraction concentration Total gas pressure

20 0

18 50

16 100

14 150
reinjection depth
12
CH4 (mol m )

200
-3

Depth (m)

10 250

8 300 extraction depth

6 350 2004
2014
4 400 2024
2054
2 450 2104
hydrostatic
0 500
2004 2024 2044 2064 2084 2104 0 10 20 30 40
Time (year) Pressure (atm)

Figure 7.5.7 Simulated CH4 extraction Figure 7.5.8 Simulated total gas pressure
concentration as a function of time for profiles for Scenario IZ5.
Scenario IZ5.

91
7.6 Scenario PR1: Reinjection into Potential Resource Zone

Design parameters - The difference of Scenario PR1

PR1 to the IZ Scenarios is mainly the depth of
the reinjection. The lower the reinjection, the Extraction flow 70 m3 s-1
smaller the density difference between the Fraction of CH4 reinjected 5.9 %
CH4-depleted reinject-water and the ambient Fraction of CO2 reinjected 60 %
lake water. As an effect, the modification to the
Withdrawal depth 475 m
vertical structure of salinity, CO2 and subse-
Withdrawal range 10 m
quently to density is smaller.
For PR1 the withdrawal (475 ± 5 m) and Reinjection depth 240 m
Reinjection range 60 m
reinjection (210 to 270 m) depth is set deeper
and the intention of PR1 is restratifying the reinject-water into the Potential Resource
Zone (without dilution-water). The layer between withdrawal and reinjection (155 km3
of downwelling water below 228 m depth) is 1.42-times replaced (221 km3) during
the 100 yr simulation time.

Stability and safety - As the volume between withdrawal and reinjection is

relatively small, the concentrations of S, gases and nutrients are quite homogenous
in the lower half of the lake (Figures 7.6.3 to 7.6.6) and the density stratification
below the reinjection depth shrinks to a minimum. Above 100 m depth the modifica-
tions to salinity and CO2 are minor and therefore the overall vertical density structure
(density difference from top to max depth) remains intact to a large extent. The
density gradient above the reinjection depth strengthens and becomes strongest in
the lake after 50 yr of extraction. As a result, after 100 yr of extraction, the Schmidt
stability is only reduced by ~20 % relative to natural.
The safety margin improves during the 100 yr of extraction over the entire
water column and will be much less critical than today. Even in the relatively homo-

Table 7.6 Evaluation of Scenario PR1

No Extraction Scenario PR1

2004 2054 2104 2004 2054 2104
PO4 upflux 80 m depth 1.9 2.4 2.9 1.9 2.4 2.9 103 tyr-1
NH4 upflux 80 m depth 15.9 21.0 26.0 15.9 21.0 26.2 103 tyr-1
Schmidt stability 8.8 8.7 8.7 8.8 7.6 7.0 1014 J
CH4 in lake (i) 64.6 81.7 96.3 64.6 35.5 30.4 km3
CH4 harvestable (ii) 42.3 58.9 69.0 42.3 5.1 4.5 km3
CH4 harvested (iii) 47.1 66.4 km3
Average energy output 279 197 MW
CO2 in lake 290 315 336 290 231 204 km3
CO2 removed from lake 86 133 km3

92
geneous deep water the safety margin reaches very high (safe) values. They are
lowest but still comfortably high at the deepest point with 1920 J m-3. Compared to
the IZ Scenarios, PR1 is on a much safer level after 100 yr of extraction, because
the overall strong density stratification remains while the higher gas concentrations
in the deepest reaches are removed.

Nutrients - The nutrient transport to the surface increases insignificantly, only

~1% after 100 yr of extraction. It is possible that it will be higher in the (long-term)
steady state if the nutrient concentrations in the layer between 100 and 200 m depth
increase due to the higher concentrations in the Potential Resource Zone.

Methane and energy - There is a benefit of reinjecting the extracted deep-

water above the main density gradient: The high CH4 concentrations of the
Resource Zone are not diluted and the available CH4 reserves can be depleted to a
lower level. The efficiency of the power plants is most probably higher as the extrac-
ted CH4 concentrations are higher.
The CH4 flux to the surface should not increase noteworthily as the density
gradient strengthened above the Potential Resource Zone replaces the function of
the present main density gradient in ~180 m depth.
The harvested amount of CH4 during 100 yr is ~66 km3 and the corresponding
electrical energy production is ~200 MW (Table 7.6).

93
 Advection Safety margin
0 0

50 50 2004

100 100 2104 no extraction

2104 Scenario PR1
150 150

200 200
Depth (m)

reinjection depth

Depth (m)
reinjection depth
250 250

300 300

350 350

400 400

450 extraction depth 450 extraction depth

500 500
-50 -40 -30 -20 -10 0 10 0 1000 2000 3000 4000 5000 6000
-1 -3
Upwelling advection (m yr ) Energy needed to lift until ptotal = phydrostatic (J m )

Figure 7.6.1 Simulated profiles of vertical Figure 7.6.2 Safety margin for Scenario PR1
velocity (+: upwelling; -: downwelling) for in 2104 (green) compared to Scenario O (red).
Scenario IZ5 (orange), compared with
Scenario O (black).

Salinity Phosphate
0 0

50 2004 50 2004
2014 2014
100 100
2024 2024
2054 2054
150 150
2104 2104
200 200
Depth (m)

Depth (m)

reinjection depth reinjection depth

250 250

300 300

350 350

400 400

450 extraction depth 450 extraction depth

500 500
0 1 2 3 4 5 6 0 0.1 0.2 0.3 0.4
-3 -3
S (kg m ) PO4 (mol m )

Figure 7.6.3 Simulated salinity profiles for Figure 7.6.4 Simulated PO4 profiles for
Scenario PR1. Scenario PR1.

94
 Methane Carbon dioxide
0 0

50 2004 50 2004
2014 2014
100 100
2024 2024
2054 2054
150 150
2104 2104
200 200
Depth (m)

Depth (m)
reinjection depth reinjection depth
250 250

300 300

350 350

400 400

450 extraction depth 450 extraction depth

500 500
0 5 10 15 20 25 0 50 100 150 200
-3 -3
CH4 (mol m ) CO2 (mol m )

Figure 7.6.5 Simulated CH4 profiles for Figure 7.6.6 Simulated CO2 profiles for
Scenario PR1. Scenario PR1.

Methane extraction concentration Total gas pressure

25 0
2004
50 2014
2024
20 100 2054
2104
150
hydrostatic

15 200
CH4 (mol m )
-3

Depth (m)

reinjection depth
250

10 300

350

5 400

450 extraction depth

0 500
2004 2024 2044 2064 2084 2104 0 10 20 30 40
Time (year) Pressure (atm)

Figure 7.6.7 Simulated CH4 extraction Figure 7.6.8 Simulated total gas pressure
concentration as a function of time for profiles for Scenario PR1.
Scenario PR1.

95
7.7 Scenario RZ1: Single reinjection into Resource Zone

Design parameters - The intention of

Scenario RZ1
this Scenario RZ1 is to draw-down the entire
Resource Zone to the deepest point - by rein- Extraction flow 70 m3 s-1
jecting the water at the upper end of the RZ
Fraction of CH4 reinjected 5%
(290 ± 15 m depth) - and thereby not to affect Fraction of CO2 reinjected 78 %
the lake body above. The modifications to the
vertical structure of salinity, CO2, density and Withdrawal depth 475 m
Withdrawal range 10 m
nutrients are small and much less than for the
IZ Scenarios. Reinjection depth 290 m
The layer between withdrawal (475 ± 5 Reinjection range 30 m
3
m) and reinjection depth (112 km of downwelling water below 275 m depth) is 2.0-
times replaced (221 km3) during the 100 yr simulation time.

Stability and safety - As evident from the Figures 7.7.3 to 7.7.6, the major
changes to the lake stratification occurs only in the lowest 200 m and the lake above
is hardly affected. Within the Resource Zone the density stratification shrinks to a
minimum (Figure 7.7.3). But the main density gradient above the RZ remains intact
and steep. Therefore the overall stability is maintained and even after 100 yr of
extraction, the Schmidt stability is still at the natural (undisturbed) level (Table 7.7).
The safety margin is minimal at 200 m depth but is comfortably high at 916 J
-3
m . Below, and particularly in the RZ, the safety margin reaches high levels (Figure
7.7.2) - much higher than today - because the gases are extracted in the RZ.

Nutrients - The nutrient concentrations and the vertical nutrient gradients

slightly increase with time (Figure 7.7.4). This increase is, however, related to the
nutrient balance of the lake (Pasche et al 2009; Muvundja et al 2009) and is not
related to the extraction process. Therefore, the nutrient transport to the surface
does not increase during the 100 yr of simulation compared to natural (Table 7.7).

Table 7.7 Evaluation of Scenario RZ1

No Extraction Scenario RZ1

2004 2054 2104 2004 2054 2104
PO4 upflux 80 m depth 1.9 2.4 2.9 1.9 2.4 2.9 103 tyr-1
NH4 upflux 80 m depth 15.9 21.0 26.0 15.9 21.0 26.0 103 tyr-1
Schmidt stability 8.8 8.7 8.7 8.8 8.9 8.7 1014 J
CH4 in lake (i) 64.6 81.7 96.3 64.6 44.1 39.3 km3
CH4 harvestable (ii) 42.3 58.9 69.0 42.3 20.9 12.1 km3
CH4 harvested (iii) 38.3 57.5 km3
Average energy output 227 171 MW
CO2 in lake 290 315 336 290 269 251 km3
CO2 removed from lake 47 85 km3

96
We conclude that for all RZ Scenarios the nutrient transport to the Biozone will not
change before reaching a steady-state. In a few hundred yr the nutrient flux to the
Biozone will increase, however, independent of the extraction process.

Methane and energy - Oxidation and gas exchange to the atmosphere are
the only ways to loose CH4. As the IZ and the PRZ are only marginally affected by
Scenario RZ1, we expect almost no additional losses due to the extraction operation.
In fact the model simulations indicate even a slightly smaller loss, as a result of
upwelling lower CH4 concentrations after some decades of extraction. A potential
disadvantage of Scenario RZ1 is evident from Figure 7.7.7 which shows that the
CH4 concentration at the intake collapses after ~25 yr within a very brief time span
by a factor of 2 from ~20 to ~10 mol m-3.
The Potential Resource Zone is not affected by this extraction method and
can be treated separately as soon as its CH4 concentration has grown to an extrac-
table level.
The benefit of this deep zone extraction method is that the low CH4 concen-
trations at the reinjection depth are being recharged with renewing CH4 while they
are moving downwards towards the withdrawal depth. The CH4 gradient which evol-
ves thereby allows an extraction of higher CH4 concentrations and a depletion of the
CH4 stock to a lower level in comparison to the “completely mixed” RZ2 and RZ3
(see below). The efficiency of the power plants will be larger due to higher CH4
concentrations.
The harvested amount of CH4 during 100 yr is ~58 km3 and the corresponding
electrical energy production is ~170 MW (Table 7.7).

97
 Advection Safety margin
0 0

50 50 2004

100 100 2104 no extraction

2104 Scenario RZ1
150 150

200 200
Depth (m)

Depth (m)
250 250
reinjection depth reinjection depth
300 300

350 350

400 400

450 extraction depth 450 extraction depth

500 500
-50 -40 -30 -20 -10 0 10 0 1000 2000 3000 4000 5000 6000
-1 -3
Upwelling advection (m yr ) Energy needed to lift until ptotal = phydrostatic (J m )

Figure 7.7.1 Simulated profiles of vertical Figure 7.7.2 Safety margin for Scenario RZ1
velocity (+: upwelling; -: downwelling) for in 2104 (green) compared to Scenario O (red).
Scenario RZ1 (orange), compared with
Scenario O (black).

Salinity Phosphate
0 0

50 2004 50 2004
2014 2014
100 100
2024 2024
2054 2054
150 150
2104 2104
200 200
Depth (m)

Depth (m)

250 250
reinjection depth
300 300
reinjection depth
350 350

400 400

450 extraction depth 450 extraction depth

500 500
0 1 2 3 4 5 6 0 0.1 0.2 0.3 0.4
-3 -3
S (kg m ) PO4 (mol m )

Figure 7.7.3 Simulated salinity profiles for Figure 7.7.4 Simulated PO4 profiles for
Scenario RZ1. Scenario RZ1.

98
 Methane Carbon dioxide
0 0

50 2004 50 2004
2014 2014
100 100
2024 2024
2054 2054
150 150
2104 2104
200 200
Depth (m)

Depth (m)
250 reinjection depth 250 reinjection depth
300 300

350 350

400 400

450 extraction depth 450 extraction depth

500 500
0 5 10 15 20 25 0 50 100 150 200
-3 -3
CH4 (mol m ) CO2 (mol m )

Figure 7.7.5 Simulated CH4 profiles for Figure 7.7.6 Simulated CO2 profiles for
Scenario RZ1. Scenario RZ1.

Methane extraction concentration Total gas pressure

25 0
2004
50 2014
2024
20 100 2054
2104
150
hydrostatic
15
CH4 (mol m )

200
-3

Depth (m)

250 reinjection depth

10 300

350

5 400

450 extraction depth

0 500
2004 2024 2044 2064 2084 2104 0 10 20 30 40
Time (year) Pressure (atm)

Figure 7.7.7 Simulated CH4 extraction Figure 7.7.8 Simulated total gas pressure
concentration as a function of time for profiles for Scenario RZ1.
Scenario RZ1.

99
7.8 Scenario RZ2: Double reinjection into Resource Zone
Scenario RZ2
Design parameters - The intention of
Scenario RZ2 is to extract CH4 over the entire Extraction flow 70 m3 s-1
RZ without causing any changes to the overall Fraction of CH reinjected 5%
4
stratification of the lake. As lake water is Fraction of CO2 reinjected 78 %
extracted and reinjected in the same layers
(from 275 m to 480 m depth) there is no 1. Withdrawal depth 400 m
additional (relative to natural) up- or down- 1. Withdrawal range 160 m
welling induced by the extraction process 1. Reinjection depth 400 m
(Figure 7.8.1). The extraction (two depth 1. Reinjection range 160 m
ranges) and reinjection (two depth ranges) is
proposed to be realized by two sets of pipes 2. Withdrawal depth 290 m
to harvest the Upper RZ (275 to 305 m depth) 2. Withdrawal range 30 m
and the Lower RZ (320 to 480 m depth) at the 2. Reinjection depth 290 m
same pace. The water flow of 70 m3 s-1 is split 2. Reinjection range 30 m
between the two zones to reach homogenous
CH4 concentrations in the two layers: 75% for the primary extraction and 25% for the
secondary extraction (Table 5.5). The lake volume between extraction and
reinjection depth (104 km3) is 2.13 times flushed during the 100 yr of extraction
(volume: 221 km3).
The lake water body above the RZ is not affected by this extraction method
and can be treated separately as soon as its CH4 concentration has grown to extrac-
table level. Therefore we do not expect any changes of the overall stratification or of
the lake internal (nutrient) fluxes.

Stability and safety - As evident from Figures 7.8.3 to 7.8.6, the changes to
the lake stratification occur only in the RZ (lowest 200 m) and the lake above 275 m
depth is not affected (see explanations for RZ1). Therefore, the Schmidt stability

Table 7.8 Evaluation of Scenario RZ2

No Extraction Scenario RZ2

2004 2054 2104 2004 2054 2104
PO4 upflux 80 m depth 1.9 2.4 2.9 1.9 2.4 2.9 103 tyr-1
NH4 upflux 80 m depth 15.9 21.0 26.0 15.9 21.0 26.0 103 tyr-1
Schmidt stability 8.8 8.7 8.7 8.8 8.8 8.7 1014 J
CH4 in lake (i) 64.6 81.7 96.3 64.6 54.2 51.2 km3
CH4 harvestable (ii) 42.3 58.9 69.0 42.3 33.9 28.1 km3
CH4 harvested (iii) 28.3 45.9 km3
Average energy output 168 136 MW
CO2 in lake 290 315 336 290 276 263 km3
CO2 removed from lake 44 79 km3

100
remains at the natural (undisturbed) level (Table 7.8). The safety margin is at the
minimum in 200 m depth at a high value of 926 J m-3 (see explanations for RZ1).

Nutrients - The nutrient transport to the Biozone (Table 7.8) does not increa-
se during the 100 yr of simulation compared to the natural fluxes to the surface (see
explanations for RZ1).

Methane and energy - As explained for RZ1, there is no additional loss of

CH4 due to the extraction operation. The effect of the RZ-wide withdrawal and
reinjection is the continuous and steady decrease of the CH4 concentrations at the
intake (Figure 7.8.7). This continuous CH4 decrease for RZ2 is in contrast to the IZ
and the PR1 Scenarios which all showed some rather discontinuous decrease in
concentration (see examples in Figures 7.2.7, 7.5.7).
If the extraction rate for RZ2 would be chosen low, such that it would not
exceed the CH4 new formation by much, then RZ2 could be run for very long,
without causing any changes to the zones above. After many decades, some of the
PRZ could become economically harvestable and could be extracted independently
of the RZ.
It is important to realize that Scenario RZ2 leads to almost homogenous CH4
concentrations which finally would approach the economic threshold (5 mol m-3
assumed in this report). Before this level is reached, the extraction Scenario needs
to be changed, such that lake water is drawn-down (at higher CH4) and the reinject-
water is stacked back into the lake above the draw-down zone. Scenarios PR1 or
RZ4 offer such options. This choice, which should also account for the PRZ CH4
concentration, can be made in several decades in the future.
The harvested amount of CH4 during 100 yr is ~46 km3 and the corresponding
electrical energy production is ~136 MW (Table 7.8).

101
 Advection Safety margin
0 0

50 50 2004

100 100 2104 no extraction

2104 Scenario RZ2
150 150

200 200
Depth (m)

Depth (m)
secondary extraction
250 secondary extraction
and reinjection 250
and reinjection
300 300

350 primary extraction 350

and reinjection primary extraction
400 400
and reinjection
450 450

500 500
-50 -40 -30 -20 -10 0 10 0 1000 2000 3000 4000 5000 6000
-1 -3
Upwelling advection (m yr ) Energy needed to lift until ptotal = phydrostatic (J m )

Figure 7.8.1 Simulated profiles of vertical Figure 7.8.2 Safety margin for Scenario RZ2
velocity for Scenario RZ2 is identical to in 2104 (green) compared to Scenario O (red).
Scenario O (no extraction), as the water is
reinjected into the same volume as
extracted.

Salinity Phosphate
0 0

50 50 2004
2004
2014 2014
100 100
2024 2024
2054 2054
150 150
2104 2104

200 200
Depth (m)
Depth (m)

secondary extraction secondary extraction

250 250
and reinjection and reinjection
300 300

350 350 primary extraction

primary extraction
and reinjection
and reinjection 400
400

450 450

500 500
0 1 2 3 4 5 6 0 0.1 0.2 0.3 0.4
-3
-3
S (kg m ) PO4 (mol m )

Figure 7.8.3 Simulated salinity profiles for Figure 7.8.4 Simulated PO4 profiles for
Scenario RZ2. Scenario RZ2.

102
 Methane Carbon dioxide
0 0

50 2004 50 2004
2014 2014
100 100
2024 2024
2054 2054
150 150
2104 2104
200 200
Depth (m)

Depth (m)
secondary extraction secondary extraction
250 250 and reinjection
and reinjection
300 300

350 350
primary extraction primary extraction
and reinjection and reinjection
400 400

450 450

500 500
0 5 10 15 20 25 0 50 100 150 200
-3 -3
CH4 (mol m ) CO2 (mol m )

Figure 7.8.5 Simulated CH4 profiles for Figure 7.8.6 Simulated CO2 profiles for
Scenario RZ2. Scenario RZ2.

Methane extraction concentration Total gas pressure

20 0
2004
18 primary extraction 50 2014
secondary extraction 2024
16 100 2054
2104
14 150
hydrostatic

12 200
CH4 (mol m-3)

Depth (m)

secondary extraction
10 250
and reinjection
8 300

6 350

4 400
primary extraction
2 450 and reinjection

0 500
2004 2024 2044 2064 2084 2104 0 10 20 30 40
Time (year) Pressure (atm)

Figure 7.8.7 Simulated CH4 extraction Figure 7.8.8 Simulated total gas pressure
concentrations as a function of time for profiles for Scenario RZ2.
both pipes of Scenario RZ2.

103
7.9 Scenario RZ3: Long-term effects of gas removal efficiency
Scenario RZ3
Design parameters - This Scenario RZ3
is identical to RZ2 except that the fraction of Extraction flow 70 m3 s-1
reinjected gases is different: For RZ3 only 46%
Fraction of CH4 reinjected 3.4 %
and 3.4 % of the extracted CO2 and CH4, Fraction of CO2 reinjected 46 %
respectively, remain in the reinject-water.
1. Withdrawal depth 400 m
1. Withdrawal range 160 m
Stability and safety - As only 46% of
the CO2 is reinjected, the deep-water becomes 1. Reinjection depth 400 m
lighter in the long-term, while CO2 concentra- 1. Reinjection range 160 m
tions in the lake decrease. Although the strati- 2. Withdrawal depth 290 m
fication structure remains, the density difference 2. Withdrawal range 30 m
between surface and deep-water is reduced 2. Reinjection depth 290 m
and therefore the Schmidt stability is signifi- 2. Reinjection range 30 m
cantly (~13%) reduced (Table 7.9.1) but still
very safe. The safety margin remains the same (929 J m-3), as at the critical depth in
200 m the changes to the stratification by RZ2 or RZ3 are negligible.

CH4/CO2 ratio in the extraction water - If RZ Scenarios are chosen as

extraction option, then the reinject-water evidently influences in the long-term the
gas composition of the RZ water. This influence becomes stronger over time, as the
fraction of recycled water is increasing, until the steady-state (long-term continuous
extraction) would be achieved. For steady-state the CH4/CO2 ratio is solely deter-
mined by the Design Parameters of the extraction plant (Scenario) and by the rate of
recharge of CO2 and CH4. Practically, the threshold concentration for extractable
CH4 (5 mol m-3) would define how much water can be extracted in order to stay in
steady-state balance with new formation, extraction and reinjection.

Table 7.9 Evaluation of Scenario RZ3

No Extraction Scenario RZ3

2004 2054 2104 2004 2054 2104
PO4 upflux 80 m depth 1.9 2.4 2.9 1.9 2.4 2.9 103 tyr-1
NH4 upflux 80 m depth 15.9 21.0 26.0 15.9 21.0 26.0 103 tyr-1
Schmidt stability 8.8 8.7 8.7 8.8 8.0 7.5 1014 J
CH4 in lake (i) 64.6 81.7 96.3 64.6 53.2 49.8 km3
CH4 harvestable (ii) 42.3 58.9 69.0 42.3 33.1 26.9 km3
CH4 harvested (iii) 29.3 47.3 km3
Average energy output 174 140 MW
CO2 in lake 290 315 336 290 233 199 km3
CO2 removed from lake 93 151 km3

104
 However, we assume that at least for the next few decades the steady-state
concept is not the envisaged harvesting option. But also for much shorter time
scales it is important to realize that the CH4/CO2 ratios are decreasing during
extraction (see Figures 7.9.1 to 7.9.6 for the Scenarios RZ1 to RZ6). Only for RZ3
(Figure 7.9.3) the CH4/CO2 ratio is maintained on a practically constant level of ~18
to ~22%. The reason for the CH4/CO2 ratio to be maintained and not to decrease is
the low CO2 reinjection percentage.
The estimation of the ideal CH4/CO2 ratio is a delicate issue: On the one hand,
CO2 helps to siphon the deep water, as the CO2-containing bubbles in the tubes sup-
port the buoyancy. On the other hand a low CH4/CO2 ratio reduces the efficiency of
the power production. Thus, the benefit of reinjecting less CO2 in the lake is a higher
long-term CH4/CO2 ratio and therefore better power plant efficiency. Finding the ideal
ratio is obviously an engineering optimization

105
 Methane / Carbondioxide ratio Methane / Carbondioxide ratio
0.28 0.23

0.21
primary extraction
secondary extraction
0.23 0.19

0.17

0.18 0.15
CH4 / CO2

CH4 / CO2
primary extraction
0.13

0.13 0.11

0.09

0.08 0.07

0.05

0.03 0.03
2004 2024 2044 2064 2084 2104 2004 2024 2044 2064 2084 2104
Time (year) Time (year)

Figure 7.9.1 Simulated CH4/CO2 ratio as a Figure 7.9.2 Simulated CH4/CO2 ratio as a
function of time for Scenario RZ1. function of time for Scenario RZ2.

Methane / Carbondioxide ratio Methane / Carbondioxide ratio

0.28 0.28

primary extraction primary extraction

secondary extraction secondary extraction
0.23 0.23

0.18 0.18
CH4 / CO2

CH4 / CO2

0.13 0.13

0.08 0.08

0.03 0.03
2004 2024 2044 2064 2084 2104 2004 2024 2044 2064 2084 2104
Time (year) Time (year)

Figure 7.9.3 Simulated CH4/CO2 ratio as a Figure 7.9.4 Simulated CH4/CO2 ratio as a
function of time for Scenario RZ3. function of time for Scenario RZ4.

106
 Methane / Carbondioxide ratio Methane / Carbondioxide ratio
0.28 0.28

primary extraction
secondary extraction
0.23 0.23

0.18 0.18
CH4 / CO2

CH4 / CO2
0.13 0.13
primary extraction

0.08 0.08

0.03 0.03
2004 2024 2044 2064 2084 2104 2004 2024 2044 2064 2084 2104
Time (year) Time (year)

Figure 7.9.5 Simulated CH4/CO2 ratio as a Figure 7.9.6 Simulated CH4/CO2 ratio as a
function of time for Scenario RZ5. function of time for Scenario RZ6.

107
7.10 Scenario RZ4: Resource Zone Reinjection

Scenario RZ4
Design parameters - This Scenario RZ4
is not a “stand-alone” Scenario, but is conside- Extraction flow 70 m3 s-1
red as the first leg in combination with RZ5,
Fraction of CH4 reinjected 6%
which would cover the second part. The idea of Fraction of CO2 reinjected 60/55 %
RZ4 is to use CH4 in the RZ (Upper and Lower
RZ) as for Scenario RZ2, but to use the CH4 in 1. Withdrawal depth 410 m
1. Withdrawal range 60 m
the Upper RZ (275 to 305 m depth) at a faster
rate than in the Lower RZ (320 to 480 m depth) 1. Reinjection depth 325 m
and basically remove the CH4 in the Upper RZ 1. Reinjection range 5m
in order to create a CH4-free space (Figure 2. Withdrawal depth 310 m
7.10.5), which the follow up RZ5 could use for 2. Withdrawal range 10 m
disposing the reinject-water. 2. Reinjection depth 270 m
Another difference is that for RZ4 the 2. Reinjection range 10 m
ranges of withdrawal and reinjection are verti-
cally separated, so that no recirculation can occur from the reinjection outlet to the
intake pipe. The water flow of 70 m3 s-1 is split between the two zones (details in
Table 5.5). The lake water volume (104 km3) between extraction and reinjection
depth is recycled 2.13-times for the extraction volume (221 km3) for the 100 yr of
simulation.
The Lower RZ is intentionally extracted with a larger CO2 removal (55%
rejected, see above) in order to reduce the density of the Lower RZ water. This
preparation is advantageous for the stacking of the reinject-water while Scenario
RZ5 is in operation.
Again, as for RZ2, there is no additional (relative to natural) up- or down-
welling induced above the RZ (such as in the PRZ or the IZ) by the extraction
process (Figure 7.10.1). Only within the two RZ layers, there is dowelling between
reinjection and withdrawal (Figure 7.10.1). Therefore we do not expect significant
changes of the overall stratification or of the lake internal (nutrient) fluxes.

Table 7.10 Evaluation of Scenario RZ4

No Extraction Scenario RZ4

2004 2054 2104 2004 2054 2104
PO4 upflux 80 m depth 1.9 2.4 2.9 1.9 2.4 2.9 103 tyr-1
NH4 upflux 80 m depth 15.9 21.0 26.0 15.9 21.0 26.0 103 tyr-1
Schmidt stability 8.8 8.7 8.7 8.8 8.4 7.9 1014 J
CH4 in lake (i) 64.6 81.7 96.3 64.6 48.9 44.1 km3
CH4 harvestable (ii) 42.3 58.9 69.0 42.3 24.3 15.1 km3
CH4 harvested (iii) 33.7 53.0 km3
Average energy output 200 157 MW
CO2 in lake 290 315 336 290 243 209 km3
CO2 removed from lake 74 125 km3

108
 Stability and safety - As evident from Figures 7.10.3 to 7.10.6, the changes
to the lake stratification occur only in the RZ (lowest 200 m) and the lake above 275
m depth is not affected (see explanations for RZ1). Therefore, the Schmidt stability
remains at the natural (undisturbed) level (Table 7.10).
After 100 yr, the safety margin would be minimal at 445 m depth at only 70 J
m-3 (see explanations for Figure 7.10.2). However, this Scenario RZ4 is not intended
to run for 100 yr, as after ~25 yr the latest, the operation has to switch to RZ5 (as
otherwise the CH4 concentration would become too low for extraction; Figure 7.10.7).
Therefore, the safety- margin is not a concern and can only be judged in
combination with RZ5 (realized as RZ6).

Nutrients - The nutrient transport to the Biozone (Table 7.10) does not in-
crease during the 100 yr of simulation compared to the natural fluxes to the surface
(see explanations for RZ1).

Methane and energy - As explained for RZ1, there is no additional loss of

CH4 due to the extraction operation. Compared to RZ2/RZ3, the CH4 concentrations
are not homogenised, as the separation of extraction and withdrawal generates
vertical gradients (differences; Figure 7.10.5) and keeps the CH4 on a higher con-
centration at the intakes (Figure 7.10.7). Well before (~25 yr) the Upper RZ
withdrawal reaches uneconomically low CH4 levels (~5 mol m-3, Figure 7.10.7), the
operation has to switch to RZ5 such that lake water is drawn-down (at higher CH4-
level) and the reinject-water is stacked back into the lake above the draw-down zone.

109
 Advection Safety margin
0 0

50 50 2004

100 100 2104 no extraction

2104 Scenario RZ4
150 150

200 200
Depth (m)

Depth (m)
250 secondary reinjection 250 secondary reinjection

300 secondary extraction secondary extraction

300

350 primary reinjection 350

primary reinjection

400 400 primary extraction

primary extraction
450 450

500 500
-50 -40 -30 -20 -10 0 10 0 1000 2000 3000 4000 5000 6000
-1 -3
Upwelling advection (m yr ) Energy needed to lift until ptotal = phydrostatic (J m )

Figure 7.10.1 Simulated profiles of vertical Figure 7.10.2 Safety margin for Scenario
velocity (+: upwelling; -: downwelling) for RZ4 in 2104 (green), compared to Scenario
Scenario RZ4 (orange), compared with O (red).
Scenario O (black).

Salinity Phosphate
0 0

50 50 2004
2004
2014 2014
100 100
2024 2024
2054 2054
150 150
2104 2104

200 200
Depth (m)
Depth (m)

250 secondary reinjection 250 secondary reinjection

secondary extraction 300 secondary extraction
300

primary reinjection 350 primary reinjection

350

400 400
primary extraction primary extraction
450 450

500 500
0 1 2 3 4 5 6 0 0.1 0.2 0.3 0.4
-3
-3
S (kg m ) PO4 (mol m )

Figure 7.10.3 Simulated salinity profiles for Figure 7.10.4 Simulated PO4 profiles for
Scenario RZ4. Scenario RZ4.

110
 Methane Carbon dioxide
0 0

50 2004 50 2004
2014 2014
100 100
2024 2024
2054 2054
150 150
2104 2104
200 200
Depth (m)

Depth (m)
250 secondary reinjection 250 secondary reinjection

300 secondary extraction 300 secondary extraction

350 primary reinjection 350 primary reinjection

400 400
primary extraction primary extraction
450 450

500 500
0 5 10 15 20 25 0 50 100 150 200
-3 -3
CH4 (mol m ) CO2 (mol m )

Figure 7.10.5 Simulated CH4 profiles for Figure 7.10.6 Simulated CO2 profiles for
Scenario RZ4. Scenario RZ4.

Methane extraction concentration Total gas pressure

25 0
2004
primary extraction 50 2014
secondary extraction 2024
20 100 2054
2104
150
hydrostatic
15
CH4 (mol m )

200
-3

Depth (m)

250 secondary reinjection

10 300 secondary extraction

350 primary reinjection

5 400
primary extraction
450

0 500
2004 2024 2044 2064 2084 2104 0 10 20 30 40
Time (year) Pressure (atm)

Figure 7.10.7 Simulated CH4 extraction Figure 7.10.8 Simulated total gas pressure
concentration as a function of time for profiles for Scenario RZ4.
Scenario RZ4.

111
7.11 Scenario RZ5: Resource Zone Reinjection

Design parameters - This Scenario RZ5 Scenario RZ5

is intended as a continuation of RZ4. After the
Extraction flow 70 m3 s-1
Upper RZ has been emptied of CH4 by
Scenario RZ4 (~25 yr), the CH4-free space Fraction of CH4 reinjected 3.4 %
(Figure 7.10.5) is used by Scenario RZ5 for Fraction of CO 2 reinjected 50 %
reinjection into a narrow range at 270 m depth Withdrawal depth 400 m
at the upper boundary of the Upper RZ. The Withdrawal range 160 m
withdrawal is foreseen over the entire Lower RZ Reinjection depth 270 m
from 320 m to the maximum depth. However, if Reinjection range 10 m
this broad range causes disadvantages (too low
CH4 concentrations) at the upper boundary of the Lower RZ, the withdrawal range
can be confined to the lower reaches, without changing the concept. Also for RZ5,
the withdrawal and reinjection are vertically separated, so that no recirculation can
occur from the reinjection outlet to the intake pipe. The draw-down lake volume (104
km3) is replaced 2.13-times during the 100 yr of simulation (extraction volume: 221
km3).
As there are only slight salinity gradients in the RZ and as the CH4 extraction
increases the density of the reinject-water, it is important to remove a large fraction
of CO2 (here 50%) in order to keep the reinject-water on top of the Lower RZ and not
to cause convective mixing with the RZ. This is most crucial for this Scenario RZ6.
As for the other RZ Scenarios, the extraction does not induce additional
(relative to natural) up- or downwelling in the PRZ or the IZ (Figure 7.11.1). There-
fore we do not expect any changes of the overall stratification or of the lake internal
(nutrient) fluxes.

Table 7.11 Evaluation of Scenario RZ5

No Extraction Scenario RZ5

2004 2054 2104 2004 2054 2104
PO4 upflux 80 m depth 1.9 2.4 2.9 1.9 2.4 2.9 103 tyr-1
NH4 upflux 80 m depth 15.9 21.0 26.0 15.9 21.0 26.0 103 tyr-1
Schmidt stability 8.8 8.7 8.7 8.8 8.0 7.4 1014 J
CH4 in lake (i) 64.6 81.7 96.3 64.6 42.3 35.4 km3
CH4 harvestable (ii) 42.3 58.9 69.0 42.3 20.5 6.6 km3
CH4 harvested (iii) 40.1 61.6 km3
Average energy output 238 183 MW
CO2 in lake 290 315 336 290 216 172 km3
CO2 removed from lake 100 163 km3

112
 Stability and safety - As evident from Figures 7.11.3 to 7.11.6, the changes
to the lake stratification occur only in the RZ (lowest 200 m) and the lake above 265
m depth is not affected (see explanations for RZ1). Therefore, the Schmidt stability
remains at the natural (undisturbed) level (Table 7.11).
After 100 yr, the safety margin would be minimal at 200 m depth at 927 J m-3
(Figure 7.11.2), which is still a very safe level.

Nutrients - The nutrient transport to the Biozone (Table 7.11) does not in-
crease during the 100 yr of simulation compared to the natural fluxes to the surface
(see explanations for RZ1).

Methane and energy - As for the other RZ Scenarios, there is no additional

loss of CH4 due to the extraction operation. Due to the draw-down approach, a
strong vertical gradient of CH4 remains in the RZ (CH4-free reinject-water is stacked
on top of the CH4-containing deep-water (Figure 7.11.5). Therefore, after 30 to 40 yr
of extraction, the intake CH4 concentration drops within one decade by a factor of ~2
(from ~20 to 10 mol m-3; Figure 7.11.7). As the intake stretches over a wide vertical
range with CH4 concentrations below the threshold, the option will be to reduce the
intake range and to restrict to higher concentrations in the deeper range.
The harvested amount of CH4 during 100 yr is ~61.6 km3 and the correspon-
ding electrical energy production is ~183 MW (Table 7.11).

113
 Advection Safety margin
0 0

50 50 2004

100 100 2104 no extraction

2104 Scenario RZ5
150 150

200 200
Depth (m)

Depth (m)
250 reinjection depth
reinjection depth 250

300 300

350 350
extraction depth extraction depth
400 400

450 450

500 500
-50 -40 -30 -20 -10 0 10 0 1000 2000 3000 4000 5000 6000
-1 -3
Upwelling advection (m yr ) Energy needed to lift until ptotal = phydrostatic (J m )

Figure 7.11.1 Simulated profiles of vertical Figure 7.11.2 Safety margin for Scenario
velocity (+: upwelling; -: downwelling) for RZ5, compared to Scenario O.
Scenario RZ5 (orange), compared with
Scenario O (black).

Salinity Phosphate
0 0

50 2004 50 2004
2014 2014
100 100
2024 2024
2054 2054
150 150
2104 2104
200 200
Depth (m)

Depth (m)

250 reinjection depth 250 reinjection depth

300 300

350 350
extraction depth
400 400
extraction depth
450 450

500 500
0 1 2 3 4 5 6 0 0.1 0.2 0.3 0.4
-3 -3
S (kg m ) PO4 (mol m )

Figure 7.11.3 Simulated salinity profiles for Figure 7.11.4 Simulated PO4 profiles for
Scenario RZ5. Scenario RZ5.

114
 Methane Carbon dioxide
0 0

50 2004 50 2004
2014 2014
100 100
2024 2024
2054 2054
150 150
2104 2104
200 200
Depth (m)

Depth (m)
250 reinjection depth 250 reinjection depth

300 300

350 350
extraction depth
400 400

450 450
extraction depth
500 500
0 5 10 15 20 25 0 50 100 150 200
-3 -3
CH4 (mol m ) CO2 (mol m )

Figure 7.11.5 Simulated CH4 profiles for Figure 7.11.6 Simulated CO2 profiles for
Scenario RZ5. Scenario RZ5.

Methane extraction concentration Total gas pressure

25 0
2004
50 2014
2024
20 100 2054
2104
150
hydrostatic

15 200
CH4 (mol m-3)

Depth (m)

250 reinjection depth

10 300

350

5 400
extraction depth
450

0 500
2004 2024 2044 2064 2084 2104 0 10 20 30 40
Time (year) Pressure (atm)

Figure 7.11.7 Simulated CH4 extraction Figure 7.11.8 Simulated total gas pressure
concentration as a function of time for profiles for Scenario RZ5.
Scenario RZ5.

115
7.12 Scenario RZ6: Resource Zone Reinjection

Design parameters - This Scenario RZ6 is

Scenario RZ6
designed as the combination of RZ4 (first leg) and
RZ5 (second leg). The time scale to switch from Extraction flow 70 m3 s-1
RZ4 to RZ5 may depend on the extraction rate and
Fraction of CH4 6 % -->
is flexible. For the model calculation the switch was reinjected 3.4 %
set to 50 yr (Figure 7.12.7), when the upper intake Fraction of CO2 60/55 %
reaches non-harvestable CH4 concentrations. reinjected --> 50 %
Although a shorter timescale will most probably be
necessary to maintain the CH4 extraction efficiency. Withdrawal depth RZ4 -->
RZ5
The idea of RZ6 is to use CH4 in the Upper RZ Withdrawal range RZ4 -->
faster (than in the Lower RZ, see RZ4) in order to RZ5
create CH4-free space at the upper edge of the RZ
Reinjection depth RZ4 -->
(Figure 7.10.5). During the following extraction with RZ5
Scenario RZ6, this CH4-free layer is then used as Reinjection range RZ4 -->
the reinjection range (see RZ5). For both Scenarios RZ5
the extraction flow is 70 m3 s-1. For details of RZ4
and RZ5, see Sections 7.10 and 7.11, respectively.
The lake volume between extraction and reinjection is 104 km3 compared to
the extracted volume of 221 km3 in 100 yr of simulation. Therefore, the lake volume
would be recycled 2.13-times during 100 yr of extraction. As explained for RZ5 the
somewhat larger CO2 removal in the lower extraction (see table above) is intended
to keep the density profile stable in the Lower RZ.
Again, as for the other RZ Scenarios, there is no additional (relative to natural)
up- or downwelling induced in the PRZ or the IZ by the extraction process (Figure
7.12.1). Therefore we do not expect any changes of the overall stratification or of the
lake internal (nutrient) fluxes.

Table 7.12 Evaluation of Scenario RZ6

No Extraction Scenario RZ6

2004 2054 2104 2004 2054 2104
PO4 upflux 80 m depth 1.9 2.4 2.9 1.9 2.4 2.9 103 tyr-1
NH4 upflux 80 m depth 15.9 21.0 26.0 15.9 21.0 25.9 103 tyr-1
Schmidt stability 8.8 8.7 8.7 8.8 8.4 7.6 1014 J
CH4 in lake (i) 64.6 81.7 96.3 64.6 48.7 35.9 km3
CH4 harvestable (ii) 42.3 58.9 69.0 42.3 24.5 6.0 km3
CH4 harvested (iii) 32.1 61.5 km3
Average energy output 190 182 MW
CO2 in lake 290 315 336 290 242 182 km3
CO2 removed from lake 76 151 km3

116
 Stability and safety - As evident from Figures 7.12.3 to 7.12.6, the changes
to the lake stratification occur only in the RZ (lowest 200 m) and the lake above 275
m depth is not affected (see RZ4, RZ5). Therefore, the Schmidt stability remains at
the natural (undisturbed) level (Table 7.12).
After 100 yr, the safety margin would be minimal at 200 m depth at 970 J m-3
(Figure 7.12.2), at a still very safe level.

Nutrients - The nutrient transport to the Biozone (Table 7.12) does not in-
crease during the 100 yr of simulation compared to the natural fluxes to the surface
(see explanations for RZ1).

Methane and energy - As for the other RZ Scenarios, there is no additional

loss of CH4 due to the extraction operation. Due to the draw-down approach, a
strong vertical gradient of CH4 remains in the RZ (CH4-free reinject-water is stacked
on top of the CH4-containing deep-water, Figure 7.12.5). After ~25 and ~70 yr of
extraction, the intake CH4 concentration drops in two steps from originally ~20 to ~5
mol m-3; Figure 7.12.7). As the intake ranges over a wide vertical range with CH4
concentrations below the threshold, the option will be to reduce the intake range and
to restrict to higher concentrations in the deeper range of the lake.
The harvested amount of CH4 during 100 yr is ~61.5 km3 and the correspon-
ding electrical energy production is ~182 MW (Table 7.12).

117
 Advection Safety margin
0 0

50 50 2004

100 100 2104 no extraction

2104 Scenario RZ6
150 150

200 200
Depth (m)

Depth (m)
250 secondary reinjection 250 secondary reinjection

300 secondary extraction secondary extraction

300

350 primary reinjection 350

primary reinjection

400 400 primary extraction

primary extraction
450 450

500 500
-50 -40 -30 -20 -10 0 10 0 1000 2000 3000 4000 5000 6000
-1 -3
Upwelling advection (m yr ) Energy needed to lift until ptotal = phydrostatic (J m )

Figure 7.12.1 Simulated profiles of vertical Figure 7.12.2 Safety margin for Scenario
velocity (+: upwelling; -: downwelling) for RZ6, compared to Scenario O.
Scenario RZ6 (orange), compared with
Scenario O (black).

Salinity Phosphate
0 0

50 2004 50 2004
2014 2014
100 100
2024 2024
2054 2054
150 150
2104 2104
200 200
Depth (m)

Depth (m)

250 secondary reinjection 250 secondary reinjection

300 secondary extraction secondary extraction

300

350 primary reinjection primary reinjection

350

400 400
primary extraction primary extraction
450 450

500 500
0 1 2 3 4 5 6 0 0.1 0.2 0.3 0.4
-3 -3
S (kg m ) PO4 (mol m )

Figure 7.12.3 Simulated salinity profiles for Figure 7.12.4 Simulated PO4 profiles for
Scenario RZ6. Scenario RZ6.

118
 Methane Carbon dioxide
0 0

50 2004 50 2004
2014 2014
100 100
2024 2024
2054 2054
150 150
2104 2104
200 200
Depth (m)

Depth (m)
250 secondary reinjection 250 secondary reinjection

300 secondary extraction 300 secondary extraction

350 primary reinjection 350 primary reinjection

400 400
primary extraction primary extraction
450 450

500 500
0 5 10 15 20 25 0 50 100 150 200
-3 -3
CH4 (mol m ) CO2 (mol m )

Figure 7.12.5 Simulated CH4 profiles for Figure 7.12.6 Simulated CO2 profiles for
Scenario RZ6. Scenario RZ6.

Methane extraction concentration Total gas pressure

25 0
2004
primary extraction 50 2014
secondary extraction 2024
20 100 2054
2104
150
hydrostatic

15 200
CH4 (mol m )
-3

Depth (m)

250 secondary reinjection

10 300 secondary extraction

350 primary reinjection

5 400
primary extraction
450

0 500
2004 2024 2044 2064 2084 2104 0 10 20 30 40
Time (year) Pressure (atm)

Figure 7.12.7 Simulated CH4 extraction Figure 7.12.8 Simulated total gas pressure
concentration as a function of time for profiles for Scenario RZ6.
Scenario RZ6.

119
7.13 Summary of Attributes for all Scenarios

In the following Table 7.13 the Attributes, listed in the Tables 7.1 to 7.12, are
compiled for the purpose of comparison. The implications of the different Attributes
are discussed in the Section 7.1 to 7.12. A cross-comparison of the important
Attributes is provided in the Consequence Table 2.4.1.

Table 7.13 Design Parameters used for all Scenarios and comparison
of the Attributes after 0, 50 and 100 yr of simulations
Table 7.13 Continuation

121
122
8. Appendix

8.1 AQUASIM

In order to support environmental scientists in finding an “adequate” model

for the system investigated, often numerical software is necessary, which allows
performing simulations using different models. The tasks of such models are:
- performing simulations,
- assessing the identifiably,
- estimating the values of model parameters (using measured data), and
- estimating prediction uncertainty.
The computer program AQUASIM (Computer Program for Identification and
Simulation of Aquatic Systems) was developed to perform such analyses for
technical and natural aquatic systems (Reichert 1994).
The lake compartment describes stratification, substance transport and
transformation in the water column of the lake and at the sediment surface. The
user of the program is free in specifying any set of dynamic model variables (con-
centrations, temperature, etc) and transformation processes within the compart-
ments. Calculated results can be plotted to the screen (window interface version
only), written to a post-script file for transfer to a printer, or written to a text file for
external post-processing (Reichert 1994). Detailed descriptions, user manuals and
tutorial models (demo version) can be downloaded at (see short description):
[Link]

The relative and absolute accuracies of the dynamic model variables were
set according to the recommendation given in the AQUASIM manual (Reichert
1994), where detailed explanations are available ([Link]
The accuracies for the simulated dynamic model variables (S, T, CH4, CO2) were
set according to those recommendation and are listed in Table 8.1.

Table 8.1 Relative and absolute accuracies of dynamic model variables

State variable Value range Units Rel. accuracy Abs. accuracy

Salinity S 1 to 6 kg m-3 1x10-6 1x10-5
Temperature T 23 to 26 °C 1x10-6 1x10-4
CH4 0 to 35 mol m-3 1x10-6 1x10-4
CO2 0 to 200 mol m-3 1x10-6 1x10-3
PO4 0 to 0.5 mol m-3 1x10-6 1x10-6
NH4 0 to 10 mol m-3 1x10-6 1x10-5

123
8.2 Model parameters

Detailed explanations of the model parameters and their values, listed

below (Tables 8.2.1 and 8.2.2), are given in Section 4. Detailed explanations
concerning the assumed parameter errors (expressed in one standard deviation)
are given in the sensitivity analysis Section 8.7 (Appendix). The following symbols
are used:

z = depth [m] (positive downward)

A(z) = lake cross sectional area [km2] as a function of depth (Figure 4.3.1).

dA/dz = area gradient [m2 m-1 = m] as a function of depth shown in Figure

4.3.2. Because the cross sectional area of the lake decreases with depth,
the areal gradient dA/dz [m] is negative.

N2 = (g/ρ)*(dρ/dz) [s-2] is water column stability (N = Brunt Väisälä frequency)

g = gravitational acceleration = 9.7758 m s-2

pair = atmospheric pressure = 0.843466 atm (1460 m above sea level)

Tabs = absolute temperature = T [°C] + 273.15°C

Table 8.2.1 Values and standard deviations (assumed) of parameters

used for sensitivity analysis

Parameters for sensitivity analysis Parameter Units Standard

value deviation
Areal CH4 formation rate, CH4,formation rate 120 g-C m-2 yr-1 12
10 mol m-2 yr-1 1
Geothermal heat flux, Hgeothermal 0.02 W m-2 0.002
Turbulent diffusion K(z) above 120 m depth(1) ~1.0x10-5, m2 s-1 10%
decreasing
with depth
Energy dissipation ε below 120 m depth 1.0x10-10 W kg-1 1.0x10-11
Ratio of double-diffusive transport KT/KS 10 - 1
Maximum of turbulent diffusion for salt 1.0x10-5 m2 s-1 1.0x10-6
below 120 m depth, KS, max(1)
(1)
For the double-diffusive zone, we estimated the apparent diffusion coefficient to be different for
heat (KT) and for salt (KS). We set the ratio KT/KS = 10, based on the analysis of measure profiles
(Schmid et al 2010).

124
Table 8.2.2 Water input parameters and assumed standard
deviations for the uncertainty analysis
Water input parameters Parameter value Units Std. dev.
Surface water input: S 0.1 kg m-3 0.01
Surface water input: T 23.1 °C 0.1
Surface water input: PO4 0.004 mol m-3 0.0004
Surface water input: NH4 0.043 mol m-3 0.0043
Water input at 180 m depth: Q 22 m3 s-1 2.2
Water input at 180 m depth: S 2.1 kg m-3 0.21
Water input at 180 m depth: T 23.7 °C 0.1
Water input at 180 m depth: CH4 0 mol m-3 0.040
Water input at 180 m depth: CO2,total (1) 28 mol m-3 2.8
Water input at 180 m depth: PO4 0.017 mol m-3 0.0017
Water input at 180 m depth: NH4 0.52 mol m-3 0.052
Water input at 250 m depth: Q 15 m3 s-1 1.5
Water input at 250 m depth: S 2.7 kg m-3 0.27
Water input at 250 m depth: T 23.3 °C 0.1
Water input at 250 m depth: CH4 1.8 mol m-3 0.18
Water input at 250 m depth: CO2,total (1) 45 mol m-3 4.5
Water input at 250 m depth: PO4 0.032 mol m-3 0.0032
Water input at 250 m depth: NH4 0.65 mol m-3 0.065
Water input at 310 m depth: Q 1 m3 s-1 0.1
Water input at 310 m depth: S 3.4 kg m-3 0.34
Water input at 310 m depth: T 25.2 °C 0.1
Water input at 310 m depth: CH4 1.9 mol m-3 0.19
Water input at 310 m depth: CO2,total (1) 53 mol m-3 5.3
Water input at 310 m depth: PO4 0 mol m-3 0.0016
Water input at 310 m depth: NH4 1.4 mol m-3 0.14
Water input at 365 m depth: Q 1.5 m3 s-1 0.15
Water input at 365 m depth: S 5.5 kg m-3 0.55
Water input at 365 m depth: T 24.5 °C 0.1
Water input at 365 m depth: CH4 0 mol m-3 0.19
Water input at 365 m depth: CO2,total (1) 130 mol m-3 13
Water input at 365 m depth: PO4 0.012 mol m-3 0.0012
Water input at 365 m depth: NH4 1.4 mol m-3 0.14
Water input at 425 m depth: Q 0.8 m3 s-1 0.08
Water input at 425 m depth: S 5.8 kg m-3 0.58
Water input at 425 m depth: T 25.3 °C 0.1
Water input at 425 m depth: CH4 0 mol m-3 0.19
Water input at 425 m depth: CO2,total (1) 130 mol m-3 13
Water input at 425 m depth: PO4 0 mol m-3 0.0019
Water input at 425 m depth: NH4 0 mol m-3 0.040
Water input at 465 m depth: Q 1.25 m3 s-1 0.125
Water input at 465 m depth: S 6 kg m-3 0.6
Water input at 465 m depth: T 26 °C 0.1
Water input at 465 m depth: CH4 11 mol m-3 1.1
Water input at 465 m depth: CO2,total (1) 170 mol m-3 17
Water input at 465 m depth: PO4 0.12 mol m-3 0.012
Water input at 465 m depth: NH4 3.0 mol m-3 0.30
(1)
CO2 input is in form of CO2,total. Speciation (Section 8.3) is defined by alkalinity (and pH) and varies (but
only slightly) with time.

125
8.3 Formulas used in the model

Total gas pressure

ptotal = pCH4 + pCO2 + pN2, air [atm]
assuming nitrogen is in equilibrium with the atmosphere in the whole lake.

Hydrostatic pressure
phydrostatic = 1000 kg m-3 * g * z * 0.986923x10-5 atm Pa-1 + pair [atm]

g = 9.7758 m s-2, z = depth [m]

Partial pressure of nitrogen (N2) in the air

pN2, air = 0.78084 * pair [atm]

Partial pressure of CO2 in the air

pCO2, air = 0.000383 * pair [atm]

Partial pressure of CH4 in the air

pCH4, air = 1.75x10-6 * pair [atm]

Alkalinity
Alk = S * 11.96 [mol m-3]
(Correlation coefficient of 0.999976 for the measurements of 2006)

Partial pressure of dissolved CO2

pCO2 = H2CO3 / HCO2 [atm]

Henry constant for CO2

HCO2 = H0CO2 * Pcorrection CO2 * fCO2/pCO2 * 1000 L m-3)
[mol m-3 atm-1]

fCO2/pCO2 = exp((BCO2 * phydrostatic * 101325 Pa atm-1) / (8.3145 J mol-1 K-1 * Tabs))

BCO2 = (3.1653x10-5 * Tabs3 - 0.032796 * Tabs2 + 12.0408 * Tabs - 1636.75) * 10-6

[m3 mol-1] (Esper et al 1995)

H0CO2 = exp(-58.0931 + 90.5069 * (100 / Tabs) + 22.294 * ln(Tabs / 100) + S

* (0.027766 - 0.025888 * (Tabs / 100) + 0.0050578 * (Tabs / 100)2))
[mol L-1 atm-1] (Weiss 1974)

Pcorrection CO2 = exp((-101325 Pa atm-1 (phydrostatic - 1 atm) * vmol,CO2) /

(8.3145 J mol-1 K-1 * Tabs)) [-]
(Weiss 1974)

vmol,CO2 = 3.23x10-5 m3 mol-1 is the partial molal volume of dissolved CO2

Speciation of carbonates
aCO2 = (1 + 10(pH - pK1) + 10(2 * pH - pK1 - pK2))-1 [-]

126
Carbonic acid
H2CO3 = aCO2 * CO2,total [mol m-3]

Bicarbonate
HCO3- = H2CO3 * 10(pH - pK1) [mol m-3]

Carbonate
CO32- = HCO3- * 10(pH - pK2) [mol m-3]

pH in equilibrium with alkalinity and CO2 partial pressure

pH is calculated from the solution for the charge balance, based on alkalinity (Alk)
and the distribution of the charges of HCO3- and of CO32-.

pH = -Log10 (H [mol L-1]),

where H (concentration of protons) is given by the quadratic solution of

H= 1000 m3 L-1 * (10-pK1 * (Alk-CO2,total) + (2 * Alk)-1 * [10-2*pK1 * (Alk-CO2,total)2 -

4 * Alk * 10-(pK1+pK2) * (Alk - 2 * CO2,total)]1/2) [mol L-1]

First dissociation constant of CO2

pK1 = 3404.71 / Tabs + 0.032786 * Tabs - 14.8435 - 0.071692 * S0.5 * (200.1/
Tabs + 0.322) + 0.0021487 * S [-] (Cai and Wang 1998)

Second dissociation constant of CO2

pK2 = 2902.39 / Tabs + 0.02379 * Tabs - 6.498 - 0.3191 * S0.5 * (-129.24 / Tabs
+ 1.4381) + 0.0198 * S [-] (Cai and Wang 1998)

CO2 surface equilibrium concentration

H2CO3 surface eq. = HCO2 * pCO2, air [mol m-3]

Partial pressure of CH4

pCH4 = CH4 / HCH4 [atm]

Henry constant for CH4

HCH4 = H0CH4 * Pcorrection CH4 * fCH4/pCH4 * 1000 L m-3 / 0.986923 atm bar-1
[mol m-3 atm-1]

H0CH4 = exp(S * (0.027766 - 0.025888 * (Tabs/100) + 0.0050578*(Tabs/100)2))*

1 / (10-5 * exp(127.174 - 155.576/Tabs * 100 - 65.2553 * ln(Tabs / 100)
+ 6.16976 * Tabs / 100)) * 55.5 [mol L-1 bar-1]
(Rettich et al 1981; Weiss 1974)

fCH4/pCH4 = exp((BCH4 * phydrostatic * 101325 Pa atm-1) / (8.3145 J mol-1 K-1 * Tabs))

[-] (Weiss 1974)

BCH4 = (8.708x10-6 * Tabs3 - 0.009513 * Tabs2 + 3.73 * Tabs - 539.7) * 10-6

[m3 mol-1] (Esper et al 1995)

127
 Pcorrection CH4 = exp((-101325 Pa atm-1 (phydrostatic - 1 atm) * vmol,CH4)/
(8.3145 J mol-1 K-1 * Tabs)) [-]
(Weiss 1974)

vmol,CH4 = 3.6x10-5 m3 mol-1 is the partial molal volume of dissolved CH4

CH4 surface equilibrium concentration

CH4 surface eq. = HCH4 * pCH4, air [mol m-3]

Water density
ρ(S, T, CH4, CO2) = 1000 L m-3 * ρ(T) * (1 + 0.75x10-3 m3 kg-1 * S
+ 0.284x10-6 m3 g-1 * 44.0099 g mol-1* CO2
- 1.25x10-6 m3 g-1 * 16.04 g mol-1 * CH4) [kg m-3]
(Schmid et al 2004)
-3
S = salinity [kg m ]

ρ(T) = 0.999839 + 6.7914x10-5 * T - 9.0894x10-6 * T2 + 1.0171x10-7 * T3 -

1.2846x10-9 * T4 + 1.1592x10-11 * T5 - 5.0125x10-14 * T6 [kg L-1]
(Chen and Millero 1977)

Turbulent vertical diffusion coefficient

a) From surface to 120 m depth:
K (0 m) = 2x10-4 m2 s-1
K (18 m) = 2x10-4 m2 s-1
K (35 m) = 1x10-5 m2 s-1
K (58 m) = 5x10-6 m2 s-1
K (68 m) = 1x10-6 m2 s-1
K (120 m) = 1.92x10-7 m2 s-1

b) Below 120 m depth:

K = 0.15 * ε / N2 [m2 s-1]
(factor 0.15 taken from Wüest and Lorke (2003))

Maximal turbulent vertical heat diffusion coefficient (below 120 m depth)

Kmax = 1x10-4 m2 s-1

Geothermal heat flux rate

dA / dz * H geothermal
r geo thermal heat flux = − [°C s-1]
A( z ) * ρC p

(ρCp = 4.18x106 J m-3 K-1; Hgeothermal = 0.02 W m-2)

CH4 oxidation rate

rCH4 oxidation = 10-6 s-1 (if z < 60 m depth) or 0 (else)
(the process is self-limited as the rate is multiplied with the CH4 concentration;
the absolute value of this rate is irrelevant as long as large enough to remove most
of the CH4).

128
Primary production rate
rprimary production = 10-6 s-1 (if z < 50 m depth) or 0 (else)
(the process is self-limited as the rate is multiplied with the PO4 concentration;
the absolute value of this rate is irrelevant as long as large enough).

Methane formation rate

dA / dz * CH 4 area formation rate
r CH 4 formation = − [mol m-3 s-1]
A( z )
(CH4, formation rate = 3.2x10-7 mol m-2 s-1 is the areal rate of CH4 formation at
the sediment; in this report CH4, formation rate = 120 g-C m-2 yr-1 also given in
units of C).

8.4 Dynamic processes and stoichiometry

Dynamic processes describe transformations by their temporal rate of

change of the dynamic model variables. Usually, a biological or chemical process
transforms several substances in fixed stoichiometric proportions. Therefore, it is
advantageous to separate a “common” factor as a process rate, and to describe a
process by this rate and by stoichiometric coefficients for all substances involved in
the process. The contribution of a process to the temporal change of the concen-
tration of a substance is then given as the product of the “common” process rate
and the substance-specific stoichiometric coefficient.
This decomposition of process rates into a “common” process rate and
individual stoichiometric coefficients is not unambiguous. To make it unambiguous,
one of the stoichiometric coefficients is set to unity (Reichert 1994). We use the
following coefficients in this model:

The geothermal heat flux with the rate rgeothermal heat flux affects
T: +1

Primary production with the rate rprimary production * PO4 affects

PO4: -1
NH4: -13.9
CO2: -216
(PO4-limitation for algal growth; rprimary production contains also Michaelis-Menten
limitation in NH4)

The aerobic CH4 oxidation with the rate rCH4 oxidation * CH4 affects
CH4: -1
CO2: +1
(Self-limiting process in CH4)

Biomass mineralization with the rate rCH4 formation affects

CH4: +1

129
 CO2: +1
NH4: +0.14
PO4: +0.0089
(Values from the measured stoichiometric coefficients of the organic material found
in the sediment traps and cores in 2007 (molar ratios):
C:P = 216, and N:P = 13.9 and C:N = 15.8 (Pasche et al submit);
note that the carbon is transformed in equal parts to CH4 and CO2).

8.5 Exchange at the lake surface

The surface exchange can either be an input (inflow, gas exchange gain) or
an output (outflow, gas exchange loss), here described as a flow across the lake
surface. Note that this flux represents a mass (or temperature) flux per unit lake
surface area and per unit of time. A positive flux value represents a flux into the
lake, a negative value a flux out of the lake. The fluxes are driven by the differen-
ces of the concentrations (temperature) between the surface water (dynamic
model variables) and the concentration in the atmosphere. In the model, we use
the following parameterizations (Reichert 1994):

Gas exchange velocity

vgas exchange = gas exchange velocity = 1.0x10-5 m s-1
(typical value for an average surface wind speed of about 3 m s-1; positive
flux = flux into lake water; negative flux = flux from water to atmosphere).

CH4 surface flux (in / out)

vgas exchange * (CH4 surface eq. - CH4) [mol m-2 s-1]

CO2 surface flux (in / out)

vgas exchange * (CO2,surface eq. - CO2) [mol m-2 s-1]

Temperature surface flux

1x10-3 * (23.06 °C - Tsurface) [°C m s-1]

8.6 Standard units used

The standard units used in the model are:
[s] for time
[m] for length
[kg] for masses
[mol] for amounts (number) of a substance (gases, nutrients)
[atm] for pressures (1 atm = 101,325 Pa).

130
8.7 Sensitivity analysis

The sensitivity analysis combines the tasks of uncertainty analysis and

identifiability analysis.

(i) Uncertainty analysis - In uncertainty analysis, the uncertainty of model

parameters is propagated to the uncertainty of the model results. In the current
version of AQUASIM only the simplest error propagation method is implemented:
The linearized propagation of standard deviations of uncorrelated parameters.

(ii) Identifiability analysis - The goal of the identifiability analysis is to

check if model parameters can be uniquely determined by using the available
(measured) data and to estimate the uncertainty of the parameter estimations.

The parameter values and their assumed uncertainty (expressed as one

standard deviation) used in the model are listed in Tables 8.2.1 and 8.2.2
(Appendix). For temperature the standard deviation was set to 0.1 °C. For all
parameters defining concentrations (state variables) the following assumptions
have been made: (i) for subaquatic sources with concentration value 0 (zero) the
standard deviation was set to 1% of the measured profile value (Table 8.8) at its
corresponding depth; (ii) for all the other parameters a standard deviation of 10%
was assumed (Table 8.2.2). Similar uncertainties for all parameters were used in
order to compare the sensitivity of the model results to these parameters. We
expect that the real uncertainties, especially of concentrations in the inflow water of
the subaquatic sources, are higher than 10%. In that sense, the sensitivity plots
shown below should not be interpreted in terms of absolute uncertainties. However,
an uncertainty of 15 or 20% for all concentrations would not change the message
to be drawn from the figures.
The lower and upper boundaries of the dynamic model variables (concen-
trations, temperature) are shown in Figures 8.7.1 to 8.7.6 for the years 2054 and
2104 (100 yr of simulation). The output errors result from the errors of the parame-
ters used in the model (Tables 8.2.1 and 8.2.2).
The contributions to the errors of the dynamic model variables (concentra-
tions, temperature), which are influenced by the 10 most important factors, are
shown in Figures 8.7.7 to 8.7.12.
It is obvious, that most of the uncertainty of the simulated model variables
stems from the estimated values of the subaquatic sources. The deepest source at
465 m depth has generally the strongest influence on the simulated model variab-
les (Figures 8.7.7 to 8.7.12). The analysis shows also that the salinity and CO2
content of the deepest sources have an (indirect) influence on the distribution of
the gases and nutrients, because salinity and CO2 influence the density and
thereby the stratification.

131
 Salinity

50

100

150

200
Depth (m)

2004
250 2054
2104
300

350

400

450

500
0 2 4 6
-3
S (kg m )

Figure 8.7.1 Lower and upper boundaries for salinity (one standard deviation). The three
reaches of large uncertainties are due to subaquatic sources at the corresponding depth.

Temperature

50

100

150

200
Depth (m)

2004
250 2054
2104
300

350

400

450

500
22.5 23.5 24.5 25.5 26.5

T (°C)

Figure 8.7.2 Lower and upper boundaries for temperature (one standard deviation). The
large uncertainties in 250 m depth and at the deepest location are due to subaquatic
sources. The uncertainty at the surface is due to inadequate boundary conditions (natural
variability).

132
 Methane

50

100

150

200
Depth (m)

2004
250 2054
2104
300

350

400

450

500
0 10 20 30 40
-3
CH4 (mol m )

Figure 8.7.3 Lower and upper boundaries for CH4 (one standard deviation). Local maxima
of the uncertainties are due to subaquatic sources (error in flow) and the overall uncer-
tainty is due to the poorl-known CH4 new formation rate.

Carbon dioxide

50

100

150

200
Depth (m)

2004
250 2054
2104
300

350

400

450

500
0 50 100 150 200
-3
CO2 (mol m )

Figure 8.7.4 Lower and upper boundaries for CO2,total (one standard deviation). Local
maxima of the uncertainties are due to subaquatic sources and the overall uncertainty is
due to the poorly-known rate of degradation of organic matter.

133
 Phosphate

50

100

150

200
Depth (m)

2004
250 2054
2104
300

350

400

450

500
0 0.05 0.1 0.15 0.2 0.25 0.3
-3
PO4 (mol m )

Figure 8.7.5 Lower and upper boundaries for PO4 (one standard deviation). Local maxima
of the uncertainties are due to subaquatic sources (error in flow and PO4 content) and the
overall uncertainty is due to the poorly-known rate of degradation of organic matter.

Ammonium

50

100

150

200
Depth (m)

2004
250 2054
2104
300

350

400

450

500
0 1 2 3 4 5 6
-3
NH4 (mol m )

Figure 8.7.6 Lower and upper boundaries for NH4 (one standard deviation). Local maxima
of the uncertainties are due to subaquatic sources (error in flow and NH4 content) and the
overall uncertainty is due to the poorly known rate of degradation of organic matter.

134
 Salinity

50

100 465 S
250 S
150
180 S
200 425 S
Depth (m)

365 S
250
180 Q
300 250 Q
465 CO2
350
465 T
400 turb. surf.
450

500
-0.1 0.1 0.3 0.5 0.7
-3
S (kg m )

Figure 8.7.7 Error contributions for salinity determined by identifiability analysis: The large
uncertainty contributions are due to the errors of the concentrations of the three deepest
sources.

Temperature

50

100 surf. T
180 T
150
250 T
200 465 T
Depth (m)

465 S
250
425 T
300 465 CO2
365 T
350
310 T
400 250 S
450

500
-0.5 0.0 0.5 1.0 1.5 2.0 2.5

T (°C)

Figure 8.7.8 Error contributions for temperature determined by identifiability analysis. The
largest contribution is due to the variable surface temperature. It is interesting to note that
the temperature at maximum depth is more (indirectly) influenced by the salinity and CO2
content of the sources than by their temperatures.

135
 Methane

50

100 465 S
465 CO2
150
465 T
200 CH4 prod.
Depth (m)

425 S
250
465 Q
300 465 CH4
425 CO2
350
365 S
400 425 T
450

500
-5 0 5 10 15 20 25
-3
CH4 (mol m )

Figure 8.7.9 Error contributions for CH4 determined by identifiability analysis. The large
uncertainty contributions are due to the error of the CH4 new formation rate. Note again
the strong (indirect) influence by the salinity and CO2 content of the deepest sources.

Carbon dioxide

50

100 465 CO2

465 S
150
465 T
200 250 CO2
Depth (m)

CH4 prod.
250
425 S
300 425 CO2
365 CO2
350
180 CO2
400 250 Q
450

500
-5 5 15 25 35
-3
CO2 (mol m )

Figure 8.7.10 Error contributions for CO2 determined by identifiability analysis. The large
uncertainty contributions are due to the concentrations errors of the deep-water subaqua-
tic sources. In the deepest reaches, salinity influences the density and thereby the CO2
vertical distribution.

136
 Phosphate

50

100 465 S
465 CO2
150
465 T
200 CH4 prod.
Depth (m)

425 S
250
465 Q
300 465 PO4
465 CH4
350
425 CO2
400 250 S
450

500
-0.05 0.00 0.05 0.10 0.15 0.20
-3
PO4 (mol m )

Figure 8.7.11 Error contributions for PO4 determined by identifiability analysis. The large
uncertainty contributions are due to the rate of degradation of organic matter. Note again
the strong (indirect) influence by the salinity and CO2 content of the deepest sources.

Ammonium

50

100 465 S
465 CO2
150
465 T
200 CH4 prod.
Depth (m)

465 NH4
250
425 S
300 465 Q
250 S
350
425 CO2
400 250 S
450

500
-0.5 0.0 0.5 1.0 1.5
-3
NH4 (mol m )

Figure 8.7.12 Error contributions for NH4 determined by identifiability analysis. The large
uncertainty contributions are due to the rate of degradation of organic matter. Note again
the strong (indirect) influence by the salinity and CO2 content of the deepest sources.

137
8.8 Data profiles used in the model

Table 8.8 Vertical profiles (2004) of S, T, CH4, CO2, PO4, NH4, density
and w used for the model simulations

Depth Area Vol S T CH4 CO2, total PO4 NH4 Density w

m km2 km3 kg m-3 °C mol m-3 mol m-3 mol m-3 mol m-3 kg m-3 m yr-1

0 2’368 551.3 1.074 23.48 0 11.70 0.00037 0 998.373 1.520

20 2’256 505.0 1.078 24.29 0 11.51 0.00036 0 998.174 0.577
40 2’141 461.0 1.106 23.12 0 12.18 0.00035 0 998.490 0.612
60 2’022 419.3 1.166 23.06 0.105 13.18 0.00037 0.0039 998.558 0.648
80 1’903 380.0 1.666 23.00 0.729 21.41 0.01286 0.2733 999.036 0.689
100 1’789 343.0 2.146 23.03 1.791 31.09 0.02917 0.5706 999.488 0.733
120 1’675 308.3 2.442 23.13 2.538 37.71 0.03658 0.7047 999.754 0.783
140 1’580 275.7 2.575 23.24 2.819 40.67 0.04114 0.8075 999.857 0.830
160 1’486 245.0 2.620 23.31 2.842 42.01 0.04208 0.8643 999.890 0.882
180 1’386 216.2 2.708 23.42 4.046 44.85 0.04251 1.0949 999.943 0.784
190 1’335 202.6 2.899 23.61 4.495 51.91 0.05120 1.2124 1000.118 0.553
200 1’284 189.5 3.120 23.81 4.603 58.56 0.06034 1.3300 1000.315 0.480
220 1’205 164.6 3.166 23.87 4.297 60.06 0.06559 1.3721 1000.359 0.512
240 1’127 141.2 3.173 23.88 3.873 60.33 0.06530 1.3667 1000.373 0.547
250 1’091 130.1 3.178 23.87 4.043 60.82 0.06529 1.3666 1000.383 0.566
260 1’055 119.4 4.280 24.44 10.18 98.47 0.12213 2.5862 1001.411 0.136
270 1’019 109.0 4.990 24.87 14.80 127.5 0.14180 3.0255 1002.104 0.141
280 984 98.93 5.032 24.92 15.33 128.4 0.15585 3.3453 1002.123 0.146
300 920 79.70 5.062 24.96 15.46 129.9 0.15861 3.5097 1002.149 0.156
310 870 70.73 5.193 25.09 15.67 135.2 0.15867 3.5135 1002.276 0.149
320 820 62.26 5.525 25.39 17.39 150.0 0.17669 3.6871 1002.596 0.137
330 770 54.29 5.666 25.54 18.43 156.9 0.17965 3.8409 1002.728 0.146
340 720 46.82 5.697 25.58 18.70 158.1 0.18175 3.9932 1002.750 0.156
360 594 33.62 5.711 25.60 18.58 159.5 0.18237 3.9563 1002.775 0.157
380 459 23.02 5.781 25.70 18.47 163.4 0.18290 3.9106 1002.850 0.162
400 362 14.76 5.874 25.84 18.42 167.5 0.18326 3.9317 1002.936 0.173
420 274 8.354 5.904 25.89 18.67 169.4 0.18361 3.9546 1002.965 0.198
440 187 3.702 5.914 25.91 18.93 169.7 0.18362 3.9551 1002.965 0.246
460 84 0.939 5.932 25.93 19.10 171.3 0.18362 3.9551 1002.988 0.322
480 12 0.012 5.932 25.93 19.03 172.6 0.18362 3.9551 1003.007 0
485 0 0 0

Data sources: Schmid et al (2005a,b)

Legend: Vol is the total vertically integrated lake volume below the given depth; S = salinity;
T = temperature; CO2,total is the sum of all carbonate components as defined in Section 4.10:
w = the upwelling velocity (uplift water flow Q(z) divided by lake area A(z) at depth z).

138
9. References
Cai, W.-J., and Y. Wang (1998), The chemistry, fluxes and sources of carbon dioxide in the
estuarine waters of the Satilla and Altamaha Rivers, Georgia, Limnology and Oceanography,
43(4), 657-668.
Chen, C. T., and F. J. Millero (1986), Precise thermodynamic properties for natural waters covering
only the limnological range, Limnology and Oceanography, 31, 657-662.
Chen, C. T., and F. J. Millero (1977), Use and misuse of pure water PVT properties for lake waters,
Nature, 266, 707-708.
Damas, H. (1937), La stratification thermique et chimique des lacs Kivu, Edourd et Ndalaga (Congo
Belge), Verhandlungen der Internationalen Vereinigung für Limnologie, 8(3), 51-68.
Degens, E. T., R. P. von Herzen, H.-K. Wong, W. G. Deuser, and H. W. Jannasch (1973), Lake
Kivu: structure, chemistry and biology of an East African Rift Lake, Geologische Rundschau,
62, 245-277.
Deuser, W. G., E. T. Degens, and G. R. Harvey (1973), Methane in Lake Kivu: New data bearing its
origin, Science, 181, 51-54.
Dumont, H. J. (1986), The Tanganyika Sardine in Lake Kivu: another ecodisaster for Africa?
Environmental Conservation, 13, 143-148.
Durieux, J. (2004), Nyiragongo: the January 10th 1977 eruption, Acta vulcanologica, 14/15(2002-
2003), 145-148.
Esper G., W. Lemming, W. Beckermann and F. Kohler (1995), Acoustic determination of ideal gas
heat capacity and second virial coefficient of small hydrocarbon, Fluid Phase Equilibria, 105,
173-192.
Fan, J. (2008), Stratified flow through outlets, Journal of Hydro-environment Research 2, 3 - 18
Haberyan, K. A., and R. E. Hecky (1987), The late pleistocene and holocene stratigraphy and
paleolimnology of Lake Kivu and Tanganyika, Palaeogeography, Palaeoclimatology,
Palaeoecology, 61, 169-197.
Isumbisho, M., M. Kaningini, J.-P. Descy, and E. Baras (2004), Seasonal and diel variations in diet
of the young stages of the fish Limnothrissa miodon in Lake Kivu, Eastern Africa, Journal of
Tropical Ecology, 20, 73-83.
Jannasch, H. W. (1975), Methane oxidation in Lake Kivu (central Africa), Limnology and
Oceanography, 20(5), 860-864.
Kataoka, T., M. Tsutahara, and M. Tanaka (2000), Point sink flow in a linearly stratified fluid of finite
depth, Physics of Fluids 12(11), 2775-2786.
Kelley, D. E. (1990), Fluxes through diffusive staircases: a new formulation, Journal of Geophysical
Research, 95(C3), 3365-3371.
Kling, G. W., M. A. Clark, H. R. Compton, J. D. Devine, W. C. Evans, A. M. Humphrey, E. J.
Koenigsberg, J. P. Lockwood, M. L. Tuttle, and G. N. Wagner (1987), The 1986 Lake Nyos
gas disaster in Cameroon, West-Africa, Science, 236(4798), 169-175.
Kocsis, O., H. Prandke, A. Stips, A. Simon, and A. Wüest (1999), Comparison of dissipation of
turbulent kinetic energy determined from shear and temperature microstructure, Journal of
Marine Systems, 21, 61-84.
Komorowski, J.-C., et al. (2004), The January 2002 flank eruption of Nyiragongo volcano
(Democratic Republic of Congo): chronology, evidence for a tectonic rift trigger, and impact of
lava flows on the city of Goma, Acta vulcanologica, 14/15(2002-2003), 27-62.
Lorke, A., K. Tietze, M. Halbwachs, and A. Wüest (2004), Response of Lake Kivu stratification to
lava inflow and climate warming, Limnology and Oceanography, 49(3), 778-783.
Manins, P. C. (1976), Intrusion into a stratified fluid, J. Fluid Mech., 74(3), 547-560.
Marshall, B. E. (1991), The impact of the introduced sardine Limnothrissa miodon on the ecology of
Lake Kariba, Biological Conservation, 55, 151-165.
Matzner, R. A. (2001), Dictionary of Geophysics, Astrophysics and Astronomy, 536 pp., CRC Press,
Boca Raton.
Muvundja, F. A., N. Pasche, F. W. B. Bugenyi, M. Isumbisho, B. Müller, J. N. Namugize, P. Rinta, M.
Schmid, R. Stierli, and A. Wüest (2009), Balancing nutrient inputs to Lake Kivu, J. Great
Lakes Research, 35(3), 406-418.

139
MP 2009. Expert working group on Lake Kivu gas extraction (J. Boyle, F. Hirslund, P. Morkel, M.
Schmid K. Tietze, and A. Wüest). Management prescriptions for the development of Lake
Kivu gas resources. Report to Ministry of Infrastructure Republic of Rwanda and Ministry of
Hydrocarbons Democratic Republic of the Congo. 17 June 2009, [Link]
NCEA (2007, 2009), Safe gas extraction in Lake Kivu, Netherlands Commission for Environmental
Assessment, [Link]
Newman, F. C. (1976), Temperature steps in Lake Kivu: a bottom heated saline lake, Journal of
Physical Oceanography, 6, 157-163.
Nzayisenga, T., A. Mugisha, L. Rwandekwe, A. Umutoni, N. Pasche (2009), Monitoring of the
discharge plume of KP1 in Lake Kivu - Initial phase October 2008 to April 2009. MININFRA,
Rwanda and Eawag, Switzerland. [Link]
kivu/index_EN.
Osborn, T. R. (1980), Estimates of the local rate of vertical diffusion from dissipation measurements,
Journal of Physical Oceanography, 10, 83-89.
Pasche, N., C. Dinkel, B. Müller, M. Schmid, A. Wüest, and B. Wehrli (2009), Physical and
biogeochemical limits to internal nutrient loading of meromictic Lake Kivu, Limnology and
Oceanography, 54(6), 1863-1873.
Pasche, N., B. Wehrli, G. Alunga, F. Muvundja, M. Schurter, and M. Schmid (2011), Abrupt onset of
carbonate deposition in Lake Kivu during the 1960s: response to food web alteration and
hydrological change, J. Paleolimnol. in revision.
Pasche, N., M. Schmid, F. Vazquez, C. J. Schubert, A. Wüest, J. D. Kessler, M. Pack, W. S.
Reeburgh, and H. Bürgmann (submit), Methane sources and sinks in Lake Kivu, submitted.
Pollack, H. N., S. J. Hurter, and J. R. Johnson (1993), Heat flow from the earth's interior: analysis of
the global data set, Reviews of Geophysics, 31(3), 267-280.
Reichert, P. (1994), AQUASIM - A tool for simulation and data analysis of aquatic systems, Water
Science and Technology, 30(2), 21-30.
Reckhow, K. H. (1994), Water-Quality simulation modelling and uncertainty analysis for risk
assessment and decision-making, Ecological Modelling, 72(1-2), 1-20.
Reitsma, R. F. (1996), Structure and support of water-resources management and decision-making, J.
Hydrology, 177(3-4), 253-268.
Rettich, T. R., Y. P. Handa, R. Battino, and E. Wilhelm (1981), Solubility of gases in liquids.13.
High-Precision determination of Henry constants for methane and ethane in liquid water at
275-K TO 328-K, J. Physical Chemistry, 85, 3230-3237.
Sarmento, H., M. Isumbisho, S. Stenuite, F. Darchambeau, B. Leporcq, and J.-P. Descy (2009).
Phytoplankton ecology of Lake Kivu (eastern Africa): biomass, production and elemental
ratio. Verh. Internat. Verein. Limnol. 30, 709–713.
Schmid, M., K. Tietze, M. Halbwachs, A. Lorke, D. McGinnis, and A. Wüest (2004), How hazardous
is the gas accumulation in Lake Kivu? Arguments for a risk assessment in light of the
Nyiragongo Volcano eruption of 2002, Acta vulcanologica, 14/15(2002-2003), 115-121.
Schmid, M., M. Halbwachs, B. Wehrli, and A. Wüest (2005a), Weak mixing in Lake Kivu: New
insights indicate increasing risk of uncontrolled gas eruption, G3 Research Letter 6, Q07009,
doi:10.1029/2004GC000892.
Schmid, M., M. Halbwachs, and B. Wehrli (2005b), Report of the scientific expeditions to Lake Kivu
in November 2003 and February 2004. An investigation of physical and chemical properties
of Lake Kivu as a base for gas outburst risk assessment, 17 pp, EAWAG, Kastanienbaum,
[Link]
Schmid, M., M. Halbwachs, and A. Wüest (2006), Simulation of CO2 concentrations, temperature,
and stratification in Lake Nyos for different degassing scenarios, Geochem., Geophys.,
Geosys., 7, Q06019, doi: 10.1029/2005GC001164.
Schmid, M., M. Busbridge, and A. Wüest (2010), Double diffusive convection in Lake Kivu,
Limnology and Oceanography, 51(1), 225-238.
Schmidt, W. (1928), Über Temperatur- und Stabilitätsverhältnisse von Seen’, Geogr. Ann., 10, 145–
177.
Schoell, M., K. Tietze, and S. M. Schoberth (1988), Origin of methane in Lake Kivu (East-Central
Africa), Chemical Geology, 71, 257-265.

140
Sigurdsson, H., J. D. Devine, F. M. Tchoua, T. S. Presser, M. K. W. Pringle, and W. C. Evans
(1987), Origin of the Lethal Gas Burst from Lake Monoun, Cameroun, Journal of Volcanology
and Geothermal Research, 31(1-2), 1-16.
Sigvaldason, G. E. (1989), International Conference on Lake Nyos Disaster, Yaoundé, Cameroon
16-20 March, 1987 - Conclusions and Recommendations, Journal of Volcanology and
Geothermal Research, 39(2-3), 97-107.
Stoffers, P., and R. E. Hecky (1978), Late Pleistocene-Holocene evolution of the Kivu-Tanganyika
Basin, Special publications of the International Association of Sedimentologists, 2, 43-55.
Tedesco, D. (2004), 1995 Nyiragongo and Nyamulagira activity in the Virunga National Park: A
volcanic crisis, Acta vulcanologica, 14/15(2002-2003), 149-156.
Tietze, K. (1978), Geophysikalische Untersuchung des Kivusees und seiner ungewöhnlichen
Methangaslagerstätte - Schichtung, Dynamik und Gasgehalt des Seewassers, PhD Thesis,
Christian-Albrechts-Universität, Kiel, Germany.
Tietze, K., M. Geyh, H. Müller, L. Schröder, W. Stahl, and H. Wehner (1980), The genesis of the
methane in Lake Kivu (Central Africa), Geologische Rundschau 69(2), 452-472.
Tietze, K., and E. Maier-Reimer (1977). Recherches Mathématiques– Physiques pour la Mise en
Exploitation du Gisement de Gaz Méthane dans le Lac Kivu (Zaïre/Rwanda). Report No
76003, Bundesanstalt für Geowissenschaften und Rohstoffe, 2 Volumes, Hanover, 180 p.
Tietze, K. (1981). Direct measurements of the in-situ density of lake and sea water with a new
underwater probe system. Geophysica, 17 (1-2), 33-45.
Turner, J. S. (1965), The coupled turbulent transports of salt and heat across a sharp density
interface, International Journal of Heat and Mass Transfer, 8, 759-767.
Turner, J. S. (1979), Buoyancy Effects in Fluids, 368 pp., Cambridge University Press, Cambridge.
Weiss, R. F. (1974), Carbon dioxide in water and seawater: The solubility of a non-ideal gas,
Marine Chemistry, 2, 203-215.
Wood, I. R. (2001), Extensions to the theory of selective withdrawal, Journal of Fluid Mechanics,
448, 315-333.
Wüest, A., and A. Lorke (2003), Small-scale hydrodynamics in lakes, Annu. Rev. Fluid Mech. 35,
373-412.
Wüest, A., G. Piepke, and J. D. Halfman (1996), Combined effects of dissolved solids and
temperature on the density stratification of Lake Malawi, in The limnology, climatology and
paleoclimatology of the East African Lakes, edited by T.C. Johnson, and E.O. Odada, pp.
183-202, Gordon and Breach, Toronto.

- end of report -

141