Factors that influence the (local) water balance :

It is not only spatial but also temporal variation of the water balance components. This example shows for a
humid climate the precipitation volumes for individual
months. The precipitation is higher in the winter
months starting from September-October lasting until
about February and in summer as expected
precipitation is lower. In contrast the
evapotranspiration has its maximum in the late
summer periods where the days are longer and solar
radiation is higher. If we look at the differences
between the two curves we see that there is a water
surplus in the months of January to April and again
from October to December where precipitation is
higher than the evapotranspiration and there is a water deficit in the
summer months between May and September that leads to soil and
ground for the store depletion. The subsurface needs to be recharged in
winter and typically in November the subsurface stores are recharged.
This is the reason why in central Europe the hydrological year starts at
the first of November and ends at the 31st of October.
Variation of R and ETR depending on soil properties for constant
climate (energy, P)
Let's assume we have a constant climate so the energy input is constant and so is the
precipitation. Now the maximum evapotranspiration and the minimum runoff is
found at an average infiltration capacity of the soil. In the graph the elements of the
water balance are depicted on the y-axis, so the division of the precipitation into
evapotranspiration and runoff and on the x-axis we see the infiltration capacity
increasing from left to right and conversely the water retention capacity, so the ability
of the soil to store water decreasing from right to left. Half way down the x-axis we
find average infiltration capacity and there we also find the smallest runoff and the
largest evapotranspiration If we move along the x-axis to the right we have an
infiltration capacity that increases, so more water is actually recharged by
groundwater and therefore there is less water available for evapotranspiration .If we
move from the average infiltration capacity towards the left, so towards a smaller
infiltration capacity, we get lower infiltration and higher surface runoff and therefore
also a decrease in evapotranspiration. (better explained in pdf)
 What are the six storage
compartments for water? Draw a
picture depicting these. What are the
major hydrological processes? Include
them on the picture to show how they
are related to/how they interact with
the storage compartments.
The six storage compartments for
water in the hydrological cycle
are:Oceans: The largest reservoir
where a vast amount of Earth's water
is stored, containing about 97% of the
Earth's water. Glaciers and Ice Caps:
These store a significant portion of
freshwater in the form of ice, located
predominantly in polar regions and
mountainous [Link]:
Water stored underground in
aquifers, permeable rock formations,
or soil, serving as a vital source for wells and [Link] and Rivers: Surface water bodies that store water
temporarily, receiving inflow from precipitation, runoff, and groundwater discharge, while releasing water via
[Link] Moisture: Water held within the soil, which is crucial for supporting plant growth and replenished
by infiltration from [Link]: The reservoir where water exists as water vapor, contributing to
the hydrological cycle through processes like evaporation, condensation, and precipitation.

The major hydrological processes

interact with these storage
compartments as
follows:Precipitation: Water falls
onto the Earth's surface, directly
replenishing various compartments
such as lakes, rivers, and soil
moisture. Evaporation and
Transpiration: Surface water
bodies, soil moisture, and plants
release water vapor into the
atmosphere through evaporation
and transpiration, contributing to
atmospheric moisture. Infiltration:
Precipitation infiltrates into the soil,
recharging groundwater and
contributing to soil moisture.
Runoff: Excess water from
precipitation that does not infiltrate
the soil flows overland into rivers,
lakes, and ultimately the oceans, contributing to surface water storage. Groundwater Flow: Water moves
underground through aquifers, replenishing groundwater storage and sometimes discharging into surface
water bodies. Condensation: Water vapor in the atmosphere condenses to form clouds, leading to precipitation
and replenishing various compartments.
##Name three way to measure each of the three main components of the water balanceequations.
Precipitation (P): Rain Gauges: Rain gauges are simple instruments used to measure rainfall. They consist of a
cylindrical container that collects rainwater. The amount of rainfall is measured by the depth of water
accumulated in the gauge.
Weather Radars: Weather radars use radio waves to detect precipitation in the atmosphere. They provide
information on the intensity, location, and movement of precipitation over a wide area.
Satellite-Based Precipitation Estimation: Remote sensing satellites equipped with sensors can estimate
precipitation by measuring cloud properties, temperature, and other atmospheric variables, providing valuable
data on precipitation patterns over large regions.
Evapotranspiration (ET): Bowen Ratio Technique: This method determines ET by measuring the ratio of
sensible heat flux (related to temperature changes) to latent heat flux (related to ET) at the Earth's surface
using specialized sensors.
Eddy Covariance Method: This technique directly measures the exchange of gases (including water vapor)
between the Earth's surface and the atmosphere, providing continuous and real-time data on ET.
Satellite-Based Thermal Remote Sensing: Remote sensing satellites equipped with thermal infrared sensors can
estimate ET by measuring land surface temperatures and analyzing heat fluxes, providing spatial information
on ET at different scales.
Runoff (R): Stream Gauges: Stream gauges or flow meters are installed in rivers or streams to measure the
volume or flow rate of water passing through a specific point over time, providing data on runoff.
Hydrological Modeling: Various hydrological models (e.g., hydrologic simulation models) use mathematical
equations and data inputs (such as precipitation, soil characteristics, land use) to simulate and estimate runoff.
Tracer Studies: Tracers, such as dye or chemicals, can be introduced into a water body to track its movement
and estimate runoff by analyzing the concentration changes downstream.
##Three main components of annual water balance at the land surface:
Precipitation (P): Total amount of water, including rain and snow, that falls onto the land surface within a
specified area and time frame.
Evapotranspiration (ET): The combined process of water evaporation from surfaces and transpiration from
plants back into the atmosphere.
Runoff (R): The portion of precipitation that flows over the land surface, eventually reaching rivers, lakes, and
oceans after not being absorbed or evaporated.
##. Which physical laws are of fundamental importance for the movement of water on the land surface? Ans:
conservation of energy and conservation of mass
##2. Name the important components and processes of the hydrologic cycle. Ans: Evapotranspiration ,
Condensation, Precipitation, Runoff, infiltratiom
## Which physical laws are of fundamental importance for the description of the movement of water
Conservation of Mass: it means that the total volume of water entering an area (such as precipitation) should
equate to the sum of outflows (such as evaporation, runoff, and groundwater recharge) and any changes in
storage (such as in lakes, rivers, or groundwater reservoirs). Thus, the conservation of mass is crucial in
calculating and understanding water inflows, outflows, and storage changes in hydrological systems.
Conservation of Energy: The conservation of energy is also critical in hydrology. Energy transformations are
integral to various hydrological processes, particularly in the phase changes of water. For instance, solar
radiation provides the energy required for evaporation, where liquid water turns into vapor. This principle
helps in understanding the energy transfers involved in these phase changes. Additionally, energy
considerations are essential for analyzing temperature variations, which influence the movement and behavior
of water within the hydrological system. Overall, the conservation of energy principle aids in comprehending
the role of energy exchanges in altering water states and driving movements within the hydrological cycle.
##How can the dynamics of the hydrologic cycle be
described
Variation of solar radiation intensity with angle of
inclination The global energy input is highly viable in
space, due to the variation of the solar radiation
intensity with the angle of inclination. At the equator,
there is a much higher energy density as compared to
near the poles, because the same energy is distributed
over a larger area. This can be seen in the pictures here
towards the right. The top shows the energy input at the
outer bounds of the atmosphere, with the yellow color
indicating a value of around 400 W per square meter,
and the red and white colors indicating reduced energy
input. The energy inputs are not only variable in space,
but they are also variable in time. This is course by two
factors. First, the rotation of Earth around its own axis,
which causes days and nights. And second, the rotation of Earth around the Sun, which causes our seasons
because of the constant tilt of the Earth's axis
In the following animation we see different products for precipitation,
evapotranspiration soil moisture, vegetation index and changes in
storage. Those products are shown here for a season in denial catchment
and we start with precipitation data. The product is called TRMM. The
higher values are in the upper parts of the catchment to the right and the
lower values are towards the left in the red colors. Next is ALXI for evapotranspiration then soil moisture LDAS,
larger values towards the upper parts of the catchment, vegetation index and the VI and the changes in storage
with GRACE.
Earth's atmospheric circulation is influenced by
several factors: the Sun's energy, the planet's
rotation, and how land and water are spread across
the globe. These factors contribute to the creation of
three major circulation loops in both the northern
and southern [Link] primary reason for
these circulation bands is the uneven distribution of
solar energy received at the equator and the poles.
This imbalance sets up a transfer of heat from the
warmer equatorial regions to the cooler polar areas.
These three main circulation bands are known as the
Hadley cell, Ferrel cell, and Polar cell.
Hadley Cell: This cell is the closest to the equator and
extends up to approximately 30 degrees latitude in both hemispheres. Here, warm air rises from the equatorial
region and moves towards the poles, cooling as it [Link] Cell: Situated between 30 and 60 degrees
latitude, this cell operates in both hemispheres. Air in this region moves in a more complex manner, driven by
the interactions between the Hadley and Polar cells.
Polar Cell: Located poleward of 60 degrees latitude in both hemispheres, this cell involves sinking cool air that
moves back towards the lower [Link] the equator, warm and moist air from the northern and southern
hemispheres converges. As this air rises, it cools and condenses, forming clouds and precipitation. This rising
motion of air results in an area of high pressure known as the Intertropical Convergence Zone (ITCZ). It's
marked by heavy rainfall and atmospheric instability due to the constant ascent of warm [Link] circulation
pattern, coupled with the higher energy input from the Sun at the equator compared to other parts of the Earth,
leads to greater precipitation on average at the equator. The consistent rising of warm, moisture-laden air
contributes to the formation of rain clouds and heavy rainfall in these region
Chapter 2
Hydrological processes: transient movement of water
Precipitation P
 Evapotranspiration ETR- Interception , Evaporation , Transpiration
 Runoff R - Surface runoff , Runoff from soils , Runoff from groundwater, Runoff from snowmelt , Infiltration I
Storage S - Soil storage - Groundwater storage - Storage in rivers and lakes - Interception - Surface storage -
Snow/ice storage
purpose of the investigation.
Again, an example for the operation of reservoirs for flood protection, we need only a low level of detail in the
process understanding, and it's enough to have the runoff time series as an input of the flow rates into the
reservoir, and we can do a long analysis with a simple water balance equation. On the other hand, if we are
interested in the water quality of a drinking water bore, we need a much higher level of detail in the process
understanding. We need to understand the various flow paths in the subsurface and the residence times to
delineate the 50-day protection zone around the bore, which in turn requires a more distributed analysis of the
flow and transport in the subsurface.
Quantification of hydrological processe : Water balance calculations and hydrological models are used to
support water resources management and planning,
Quantification of hydrological processe forms the basis for
 availability of water resources in space and time.
 They are required for the design of water supply and distribution facilities, such as drinking water
reservoirs.
 They are useful for regional landscape planning, where different values are traded off against each other,
such as sustainability and economical versus ecological values. For instance, when the decision has to be made,
whether or not a dam can be built or needs to be built in a national park.
 Empty quantification of hydrological processes is also required for the development of regionalization
methods, that means when we want to divide aerial estimates from point measurements, for example for
precipitation, or if the infer characteristics of gauged catchments and apply those characteristics to ungauged
catchments. The example here would be the prediction of runoff from an ungauged catchment.
transformation of precipitation into runoff. (core hydrological process)
Runoff Generation Phases: Includes three phases—runoff formation, runoff concentration, and flood
propagation—each contributing to
the overall response.

Runoff formation describes how

much rainfall is turned into direct
runoff. Direct runoff is the immediate
response of the flow hydrograph to a
rainfall event. Intuitively, this is the
part of the precipitation that runs off
at the surface and ends up directly
into the streams. The land surface
properties determine how much of
the precipitation is infiltrating into
the subsurface and whether or not
there is some direct runoff at the
land surface. In most environments,
the direct runoff at the land surface
is less than 10% of the precipitation
volume. But there are also
exceptions. For example, if the infiltration capacity, so the hydraulic conductivity of the soils, is particularly low,
such as urban areas where the percentage of sealed surfaces is relatively high, or for soils that are hydrophobic,
like after a prolonged drought or after a forest fire. But there are also other mechanisms that lead to a fast
response in the runoff from a precipitation event. In the presence of macropores that are depicted in the
schematic here as preferential flow paths, water of a precipitation event can reach greater depths of the soil or
the subsurface relatively quickly. And if it hits a layer of lower permeability as indicated here, it can produce
lateral flow along the slope here, and the response in the runoff hydrograph will also be relatively quick.
Therefore, vertical and lateral bypass flow or preferential flow can also lead to a fast runoff response to the
rainfall. This is also sometimes called hypodermic flow. Another mechanism is that infiltrating water, not only
preferentially, but also within the soil matrix, can lead to a pressure change in the groundwater system. And
then old groundwater is pushed into the river, and again we observe this as a fast response of the precipitation
event. This mechanism is sometimes called the old water paradox.
To illustrate the old water paradox, let's consider a hill slope with a steep angle and a river at the bottom, and
just one particular preferential flow path, with a groundwater reservoir underneath, like this. And there is some
water level in the river, like this, which then roughly equates to the groundwater level over here. Now I just
indicate that this is all water, and now we have a steady state situation where the river and the adjacent
groundwater level are connected. Now let's consider that there is a precipitation event, so it's raining at the
surface, and water is moving into that preferential flow path, in here, and filling up this pipe. Then we have a
new water level in this pipe, and the head difference between the existing water level and the new water level,
and therefore the system tries to equilibrate by pushing the water down here and up here. And as this happens,
the old water that has been sitting here for some time is pushed out of the system, and the new water here is
entering the system at the top. And therefore the water that appears in the river doesn't seem to be water that
originates from the precipitation event, but seems to be older water. And the old water paradox is that we see
an increase in the water levels and the runoff in the streams as a result of a precipitation event, but the water in
the stream is rather old. And that's why we have a new water level in this pipe
##Runoff formation hydrological processes of hill slopes.
 Precipitation division and water movement on the hill slope: When precipitation occurs on the depicted
hill slope, it gets partitioned into various components:Surface storage: Some water is retained at the land
surface, possibly in depressions or as surface water [Link] runoff: A portion of the water runs off
directly on the land [Link] subsurface infiltration: Another part infiltrates into the shallow soils and
moves laterally along the hill slope, referred to as hypodermic flow.
 Different pathways and residence times of water: The water that infiltrates the shallow subsurface can
take various pathways:Hypodermic flow: This is relatively quicker than percolation but slower than surface
runoff. It moves swiftly but not as fast as water running off directly at the [Link] into deeper layers:
Some water percolates deeper and becomes groundwater, taking weeks or even years to reach the stream.
 Arrival times of water at the stream: Due to the different pathways and residence times:Surface water
arrives quickest: Water running off directly at the surface reaches the stream [Link] flow arrives
relatively quickly: It's faster than percolation but slower than surface [Link] takes the longest:
Water that infiltrates deeper layers and becomes groundwater can take a significantly longer time to reach the
stream, sometimes years.
 Mobilization of water by precipitation: The response seen during a precipitation event might not always
be the direct water from that event. It can also mobilize stored water from various sources and pathways (like
hypodermic flow or deeper groundwater), resulting in older water contributing to the runoff.
 Run-off concentration and its complexity: The complexity of natural systems leads to a wide range of
water ages entering the stream, each with its own characteristic arrival times. Natural tracers like isotopes help
identify water origins and residence times in this system.
 Factors influencing water arrival times: Each hydrological system is unique, influenced by factors like
morphology, soil types, geology, vegetation, and transient soil moisture conditions during precipitation events.
These factors initiate various hydrological processes, contributing to different water arrival times in the stream.
 Stream as the final point of convergence: All these different quantities and timings of water entering the
stream, with their diverse origins and ages, collectively contribute to what's known as the run-off
concentration.(slide 3 page 8)
##the run-off concentration
The run-off concentration is like a mix of different waters coming together in a stream. These waters take
different paths, come from different places, and have different ages. It happens because of how nature works
and how it reacts when it rains Run-off concentration relies on various catchment characteristics and additional factors
like its [Link] of catchment shape on run-off hydrograph: Comparing two catchments of the same area but
different shapes—elongated (Catchment A) and round (Catchment B)—under the same rainfall event (same
volume, intensity, and duration):
Catchment A (elongated shape): The blue curve in the graph represents this catchment. Run-off from the upper
part of the catchment takes longer to reach the outlet than run-off from the lower part. This creates a longer,
flatter hydrograph due to the longer travel time along the stream from the farthest end to the [Link]
B (round shape): The red curve in the graph represents this catchment. Flow paths and travel times from all
parts of the catchment to the outlet are more similar. Consequently, the run-off arrives more uniformly,
resulting in a quicker rising, peaky hydrograph that recedes faster than Catchment [Link] in response
explained by shape: Catchment shape influences the timing of run-off reaching the outlet. In the elongated
catchment, the distance water travels along the stream from the farthest part leads to a more spread-out,
prolonged response. On the other hand, the round catchment's uniform flow paths create a quicker,
concentrated response at the outlet.(pdf 3slide 9)
##Flood protection
The third phase of run-off generation is flood propagation. Flood propagation describes processes that are
happening solely in the channel of a river such as retention and storage. Depicted in the upper panel of this slide
are two gaging stations along a river. Station A is upstream, station B is downstream and the run-off over time is
measured at both stations and plotted in the figure at the bottom right. Due to the friction and resistance of the
riverbed and the retention of the channel, we observe that the relatively steep peak of the run-off observed at
the upstream gaging station is then at the downstream location B. The peak arrival time is later, the peak is
lower and the hydrograph is more elongated.
Example of data issue and scale
 Catchment Sensor Distribution:A 27-hectare research site, Wüstebach, features various sensors
distributed across the catchment [Link] moisture probes installed at different depths provide data for
spatial interpolation, creating soil moisture maps at different [Link] viability of soil moisture at
 smaller scales, but limitations in instrumenting larger catchments due to cost, maintenance, and missing
important processes like preferential flow despite high-density sensor networks.

 Scale of Measurements and Upscaling Challenges:Point-scale measurements, like lysimeters with high
accuracy at a smaller scale, pose challenges in upscaling to larger scales of fields, hydrotopes, or
[Link] techniques (soil moisture monitoring networks, geophysics, cosmic ray probes,
remote sensing) provide information at larger scales but face challenges of temporal continuity and
reduced accuracy with increasing scale due to indirect methods reliant on transfer functions.A trade-off
exists between measurement scale and accuracy, suggesting the use of spatially distributed
measurements alongside point-scale measurements to improve accuracy through assimilation and bias
correction.

 Precipitation Measurements at Different Scales:Various precipitation measurement techniques include

rain gauges, radar measurements, and remote sensing products like TRMM, available at different scales
and offering varying accuracy (local to global).

 Information Content of Data: Redundancy exists in heavily instrumented sites, where similar
measurements from close sensors or similar soils yield redundant information without providing
additional insights. Low information content may result from redundancy or irrelevance of certain data
types for specific hydrological analyses. Measurement type, density, frequency, scale, and accuracy
influence information content.

Chapter 3
Role of Hydrological Modeling
Hydrological models integrate measurements at different scales, aiding in hypothesis testing, understanding
hydrological processes, identifying their significance, and facilitating upscaling and regionalization. These
models help in assessing the value of measurement techniques, observation networks, and guide future
investigations to derive meaningful insights.
1. Fit-for-Purpose Modeling:
 Scale and Relevance: Models in hydrology must be tailored to the specific investigation to answer
pertinent questions effectively.
 Simplicity vs. Complexity: Models should strike a balance—being as simple as possible yet as complex as
necessary—to describe only the relevant hydrological processes based on the intended purpose of the
model.
2. Illustrative Examples - Climate Change Impact Assessment:
 Sea Level Rise Prediction: To estimate sea level rise by 2050 due to climate change, a global scale energy
budget may be used. Predictions involving annual averages of global greenhouse gas emissions, energy
retardation in the atmosphere, and temperature rise might be considered. These, along with other
factors, help predict glacial melt and subsequently forecast sea level rise by 2050.
 Fir Trees' Vulnerability to Drought: Assessing the vulnerability of fir trees to drought involves a different
set of scales and processes. It requires considering local-scale dynamics such as energy, moisture,
carbon, and nutrient cycles over shorter time scales (days or hours). Climate drivers like solar radiation,
temperature, wind velocity, precipitation, and CO2 concentration become essential inputs. Specific
hydrological processes like canopy interception, throughfall, stem flow, soil infiltration, soil water
movement, root water uptake, photosynthesis, and respiration need to be incorporated.
 The examples highlight that different questions demand different scales—global vs. local—and varying
temporal resolutions—annual averages vs. hourly/daily timeframes.: The hydrological processes
required to answer distinct questions vary. For sea level rise, global-scale climate dynamics are relevant.
In contrast, fir trees' vulnerability to drought demands localized, intricate ecological and hydrological
interactions.
Model Scales
1. Micro Scale:
a. Spatial Domain: Ranges from dizometers (tens of square meters) up to [Link]: Deals with small-
scale features such as fields, soil heterogeneity, preferential flow paths, individual plants or trees, and small
 groups of [Link] Emphasis: Emphasizes spatial heterogeneity in soils and individual
plant interactions.
2. Meso Scale:
a. Spatial Domain: Spans from hectares to a few hundred square [Link]: Encompasses catchments,
aquifers, and similar [Link] vs. Homogeneity: Considers the homogeneity of landforms within
this scale and generally assumes less spatial variability in soils compared to the micro scale. It addresses
heterogeneity at a higher, more generalized level.
3. Macro Scale:
a. Spatial Domain: Ranges from a few hundred square kilometers to millions of square [Link]:
Encompasses large-scale features like river basins and global climate [Link] of Heterogeneity:
Resolves heterogeneity at the scale of degrees of latitude and longitude, focusing on broader climate
patterns and global-scale hydrological processes.
Models at different scales might describe hydrological processes using different physical relationships or models. For
instance, transpiration rates could have distinct physical representations at the microscale (individual tree) compared to
the macro scale (global climate models). Models must consider subscale effects if relevant processes at smaller scales
significantly impact the larger-scale dynamics being modeled. These subscale effects ensure that processes not fully
resolved at the chosen model scale are appropriately considered in the model representation.

The examples provided illustrate how hydrological models vary not only in spatial scale but also in the level of
detail regarding the description of hydrological processes.

Macro Scale - River Flow Prediction: Predicting river flows at a larger catchment scale (macro scale).
Uses a macro-scale model employing linear stores or a rainfall-runoff relationship. Focuses on predicting overall
river flow behavior without intricate spatial details. Represents the catchment-scale behavior of water
movement and runoff generation.
Detailed Groundwater Model - Self-Cleaning Mechanisms of Rivers Studying how rivers clean themselves
naturally.
Uses a detailed groundwater model for a specific part of the river. Looks at how long water stays in the ground
and moves between the surface and underground water. Studies how quickly reactions happen underground to
understand how the river cleans itself, focusing closely on specific locations and timeframes.

Micro Scale - Crop Productivity Model: Assessing crop productivity. Describes soil-plant-atmosphere processes
in detail, including water, nitrogen, carbon fluxes, and plant growth-related processes. Provides a high level of
detail with a focus on the soil-plant-atmosphere continuum. Describes processes such as evapotranspiration at a
higher temporal and spatial resolution compared to larger-scale models.

Modelling concept
A model is a simplified representation of reality. Hydrological models are abstractions of natural hydrological
systems and should describe their dominant processes. First step to move from qualitative to quantitative
description of hydrologic processes: conceptual outline of the connection and interplay of compartments and
processes.
Explanation of flow chart very important (slide 4 pg 9)
We have the driver of the hydrological processes, the energy sun, and on top the precipitation as the main input. Precipitation might be
divided into snow, so solid precipitation and liquid precipitation that goes first into interception storage. Some of it bypasses the
interception storage and goes to surface storage. Depending on the energy input, some of the storage in the interception might be lost
by evaporation in this case. Some of the solid precipitation is accumulating in the snowpack. Then from the interception storage there
is also throughfall in both solid and liquid form. At the surface we have storage, and then from that storage there is water moving along
as infiltration into the soil store. That leads to a change in soil moisture. Some of the water at the surface leaves the surface storage as
overland flow and feeds directly into the streams or lakes. In the soil store water can be uptaken by plants, so that is the transpiration
part. Some of the water can also percolate in deeper layers and form groundwater. This is the groundwater store here, which can feed
streams and lakes or can cause runoff. The runoff comes either from the groundwater store or from the streams and lakes. This is one
of the components of the water balance shown here on the left. The runoff, evapotranspiration at the top here originates from
interception, from soil, directly or from plants by transpiration, or evaporation from streams and lakes. So from open water surfaces.
This is a rough outline of the conceptual interaction of hydrological processes that are typically considered in hydrological models.
General structure of (hydrological) models
Hydrological modeling involves a structured approach encompassing inputs, system description, and outputs.
Understanding the complexities within each of these components is crucial as it helps in quantifying and
managing uncertainties inherent in hydrological models.
1. Inputs:
The inputs to hydrological models typically include variables such as precipitation, potential
evapotranspiration, and temperature. These factors significantly influence the behavior of water within a
hydrological system. However, these inputs can have various associated errors:
 Measurement Errors: Mistakes or inaccuracies during data collection or recording.
 Resolution Errors: Limitations arising from the granularity or accuracy of data in terms of time and
space.
 Interpolation Errors: Occur when data from point-scale measurements are extrapolated or upscaled to
cover larger areas.
For instance, precipitation measurements might have inaccuracies due to the sampling method, gauge
placement, or interpolation techniques when extrapolating from point measurements to broader areas.
2. System Description:
The system description involves several choices regarding model equations, parameters, model structures, and
solvers:
 Model Equations: These can range from linear models to differential equations (ordinary or partial), and
the dimensionality of these equations might be 1D, 2D, or 3D.
 Model Parameters: These include characteristics such as hydraulic conductivity, storage capacity,
infiltration rates, etc. These parameters may be deterministic or could incorporate stochastic models, such as
geostatistical methods.
 Model Structure: This encompasses the framework of the model, whether it's a bucket-type model,
distributed model, mechanistic model, deterministic, or stochastic.
 Solver: The method employed to solve the set of equations, which could be numerical (finite difference,
finite element, finite volume schemes) or an analytical solution.
Errors within the system description arise from various sources:
 Model Structure Errors: These result from the simplification or abstraction of natural systems, choosing
which processes to include or neglect, and how to represent these processes using specific equations or
parameters.
 Parameterization Errors: Stem from uncertainties or lack of knowledge regarding specific parameters,
including their spatial and temporal variability.
 Scale Problems: Discrepancies when transitioning processes across different spatial or temporal scales.
 Numerical Errors: Associated with the chosen solution scheme or numerical methods used.
3. Outputs:
Model outputs represent the quantities of interest, such as runoff, infiltration, evapotranspiration, storage
within different compartments of the hydrological system, travel times, etc. Deterministic models provide a
single set of outputs for a given parameterization or model structure.
Dealing with Uncertainty:
To effectively manage and quantify uncertainties associated with hydrological models, various strategies can be
employed:
 Uncertainty Analysis: Employ statistical methods to quantify uncertainties arising from different sources.
 Sensitivity Analysis: Assess the impact of variations in inputs or parameters on model outputs.
 Monte Carlo Simulations: Generate multiple model runs with random parameter variations to explore
output uncertainties.
 Ensemble Modeling: Use multiple model structures or parameter sets to cover a range of possible
outcomes.
Understanding and acknowledging these uncertainties is crucial. Treating and incorporating these uncertainties
in model simulations can provide insights into the reliability and limitations of model outputs, enhancing the
credibility and utility of hydrological model predictions for decision-making in various water-related
applications.

The modelling process: Good Modelling Practice (GMP)

1. Research question -hypothesis
2. Conceptualization - type of information required, main processes, accuracy, time intervals
3. Model selection and development - purpose, data availability, run-time, alternatives
4. Parameter estimation (calibration) - formulation of the objective function - search method: manual,
local/global, stochastic
5. Model evaluation (testing)- independent data
6. Uncertainty assessment
7. Model application
It starts with a clear identification of the research question and also with a basic hypothesis about the course
and effect relationship. Next we derive the conceptual outline as previously explained and we should establish
what type of information is required at which accuracy and what time intervals. Based on the conceptual outline
we then select a model or several models and implement the model equations. Most models employ parameters
that can't be directly measured and therefore have to be estimated in some way or other. So the next step is
then the estimation of those parameters which is also called parameter calibration. Here again a few choices
have to be made. For example the choice of the objective function, one or several objective functions and the
choice of the search method which can be a manual calibration or some other techniques like local, global or
stochastic calibration. After the model parameters are calibrated the model has to be tested with independent
data. And depending on the outcome of the test we might have to backtrace the modeling process. If for example
the performance of the model in the test period is poor then we have to think about what kind of processes
might be missing. Do we need to select a different model and correspondingly move to step 2 or 3 and revise our
conceptual outline of the model or choose a different model which considers the missing hydrological processes.
So modeling is typically not a one-way process. It is an iterative process especially if there is new information
that comes in and we need to revise our current thinking and develop our model in an iterative way. Once the
model reproduces or predicts the quantity of interest at the required accuracy we can then move to the next
step which is an uncertainty assessment that is always required and finally to the model application.
Selecting the most suitable hydrological model involves considering various factors and making choices that
align with the study's objectives, available data, and resources. Here's a breakdown of the key considerations:

Model choice

 When selecting a the model type. There are several available like empirical models, conceptual models,
mechanistic or physical based models or stochastic models.
 Then we have to decide about the solution method whether we choose an analytical model or a
numerical model is required or something that is in between a hybrid analytical numerical model.
 Then we have to make choices about the parameterization whether it is a lump parameterization or a
distributed parameterization or a grid based parameterization.
 Also we have to make choices about the temporal representation of the model be it a steady state model
so long term averages only or an event based model, a shorter time period or a transient model. That allows us
to simulate time variability of state variables and model outputs. These choices to be made should always be
motivated by the model purpose.

Model Purpose
 There are different model outputs of interest. There is a difference between whether we want to
reproduce historic data which is called hind cast
 whether the model should be fit for predictions to be made for unknown future states.
 The purpose can also be support in water resources management
 simply process understanding and hypothesis testing
 an integration of complex data sources and data sets of different types and different time resolution.

Points to be be considered on selection of a model (can be written as limitattions encountered on choosing a

model)
1. Fit-for-Purpose Model Choice: The model chosen should align with the intended purpose. Its selection
should not be solely driven by software availability or past experience but by its suitability for the
specific task.
2. Simplicity and Complexity Balance: The selected model should be as simple as possible while being
complex enough to address the targeted scale, required accuracy, and level of detail.
3. Alignment with Scale and Accuracy Needs: Ensure that the model aligns with the scale of the problem at
hand and can deliver the required level of accuracy and detail.
4. Data Availability and Quality: Data availability is crucial; insufficient data might lead to uncertain model
outputs. In such cases, opt for simpler models or procure additional data to inform more complex ones.
5. Resource Constraints: Consider both financial and time resources. Operational models requiring real-
time action may need shorter computational times. Long run times in complex models can limit model
calibration and uncertainty analysis.
6. Trade-offs Between Optimal Models, Data, and Resources: Recognize the trade-offs between the ideal
model for the purpose, data availability, and resource constraints. Balancing these factors is crucial.
 7. Start Simple, Increase Complexity as Needed: Initiate with a simple model and incrementally increase
complexity as warranted by the task or as more data becomes available.
8. Types of Models: Understand the nuances and advantages of different model types (e.g., statistical,
machine learning, mechanistic) and their suitability in different scenarios.
9. Consider Model Run Times: Real-time operational models necessitate shorter computational times,
whereas longer run times can hinder model calibration and certain analytical techniques.
10. Continuous Evaluation and Adaptation: Regularly assess the model's performance against its intended
purpose and adjust complexity or data inputs as needed for improvement.

Classify model types by the level of mechanistic insight into the system.
Black box model
Black box models that have next to no insight into the mechanistic of the hydrological system. The empirical
models or black box models describe the relationship between the inputs and outputs with the transfer
function. And this transfer function is solely derived from observations of the natural system. That transfer
function is also sometimes called the system operator. It can be any mathematical function that relates the input
to the output and describes solely course and effect.
The advantages of empirical models are
 typically good reproduction of the system behavior
 an easy application.
 Those models are fast and easy to set up.
The disadvantages of those model types are
 that the models are only valid for the data set on which the models are trained and not valid beyond the
historical time series.
 Empirical models are not transferable to other systems and no hypothesis testing is possible
 no process understanding can be derived from those models.
Examples :
 Examples are if we have a gaging stage in the river that measures runoff and we have a downstream
gaging station where we like to know the runoff and have an incomplete time series. We can establish a
relationship between the upstream and the downstream gaging station for example by a regression
function. And use the runoff time series from the upstream gaging station as an input to that model to
predict or simulate the runoff at the downstream gaging station.
 Another example are so-called artificial neural networks which can be trained with the data from the
upstream gaging station to predict runoff at the downstream gaging station. But again there is no
physical understanding in those types of empirical models.
Grey Box or conceptual model
The next model type are conceptual models or gray box models and they use a mixture of physical concepts
together with empirical components in the model. And most hydrological models that describe the rainfall
runoff relationship fall into the category of conceptual models
The advantages are that the CPU run times are typically still relatively small and we can apply those type of
models in viable spatial resolution. So either as a lump model but also as a semi-distributed or a distributed
model.
The disadvantages are
 that the parameters of the model are only limited in terms of what they represent in the natural world
and they need to be calibrated.
 those models might also be restricted in terms of the accuracy of their predictions for unknown states
 depending on the concepts that are employed, care should be taken when applied beyond the calibration
data range.
Example: The prime example for a conceptual model are the linear storage models or bucket type models. The
hydrological system or a compartment of the hydrological system might be considered or expressed as a linear
reservoir with a storage s, an input p and an output q and then the linear storage model can be expressed by the
storage equation shown here
White box
third category of model types are physical based models or white box models, sometimes also called
mechanistic models and they are more or less built on physical laws for the description of hydrological
processes.
 The advantages are that they employ measurable parameters, so parameters that can be related to
observations, at least some of the parameters, and they are or can be spatially explicit.
 Since they are based on physical laws, they can be applied to system states that are beyond the scope of
historical times years. That means for future drivers of the hydrological system that are yet not observed.
That makes these types of models the primary choice for making model predictions,
For example for climate change impact studies. The model concepts are usually transferable to other systems,
given that the same type of data is available to inform the models.
Disadvantages.
 The first is that they are usually very data hungry, so the parameterization effort is typically very large.
 There might be sub scale effects that are not considered but play a role and have an impact on the model
outputs.
 And often they come with large CPU runtime requirements, which restricts them with respect to the
choice of model calibration and uncertainty estimation techniques.
Examples of models that
 are based on partial differential equations, like a spatially explicit finite element model to describe the
combined coupled surface water, subsurface water flow at the small catchment scale,( Imagine a computer
program that uses math to understand how water moves on the surface (like rivers, lakes) and underground in a small area,
like a small watershed or catchment area. This model pays attention to details about different places and how water flows
between the surface and below the ground.)
 detailed ground water or hyperic exchange model to describe river flow, hyperic exchange, but also
residence times and contaminant transport and turnover. Physically based models, conceptual models
and also empirical models all belong to the group of deterministic models. (This is another computer program,
but this one focuses on how water moves in rivers and the area just under the riverbed (which is called the hyporheic zone).
It's really detailed and looks at things like how fast water moves, how long it stays in certain places, and how things like
pollutants might move around in the water.)

Stochastic models
There is another category called stochastic models. Stochastic models estimate probability distributions of
model outcomes by random variation of one or more model inputs. This can involve the random variation of
drivers of the hydrological systems like precipitation, but it could also random variation of unknown
parameters. These random variations are sometimes also called Monte Carlo methods.
Advantage of stochastic models is the implicit assessment of uncertainty.
The disadvantages are that with stochastic models many model runs are required and we have to deal typically
with large CPU time requirements, especially if stochastic models are used in combination with mechanistic
models

Examples. The first is the design of a reservoir for flood protection. In this case we have a stochastic
precipitation time series model as in driving input of the reservoir model and we could derive probability
functions for the spillover of the reservoir which helps us to design the volume that needs to be hold back by the
reservoir.
Another example is the contaminant plume arrival time that uses stochastic fields for the hydrologic
conductivity in the subsurface depicted here at the bottom left. Together with stochastic representations of the
reactivity in the subsurface here in the middle and the fitting flow network on the right, we can derive the
probability of the plume arrival by random variation of those fields.
Conceptual model :
  most frequently used in hydrology
 they are applied to describe the rainfall runoff relationship for operational flood forecasting,
 for water balance calculations
 using climate change impact studies
 the estimation of design values for reservoirs and similar applications

Physically based models or mechanistic models

 are applied more in applications where we deal with surface water-ground water interactions with the
soil-plant atmosphere continuum in applications where contaminant transport is of interest
 salt water intrusion studies
 Well capture zone delineation
 studies that relate to hydrological change
 as well as studies that are directed towards process identification and process understanding.

(Parameters are values or factors in a model or system that determine or influence its behavior, output, or characteristics. Parameters
can represent various things such as physical properties, settings, conditions, or characteristics of a system or model.)

Parameter Types
Empirical Parameters: These parameters are based on observations or data but lack a direct connection to physical
principles or properties. They're determined from observations without a clear physical interpretation.

Physical Parameters: These parameters have a direct connection to physical properties or characteristics of the system.
They can be measured or observed in the real world and have specific physical meanings.
 Their reliability or transferability depends on how they're measured. Direct measurements usually provide more
reliable values compared to indirect measurements. Errors in measurement methods can affect their reliability.

 They might change based on the scale at which they're observed or measured. For example, if we take pump test
results of hydraulic conductivity and use them in a model that has a grid size of one by one kilometers, for
example, then we have a disparity between the scale of the measurement of the pump test and the scale at which
we employ this power meter, a larger scale in this case. And then again, we might resolve to determine those
parameters indirectly as effective parameters through model calibration, because the point scale measurement is
not representative of the scale at which the model is operating.
 We also have to think about the validity of the postulated model equations and whether or not the measurements
fit to those equations. There are, for example, a wide range of different models available to compute the actual
evapotranspiration from the potential evapotranspiration, and the choice has an impact on the model outputs.
And particularly, if we use inadequate model equations for the forcing of the hydrologic model, then we might get
biased results. If we use the correct measurement in an inadequate equation, and similarly, if we use an incorrect
measurement in a correct model equation, we also get wrong results

Hyperparameters: Hyperparameters are not the typical parameters seen within a model. Instead, they often govern the
behavior or settings of the model itself. They can take various forms like real numbers, integers, or boolean (true/false)
values, and they might not directly represent physical properties.
Model calibration.
There are plausible ranges of parameters and model calibration is typically associated with estimating those parameter
values in a way that remodel fits to a historic time series in the best way possible. Models are also sometimes classified
into lumped and distributed models

Parameterization methods of models.

Distributed models These models employ spatially explicit parameters, assigning values to different locations within the
model domain. In grid-based models, each grid cell possesses specific parameter values (e.g., hydraulic conductivity),
reflecting spatial variability. It can also be based on hydrological zones or subcatchments with varying [Link] a
distributed model, each grid node or spatial division contains specific parameter values (e.g., soil properties like porosity
and soil thickness). This method accounts for spatial heterogeneity, reflecting diverse properties across the landscape.
Lumped models with the linear reservoir employ effective parameters. Such as the groundwater store for of a hole
catchment is conceptualized by a single store or a sequence of a path and a flow ground water store for. Similarly, soil
moisture in the whole catchment can also be expressed as a lumped linear reservoir.

Lumped models offer simplicity but may overlook spatial variations, while distributed models capture spatial variability
but can be more complex to calibrate due to the increased number of parameters. The selection between these approaches
relies on the specific objectives of the modeling study and the available data and computational resources.
Let's further assume that the true inputs to the system are known, like the spatial-distributed precipitation, temperature,
and so forth. There is a true response of the hydrological system. Let's focus in this example on the runoff from a
catchment. We typically don't know the true response, the true runoff, but we make observations at different points in
time, so we have an observed response of the hydrological system. (1)
The true input to the system is also not known, but we do observations in space and time, for example, precipitation at
different rain gauges, and then extrapolate this as an input of the system. Then we have chosen a hydrological model that
represents the dominant processes in the hydrological system, and we use our observed input to force that hydrological
model. That model employs a set of parameters, and you might have some initial idea about the values for those
parameters, which we put in here as prior information. With those parameters and the observed inputs, we get a model
response, a simulation of the model, and we then can plot again the observed response in the graph to the right. Then we
can compare the difference between the observed and the simulated response, which is typically done by some form of a
distance measure, and we can calculate from that the model fit to the observations. Model calibration, then, is the process
of tuning the parameters of the model with the goal to minimize the residuals between the simulated and observed
response of the hydrological system. In other words, model calibration is the search in the parameter space for the best
possible combination of parameters with the goal to minimize the residuals. We can do this search either manually by trial
and error, or we can employ search algorithms that help us finding the best parameter combination.

Model calibration process in a more abstract way as an optimization problem.

In very general terms, an optimization problem can be described by finding a value of x for which a mathematical function
f of x takes on the smallest or largest possible value. Corresponding to model calibration, that means finding the best
parameter set that leads to the best performance of the model with respect to the data. And the performance of the model
is described by a error function or a fitness criteria, or sometimes it's also called the objective function or model misfit.
And we have to make a choice between one or several of those objective functions.

The parameter space

The parameter space is the set of all possible combinations of values for all parameters of the model that are considered to
be uncertain and need to be calibrated. Typically, the parameter space is constrained. So we have some initial ideas about
the parameter ranges that individual parameters can obtain.
1. In linear regression, the model equation f(x)=ax+b involves two parameters: 'a' and 'b', where 'a' represents the
slope and 'b' is the intercept of the line.
2. To find the best values for 'a' and 'b' that make the model match observed data, a parameter space is explored.
This space includes all possible combinations of 'a' and 'b' within defined ranges.
3. For example, let's say 'a' is set to range between 0 and 3, and 'b' ranges from 0 to 2. This creates a two-dimensional
parameter space, illustrated as a rectangle with 'a' on the x-axis and 'b' on the y-axis.
4. Searching through this parameter space involves testing different combinations of 'a' and 'b' to see which
combination fits the data points the best. However, as the number of parameters increases, the complexity of this
search grows exponentially.
 5. Model calibration aims to find the most suitable 'a' and 'b' values that make the model fit observed data closely.
This search in the parameter space is facilitated by an objective function, measuring how well the model matches
the observed data.
6. The objective function helps project the parameter space into what's known as the objective space, allowing
evaluation of the model's fit to observations based on different criteria. The goal is to find the combination of
parameters that best represent the real-world data.
Objective Function
In hydrology, there are several fitness criteria used to express the model fit or the difference between the observed and
simulated values. One of the most traditional ones are the Nash-Sutcliffe efficiency., the sum squared error, the root mean
square error, but also others like the mean error, the maximum absolute error, the bias in time to peak, the coefficient of
variation. And more recently, the Kling-gupta efficiency, which is an unbiased estimate or an unbiased variant of the Nash-
Sutcliffe efficiency.
 Let's take the Nash-Sutcliff efficiency here as the fitness criteria. We see on the left the discharge over time, so the
hydrograph. In blue is the observed data. In red is a curve that has been generated after fitting the model with the
Nash-Sutcliffe efficiency.
 The Nash-Sutcliffe efficiency is the ratio of these graph residuals divided by the sum of the difference from the
mean. The Nash-Sutcliffe efficiency has a range between 1 and minus infinity with 1 expressing the perfect fit and
smaller values expressing a decreasing fit.
 It can also be expressed as 1 minus the mean squared error divided by the variance of the observation.
 In terms of model performance of our calibrated model, we see that the peaks are reasonably well fitted here and
here, but during the periods of low flow, the model shows some deviation from the observed values.
 Now, if we use in the model calibration a Nash-Sutcliffe efficient objective function with the log-transformed flows
instead of the non-transformed flows.
 Here we see a much better fit to the lower observed values, particular in this period, but the model doesn't fit to
the peaks very well.
 So you can see even with the same search methods or calibration methods and the use of the different objected
functions you can get quite different results and the choice of the fitness criteria should be closely aligned with the
model purpose.
 If you want to do flat forecasting and we are interested in the maximum peak then we might actually use the Nash-
Sutcliffe efficiency with the untransformed Q values as shown in the upper graph, but if we are interested more in
the prediction of drought and the maintenance of ecological flows in a river for example, where the focus of the
investigations are more on the low flows then it would be more suitable to log transformed Nash-Sutcliffe
efficiency or any other fitness criteria that fits to lower flow observations.
(Slide 4 pg 28)
The objective space in model calibration.
 The objective space sometimes also called problem space or search space is the region defining the set of all
possible model outcomes.
 We have the parameter space on the left here with our two parameters a and b and possible combinations in the
space. Then we have the model that is defined by this function the linear regression function and we also have
chosen an objective function or fitness criteria in this case that this be the sum squared error SSE.
 The objective space is evaluated by computing all possible combinations of the parameters a and b in this case and
evaluating the sum squared error for those combinations.
 This adds a third dimension to the parameter space a and b and for each of the different combinations we get a
value which is in this plot to the right here expressed as contour lines of the objective function value sum squared
error.
 The red colors are high values and the blue colors are lower values. The best model calibration can be found by
finding the minimum value in this surface and for that particular case the minimum is found here in the middle of
this graph.
 In this particular example we might be able to find the optimum of the parameter combinations by trial and error
since the objective space is rather smooth so we are going downhill from all corners of the parameter domain and
there is a clear identified optimum in the middle.
 However there might be also parameter combinations that are a little bit too developed in the right that are at
least nearly as accurate or as good as the optimum parameter indicated here in the middle. A complicating factor
in model calibration is that we typically don't know the shape of the objective space.
 Optimum ... the best solution(s) in the objective space.
(Slide 4 page 29)
Optimum
(Slide 4 page 30)
1. Complexity of Objective Spaces: Objective spaces, representing all possible model outcomes, can have intricate
shapes and structures, impacting calibration results. They might contain global maximums, local minimums, hills,
flat regions, and more complex forms, making it challenging to find the best fit.
2. Search Methods Impact Calibration: The choice of search method is crucial. Gradient-based methods look for
decreasing directions in the objective function. They may get stuck in local minimums if not started correctly.
Stochastic approaches that explore multiple initial parameter values, evolving them over iterations, might reach
global minimums more effectively.
3. Dependency on Initial Parameters: Certain search methods might heavily rely on the starting point or initial
parameter values. Depending on where the search begins, it could lead to different results, especially in complex
objective spaces.
4. Unknown Objective Space Shape: Often, the shape of the objective space is unknown before calibration. It could
exhibit various complexities, discontinuities, or undefined areas, making it challenging to predict the best strategy
for finding the global minimum.
5. Regions of Similar Model Performance, Different Parameters: Objective spaces might contain regions where model
performance is similar, yet parameter values differ significantly, adding to the challenge of identifying the best fit.
6. Model Dependency and Potential Discontinuities: Depending on the model used, the objective space might have
discontinuities or undefined values, complicating the calibration process.
7. Assumed Complexity and Pitfalls: Due to the unpredictability of objective space shapes, it's essential to anticipate
complexity and potential pitfalls during the calibration process

Model complexity
Another important aspect with model calibration is the relation between available data and model complexity.
1. Overfitting with Complex Models: Overfitting occurs when a complex model, such as a high-degree polynomial,
perfectly fits available data points but fails to generalize well to additional or unseen data points. It appears to
perform well with existing data but poorly with new observations, indicating that the model may capture noise
rather than the underlying patterns.
2. Indicator of Overfitting: Overfitting can be identified through split-sample analysis, where data is divided into
sections for calibration and validation. If the model performs significantly worse on the validation dataset
compared to the calibration dataset, it's an indication of overfitting. This happens when the model has too many
parameters or degrees of freedom relative to the available information in the data.
3. Oversimplification Pitfall: Conversely, an oversimplified model (e.g., a simple linear or quadratic function) may fail
to capture the true complexity of the system. While it might perform adequately with limited data, it performs
poorly when additional or more diverse data points are considered.
4. Need for Validation with Independent Data: To avoid both overfitting and oversimplification, it's essential to
conduct split-sample calibration, where the model is trained on one portion of the data and evaluated on another.
This helps in assessing the model's performance with independent datasets and ensures it captures the underlying
patterns rather than just the noise or limited aspects of the system's behavior.
5. Balance between Model Complexity and Available Data: Understanding the balance between model complexity
and available data is crucial. A model should be complex enough to capture essential patterns but not excessively
complex to fit noise or lack generalization capabilities when faced with new data.
6. Importance of Generalizability: The aim is not just to fit the available data perfectly but to create a model that
generalizes well to new, unseen data. A good model captures the essential relationships within the data without
being overly complex or oversimplified.

1. Manual vs. Automatic Calibration: Manual calibration, often by trial and error, suits simple optimization problems
with very few parameters. However, for more complex scenarios with numerous parameters, automatic
calibration is preferred, employing search algorithms to find optimal parameter combinations.
2. Optimization Algorithm Selection: The choice of optimization method relies on various factors: problem
characteristics, computational requirements of the hydrological model, and pragmatic constraints. The method
selection is crucial and depends on the optimization problem's nature.
3. Categories of Optimization Algorithms:
 Deterministic Optimization: Categorized by continuous or discrete problems.
 Continuous Deterministic Optimization: Could be constrained or unconstrained.
 Unconstrained, Nonlinear Least Squares: Includes gradient-based methods like Gauss-Newton,
Levenberg-Marquardt, and steepest descent. These methods are computationally efficient but
sensitive to initial parameter values and might struggle in complex objective spaces.
 Global Optimization Techniques: Suitable for more complex objective spaces, as they avoid getting
stuck in local optima.
4. Multi-objective Optimization: Helps in weighing tradeoffs between different objective functions. It aids in finding
Pareto optimal solutions, allowing modelers to select compromises between various fitness criteria. It's not only
useful for balancing fitness criteria but can also detect structural errors in models.
 5. Robustness and Efficiency Tradeoff: There's a tradeoff between the robustness (ability to find the global optimum)
and efficiency (speed to converge to near-optimal solutions) of an optimization method, referred to as the "no-free
lunch theory." Choosing an algorithm often involves balancing between these factors in practical applications.
6. Optimization Method Selection Criteria: For practical applications, modelers seek algorithms that are reasonably
robust, capable of finding near-optimal solutions efficiently, and can avoid getting trapped in local optima.
7. Efficiency and Robustness Consideration: Especially in complex calibration problems, choosing an optimization
method involves considering its robustness in finding the optimal solution and its efficiency in finding that
solution within a reasonable computational time.

Model caliberation important points

 Firstly the model structure should fit to the data available for model calibration. It is better to start with a
simple model to avoid overfitting and then add complexity as necessary to explain the data.
 Hydrological models are typically higher dimensional which means they employ a relatively large set of
parameters and therefore the optimization problem or the parameter calibration problem is also higher
dimensional. Secondly it is important to choose the objective function or the fitness criteria in a way that it fits
to the model purpose.
 And lastly here should be taken by the choice of the calibration algorithm. The objective space is often complex
and non-continuous so if possible global optimization algorithms should be employable parameter calibration.
Chapter 5
precipitation also comes in different forms, like fog, snow and sleet, hail, but also less common forms like freezing rain.
quantify precipitation by volume, duration and intensity, but also the return period of particular precipitation events.

Formation of precipitation
1. Moist Air and Cooling: Precipitation formation starts with the presence of moist air containing a significant
amount of water vapor. Cooling of this air is essential to reach the dew point temperature, which can happen due
to heat loss through radiation, conduction, mixing with cooler air, or rapid adiabatic cooling.
2. Cooling Rates and Uplift: High cooling rates, critical for precipitation formation, are often associated with rapid
vertical uplift of air. Faster cooling rates generally result from swift vertical motion.
3. Condensation and Droplet Formation: Once the air reaches the dew point temperature, condensation occurs,
leading to the formation of small liquid water droplets. This process requires the presence of small nuclei such as
aerosols or dust. The initial droplet size ranges from 0.01 to 0.2 millimeters in diameter, with velocities under 0.7
meters per second.
4. Droplet Growth: As these droplets move and collide in the air, they grow in size to 0.4 to 4 millimeters. The larger
droplets can attain velocities greater than 1 meter per second due to their increased size.
5. Continuous Growth and Vapor Input: Further growth of droplets requires a continuous supply of water vapor from
the surrounding air, enabling them to reach precipitation size.
Uplift Scenarios: There are three main meteorological situations conducive to rapid uplift of air, facilitating the necessary
high cooling rates for precipitation:
 Convection: Vertical motion driven by heating of the Earth's surface, causing air to rise rapidly.
 Orographic Uplift: Air forced upwards when encountering elevated terrain like mountains, leading to
cooling and precipitation formation on the windward side.
 Convergence: The meeting and lifting of air masses with different characteristics, often leading to upward
movement and subsequent cooling conducive to precipitation.
Convection
1. Convection occurs quite often in summer when the surface of the earth is heated up and warm, moist air is uplifted
up to the condensation level. Whereby the cooling of the air is adiabatic.
2. It is typically short-lived, localized, and exhibits variable intensities, ranging from light showers to intense
thunderstorms with heavy downpours and sometimes hail.
3. Starts with the formation of droplets and smaller raindrops, initially moving upwards in the air mass. As they grow
larger, they begin a downward motion, cycling within the cloud until reaching a critical size. The resulting heavy
rain leads to cloud dissipation.
 4. Convection can lead to extreme events, as seen in a supercell storm in northern Italy in November 2012. This
storm caused an "overshooting top," reaching beyond the troposphere into the tropopause, with precipitation
intensities reaching up to 150 mm/day, significantly high for short time spans, like 52 mm/hour or 30 mm in five
minutes.
5. Convective precipitation poses challenges in prediction and can result in extreme volumes and intensities, leading
to flash floods, mudslides, and landslides. The intensities may increase in Europe due to climate change, with rising
temperatures adding energy to the atmosphere, potentially heightening convective precipitation events.
Convective precipitation is significant in hydrology due to its unpredictable nature, potential for extreme impacts, and the
likelihood of intensification with climate change, highlighting its importance in climate and hydrological studies.
Orographic rainfall
1. Moisture content and wind speed: Higher moisture content in the air masses, along with varying wind speeds,
influences the amount of moisture that can be carried and subsequently condensed upon ascent over the barrier.
2. Elevation and steepness of the barrier: Higher elevations and steeper slopes tend to enhance the upward
movement of air, leading to more pronounced cooling and increased precipitation.
3. Geographical location: Mountain ranges close to the coast often experience higher precipitation due to the
interaction of moist oceanic air with the elevated terrain. This results in a correlation between average annual
precipitation volumes and elevation.
4. Leeward (downwind) side effects: On the leeward side, known as the rain shadow, descending air creates drier
conditions, often leading to reduced precipitation and sometimes creating arid regions or "rain shadows."
In the German Alps, the phenomenon you mentioned, known as "foehn" or dry conditions on the leeward side, is a classic
example of the rain shadow effect where the descending air on the far side of the mountain range leads to drier conditions
and lower precipitation volumes.
This process creates a correlation between topographic elevation and mean annual precipitation volumes, where higher
elevations often receive more precipitation due to orographic uplift, while the leeward side experiences reduced rainfall
due to the rain shadow effect.

Convergence
Convergence occurs when air masses with different temperatures, moisture levels, or directions come together, leading to
upward movement of air and potential precipitation. Two primary types of fronts associated with convergence rainfall are
warm fronts and cold fronts.
1. Warm fronts: In a warm front scenario, warm air slides over a mass of cold air, gradually rising above it. As the
warm, moist air ascends, it cools, leading to the formation of precipitation. Typically, the precipitation intensities
along warm fronts are relatively uniform and moderate in nature.
2. Cold fronts: Conversely, in a cold front situation, dense cold air moves underneath a warmer, moist air mass. The
steeper angle of the cold front, especially near the front, leads to a faster uplift of the warm air. Consequently, the
precipitation intensities along cold fronts are typically higher and more intense than those along warm fronts.
Aside from frontal systems, convergence can also occur when two air masses converge from different directions, forcing
each other upward and leading to rising air. The Intertropical Convergence Zone (ITCZ) at the equator is a significant
example. Here, warm, moist air masses from both the northern and southern hemispheres converge near the equator due
to the trade winds. As these air masses converge, they rise due to the convergence, leading to low pressure at the surface.
The rising air cools, expands, and results in precipitation. This process contributes to a continuous cycle where air flows
towards the lower pressure system, creating a loop.
This convergence and uplift of air masses in various scenarios contribute significantly to the formation of weather
patterns, distribution of precipitation, and the establishment of pressure systems, thereby impacting regional climates and
weather conditions.(slide 5 pg 11,)

Forms of precipitation – slid5 page 12

Next
Köppen-Geiger climate classification

1. Tropical (A): These climates are typically found near the equator and are characterized by high temperatures
year-round. They often have distinct wet and dry seasons.
2. Dry (B): Dry climates are divided into desert (BW) and steppe (BS) subtypes. Deserts are extremely dry with very
low precipitation, while steppe climates have more moderate dryness.
3. Temperate (C): These climates are found in the mid-latitudes and generally have moderate temperatures. They
include various subtypes based on the distribution of precipitation throughout the year.
4. Continental (D): These climates experience distinct seasons, with hot summers and cold winters. The subtypes
vary based on the amount and distribution of precipitation.
 5. Polar (E): Polar climates are characterized by extremely cold temperatures and generally low precipitation. They
include tundra (ET) and ice cap (EF) subtypes.

Precipitation regimes - slid5 page 16

measurement of precipitation using rain gauges provides several crucial characteristics:

1. Rainfall Depth: This refers to the amount of precipitation observed over a specified time frame or event. It is
typically measured in units like millimeters (mm) and is crucial for understanding the total volume of water
received during a specific period, be it daily, monthly, or annually. Rainfall depth is essential for water resource
analysis and understanding overall water availability.
2. Rainfall Intensity: This parameter measures the rate at which precipitation falls, expressed as a volume per unit of
time. It could be measured in millimeters per minute, per hour, or per day. Understanding rainfall intensity,
especially in correlation with rainfall depth, is vital for assessing flood potential, predicting inundation, evaluating
hill slope stability, and other aspects related to water-related hazards.
3. Rainfall Duration: This denotes the length of time over which a precipitation event occurs. It is a valuable measure
when combined with rainfall depth and intensity, providing insights into the duration of the rainfall event. This
information is useful for various analyses and predictions related to hydrology, including flood modeling and
stormwater management.
4. Yield Factor: This factor, expressed in units of volume per time and area, is computed to estimate resulting
stormwater volumes. It assists in determining the amount of water that will run off a specific area over a given
time, aiding in stormwater management, drainage system design, and urban planning related to water runoff.

To install a rain gauge, we need to consider different factors.

First factor : Resolution
 The frequency at which precipitation data is measured (e.g., hourly, daily, weekly) significantly affects the
information captured. For long-term analyses, such as studying climatic patterns, daily, weekly, or monthly
measurements might suffice. However, for operational tasks like flood forecasting, higher temporal resolutions
(e.g., every 10 minutes) are crucial.
 When looking at a distribution of precipitation over a day, if the measurement intervals are fine (e.g., hourly), the
data representation is close to the actual precipitation events observed. However, as the temporal resolution
becomes coarser (e.g., 3 hours, 6 hours, 12 hours), the original signal of precipitation gets diluted due to averaging
over larger time scales. This dilution leads to a loss of information, especially regarding the intensity of
precipitation events.
And depending on the purpose of the hydrological analysis, this can lead to biased results.

 Knowledge of precipitation intensity is particularly vital for hydrological analyses, especially in small catchments
with rapid responses. Precipitation intensity directly influences soil infiltration capacity and initial soil moisture,
determining the amount of precipitation that becomes immediate runoff. Inaccurate intensity data due to low
temporal resolution can lead to biased results, affecting flood forecasts and overall water management strategies.
The second factor for the installation of rain gauges is the selection of the type of the rain gauge
We distinguish between storage gauges, conventional rain gauges, and unconventional rain gauges.
Factors on which the selection depends –
 access to the site
 climate is an important factor.
 We have to consider rainfall intensities and expected volumes,
 the available budget place

1)Non-recording storage gauges

These devices are probably the simplest rain gauges, which comprise a funnel, as seen on the picture in the top right here,
and a collecting vessel underneath. Such devices have to be emptied manually, so from time to time somebody goes by and
takes the collecting vessel and measures what's in there and takes a record of that volume. With those storage gauges we
collect typically averages over long-term periods, and they are deployed in regions that are less accessible, like mountains,
where we just are interested in getting an idea about the total volumes. The downside of those storage gauges is that they
measure long-term averages, so we don't get any information about extreme values and about intensities of precipitation.
But they are extremely cheap alternatives, a
2)The conventional recording gauges. - Float recorders , Tipping-bucket gauges ,Weighing gauges
Helman Rain Gauge, which is a float recorder.
Depicted here is a float recorder that operates autonomously and can be
boarded in this setup still today, and it consists of the panel on top with a
receiving area of 200 square centimeters. The precipitation is directed in a
float vessel. Depicted here, in the float vessel there is a float attached to a pen
arm, and the pen arm right on a rotating drum depicted here in orange, or on
the right here. And the pen records the level of the float in the float vessel on
this piece of paper here. Once the float vessel is full, the float vessel is emptied
by a siphon into a bucket below. The drum rotates by a mechanical spring, so
no external power is required to operate the gauge, and the drum rotation
could be once a day. Then the paper on the drum has to be replaced daily, or
once per week, which lowers the effort to obtain the data. Although this
mechanical gauge is quite ingenious and robust, the downside is that the data
is collected on a piece of paper and needs to be digitalized. Therefore, most of
those gauges are replaced by automatic gauges, but there are still some of
those gauges around, and especially if you look at historic data sometimes it is useful to at least know how these gauges
are working and how to read the measurements from a float gauge recorder.

Tipping bucket gauge.(slide 5 pg 24)

 Precipitation measurement at increments of 0.2 millimeters based on bucket size.
 Allows measurement of various intensities with good accuracy.
 Maintenance effort for the rain gauge is relatively low.

Weighing gauges
They don't have any moving parts and instead measure the precipitation just by the weight. This allows to measure the
volume and intensity of rain at a high temporal resolution but it also allows to measure snow and other forms of
precipitation given that there is a heating ring at the collection vessel that prevents the build up of snow that is quite
common in environments where it is snowing in winter. If the device is equipped with a solar panel for power supply and
remote data access those devices can be autonomios and are more or less maintenance free. but all those advantages come
with a significantly higher cost of the device. These devices are installed in areas that are less accessible but still where
there is cell phone coverage to transport the data into the office.

Unconventional recording gauges

First there are capacitance rain sensors that more or less measure whether or not there is precipitation but no intensity
and no volumes. These are sometimes applied in agronomy.
And secondly there are distrometers that measure raindrops or raindrop size distributions with different methods.
 Acoustic distrometers or hyperphones record the sound of raindrops on the surface for example on an open water
surface in the ocean or in a lake and try to establish a relationship between the sounds and the raindrop size and
then basically counting the raindrops from the recordings and from that derive rainfall volumes.
 Another example are displacement distrometers which are schematically depicted in the bottom here. When a
raindrop hits the top of the device the top is pushed down and as there is a magnet at the top that runs through a
coil there is magnetic induction which is then measured as an electric impulse and the frequency and magnitude of
the displacement again is then related to raindrop size and frequency.

Quite often it is difficult to establish a unique and accurate relationship between the actual measurement and raindrop
size or precipitation volumes and therefore the precipitation measurements with distrometers are typically not very
accurate and for the displacement distrometers they often fail especially during heavy rainfall.
Optical distrometer
Another type of distrometer is the optical distrometer. It is based on a laser or infrared sensor array with emitter and a
receiver and essentially if raindrops pass through the laser beam they are counted. This measurement principle allows for
the measurement of true rates and all precipitation types. Laser distrometers can result precipitation droplets down to the
size of 0.001 millimeters and the accuracy for accumulated rainfall is in the range of 5% which is pretty good for
precipitation gauge. Other advantages are that again it is maintenance free, it has no moving parts but it requires a
constant power supply and it comes also with a much higher cost.
Systematic errors:
 related to site and
instrument type,
 evaporation,
 splashing, wetting
 wind, snow
Random errors:
 error in reading

Chapter 5
Analysis and correction of precipitation data
 The correction of point scale precipitation measurement is an important step and needs to be conducted before
spatial interpolation of this point scale measurement.
 This schematic shows the flowchart of the analysis and correction of precipitation data. It starts at the top with the
raw data of precipitation that comes from a monitoring network with its specific site characteristics and climate
conditions, which dictates how the precipitation is measured.
 Let's consider here that the precipitation is measured by a conventional rain gauge and we get a raw data set by a
download from the device or it might be telemetered back into the office.
 Now the first step is a time series analysis and we test the time series of precipitation for first consistency and
second homogeneity. If the time series is not passing this test, meaning it's not consistent or not homogeneous,
then we have to deal with that problem
 The next step is the correction of precipitation time series for various errors and
 the third step is to compute the catchment average precipitation from the records of several rain gauges that might
be in the area of the catchment
 average precipitation which all have been corrected for precipitation and before they're tested for consistency
and homogeneity.

Time series analysis which includes the consistency and homogeneity test.
The consistency test includes the identification and removal of non-plausible values or in other words of outliers, also to
fill gaps in the data if they should be present.
 Examine the time series of precipitation data spanning 60 days.
 Identify anomalies in the data, such as negative precipitation values, which are physically implausible.
 Additionally, spot unusually high values compared to others in the time series.
 To handle negative values, remove them from the time series as they do not make physical sense.
 For the unusually high values, consider them as potential outliers. A simple test involves checking neighboring
stations to see if they also recorded similarly high precipitation volumes. Alternatively, statistical tests can be
applied to assess their outlier status.
 Assume that the identified negative value and the unusually high value are outliers, requiring removal from the
time series.
 The corrected time series now reflects physically plausible precipitation values without outliers.
 Observe any gaps in the time series, highlighted in yellow on the graph.
 Fill these data gaps using information from neighboring stations if the time series has daily or smaller time steps.
 Alternatively, if the time steps are larger (monthly or yearly), gaps can be filled with long-term averages.
 After removing outliers and filling data gaps, the time series is considered consistent, providing a more accurate
representation of precipitation over the 60-day period.
The check for consistency comprises the removal of outliers from the time series and the supplementation of data gaps.
1)Outliers
Outlier detection methods should be chosen based on the characteristics of the data and the specific context.
1. Common Causes for Outliers:
 Mistakes made during the data entry or processing phase, especially when downloading and entering the
data into a database.
 Outliers may occur due to faults or malfunctions in the measuring instrument itself.
 In some cases, data analysts intentionally introduce outliers. For example, when the instrument is not
measuring for a certain period, default values (e.g., minus 999) might be inserted into the time series.
 Outliers may not always be errors. They can also represent extreme values or values with very low return
periods, indicating rare but valid occurrences.
2. Methods for Detecting Outliers:

 Extreme Value Analysis (EVA) -Z Score: This method involves calculating the Z score, which measures how
many standard deviations a data point is from the mean. Values with Z scores beyond a certain threshold
are considered outliers. The Z-score for a data point xi in a dataset is calculated using the formula: Zi=xi−μ
/σ, where μ is the mean and σ is the standard deviation of the dataset.
 Probabilistic or Statistical Modeling: These models capture the expected behavior of the data, and
deviations from the model can be indicative of outliers. Various probabilistic models, such as Gaussian
distributions or mixture models, can be employed for outlier detection.
 Linear Regression Methods: Techniques like linear regression and principal component analysis can be
employed to identify outliers by analyzing the relationships and patterns within the time series.
 Clustering : Outliers can be detected by examining data points that do not belong to any cluster or are in
sparsely populated clusters. Deviations from the typical groupings can signify outliers.
  Distance or Density-Based Outlier Detection: Distance-Based Methods: These methods involve measuring
the distance of data points from each other. Outliers are often characterized by being isolated from the
majority of data points.
Density-Based Methods: These methods focus on the local density of data points. Outliers are points that
have significantly lower local density than their neighbors.

Analysis and correction of precipitation data

Data gaps: methods for data supplementation

The specific method chosen depends on the characteristics of the data and the surrounding stations.
Each method has its own assumptions and considerations.

Situation 1: Only One Station Available

1. Limited Nearby Stations:
 In this scenario, there are no other precipitation stations nearby.
 Gaps can only be filled for longer time intervals, such as annual or monthly means.
 Compute means for the months or years without data gaps.
 Utilize these means to fill in the missing values for the respective time intervals.

Situation 2: Multiple Stations Available

1. Availability of Multiple Stations:
 Multiple precipitation stations, including the one where gaps need to be filled, are available.
2. Various Gap Filling Methods:
 Average from Surrounding Stations:
 Compute the average from the records of surrounding stations.
 Regression:
 Use regression techniques to estimate values based on the relationship with data from nearby
stations.
 Normal Ratio Method:
 Weigh or scale values from surrounding stations by an annual average precipitation ratio.
 Inverse Distance Rating:
 Fill gaps by considering the inverse distance to surrounding stations.
 Grating and Interpolation:
 Utilize other interpolation methods using surrounding stations.
Homogeneity

Homogeneous precipitation time series means the variations in that time series are only caused by weather or
climate.
Systematic changes in the time series that are not caused by weather and climate.
(These changes need correctin)
 If an obstacle has been built or removed in the vicinity of a precipitation gauge .It must be ensured that
any obstacles are far enough away from the precipitation gauge.
 There can be vegetations such as trees are growing in the vicinity of the gauge which will have an
influence on the precipitation measurement.
 It can be caused by changes in the model and make of precipitation gauges or if the gauge has been
moved to another location.
Therefore it is important to keep a record of this station history and if something unusual is being found in the
time series one could check in the metadata for the time series whether some of those changes have been
occurring.
Changes that are caused by weather and climate and those changes don't need to be corrected. In fact they must
not be corrected.
 There are local or regional changes that can also have anthropogenic courses such as the establishment
or expansion of irrigation areas, the construction of a hydroelectric power plant, the storage dam or
other major land use changes. These changes can lead to shifts in local climate for example through
enhanced evapotranspiration from open water surfaces or from irrigated areas which in turn can lead to
differences in the precipitation volumes.
 These changes can also have natural courses like natural disasters or slower changes in the hydrological
regime for example as caused by climate change.
 Homogeneity test
Homogeneity tests categorize into absolute and relative tests.
 Absolute tests focus on a single time series for homogeneity assessment.
 Examples of absolute tests include Petit's breakpoint analysis and the Mann-Whitney U-test.
 Breakpoint analysis involves dividing the time series into parts and computing trends for each.
 In the given example Time series split around a point, revealing a decreasing trend in the first part and
an increasing trend in the second part.
 Observed shifts, represented by linear trends, may indicate systematic changes.
 Validation through significant tests determines if trends are statistically significant

Relative Tests:
 Group of homogeneity tests distinct from absolute tests.
 Key Feature:
 Comparison of multiple time series, contrasting with absolute tests that focus on one.
 Utilization of a reference time series, assumed to be homogeneous.
 Examples of Relative Tests:
 Craddock Test, Abbe – Test.
 Determine sign and magnitude of deviations from the average for both tested and reference time series.
 Graphical Representation Example:
 Example Scenario:
 Station 1 (Reference):
 Fluctuates around the mean, indicating homogeneity.
 Station 2:
 Consistent, systematic deviation exceeding 2000 millimeters in 40 years.
 Precipitation anomaly returns to zero after a breakpoint in 1930 over a similar 40-
year span.
 Prolonged and extreme deviations from the mean suggest inhomogeneity.
 Breakpoint identified at the maximum deviation.
 Additional Relative Tests:
 Double Mass Curve and Autocorrelation Method.
 Purpose:
 Offer alternative approaches for assessing homogeneity.
 Examples of methods beyond Craddock and Abbe tests.
Double mass Curve

 Catchment and Precipitation Stations:

 Consider a catchment with four precipitation stations (A, B, C, D) shown in the green patch.
 Compute mean annual precipitation sums for each station.
 Test the homogeneity of the precipitation time series at station B.
 Procedure:
1. Compute Mean Cumulative Precipitation:
 For stations A, C, and D, calculate the mean annual precipitation for each year and sum it
up consecutively, creating a cumulative precipitation function.
 Do the same for station B, creating another cumulative precipitation function.
2. Plot Cumulative Precipitation Functions:
 On the x-axis, plot the mean cumulative precipitation from stations A, C, and D.
 On the y-axis, plot the cumulative precipitation computed only from station B.
3. Analysis of the Plot:
 In this case, the plot forms a linear function that aligns with the 1-1 line, suggesting
homogeneity.
 A linear function is indicative of a consistent relationship between the cumulative
precipitation of station B and the mean cumulative precipitation from stations A, C, and D.
4. Interpretation:
 A linear function implies that the precipitation time series at station B is likely
homogeneous.
 If all four stations showed a simultaneous systematic shift, it might not be detected
through this method.
  Different Scenario with Station E:
 Consider testing the homogeneity of the precipitation time series at station E, not previously
shown in the graph.
 Procedure for Station E:
1. Compute Double Mass Curve:
 Similar to the previous process, calculate cumulative mean precipitation from stations A to
D and station E separately.
2. Plot the Double Mass Curve:
 Plot the cumulative mean precipitation from A to D on the x-axis and station E's cumulative
precipitation on the y-axis.
3. Analysis:
 Observe a change in slope around the year 1981.
 A flatter slope before 1981 and a steeper slope after indicate a systematic change in the
precipitation record in that year.
4. Possible Causes:
 The change could be due to alterations in the precipitation measurement device, relocation
of the station, or other factors influencing the data.
 Conclusion:
 A change in slope in the double mass curve suggests a systematic alteration in the precipitation
record, necessitating further investigation and potential correction
When is the change in slope significant?

1. Time Period for New Trend:

 A new trend needs to persist for at least 5 years to be considered for evaluation.
2. Statistical Significance of the Trend:
 Assess whether the trend or change in trend is statistically significant.
 Apply tests of variance and covariance to determine statistical significance.
3. Consideration of Data DIFGArd:
 If applicable, consider DIFGArding the data.
 Reasons for DIFGArding may include an inability to identify the actual causes of changes.
 DIFGArding data is a practical but crude correction method, still preferable to using unreliable
data.
 In cases where the time series is long enough, cutting a few years may be affordable if the trend
occurred in the beginning.
Correcting Data in Case of Significant Changes:
 Significant Change Criteria:
 The change in slope is deemed significant.
 Correction is necessary if the changes are associated with alterations in measurement locations or
conditions, not due to weather or climate changes.
 Correction Factor (k):
 Compute the correction factor (k) as the ratio of the slope of the period after the change to the
slope of the period before the change.
Example Scenario - Testing Homogeneity of Station E:
1. Identifying Two Different Slopes:
 In the homogeneity test for Station E, observe two different slopes in the double mass curve
analysis.
2. Calculating Correction Factor (k):
 Compute k using the slopes before and after the identified change.
3. Applying Correction to Time Period Before Change:
 Apply the correction factor (k) to the time period before the identified change in slope.
4. Resultant Double Mass Curve:
 After applying the correction, recompute the double mass curve for the adjusted time series.
 The goal is to achieve a continuously and constant slope throughout the entire time period.
Importance of Double Mass Curve Test:
 Convenience in Testing Homogeneity:
 The double mass curve test is a convenient method for evaluating time series for homogeneity.
  Sequence of Tests: Always conduct the double mass curve test after the consistency check, where
outliers are removed, and data gaps are filled with station data from other stations

Correction of Precipitation Values: Richter's Method

1. Systematic Errors in Precipitation Measurements:
 All precipitation measurements inherently have systematic errors.
 Example: Helman rain gauge, a standard gauge used by DWD (German Weather Service).
2. Identifying Errors for the Helman Rain Gauge:
 Wetting Error: 5-10%
 Evaporation Error: 1-3%
 Wind-Induced Error (Rain): 2-15%
 Wind-Induced Error (Snow): 15-55%
 Total precipitation underestimation can easily reach 50% or more due to these errors.
3. Richter's Correction Method:
 Developed by Richter in 1995 to correct precipitation for different errors.
 Specific to the Helman rain gauge but provides insights into systematic error correction for other
gauges.
4. Three-Step Correction Process by Richter:
 Calculate wetting and evaporation errors based on empirical data.
 Calculate wind-induced errors for both rain and snow.
 The total error is the sum of wetting/evaporation error and wind-induced error.
Experiment :
 Richter used data from 26+ stations in Germany with Helmholtz standard observation setup.
 Minimum 10 years of data at each station.
 Additional investigations included rainfall times, ground-level measurements, and
evapotranspiration loss assessments using an evaporation pads.
5. Three-Step Correction Method Summary:
 Step 1: Compute wetting and evaporation errors.
 Step 2: Compute wind-induced errors for rain and snow.
 Step 3: Total error is the sum of wetting/evaporation error and wind-induced error.

Wetting and Evaporation Error Calculation:

1. Nature of Wetting and Evaporation Error:
 This error arises as some rain is retained on surfaces of the precipitation gauge (collecting device
and funnel).
 Water on these surfaces can evaporate and is consequently not measured by the device.
 In regions with frequent small precipitation events, this error can be significant.
2. Small Precipitation Events and Potential Non-Recording:
 In humid climates, numerous precipitation events have small volumes.
 This situation might lead to no recorded precipitation if it evaporates from the gauge surfaces.
3. Calculation of Wetting and Evaporation Error:
 Typically computed for monthly precipitation sums and days with precipitation > 0.1 millimeters.
 Mean loss of precipitation is calculated using the empirical formula:
  Delta B is an empirical percentage determined
 Calculated separately for each month.
4. Empirical Percentage (Delta B):
 Delta B values determined for each of the 12 months and two situations:
 Plains (Lowlands) - Blue line in the graph.
 Hill Country - Brownish curve in the graph.
 Loss is generally smaller in winter months and higher in summer due to increased energy and
temperatures.
 Plains exhibit higher losses compared to hill country throughout the year due to higher
temperatures.
5. Average Wetting and Evaporation Loss:
 Average loss in the hill country is approximately 3.7% over the entire year.
 In the plains, the average loss is around 6.7%.

Wind-Induced Error in Precipitation Measurements:

 Each precipitation gauge acts as an obstacle in the wind field, leading to deformation and
disruption of the wind flow around the gauge.
 The provided graph depicts anomalies in the wind field for precipitation gauges with different
geometries.
 Blue color indicates that the vertical air velocity is consistently lower than normal within the
catch of the rain gauges.
 In extreme cases, such as the precipitation gauge shown in the upper left corner of the graph,
precipitation is entirely deviated from the gauge's catch
 When turbulence occurs in the funnel of a precipitation gauge, it disrupts the smooth flow
of air and can lead to the creation of circulating air flows within the gauge. Ths leads to
gauge's inability to capture and measure the full amount of precipitation that falls.
 Different geometries of rain gauges can lead to turbulence and distortion, affecting the
accuracy of precipitation measurements.
 Severe undercatch occurs not only in situations with strong horizontal winds (as shown in
the graph) but also when the horizontal wind component is less intense.
2. Recognition of Wind-Induced Error:
 The error resulting from the distortion of the wind field by precipitation gauges has been
acknowledged for some time.
 Manufacturers have responded by experimenting with various gauge geometries to minimize
wind-induced errors.
 Adjustments in gauge geometry aim to reduce turbulence and enhance the accuracy of
precipitation measurements.
Wind-induced error
 Commonly used to minimize the effect of wind in a rain gauge are so-called wind alters that consist of a
ring around the precipitation device on which small metal or plastic strips are mounted in a way that
they can move with the wind and thus lead to a more homogeneous wind field.
 But most precipitation gauges are not equipped with such a windshield. The wind induced error can be
computed by comparison of the unprotected rain gauge with a gauge with windshield or a gauge
installed at ground level.
 It is more robust to use the precipitation data from a gauge installed at ground level for this comparison.
 1. Factors Affecting Wind-Induced Error:
 The magnitude of the wind-induced error depends on multiple factors.
 The primary factor is the type of precipitation: liquid precipitation, snow, or a mixture of both.
 The second important factor is the exposition of the meteorological station, indicating how
exposed or sheltered it is to wind.
2. Exposition Categories:
 Four exposition categories are distinguished based on fisheye images:
1. Exposed to wind (surrounded by few or no obstacles).
2. Slightly sheltered (more obstacles than exposed).
3. Moderately sheltered.
4. Strongly sheltered (relatively low exposure to wind).
3. Graph Illustrating Wind Error:
 The graph shows the wind-induced error based on exposition and precipitation type.
 The x-axis represents the exposition types (exposed, little sheltered, moderately sheltered,
strongly sheltered).
 Three color bars indicate precipitation types: rain, mixed, and snow.
 Consistently, wind error increases with the exposure of the station. Higher values are seen in
exposed situations compared to strongly sheltered situations.
 Wind error is smallest for rain, intermediate for mixed precipitation, and largest for snow due to
the ease with which snowflakes can be transported by the wind.
 A wide range of wind-induced errors is observed, ranging from a few percent in strongly
sheltered situations for rain to up to 60% for exposed stations and snow.
4. Monthly Mean Wind Error for Rain:
 Monthly mean wind error for rain is shown in a graph with four lines representing the four
exposition classes.
 In strongly sheltered conditions, the wind error for rainfall is consistently less than 4%
throughout the year.
 The error increases with the exposition of the precipitation station.
 Seasonality is observed, with larger errors in winter.
5. Use of Correction Graphs:
 Correction graphs, along with corresponding tables for wind-induced error and
wetting/evaporation error, are used to correct time series of precipitation data.
 Corrections are conducted for all stations in the catchment.
 Corrections consider seasonality, precipitation type, and the exposure of precipitation stations.
6. Catchment Average:
 After correcting individual precipitation data for each station, the process proceeds to compute
the catchment average

Spatial interpolation methods

After the station data of precipitation has been corrected, we can proceed to compute the catchment averages.
 Rainfall is a dynamic process that varies both in space and time.
 For various applications such as water balance computations or modeling, precipitation data
needs to be integrated in both dimensions.
2. Temporal Integration:
 Point or station data is commonly integrated over time periods, such as 10 minutes, 15 minutes,
or daily averages.
 For water balance computations, this may be further integrated into monthly or annual averages.
 Temporal integration involves summing up precipitation data for the specified time intervals.
3. Spatial Integration Challenges:
 Spatial integration of point measurement precipitation data is challenging due to significant
spatial variability,
 Achieving a good estimate of the observed rain field requires dense sampling.
4. Importance of Spatial Analysis and Correction:
 Accurate spatial analysis, correction, and interpolation of precipitation data are crucial for
hydrological investigations.
 Precipitation serves as the primary driver of hydrological systems, and errors in input data can
introduce biases in the analysis results.
 5. Factors Influencing Spatial Variability:
 Weather Patterns and Precipitation Type:
 The type of precipitation and the general weather pattern influence spatial variability.
 Topography:
 The terrain or topography of an area plays a significant role in shaping precipitation
patterns.
 Wind Direction:
 Wind direction contributes to the spatial distribution of rainfall.
 Time Interval Consideration:
 The considered time interval also affects spatial variability. Longer integration periods
tend to smooth out smaller-scale differences.
6. Effects of Temporal Integration on Spatial Patterns:
 Longer temporal integration, such as annual averages, tends to smooth out spatial patterns seen
in shorter time intervals.
 Patterns observed in annual averages are more related to climatic differences and geographical
features like mountains.
7. Hydrological Modeling Challenges:
 Hydrological modeling often requires daily or sub-daily time steps, maintaining the challenge of
integrating highly variable point measurements in space.
 The density of the observation network is crucial for obtaining accurate spatial estimates.
 Accurate estimates depend on both the density of observations and the quality of the data used
for interpolation.
 Incorrect or biased data for interpolation can lead to inaccuracies in spatial estimates

Spatial measurements

Radar and Remote Sensing Products:

Radar and remote sensing products provide spatial measurements of precipitation. These methods are indirect
and may not be as accurate as ground-based station [Link] is widely used for forecasting and reanalysis of
precipitation data.
Spatial Interpolation Methods:
Spatial interpolation involves estimating precipitation surfaces from point measurements on the ground.
Methods are categorized into deterministic and statistical approaches.
1. Deterministic Methods:
 Thiessen Polygons:
 Thiessen polygons are a relatively simple deterministic method for spatial interpolation.
 They involve dividing the study area into polygons based on the proximity of observation
points.
 Precipitation values within a polygon are assumed to be uniform.
2. Statistical Methods:
 Stochastic Interpolation:
 Stochastic interpolation methods are more complex and involve statistical techniques.
 These methods consider the spatial variability of precipitation fields.
 They require more computational effort and may involve training processes.
3. Method Complexity:
 The complexity of interpolation methods varies.
 Example: Thiessen polygons are relatively simple, while stochastic interpolation is more
complex.
4. Method Selection Factors:
 Accuracy Requirements:
 The choice of interpolation method is guided by the required accuracy of the precipitation
fields.
 The level of accuracy depends on the aim of the investigation.
 Consideration of whether the goal is to represent the spatial variability of a single
rainstorm in short time intervals or annual mean precipitation over a larger region.
 The distribution of rainfall stations is a crucial factor in method selection.
  Sophisticated stochastic interpolation methods may be less useful if the dataset is sparsely
distributed in space.
 Physical Factors:
 Physical factors, such as topography, play a role in method [Link] differences in
elevation in the considered catchment can impact the choice of interpolation method.

Spatial interpolation methods

 Preparing precipitation maps where the spatial interpolation has been done before for long-term
averages typically.
 For shorter time periods we can use either deterministic methods such as the arithmetic mean or the
geometric mean, inverse distance weighting, Thiessens polygons, the the method or the hypsometric
curve.
 we can use stochastic methods like geostatistical interpolation or regionalization techniques.

Arithmetic mean method (check slide)

We start with the arithmetic mean. To illustrate this and the following methods, we take again our example here
of a catchment with four stations in there. And for a particular time period, say for a given day, we get four
different estimates or four different measurements from each of the stations. And we simply sum the
precipitation values up and divide them by the number of the stations. The arithmetic mean essentially gives
equal rates to all of the stations and should only be used if no other information is available and only in
catchments where there are no large differences in the topography.
Thiessens polygons
Here each station is weighted differently by the area of the surrounding polygon. And the polygons are
constructed in a way that its boundaries define the area that is closest to each point relative to all other points.
Thiessens polygons are a geometric method that can easily be constructed. And you find this as an option in
most of the GRS toolboxes for spatial interpolation. To compute the spatial estimate of precipitation, each
station is weighed by the area of its polygon.
Isohyetal method. (check slide)
It is similar to the Thiessens polygons method as it weighs the station precipitation data by the area of
influence. The main difference is that the area is computed from precipitation contours. Drawing those
precipitation contours is no trivial task to begin with. Once this is done, we can compute the area that lies
between the lines. Then the spatial interpolation of the point measurements of precipitation is done in the same
way as explained for the Thiessens polygons. The construction of those precipitation contours requires that
there are many stations available and as this is often not the case, the method is not very often applied.

Inverse distance weighting. (check slide)

 As the name suggests, the surrounding weather stations are weighted based on the inverse of the
distance through the respective stations.

 Closer stations are considered more influential in estimating values for a specific location.
 The weight assigned to a station is inversely proportional to the distance from that station.
 Unlike the Thiessen Polygons method where the area was a normalizing factor, IDW uses the
inverse distance as a normalizing factor.
 This normalization ensures that the weights assigned to each station are appropriately scaled.
 IDW is particularly useful in grid-based interpolation.
 In grid-based interpolation, spatially distributed precipitation fields are computed by applying the
IDW equation to each grid cell.
 For each grid cell, the IDW equation is evaluated by considering the precipitation values and
distances from all surrounding weather stations.
 The result is an estimate of precipitation for each grid cell based on the weighted contributions
from nearby stations.
Geostatistical methods - Kriging:

 geostatistical methods such as Kriging consider the spatial variance of the precipitation.
 Stations are again weighted and the station weights are determined by optimization using probability
functions.
  For that purpose a virogram is used which is a function that describes the spatial dependence in the
precipitation field.
 Kriging has been proven to offer the most accurate linear unbiased predictions for locations between
observed stations, giving values which fall under Gaussian confidence interval. It has demonstrated
superior estimation performance compared to methods such as spline interpolation.
 . Kriging is also often applied for grid-based interpolation.
 Some of the geostatistical methods don't only use the precipitation values itself for interpolation but also
secondary information. For example external drift Kriging is often used to describe the dependency of
precipitation to the topography or in other words to the elevation

the choice of interpolation method is subjective and can significantly influence precipitation sums for spatial
[Link] methods yield different results, affecting the representation of precipitation [Link] is
advisable to test multiple interpolation methods to understand their impact on results. Verification with
independent data or using techniques like the leave-one-out method enhances the reliability of the chosen
method.

Chapter 6

BASIS FOR THE DESIGN OF A STORMWATER RESERVOIR.

Design a reservoir for floodwater protection:
 We started with characterizing the catchment, in particular by the delineation of the surface water and, if
required, also the subsurface water catchment.
 Then we identified sites with rainfall records in and around the catchment, collected the data and
analyzed and if required corrected the time series.
 to get an idea about the reliability of the data it is always good to check if and how it has been corrected.
And the homogeneity test is always recommended.
 The next step in the task list now is to estimate extreme values of precipitation. And for that purpose we
have to specify the time interval of interest or in other words the precipitation duration and the area of
interest as well. For our reservoir design task the region of interest is the entire catchment.
 Typically the analysis is performed for all the station data before the spatial interpolation to the
catchment area.
  But in general the analysis can also be performed after the interpolation to spatial averages. The chosen
time interval depends on the specifics of the design tasks.
 Fast responding hydrological systems require typically shorter durations. An example for this fast
responding systems is the stormwater analysis in urban catchments. But often the extreme rainfall
analysis is done for several specified rainfall durations. Let's start with some data about largest ever
recorded rainfall amounts globally.

This graph shows rainfall durations on the x-axis in loG scale and rainfall in millimeter at the y-axis also in loG
scale. Small durations are seen on the left.

PROBABLE MAXIMUM PRECIPITATION (PMP)

 Globally there seems to be a natural limit for maximum rainfall

 a reasonable good relationship between the maximum precipitation and rainfall duration.
 However, those global rainfall maxima are not necessarily transferable to specific regions.
Definition
 It uses the analysis of flood producing storms and meteorological reasoning to find out the greatest
depth of precipitation for a given duration that is physically possible at a particular geographical location

Uses of Pmp

 PMP can be used in rainfall runoff models to derive PMF, the probable maximum flood.

 risk assessment, reservoir design

key meteorological reasoning behind the estimation of PMP: Is there an upper limit for rainfall depths for a given
duration?
1. Maximum Storm Conditions:
PMP assumes the occurrence of a meteorological event that represents the most extreme conditions for
rainfall generation. This includes factors such as the maximum values of precipitable water (the amount
of water vapor present in the atmosphere that can potentially be turned into precipitation), vertical
uplift rates (the upward movement of air that leads to cloud formation and precipitation), and
convergence of moisture-laden air masses.
2. Saturated Catchment: The estimation takes into account the condition of a saturated catchment, meaning
that the soil and surface are already saturated with moisture. This condition maximizes the runoff
potential and contributes to the intensity of rainfall and the resulting flood.
3. PMF (Probable Maximum Flood): The PMP estimation is closely related to the concept of Probable
Maximum Flood (PMF), which considers the maximum flow that could occur in a river basin due to the
theoretical maximum precipitation. The PMF is used in designing hydraulic structures to ensure they can
withstand the most extreme flood events.
4. Meteorological Reasoning: PMP calculations are based on meteorological reasoning and modeling,
considering historical weather data, atmospheric dynamics, and climatic conditions specific to the
region. It involves understanding the physical processes that lead to extreme precipitation events.

ESTIMATION OF PMP
PMP estimation based on:  Storm area  Catchment area

It is a crucial factor in the design of hydraulic structures such as dams and spillways to ensure their safety and
resilience against extreme weather events. But considering this would mean high cost as the amount of data
needed to be collected but it’s a 100 percentage design method.
The local estimation of storm rainfall, especially at a specific location using a rain gauge, often involves methods
that focus on the probability of exceedance, return periods, and depth-duration-frequency (DDF) or intensity-
duration-frequency (IDF) analyses.
Unlike Probable Maximum Precipitation (PMP) or Probable Maximum Flood (PMF) that deal with theoretical
maximums, these local analyses are more practical and useful for design and risk assessment of infrastructure.

1. Exceedance Probability and Return Period:

 Exceedance Probability: This is the probability that a particular rainfall event will be exceeded in
any given year. For example, a 1% annual exceedance probability (AEP) means that there is a 1%
chance of the rainfall amount being exceeded in a given year.
 Return Period: The return period is the inverse of the exceedance probability. A rainfall event
with a 1% AEP corresponds to a return period of 100 years.
2. Depth-Duration-Frequency (DDF) Analysis:
 DDF analysis is a method used to estimate the rainfall depth for various storm durations and
specific return periods.
3. Intensity-Duration-Frequency (IDF) Analysis:
 IDF analysis focuses on the intensity of rainfall, which is the rate of rainfall per unit of time
(intensity = depth / duration).
 Similar to DDF, IDF analysis provides relationships between rainfall intensity, storm duration, and
return periods.
 IDF curves are often used for designing stormwater management systems and infrastructure that
are sensitive to the rate of rainfall

Aim: estimation of rainfall depth for various storm durations and a given reoccurence period
 Depth-Duration-Frequency (DDF) analysis
 Intensity-Duration-Frequency (IDF) analysis (intensity = depth / duration)

FEATURES OF DDF CURVE

 Its Formal statistical method to derive extreme rainfall from maximum rainfall depth for given duration and
return period
 Each line signifies a DDF curve for a specific return period.
 For a given return period, rainfall depth intuitively increases with duration, but the increase is not
proportional. Short rainfall events tend to be more intense, for example, convective thunderstorms in
summer. Because athe rainfall depth seems constants over some duration indicating small storm
 The DDF curves stack up on each other with increasing return periods and they are bounded by the
theoretical maximum, which is the probable maximum precipitation line that is indicated as the dashed
line in the graph.
ASSUMPTIONS
 The most limiting assumption and probably the most common in engineering hydrology is the
assumption of stationarity. That means that no changes, natural or anthropogenic, of weather or climate
are assumed. As we know, this assumption is not entirely true during accelerated times of climate change
 .Another assumption is that rainfall depth for different durations is assumed to be statistically
independent

To derive a DDF curve, we first need to get or derive the time series of consistent and homogeneous data. The
record should ideally comprise 30 years or more. Once we got this corrected and homogeneous record
then we aggregate the data for the different durations that we are interested in, for example 1 hour, 3 hours, 6
hours and so on. And then generate a table where we put in each year and the chosen duration the maximum
rainfall depth of that particular year as demonstrated here.

Maximum rainfall depth were identified for the individual years and are listed in the columns of this table for
the five durations.
we take the annual values of maximum rainfall depth with duration of 1 hour and plot them as a histogram that
is shown here.
For convenience, the values on the y-axis are normalized, that is they are divided by the total number of the
values, which in this case is 31 for each year in the record. T
(This means all bars in the histogram taken together will sum up to one. The histogram in other words is an
expression of the empirical distribution of annual rainfall extremes for the duration of 1 hour. It tells us that
30% of all values lie in the bin that is centered at about 12 mm rainfall depth.) If we now integrate this
distribution by ranking or sorting the values in ascending order and summing those values up, we obtain the
cumulative distribution function depicted here in blue. This empirical cumulative distribution or cumulative
density function is useful because we can read from it a few interesting numbers.
The y-axis is now ranging between 0 and 1 and signifies the non-excedence probability f. As an example, let's
pick the 80% probability of non-excedence that is indicated by the value of 0.8 at the y-axis. We draw across to
the function and down at the intersection and read from the x-axis that in 80% of all cases in the data 19 mm of
precipitation was not exceeded. This signifies also that in 20% of all years 19 mm rainfall was exceeded.
Therefore, we can define the probability of exceedance as 1-f. This analysis can now be done for other
significance levels
. However, as the rare events are typically underrepresented in the time series, we can't reliably extrapolate the
empirical distribution depicted here. For that purpose, we need to fit an empirical distribution to the data that
allows for this extrapolation.
There are several theoretical distribution functions that could be used to fit the data. For example, test fit, e.g.,
Kolmogorov test, chi² - test.
We will assume here that the data is gumbel distributed as this often fits the data well. The gumbel distribution
is shown here in its general form. It has two parameters u and alpha.
To derive these parameters, we will use the so-called methods of moment.
The theoretically derived gumbel distribution is shown here in the graph in red.
We can now use the fitted distribution to derive quantiles for selective return periods, where the return period
is related to the inverse of the exceedance probability. Or mathematically, 1-f equals 1 over r, which gives if
rearranged the return period r equals 1 over 1-f. To get rainfall depth for given return periods, we take the
equation of the gumbel distribution depicted here again at the bottom left and substitute r with f. After
rearrangement for p, we get the equation at the bottom right that allows us to compute rainfall depth for given
return periods. In our example, we choose return periods of 2, 10, 50 and 100 years, evaluate the equation that
was just derived and get the values that are depicted in this table here. I hope you still remember that the
computations that we just did were conducted for the rainfall depth with duration of 1 hour. We now have to do
the same steps for the other durations. Create the empirical cumulative distributions, fit the gumbel
distributions with yields and new set of parameters, u and alpha, which are then used to compute rainfall depth
for the individual return periods and rainfall durations. This will then fill the table here with rainfall depth for
different return periods and durations and as a last step, we can plot the results. This is shown here with the
color coded circles. Each color is a row from the previous slide that signifies the rainfall depth for a given return
period and the different durations. We can now fit a curve through the sets of data points to get the depth
duration frequency curves for the individual return periods. As shown here, a power law function usually fits
these curves well. It is easier to fit the curves when the data is plotted in log-log space as shown on the right of
the slide here because then the power law becomes a linear function. Now the depth duration frequency curves
can be used to read from the graph's rainfall depth for an arbitrary duration and the given return period. For
example, as shown in the upper most graph here, the rainfall depth for a 100 year return period is 100 mm in 20
hours, about 78 mm in 10 hours and about 59 mm in 5 hours.

Applications in water management for ddf curve

 design of reservoirs and stormwater retention tanks
 design of sewerage systems
 discharge capacity of river channels
 capacity of pumping stations
 insurance of floodwater damage
 … → Aim: to prevent flooding, thereby reducing loss of life and property, and pollution of surface waters
SNOW AND SNOW MELT

Snow significantly influences the hydrological water balance through various mechanisms, particularly in
regions with seasonal snow cover. Two key factors in this influence are:
1. Water Storage (Seasonality of Runoff):
 Snow acts as a temporary reservoir for water. During the winter, precipitation falls as snow and
accumulates on the ground.
 This stored snowpack represents a delayed release of water. As temperatures rise during the
spring and summer, the snow melts, contributing to runoff and filling rivers and streams.
 The gradual release of water from the melting snow helps regulate the flow of rivers and can have
significant impacts on downstream water availability.
2. Influence on the Energy Balance:
 The albedo of snow typically ranges from 0.8 to 0.9, indicating that it reflects 80% to 90% of
incoming solar radiation. This high reflectivity contributes to the cooling of the surrounding
environment.
 In contrast, forests have a much lower albedo, typically ranging from 0.05 to 0.18. Forests absorb
more solar radiation, leading to heating of the surface.

 The interaction between snow cover and land surface albedo influences local and regional climate
patterns, affecting temperature, precipitation, and atmospheric circulation.
CHECK RELEVANCE IIN PDF
CONSTITUENTS OF SNOW
Snow Crystals:
 Snow crystals are ice particles that form in the atmosphere when water vapor freezes onto ice nuclei.
They make up approximately 10-40% of the volume of snow.
 Air ~ 60 - 90 %  Liquid water ~ 0 - 30 %

 < 0°C: 2-phase system only snow and air

Characteristic of typical cold snow conditions, where the snow is in a frozen state.
 > 0°C 3-phase system
Snow ,liquid and air . Characteristic of melting or thawing conditions, where the presence of liquid
water alters the physical properties of the snow.

SNOW CHARACTERISTICS:

1. Snow Height (hs):

 Snow height, measured in centimeters (cm), represents the depth of the snow cover at a specific
point.
 It is a point measurement, meaning it provides information about the snow depth at a particular
location.
 Snow height can vary significantly from one location to another, as snowfall is often spatially
heterogeneous.
2. Water Equivalent (hm):
 Water equivalent, also measured in centimeters (cm), represents the amount of water that would
result from melting a column of snow.
 When snow melts, it turns into water, and the water equivalent gives an indication of the liquid
water content within the snowpack.
 This measurement is valuable for understanding the potential contribution of the snowpack to
water resources when it melts.
3. Specific Water Content:
 Specific water content is measured in millimeters of water per centimeter of snow (mm/cm).
  It represents the amount of water contained within a unit height (1 cm) of snow.
 This measurement provides information about the density of the snowpack and how much water
is stored within a certain volume of snow.

Snow density is a measure of mass per unit volume and is typically expressed in units of kilograms per cubic
meter (kg/m³) or grams per cubic centimeter (g/cm³). Snow density varies depending on factors such as the
snow age, and depth.
 now freshly fallen ~ 10 – 150
 settled snow ~ 35 – 600
 firn 500 – 850 
glacier ice 700 – 900

TYPES OF SNOW MEASUREMENT

Invasive vs. Non-Invasive Measurements:

 Invasive Measurements: These involve physically interacting with the snowpack, such as inserting
instruments or probes into the snow. Invasive methods can disturb the snow's natural structure.
Rods, which are mentioned, fall under this category, as they are inserted into the snow to measure
its depth.
 Non-Invasive Measurements: These methods do not disturb the snowpack. They typically involve
remote sensing or other techniques that allow measurements to be taken without direct physical
contact with the snow. Sonar, mentioned in this context, is an example of a non-invasive method.
Portable vs. Stationary Instrumentation:
 Portable Instrumentation: These are instruments that can be easily carried and used in various
locations. Portable instruments are useful for fieldwork, allowing measurements to be taken at
different points or locations. Rods are often portable, as they can be carried to different sites for
depth measurements.
 Stationary Instrumentation: These are fixed or installed instruments that remain in one location
for an extended period. Rulers, when used as snow stakes, are typically stationary, being installed
in the ground at a specific location to measure snow depth over time.

1. Rods, Rulers, and Sonar for Snow Depth Measurements:

 Rods: Rods, or snow probes, involve physically inserting a rod into the snow until it reaches the
ground. This method is destructive and invasive as it disturbs the snow structure.
 Rulers: Rulers, when used as snow stakes, are often installed in the ground, and the snow depth is
measured relative to the ruler. This method is typically stationary and non-invasive, as it does not
require inserting anything into the snow.
 Sonar: Sonar systems use sound waves to measure the distance between the sensor and the snow
surface. Sonar is non-invasive and can provide continuous measurements along transects. It
allows for efficient and continuous monitoring of snow depth without physically disturbing the
snowpack.
Role in Hydrological Models:
 Measurements from sonar provides essential input data for hydrological models. The combination of
snow depth and density helps estimate the water equivalent of the snowpack, which is crucial for
understanding water availability, runoff, and potential flooding

SNOW TUBE
 Researchers typically insert the snow tube into the snowpack at the location of interest, ensuring that the
tube collects a representative sample. This is called gravimetric sampling
 The snow core extracted from the tube is carefully weighed to determine its mass.
 Additional measurements, such as the length or volume of the core, may be recorded to facilitate further
calculations.
 SWE is calculated by converting the mass of the snow core to its water equivalent.
 The depth-averaged snow density is determined by dividing the mass of the snow core by its volume.
SNOW PIT
Snowpit observations are crucial for avalanche forecasting and risk assessment. They provide on-site
information about the snowpack structure and conditions. Mechanical properties of the snow, including its
hardness, strength, and cohesion, play a significant role in avalanche formation and [Link]. Avalanche
prediction

PRESSURE AND LOAD CELL SENSORS TO MEASURE SWE AT A FIXED LOCATION

 Snow pillow: The snow pillow measures how much water is in the snowpack by weighing the snow
with a pressure transducer. Automatic measuring devices in the shelter house convert the weight of the
snow into an electrical reading of the snow's water equivalent -- that is, the actual amount of water in a
given volume of snow. higher installation and maintenance costs

 (hanging) lysimeters
lysimeters are sometimes used to measure the amount of water released as the snow melts.
Hanging lysimeters are devices that collect meltwater as it drips from the snowpack.
Snowmelt Lysimeters:
 Description: Snowmelt lysimeters are devices designed to collect and measure water from the melting
snowpack. The water is funneled to a collection pan, and its quantity can be measured, for example, using
a tipping bucket.
 They provide valuable data on the quantity of water released during snowmelt, helping researchers
understand the timing and magnitude of runoff.

Snow Tensiometer:
 A snow tensiometer measures the pressure of liquid water and air in the pore spaces of the snowpack. It
provides information about the potential for water movement within the snowpack.
 Snow tensiometers are particularly useful for observing diurnal meltwater fluxes through the snowpack
and the rate of movement of water within the snow.
Dielectric Sensors (Volumetric Water Content):
 , dielectric sensors provide continuous measurements of volumetric water content within the snowpack.
They are commonly used to monitor changes in water content during snowmelt.
Radar (Snowpack Layering):
 Principle: Radar technology can penetrate the snowpack and provide information about its internal
layering.
 Use Case: Radars are used to study the stratigraphy of the snowpack, including layer thickness and
density. This information is valuable for avalanche forecasting and understanding snowpack stability.
Satellite Products (Snow Cover for Large Areas)

SNOW PACK GENERATION AND MELTING IN 3 DIFFERENT PHASES:

1. Accumulation Phase:
 Snowpack growth occurs during the accumulation phase, typically in the winter.
 Snowfall contributes to the growth of the snowpack.
 Measurements involve assessing the depth and density of the accumulating snow.
 Wind can relocate snow, causing variations in snow depth and creating wind-deposited features
2. Settlement, Metamorphosis:

Following the accumulation phase, the snow undergoes settlement and metamorphosis.
Settlement refers to the compression of snow layers under their own weight. As more snow
accumulates, the layers beneath experience increased pressure, leading to a decrease in pore
space and an increase in snow density
Increasing Snow Density and Generation of Layers:

 Metamorphosis involves changes such as the transformation of snowflakes into rounded grains or
the development of faceted crystals.
 Layers form due to variations in snowpack properties.

Several process occur during the settlement and metamorphosis phase :

  Gravitation:
As additional snow accumulates on top of existing layers, the weight of the overlying snowpack
increases density and settlement
Pore Water Vapor Pressure:
 In the presence of temperature gradients within the snowpack, capillary forces can lead to the
migration of water vapor. This migration can cause the redistribution of water vapor within the
snow layers, contributing to changes in snow crystal structure and density

 Agglomeration During Melt-Freeze Cycles:

Temperature variations, especially during daytime melting and nighttime freezing, contribute to
the formation of melt-freeze cycles. During the daytime, the snow surface may melt, and water
percolates into the snowpack. Subsequent nighttime freezing causes refreezing of this water,
leading to the creation of ice layers within the snowpack.
This agglomeration process contributes to the formation of ice layers and a denser snowpack.

 Temperature and Humidity:

Metamorphic Changes: Temperature and humidity variations influence metamorphic changes
within the snowpack. The migration of water vapor between snow crystals can lead to
recrystallization, where existing snow crystals change in size, shape, and arrangement, ultimately
affecting the overall snowpack structure and density.

 Wind Effects:
Wind not only redistributes snow during the accumulation phase but also contributes to
compaction during the settlement phase. Wind can compress loose snow, leading to increased
density and the formation of wind-packed layers within the snowpack.

3. MELTING:
 Melting is the phase change of water from solid (snow) to liquid.
 It typically occurs during warmer periods, leading to the release of meltwater.
 The net energy input becomes positive when incoming solar radiation and other heat sources
exceed heat losses from the snowpack.

 Three Sub-Phases:
 a) Warming: Snow temperature rises, bringing it closer to the melting point.
 b) Ripening: The snowpack becomes saturated with liquid water as temperatures further rise.
 c) Output: Meltwater is generated as snow undergoes a complete phase change.
 Main Driving Factors:
 Temperature is a key factor, influencing the transition from the warming phase to the ripening
phase.
 Air humidity affects the availability of water vapor for the snowpack.
 Wind can influence the distribution of heat within the snow.
 Radiation (solar and terrestrial) plays a crucial role in providing energy for the melting process.

ESTIMATION OF SNOWMELT RUNOFF MAIN CHALLANGES

1. Spatial and Temporal Heterogeneity:

 Issue: Snow cover varies significantly across space and time.
 Implication: Traditional models may struggle to capture the spatial and temporal dynamics of
snow distribution accurately.
2. Anisotropy of Snow Cover:
 Issue: Snow is anisotropic, meaning it has different properties in different directions.
 Implication: Modeling snowmelt requires considering not only the distribution of snow but also
its orientation, which adds complexity to the modeling process.
3. Differential Distribution of Snow and Liquid Precipitation:
 Issue: Snow and liquid precipitation do not distribute in the same way.
  Implication: Models need to account for the distinct patterns of snow accumulation and
distribution compared to liquid precipitation.
4. High-Resolution Data Requirement:
 Issue: Traditional models might not be sufficient; high-resolution data is needed.
 Implication: More detailed data, such as high-resolution spatial and temporal data, is required to
accurately model the complex processes of snowmelt.
5. Regionalization of Point-Scale Model Results:
 Issue: Point-scale models may not be directly applicable to larger catchment scales.
 Implication: Empirical relationships and regionalization techniques are necessary to scale up
point-scale model results to cover larger catchment areas.

CHECK THE PDF FOR REST

6. Name different concepts of snowmelt modelling.
Energy Budget Physically based;
 balances the available energy with the energy required to for the melting process. o
High data requirement
 High temporal resolution Data Requirement o
 Warming Phase : Estimate the amount of energy required to raise the average temperature of the
snowpack to the melting point
Ripening Phase : Estimate the amount of energy required to saturate the snowpack
Output phase: Estimate the amount of meltwater that is produced by additional energy input
 mechanistic

Empirical relationships based on the heat budget

Temperature index approach
 Snowmelt is calculated as a linear function of average air temperature
 Used in most models to predict snowmelt
RUNOFF

 Runoff is the surface and subsurface flow from a catchment typically measured at a river cross-section.
 The time series of run-off is the run-off hydrograph and with the assumptions made before, it is equal to the
flow hydrograph. Hydrographs have a distinctive shape which contains information about catchment flow
processes, catchment storage and also about the different components and origins of the water in the catchment.

IMPORTANCE OF (ACCURATE) DATA COLLECTION

For Accurate Quantification Of Available Water Resources
 Precise measurements and modeling techniques are necessary to estimate the quantity of water resources,
including surface water, groundwater, and the contribution from sources like snowmelt.

 For Adequate System Characterisation

Detailed system characterization involves gathering data on geology, topography, land cover, and hydrological
processes. This information is then used to develop comprehensive models that simulate the behavior of the water
system.

For The Sustainable Management Of Water Resources

 Integrated water resource management strategies should be developed based on a thorough understanding of the
water system. This includes the implementation of conservation practices, efficient use of water, and measures
to protect water quality.

Data integration and decision support based on hydrological modelling:

 Advanced data integration techniques, including remote sensing, GIS (Geographic Information System), and
hydrological modeling, are crucial for synthesizing information. Decision support systems assist water
managers, policymakers, and other stakeholders in making well-informed decisions.

CONSIDERATIONS FOR MEASURING DISCHARGE

The ideal site for flow gauging is a straight section of the river with a constant slope.
 Constant slope ensures laminar and uniform flow, unaffected by turbulence or variations in channel bed
roughness (e.g., caused by obstacles like boulders).
2. Importance of Reliable Reference Point:
 Having a reliable point of reference for stage measurements is crucial for accurate and consistent data.
3. Applicability Independent of Measurement Frequency:
 Site selection criteria are independent of the intended frequency of measurements (e.g., continuous
monitoring during floods or dried periods).
4. Continuous Data vs. Single Flow Measurements:
 In many cases, continuous data is desired for monitoring flow hydrographs. However, hydrologists may
also conduct single flow measurements, such as differential gaugings for assessing exchange rates
between river channels and underlying groundwater.
5. Differential Gaugings for Specific Purposes:
 Differential gaugings involve measuring flow concurrently at two cross sections in a stream to infer
whether the river section between the gauging sites is losing water to the subsurface or gaining water
from groundwater inflows

METHODOLOGY.

 First we need to establish the site to continuously measure river stage.

 Secondly stream flow is measured for a discrete number of stage levels that ideally cover a wide range of
different river stages.
 And lastly from the concurrent measurements of flow and stage we can derive the stage discharge relationship
that allows us to convert the continuous measurements of stage into continuous measurements of discharge.
1) STAGE GAUGES
Stage - the river stage is a water level above a fixed reference point. This could be meters above sea level, but also any
other fixed point. It is only important that the fixed point is stable and therefore it should be resurred from time to time,
particularly after major floods or when the gaging station might have experienced physical damage
The most common and easiest way to measure stage is by a so-called staff gauge.

A staff gauge is like a ruler that is mounted on a structure on the riverbank and comes in various forms.
Manual gauges
 Nilometer,surgometer,
In Germany it is very common to see the E-shaped gauges. The height of each E is 10 centimeters and each step in the
E is 2 centimeters high. An E has 3 bars and 2 gaps, if you like, that is 5 times 2 equals 10 centimeters. This design
allows for easier reading even from further away.

Mechanical float gauge recorder

.In contrast to manual gauges, stage recorders provide a continuous and automated record of river stage. Usually the
construction consists of a stilling well that is hydraulically connected to the river as shown in the picture at the bottom
right. Mechanical float recorders works with the principle very similar to the one used in the Hellman Rain Gauge.
There is a flow in the stilling well that is connected to a counterweight by a string or cable that runs over a rotating reel
or pulley and the rotation raises and lowers a pen on a rotating drum with paper attached to it.

Pneumatic gauge / pressure sensor

 The most widely used technique to measure river stage today is by a pressure sensor which is again installed in
a stilling well or directly at the bottom of the river.
 It measures the column of water above the sensor which is the hydraulic pressure. Pressure sensors are less
vulnerable to extreme winter conditions when the water in the float chamber might freeze and are relatively
robust also in rivers that carry some but not too much sediments.

Bubble Gauges
 In ,mountain regions sensors submerged in the river are frequently lost in flood cases .In such rivers so- bubble
gauges can be installed.
 Here the equipment is placed in a shelter safely up at the river banks and a simple gas tube runs from the shelter
down to the river and a fixed reference point.
 A compressor creates a specific air pressure in a reservoir, forces the pressure down the gas tube to the river.
Using the bubble unit and a pressure control the pressure is set up high enough that air bubbles out of the lower
end of the gas tube in the river.
 The pressure in the gas tube is measured by a pressure transducer next to the unit safely in the shelter. The
pressure in the tube changes proportionally to the height of the water column over the tube in the river which
can be converted to stage measurements.
 After a flood only the gas tube needs to be replaced in case it was destroyed which is much more cost effective
and easy to fix.

Non-Contact Water Level Sensors

Another way to measure river stage is by non-contact water level sensors. These sensors conveniently measure the
height of the water surface without direct contact with the water by acoustic, radar and optical methods. However they
usually require a bridge across the flow channel or a beam where the sensor can be mounted.. Here a radar sensor has
been installed in the middle over an artificial flow channel to establish a stable stage discharge relationship for this
very high energy river with a very large sediment load.

.2) Flow Measurements For A Range Of Stages

 The second step in the procedure to continuously measure river discharge is flow measurements for a range of
stages. The flow velocity in natural river cross sections varies with depth and also width. Traditionally, flow
velocity is measured with current meters. Flow velocity is measured using current meters by counting the
number of rotations during a period of time (alternatively: electromagnetic sensors)

 Other measurement techniques comprise weirs and flumes for measurements in creeks or small streams,

 tracer dilution methods with sold or die tracers in channels with complex geometry

 and radar measurements.

  An emergent technique is the use of optical camera-based systems where particles floating on the water surface
are tracked on video, which is then converted into flow velocities by computer algorithms.
Engineering Method For Discharge Measurement
 The classical engineering method for discharge measurement uses the area velocity method to derive
discharge of steady uniform flow by the continuity equation as the product of vetted area times flow velocity.

Area Velocity Method

 Since the flow velocity may vary in given cross sections, we divide the cross sections into subsections or
segments and evaluate the continuity equation for each of the segments and then sum up the discharge for all of
the segments.
 Flow velocities are typically largest in the middle of the stream at the surface. Flow velocities decrease with
depth and towards the banks of the stream due to friction at the channel perimeter.. Again, velocity is largest at
the water surface and decreases to zero at the bottom of the profile.

 . In technical, very smooth channels such as rectangular concrete channels, the velocity is almost constant with
depth. In smooth channels, the velocity profile resembles a logarithmic function. In natural rivers with rough
uneven channels, the velocity decreases with depth more rapidly. And in very rough channels of mountain
rivers that have boulders, trees and other obstacles in it, the velocities near the channel base are smaller still

Midsection Method
simply and as the name suggests, the velocity profile is measured in the middle of each chosen subsection. The
assumption then is that the velocity profile is constant across the width of each segment. Note that the segments can
have different widths and that the width can be adapted to expected flow velocity profiles. In areas with high
velocities, we would choose smaller segments and in low velocity regions near shallow banks, we could space the
segments further apart. Depending on the number of measurements per profile, we can distinguish the 3-point method,
the 6-point method and the 2-point method. The 3-point method takes measurements at 20, 40 and 80% of the depth,
the 2-point method only at 20 and 80%. In the 2-point method, the two measurements could be simply averaged to get
the mean flow of the segment.
Other 2-point measurement methods include the method by Krebs with measurements at 40% depth and at the channel
bed, which is often used in practice because of its easier adjustment of the measurement depth.

A single measurement with the current meter can take between 2 and 5 minutes approximately. Therefore, the
complete set of measurements can take a long time in an average sized river. If we take 10 sections with 3
measurements, it will take at least 1 hour to complete all measurements. During that time, the river discharge could
have already changed. This would be problematic because we assign a discharge measurement to a specific river stage,
which is the last step in the procedure to continuously measure river discharge. Therefore, we always need to check the
river stage prior, after and if it takes longer, also during the flow velocity gaging.

STAGE - DISCHARGE CURVE (FLOW RATING CURVE)

Its is the the functional relationship between water stage and discharge
If we have conducted a sufficiently large number of discharge measurements for a range of stages at the gaging
station, we can plot the concurrent measurements into a stage discharge graph depicted here as the black dots and then
fit a function through the measurement points to derive the stage discharge relationship.

With a stable stage discharge relationship we can derive discharge as a function over time which we call the flow
hydrograph. The analysis of the flow hydrograph yields useful statistics for water resources management and flood
protection.
Plotted in log-log space, the stage discharge curve often resembles a slide S shaped curve as illustrated here. The shape
of the curve depends primarily on the geometry and the hydraulic properties, such as roughness of the river cross-
section. Even with multitiude clompx channel the curve might still produce an s shape
The wetted parameter of a cross-section, often increases continuously with the [Link], such a cross-section
would be far from ideal for establishing a stable stage discharge relationship, mainly because the characteristics change
in rivers with mobile riverbeds. Therefore the geometry of the cross-section and the stage discharge relationship should
be regularly checked.

ASSUMPTIONS IN STAGE DISCHARGE CURVE

 A unique relationship between the two quantities assumes unsteady uniform flow in the river channel. Unsteady
means that the flow is variable in time and uniform means that the hydraulic parameters such as the velocity, the vetted
area and acceleration terms remain constant with discharge along the flow path. However, it has been observed that the
rising stage of the flat wave propagation is much shorter than the falling stage, hence the associated acceleration rate
are different from the two limbs of the hydrograph. From that follows that flow in a river is unsteady and non-uniform
and the stage-discharge relationship is not unique but actually histerisis

HYSTERISIS
Let's consider a longitudinal section through a river as depicted here. There is the riverbed at the bottom and the water
surface on top and the flow runs from left upstream to right. Let's further consider that we observe stage and discharge
at a river cross section depicted by the vertical dotted line. If we were to continuously measure discharge at the gaging
location during the course of the flat, we would record the hydrograph depicted here. Now let's consider the snapshot at
time t1 depicted in blue in the graphs. In the top graph we see the approaching flat wave. The peak of the wave is still
to come and we observe the water level h1 and the discharge q1 at the gaging station as indicated by the yellow circles
in the upper and lower graphs respectively. We can also plot that pair of values into the stage-discharge graph shown
here with the y-axis and the discharge on the x-axis. Now let's consider another snapshot at time t2 when the peak has
passed our observation location and the water level is exactly the same as it was at time t1. And we will observe that
the discharge of the receding flat wave is smaller than the discharge of the approaching flat wave. This illustrates that
the stage-discharge relationship is not unique, which can be explained by the differences between accelerated and
decelerated flow. For simplicity let's take the Manning's equation for open channel flow. This equation depicted on the
bottom left here we find coefficients and hydraulic power meters like the non-dimensional friction coefficient, the
cross-sectional area and the hydraulic radius which are the same for our considerate two snapshots in time

. However, the energy slope s is different for the two cases. The slope is larger for accelerated flow when the flat wave
is approaching compared to the decelerated flow when the wave is receding. It follows that the discharge of the
accelerating wave is larger than the one of the decelerating wave. From that it also follows that the discharge peak, the
largest discharge of a flood event, arrives prior to the stage peak. The largest observed water level as depicted in the
hysteresis loop in the stage-discharge relationship.

FLOW MEASUREMENTS – OTHER METHODS

1) Weirs, for example, are applied in small to medium sized streams and have the advantage that there are standardized
constructions and formulas that relate stage to discharge.

2)Tracer dilution technique. It can also be applied in small streams and the main advantage is that it does not require
the installation of any other equipment. Typically, salt is applied as a tracer, some distance upstream of the
measurement location where an electric conductivity sensor registers the tracer breakthrough curve. The tracer
attenuation or the shape of the breakthrough curve is related to discharge. We distinguish two tracer injection methods.

 For continuous injection, a tracer is injected at a constant rate Q1 and a constant concentration C1,
continuously until the concentration at the sampling location reaches a stable constant value. The stream
discharge is then computed by the proportionality equation.
 In instantaneous injection method the tracer of a known concentration C1 and the known volume V1 is
instantaneously applied to the stream and discharge is computed by the conservation of mass equation
When streams get bigger, we typically cannot use tracer dilution methods because of the great volumes of tracer
required and because of the associated environmental consents that have to be applied for.

ADCP technique

 Current meters can be used, but it takes a long time to get a complete discharge measurement done. Therefore,
in larger rivers with a more complex geometry, the ADCP technique can be used. ADCP stands for Acoustic
Doppler Current Profiler.
 The acoustic signal is emitted at a high frequency and backscattered at particulate matter. The shift in the
reflected signal is proportional to the flow velocity.
 The ADCP sensor can be mounted on floats or boats and dragged or driven across rivers. The unit comes
typically with a high precision GPS which tracks the course of the sensor at any given time.
 The main advantage of the method is that it yields the complete velocity profile. It also measures the cross-
section geometry as a side effect and the measurements are very rapid.
  The method has its limitations in very shallow streams and in rivers with a lot of sediment transport at the base.
But the latter applies to almost all other methods too except perhaps to the tracer dilution method
 . Another interesting feature of the ADCP method is that it can be applied in principle not only to the
downstream or horizontal velocity in a river, but also to the vertical velocity component in any given cross-
section.
 This can give interesting insights into the complexity of flow patterns in rivers which could be important for
contaminant transport and mixing for example.

Camera Based method

 Emerging field of camera-based methods to measure surface flow velocities in rivers.
 The velocities are derived from particle tracking using automatic detection algorithms. River discharge can then
be estimated using the surface velocities and wetted cross-sectional area that is derived at any given time from
3D surface models of the site.
 These 3D models can be derived from images using multi-media photogrammetry methods.

HYDROGRAPH ANALYSIS

Hydrograph analysis is the statistical analysis of discharge hydrographs or timescales of discharge as depicted in the
graph on this slide.

From the flow hydrograph we can calculate simple statistics like the peak flow, low flow and mean discharge values,
which are the maximum, minimum and mean values of the times series respectively
1)
 The mass curve denotes the time integral of the hydrograph. The slope of the mass curve is the time integral of
the mean discharge
 mass curve of differences to the average is computed by subtracting the slope of the mass curve from the mass
curve itself. The result is the mean discharge. This type of analysis finds its application for example for the
design of reservoir volume or of control rules for reservoir operation.
2)

 Let's consider the hydrograph is the inflow to a storage reservoir and the dashed line is a constant outflow from
the reservoir that has to be maintained. For example to maintain environmental flow in order to protect a
downstream habitat with rare aquatic floor or pylons. Then we can read from the graph directly when there are
times of water surplus in blue and water deficit in red. The inflection points of the mass curve of the differences
to the mean signify then the local minima and maxima of reservoir storage given the in this case constant relief.
Those inflection points are depicted in the graph here by those vertical black dashed lines. These in turn allows
to compute the volume of storage needed to compensate times of water deficit.
 And this analysis can again be tied to risk for example to compute the storage that compensates flow deficit in
99% of all cases observed in the time series. But then the time series must be sufficiently long or even better it
is extended by stochastic time series simulation that could in principle also account for future changes in the
hydrological system.

flood volume.
Flood volumes are computed individually for flood events above a certain threshold. For example there is a critical
flow or water level that should not be exceeded because otherwise the flow will spill over flood protection barriers or a
dam might breach or a highway gets flooded.

FLOW DURATION CURVE

It is defined a cumulative frequency curve showing the percent of time during which discharge were equalled or
exceeded in a given period
can be computed by integrating the flow frequency diagram. These diagrams can be computed for flow data of
individual years as depicted by the blue lines here and for the entire time series as depicted by the red line. A flow
duration curve shows the exceedance probability on the x-axis and the corresponding flow value at the y-axis. axis.
These probability values are relevant as statistical criteria, for example, flood-reoccurrence probability and
probabilities for the navigability of waterways, as economical criteria, for example, in power generation schemes or in
general for water usage and as ecological criteria, for example, to determine minimum discharge and inundation areas
to name just a few. And since flow-duration curves have such a wide range of applications,

ASSUMPTIONS in FLOWDURATION CURVE

 Main assumptions: continuous data record (bias, variance)
 observed fluctuations are randomly distributed around a stable mean
 annual variation and periodic processes caused by climate alone

Steps to Plot the Flow Duration Curve:

1. Data Preparation:
 Collect a dataset of daily flow values for the chosen year, and ensure that there are no missing values in
the time series.
2. Sorting or Binning:
 Sort the daily flow values in descending order, or alternatively, bin them into specified flow ranges. In
this example, bins of 20 cubic meters per second are used, with boundaries at 20, 40, 60, and so on.
3. Calculating Absolute Frequency:
 Count the number of flow values within each bin. This is the absolute frequency, as shown in column 3
of the table.
4. Calculating Cumulative Frequency:
 Sum up the absolute frequencies to get the cumulative frequency (column 4).
5. Calculating Relative Frequency:
 Normalize the absolute frequency by dividing it by the total number of days (365 in this case) to get the
relative frequency (column 5).
6. Calculating Relative Cumulative Frequency:
 Sum up the relative frequencies to get the relative cumulative frequency (column 6). Ensure that the
sum equals 1 to verify the accuracy of the calculations.
7. Plotting the Flow Duration Curve:
 The flow duration curve is typically plotted with flow values on the x-axis (normalized to probabilities)
and the cumulative frequency on the y-axis. The curve starts from the highest flow value (far left) and
descends to the lowest flow value (far right).
8. Plotting in Log Scale (Optional):
 For convenience, the y-axis may be shown in log scale. This is done in the example for the graph on the
right side.
9. Calculating Exceedance Probability:
 The exceedance probability is calculated by subtracting the cumulative frequency from 1. This results in
a separate curve, often shown in red, representing the probability of exceeding a specific flow value.
10. Reading Values from the Curve:
 Median flow: The point where the non-exceedance probability curve intersects the exceedance probability
curve represents the median flow. In the example, it is approximately 30 cubic meters per second.

I will subsequently walk you through an example and briefly explain the different steps to compute it. Let's consider
we like to compute the flow-duration curve for a single year of flow data. We start with a table of the 365 daily values
of flow and sort them by size as shown in the last column in this table here. Alternatively, we can bin the data into bins
of specified range. This is typically done when we do the statistics for multiple years. As shown in the second column
from left, here the bins are chosen to be 20 cubic meters per second wide, so the boundaries are 20, 40, 60 cubic meters
per second and so on. We sort the values into the bins and count the number of values in the bin. This is the absolute
frequency listed in column 3. If we sum the numbers in the bins up, we get the cumulative frequency in column 4. And
if we then normalize the frequency by the number of days, 365 in this case, we get the relative frequency and the
relative cumulative frequency in columns 5 and 6 respectively. If the last column sums up to unity, we have done
everything correct. To illustrate graphically what we have just done, consider a year of discharge data on the left. If we
imagine horizontal lines across this graph at 20 cubic meters per second intervals and count the number of flow values
that lie between those lines, we get the histogram or frequency diagram on the y-axis of the graph on the right. And if
we now sum up these frequencies, we get the orange line, which is the non-accidents probability. There are a few
assumptions that actually have to be considered before we do this type of analysis. The first is that we have a
continuous data record and no missing values in the time series. This would otherwise skew the results and produce a
bias or error in the variant. Another assumption is that the hydrograph fluctuations are randomly distributed around the
mean and not influenced by human activities such as flow diversions, for example. And similarly it is assumed that
annual variations and other temporal variations are caused only by climatic conditions. But back to our cumulative
frequency graph on the right side of this slide. As mentioned before, it signifies the non-accidents probability of flows.
Let's move this graph to the left of this slide here. The graph shown on the right is essentially the same graph, only that
now the y-axis is shown in log scale for convenience. And if we normalize the x-axis by the number of flow values
365, it now scales from 0 to 1 as we would expect for probabilities. Finally, we also can compute the accidents
probability by subtracting the non-accidents probability from 1. This curve is shown in red. At the exact of the two
curves, we can read the median flow from the graph. In this case, it is about 30 cubic meters per second. And we can
read the graph also the other way. For example, take the flood value of 100 cubic meters per second and read from the
graph that in about 325 days of the year or 90% of all cases, this flow level has not been exceeded. As mentioned
before, this analysis can be done for single years and for the entire time series, which allows us to determine whether a
year was particularly red or dry. We could also detect trends in the time series and infer from the shape of the flow
duration curve characteristics of the hydrological regime of the Breva. In conclusion of this chapter, I like to reiterate
that hydrograph analysis comprises a wide range of statistical methods borrowed from time series analysis and beyond.
There are several other techniques that I have not mentioned here, such as the analysis of linear and periodic trends, for
example, or powerful, more sophisticated frequency analysis methods like Waflet. Hydrograph analysis techniques are
also featured in our next chapter, which is Plates and Trials.

FLOODS

Floods are mainly caused by meteorological extremes, most notably

 extreme precipitation, but also by
 rapid and massive snowmelt events caused by jumps to higher temperatures in higher altitude
 floods are ice jams, when conduits or bridges are blocked by an build-up of ice floats or similarly,
jams by floating trees or wood.
 In coastal regions, floods can be caused by storm surges or by an extremely large tide called spring tide or king
tide. The latter isn't exactly a meteorological phenomenon, but becomes a problem when coinciding with
storms.
 But floods can also be caused by events that may not directly be related to weather, such as the failure of dams
or dykes and by earth or sea quakes that cause tsunamis and catastrophic flooding of coastal regions

FLOOD PROTECTION

 Allowing rivers more space helps retain flood volume and reduces peak flows. This involves preserving or
restoring natural floodplain areas, which can absorb excess water, slow down river flow, and elongate flood
hydrographs.
 The presence of natural shrubs and trees along riverbanks introduces friction, slowing down water flow and
aiding in flood control.
 Constructing retention reservoirs in steep tributaries, where floods often originate, helps to capture and store
excess water before it reaches downstream areas. This can prevent downstream flooding and reduce the impact
on vulnerable regions
 Building stop banks (also known as levees) and dykes along riverbanks provides structural defenses to protect
low-lying land from inundation during floods
 Integrating smart city design involves planning recreational areas, parks, and high-infiltration capacity zones
close to rivers. In this design, lower value industrial areas are strategically placed in the next tier of flooding,
and flood-resilient residential areas are located in regions protected by mobile retention walls.
 Constructing flood-resilient houses involves elevating living spaces above potential flood levels. Features may
include having no basements, with garage and utility spaces at the ground level, and living areas on upper
floors. This vertical design minimizes the footprint and reduces resistance to floodwaters.
DESIGN OF FLOOD PROTECTION STRUCTURES:
CRITERIA:

 One design criterion is the peak water level. It is a criterion for the design of flood retention walls, used for
drawing flood inundation maps or for the design of flood resilient houses.
 Another criterion is the peak discharge, used for the design of canal capacity, riverbeds, diversions and others.
The duration of a flood that exceeds a certain water level is one of the criteria to ensure dike stability or the
stability of hydraulic structures in general.
 And the flood volume is required for the design of retention basins or flood protection dams.
 But probably the most important design criterion is the design flood.
 A design flood is the hypothetical flood hydrograph or peak discharge, used for the design of a hydraulic
structure or river control. It is the maximum flood that could be passed without any damage to the structure. A
common example for a design flood is the flood hydrograph for a 100 year return period. The estimation of
such extreme events can be challenging. Often the historic flow records are not long enough and as mentioned
earlier, hydrological systems are not stationary.

ESTIMATION OF DESIGN FLOODS (HQ)

Estimation of design floods (HQ) for short data records

 Flood frequency analysis assumes the stationarity of the hydrological system. If this assumption is violated, we
need to apply predictive hydrological models. Flood frequency analysis also requires a sufficiently long flow
record. Ideally, it would be 100 years plus, but this is often not available. The statistics become unreliable,
however, if the record is shorter than approximately 30 years.

 In some circumstances, it is possible to extend a short time series with records from neighboring catchments or
from gaging stations in the same river but at an up or downstream location. This can be done through
correlation or regression analysis.
 In this case, we can derive design flood statistics from neighboring catchments using the reference HQ method,
where HQ comes from German and stands for high discharge. HQ 100, for example, is a design flood discharge
with a 100 year return period.
 The reference HQ method is applied to transfer the statistics from an observed catchment with an existing flow
record to an ungauged catchment.
Reference HQ-method assumes:
 This requires similarity of the catchments with respect to morphology, geology, soils, land use, etc.
 • similar characteristics to a gauged catchment (morphology, soils, land-use, ..) •
 similar flood probability (flow duration curve) • a
 simple relation between reference flood and characteristic catchment parameters

Estimation of design floods (HQ) for longer data records

 They are essentially the same as we have been using to derive the depth duration frequency curve, where we
computed exceedance probabilities for extreme precipitation.
 It starts with data pre-processing to select largest discharge peaks for all years in the time series or alternatively
for flood peaks above a certain threshold.
 We need to remove obvious outliers and then compute the empirical non-accidence probability fit a theoretical
probability function and can then directly read from the graph or compute with the distribution function the
design-flat discharge for a given return period R
.
STATISTICAL ANALYSIS OF THE HYDROGRAPH:
 • determine maximum annual characteristics (peak discharge HQ, flood volume) from observed or simulated
runoff time series
 • consistency test
 • empirical non-exceedance probability, F •
 fit theoretical prob. Distribution
 select return period, R R = 1/(1-F)
   determine design flood HQ (R)

DROUGHT
I like to present some hydrological drought criteria that are derived from the analysis of the flow hydrograph.

 One obvious criterion, which is the RELATED to peak discharge, is the lowest recorded discharge.

 However, for hydrological drought analysis, it is equally important how long the flow is below a certain
critical threshold. This time period is the low flow duration D.

 We can, for example, compute the lowest mean flow during a period of seven days. If this value is
computed for every year in the flow record, we can derive the 7q10 criterion, which is the seven-day low
flow with the return period of 10 years. 7q10 is a very common criterion for hydrological drought
analysis.

 duration and reoccurrence of low or lowest water levels, which is an important criterion for the
navigability of rivers, but also for wetland protection, habitat vulnerability assessments and others.
certain water demand needs to be satisfied or minimum flow regulation needs to be balanced by a
reservoir.

 frequency of flow below a certain threshold, which can be computed from the flow duration curve. In this
category, a common design criterion is the Q95, which is the 95 percentile of the flow.

 The drought severity S is related to the deficit volume and signifies the cumulative deviation from a
critical flow level, which is again useful for the design and operation of freshwater reservoirs.

 drought intensity is a measure of the magnitude of a drought and is defined by the ratio between severity
and duration.

And as already mentioned, these criteria are used for design purposes, but also for operational management.
For example, to impose restrictions on allocated surface water abstractions based on critical flow levels.

This statistical analysis of design drought is done in the same manner as explained for design floods and uses
the same method. However, we have two dependent quantities in the hydraulic drought analysis, the flow value
and the duration. Therefore, we can compute exceedance probabilities either for the mean low flow during a
given period, for example the annual seven day minimum flow, or for the duration, which is equivalent to the
number of days below a given critical flow threshold. This is a one-dimensional statistical analysis. We can of
course do the analysis also for flow and variable duration simultaneously, which is then a two-dimensional
probability analysis. But this doesn't necessarily require different methods. For un-gauge basins, we have to
once more use original analysis methods that rely on assumptions of catchment similarity, such as the
correlation analysis or the reference low-q method. And this was our brief excursion to the topic of clouds. I like
to recommend for further reading the comprehensive report by the WMO on low-flow estimation and
prediction, which also includes some practical concepts on managing stream-flow abstraction using the flow-
duration curve. And with that, we reach the end of the topic, Flaps and Droughts. The next topic will be the
runoff component.

RUNOFF COMPONENTS

 main components of runoff are groundwater flow and streamflow.

 Streamflow is a combination of base flow, interflow and saturated overland flow
 Baseflow refers to the portion of streamflow that is sustained between precipitation events, often
contributing to a river's flow during dry periods.
Interflow:
  Interflow, also known as subsurface or hypodermic flow, is the flow of water that occurs within the soil
profile, originating from near the surface.
 It results from the movement of water through the soil, typically in response to rainfall or snowmelt.
Interflow is faster than base flow but slower than overland flow.

 Saturated overland flow, also referred to as quick flow or surface runoff, is the movement of water
surface when the soil is saturated and cannot absorb additional water.

FACTORS AFFECTING RUNOFF

 Bare rock can be impermeable, causing precipitation to run off at the surface and leading to a fast
streamflow response.
 Weathered rock, such as caustic limestone, can have greater subsurface depth for precipitation
infiltration, resulting in different discharge responses.

 Dry Condition: In the dry situation, the area of saturation (depicted in blue) is smaller, indicating
that less of the catchment is saturated. The stream network (brownish line) is also shorter. Hence
more precipitation can potentially infiltrate into the soil, reducing overland runoff. The shorter
stream network implies a less extensive flow path.

 Wet Condition: In the wet situation, a larger area of the catchment is saturated, The stream
network is longer, illustrating the presence of more connected pathways for water flow. The
longer stream network signifies an increased potential for runoff to travel over a greater distance.

Catchment Shape and Catchment size

.
1. Catchment Shape:
 Effect on Runoff: The shape of a catchment influences the way water is distributed and collected
across the landscape. Different shapes can result in variations in runoff patterns and the timing of
streamflow.
 Concave vs. Convex Shapes: Concave-shaped catchments may efficiently collect and channelize
water, potentially leading to increased runoff. In contrast, convex-shaped catchments may
disperse water over a larger area, reducing the concentration of runoff.
 Slope Orientation: The orientation of slopes within a catchment, such as north-facing or south-
facing slopes, can impact solar radiation exposure. This, in turn, affects evaporation rates and,
consequently, runoff.
2. Catchment Size:
 Effect on Runoff Volume: The size of a catchment is a critical factor in determining the overall
volume of runoff. Larger catchments generally capture and accumulate more precipitation,
resulting in increased runoff.
 Flow Path Length: Larger catchments often have longer flow paths for water to travel from the
upper reaches to the outlet. This extended flow path can influence the time it takes for runoff to
reach the stream.
 Land Use Changes: Human activities, such as deforestation or urbanization, can alter the size and
shape of catchments. Urbanization, for example, often increases impervious surfaces, leading to
more rapid and concentrated runoff.

1.
Slope and Aspect: The slope of the land surface influences the speed of runoff. Steeper slopes
often lead to faster runoff. Aspect, or the direction a slope faces, can impact solar exposure and
evaporation rates.
 Aquifers and Aquicludes: The presence of aquifers (permeable rock layers that hold water) and
aquicludes (impermeable rock layers) affects the movement of water underground, influencing
the base flow component of runoff.
2. Vegetation Cover:
  Interception and Uptake: Vegetation intercepts and retains precipitation on leaves and branches,
delaying its entry into the soil and contributing to canopy storage. Vegetation also takes up water
from the soil, acting as a temporal store and influencing the overall water balance.
 Infiltration Enhancement: Biotic activity in the topsoil, facilitated by vegetation, enhances
infiltration. Vegetation roots create channels and improve soil structure, promoting water
infiltration into the soil.
3. Catchment Shape and Size (Reiterated):
 Flow Path Length: The length of the flow path within a catchment affects the time it takes for
runoff to reach the stream. Larger catchments may have longer flow paths, potentially delaying
the arrival of runoff at the outlet.
4. Meteorological Factors:
 Type of Precipitation: The form of precipitation (rain, snow, sleet) influences its interaction with
the land surface. Snowmelt, for example, introduces water gradually into the catchment over time.
 Wind Speed: Wind can impact the distribution of precipitation, potentially enhancing or reducing
interception by vegetation. Wind can also affect evaporation rates.
 Potential Evapotranspiration: The potential loss of water to the atmosphere through evaporation
from the soil surface and transpiration from vegetation affects the overall water balance in the
catchment.
 Path of Precipitation Event: The trajectory of a precipitation event across the catchment
influences the spatial distribution of rainfall, affecting the timing and magnitude of runoff
components

Techniques used for hydrograph/hydological separation/base flow seperation

Hydrograph separation refers to the process of separation of quick flow from base flow and thus to determine
the runoff response to a rainfall event.
 Graphical (semi-log plot recession lines)
 analytical
 Tracers (chemical, isotopes, … EMMA)
 Filtering (fixed interval, sliding interval, local minimum methods: Hysep 1-3)
 Recursive filtering (low-pass filter)
 Unit hydrograph method
 Rating curve method (relationship between gw levels and stream flow

METHODS OF IDENTIFICATION OF RUNOFF SOUCE COMPONENETS

1. hydrograph separation
 The first is the category of hydrograph separation method, sometimes also called base flow separation
method. As the name suggests, they aim to separate the slow base flow components from the direct or
quick flow components
 This can be done graphically for the hydrograph of single flood events. For longer times conceptual
storage models can be set up
 In each time step, precipitation is separated into interception and quick flow
 parts that contribute to the quick flow store SG1 and the base flow store SG2. The outflow from these
stores makes up the runoff component.
 In fact, these building blocks are essentially the same that make up a lumped conceptual rainfall runoff
model. Example DIFGA or DIFGA 2000.

2. CHEMICAL SIGNATURE

Another category to discriminate between runoff components is to use their chemical signature. Precipitation
water, overland runoff, soil water, springs and deep ground water all might have distinct hydrochemical
fingerprints. Example : Stiff diagram: to display the major ion composition of a water sample. The different
chemistry allows us to distinguish different water sources, which then can be used in mixing models
3. EMMA
 The different chemical fingerprints of source waters are called end-members, and the
corresponding analysis is called end-member mixing analysis, or in short, EMMA
 These end-members should be clearly distinguishable from each other. for example, through
principal component analysis or PCA.
  This type of analysis can be put in a mathematical framework using simple mixing formulas to
derive fractions of the different end-members in any given stream water sample.

 not only chemistry is used to determine the fingerprints of source waters, but also environmental
GASES, like deuterium and oxygen-18, as well as a wide range of other isotopes and water-age
tracers.

4. COMBINATION OF HYDROLOGICAL AND CHEMICAL ANALYSIS

The individual strength of both a conceptual hydrological attachment model and the chemical analysis can also
be combined.
For example stream flow and nitrate concentration data in stream water can be used to derive the transient
contributions of fast and slow groundwater flow, a quick flow component called near surface flow, as well as
nitrate concentrations and discharge. The method has higher data requirements

5. HYDROLOGICAL SYSTEM ANALYSIS APPROACH,

Hydrological system analysis approach distinguishes three phases to turn precipitation into runoff.
 The first phase is runoff formation and describes the proportion of the aerial precipitation that is
becoming direct runoff.
 The second phase, runoff concentration, describes the temporal distribution of stream flow at the
catchment outlet and thus represents the integration of the water moving from different contributing
areas to the stream along various surface and subsurface flow paths
 And the third phase is flow routing, which describes the wave propagation and retention processes in the
stream channel. This phase describes the stream flow routing processes from subcatchments to the
catchment outlet or generally between upstream and downstream cross-sections of the stream.

Runoff formation
The first phase of the approach is Runoff formation
. In natural catchments, it is literally impossible that all rainfall becomes Runoff. Precipitation is temporarily
stored at vegetated surfaces and is infiltrating and stored in soils.
Some of that water is later evaporating, taking up by plants or percolating into deeper layers to eventually form
base flow.
Runoff formation includes all of these processes and describes the proportion of precipitation that becomes
direct Runoff. This proportion is also called the effective precipitation PE.

Effective Precipitation depending on the landform and vegetation cover.?

 Different landforms yield different effective precipitation. Densely vegetated lowland areas can usually
store a lot of water and thus the effective precipitation is small.
 In contrast, in steep mountain country with no vegetation cover and bare rocks, the proportion of
precipitation that is turned into Runoff is very large.
 The most important one is pre-wetting or catchment saturation. The effective precipitation is larger
when the soil and interception stores are already full.
 Dry bare soils become hydrophobic after an extended dry period, which causes almost 100% surface
runoff.
.
 In polar and high altitude regions, precipitation usually falls in the form of snow, in which case the
effective precipitation is virtually zero as all precipitation goes into the store of a snowpack.

ESTIMATE RUNOFF FORMATION IS THE SCS CURVE NUMBER METHOD

 is easy to use and it has a low data requirement\

  The SCS method assumes an empirical relationship between the storage capacity of a catchment and
direct runoff
 The SCS method assumes an empirical relationship between the storage capacity of a catchment and
direct runoff. This is illustrated here again with the bucket storage concept. If the bucket is small or full,
there is little capacity to store precipitation in the catchment, and a large proportion becomes direct
runoff as depicted in this schematic. If the bucket is large or empty, on contrast, the storage capacity is
larger and a smaller proportion of the same event becomes direct runoff


SCS CURVE NUMBER METHOD

. Different calculation steps are performed when just the peak discharge is required or when the full tabular
runoff hydrograph is required. Separate sub-areas can be considered as well as storage effects and routing
between sub-catchments.

The steps to compute effective precipitation with the SCS method are as follows.\
The effective precipitation is assumed here to be the same as direct runoff.

 First, we estimate the curve number Cn for different soils, land use types, etc.
 Secondly, we need to check whether the curve number needs to be adjusted for other conditions of
anticedent soil moisture.
 In the third step, we compute the storage capacity S, and lastly, we determine the effective precipitation
Pe.

How does the SCS curve number method work in practice? I will first walk you through the theory of the meth

the goal of the method is to determine for discrete time steps the proportion of rainfall that becomes runoff.

This is depicted schematically in the graph at the top right, which shows the rainfall rate P on the y-axis in units
of, for example, mm per hour or inches per hour. The x-axis shows time. At the start of the rainfall event, some
rain goes into storage without any generation of direct runoff. This could be interception storage and is termed
as the initial abstraction Ia. Once that initial store is full, rainfall is divided into infiltration F that lies below the
dashed whblack line and direct runoff Q above the line indicated by the hashed area. This is the effective
precipitation that we are interested in. We can also see in the graph that the infiltration rate is initially higher
and declines progressively with time as we would expect from our knowledge of soil physics. Mathematically,
the relationship between the different variables can be expressed by the continuity equation.

P  I a  F  QD

If we now assume proportionality between infiltration and the storage capacity of the catchment or
subcatchment, we can postulate that the duration between F and S is the same as the duration between direct
runoff and the total precipitation, less the initial losses.

The assumption that the initial loss or initial abstraction is a fixed percentage of the storage capacity removes
the initial abstraction from the equation as shown here.. With this assumption of the initial abstraction to be
20%, only the storage capacity S remains unknown.
S is related to a normalized number, the curve number, Cn, by the equations
. From that equation we can further delineate that the upper limit of the storage capacity is obtained for a Cn
value of 1 and that the storage capacity is 0 for a Cn value of 100. This is illustrated here by the plot of direct
runoff as given by the equation derived previously as a function of rainfall depth on the x-axis. Each line
represents the evaluation of the equation for a fixed curve number. For a Cn value of 100, the storage capacity is
0 and all rainfall becomes runoff. Therefore, the function runs along the 1 to 1 line in this graph. With decreasing
Cn value, which means increasing storage capacity, more and more rainfall goes into storage and less rainfall
becomes direct runoff.
  The initial loss is increasing with decreasing curve number since the initial abstraction is at fixed 20% of
the storage here. This is the reason why the curve numbers with a lower Cn value take off from the x-
graph later than the curve with a higher Cn value. The curve number for a given area is determined by
soil type, vegetation and other criteria.
 In the eqn if u put 0 for storage u can see we get 100 as cn [Link] when the curve is decreasing due to
the initial abstraction at the beginning of rainfall the curve starts after some rainfall depth is over

Another important factor of the estimation of the curve number is the initial saturation of the catchment. Three
antecedent soil moisture conditions or AMC classes are distinguished that signify dry average and wet
conditions respectively. These classes are defined by the amount of rainfall that has been recorded in the five
days prior to the particular precipitation event under consideration. Further, it is distinguished whether the
precipitation event occurs during the growing season with crops on the peels or in the dormant season where
crops are largely absent and the soils are bare. The rationale is that during the dormant season less rainfall is
required for saturated conditions because there is no interception by the vegetation and no plant water uptake
from the soils. It is important to note that the CN value provided in the tables for the different land cover and
soil types are all given for the average antecedent moisture conditions or in other words for the AMC class 2.
They are denoted here as CN subscript 2 values. For particularly dry or particularly wet conditions, these CN
values have to be adjusted. The corresponding empirical equations for CN1 dry conditions and CN3 wet
conditions can be found here on the slide. The CN1 value is always consistent with the fact that a catchment that
is dry yields less direct runoff and a saturated catchment yields more direct runoff compared to the direct runoff
under average saturation.

The question now is how to determine that number. Based on experimental data from the 1970s and 80s, the US
Soil Conservation Service has developed a series of tables for Cn values for different land use classes, soil types
and hydrological conditions. Here is one example. The left column lists the land cover type. The four columns on
the right signify four soil groups. They range from sands in class A to clay soils in class D. Some of the land cover
types are subdivided by three hydrological conditions, which describe essentially how well the soils are
draining, but subsume several factors such as the tendency to soil cracking or sealing or the depth to
underlaying confining soil layers or rock. Poor denotes factors impairing, hindering infiltration, and good
denotes factors enhancing infiltration. Fair is something in between. Curve numbers are increasing with soil
type class, as the sand has a higher infiltration capacity as a clay. Curve numbers are decreasing with the
hydrological condition, since infiltration into a poorly drained sand is smaller than infiltration into a well
drained sand. For example, a pasture under a poorly drained silt that has a soil class A has a higher CN value, 68,
then a pasture under well drained sand that has a curve number of 39, which correctly indicates that the storage
capacity for the well drained pasture is larger and the direct runner from it is smaller. There is a very long list
with soil profile data from all over the United States in which the corresponding soil classes can be found. The
four soil classes are called hydrological soil groups or HSG, which can also be related to soil texture classes as
depicted in the slide. In fact, this reduced table is used worldwide for soils outside of the U.S. However, if you
have looked closely at the table on the previous slide, you might have noticed that the jumps of CN values
between soil classes are sometimes very large. And since the CN value is very sensitive to the computed direct
runner, some researchers have looked into a finer discrimination of texture classes

Once the final CN value is determined, it can be used to compute the storage capacity and the effective
precipitation by the equations previously described. For further detail on the SCS method, I like to refer you to
the USDA Handbook TR50 Pi. It is now relatively old but still the reference for the SCS method and it is freely
accessible. A few remarks on the applicability of the SCS method. The main advantage is that it is easy to use and
has low data requirement. Land use and soil type maps are typically available in electronic form. It is a useful
engineering approach particularly for engaged catchment. On the downside, it has been shown that the CN
values are between 10 and up to 30% underestimated for some parts of Europe, which means that the effective
precipitation is underestimated too. This is a problem because it underestimates direct runoff and thus
potentially underestimates the risks of flooding. Some studies suggest that the initial abstraction of 20% is too
high and I like to remind you on the large uncertainty of that number that could be seen in the data. Another
disadvantage of the method was found in a study by Forton quite some time ago. He stated that a major
weakness is the sensitivity of estimated runoff to errors in the selection of the curve number. Changes of about
15 to 20% in the curve number doubles or halves the total estimated runoff. This is indeed a major problem and
therefore the sensitivity of the result to choices in the curve number should always be analyzed and
appropriately documented. In the following slides, I like to walk you through an example to compute effective
precipitation. The first task is to calculate the ECN value for a given catchment, then to use classification and soil
type is also given. The precipitation of the 5 days prior to the event was given to 38 mm in this case. The second
question is how does the ECN value change if the precipitation of the 5th preceding days is assumed to be 76
mm? So that means what is the ECN value if the catchment is saturated? And finally, the third task is to calculate
the cumulative and actual effect of precipitation for the scenario and average initial soil moisture. We will start
with the first task, which is the computation of the ECN value. In the catchment given here, we find 5 different
land use types as listed in the first column of the table. We also have given the fraction of the total catchment
area for each land use type. The first task then is to look up the USDA tables for the land use and hydrological
conditions. For convenience, the corresponding CN values are listed in the table for the 4 soil classes. On the
previous slide, we learned that the entire catchment is covered with the same soil type, which is a silt loam that
falls into the hydrological soil class B. Now we have different land use types in the catchment, which have
distinctively different CN values. To obtain a single representative CN value for the entire catchment, we have to
determine the CN values for the individual land use types and average them proportionally to their respective
fraction of the total catchment area. We can We can We total catchment total catchment the first equation under
the table. As a result, we obtain a rounded CN value of 72, which we now can plug into the equation to compute
the storage capacity. This yields a value of s equals 98.8 millimeters. Further, we assume in this example that the
initial loss is 10% of the storage capacity, which allows us to obtain a value for IA, namely 9.9 millimeters. And
with that, we have completed the first task. The obtained CN value of 72 was for the average AMC class. In the
second task, we have to determine the CN value for wet conditions, which is done simply by inserting the CN2
number in the appropriate equation shown here. The resulting CN number is 86, and correspondingly, the
storage capacity and initial abstraction becomes smaller compared to the consideration of average soil moisture
conditions, which would eventually result in larger values of direct runoff. And with that, we have completed
also task 2. Now we have all information available to compute the effective precipitation, which is the third task.
The precipitation event is given in tableau form here as a times series of half hourly values. The equation for the
computation of the effective precipitation is shown in the blue boxes, and the power meters s and IA are listed
again in the little table on the right side of the slide. Since the equation for PE, the effective precipitation, holds
for the total values of precipitation, we need to compute the totals for each time step. Therefore, we first
accumulate the precipitation values that are given for the individual time steps. Then we can apply the equation
on the right to obtain the cumulative effective precipitation P subscript EC. Here we have to distinguish two
cases. If the cumulative precipitation is smaller than the initial abstraction, the effective precipitation is zero. If
it is larger, then we can apply the standard equation. In our example, the cumulative precipitation exceeds the
initial abstraction only in the fifth time step at 8.30pm. In the final step, we have to disaggregate the cumulative
effective precipitation values for each time step, simply by subtracting the value of the previous time step as
shown in the equation at the bottom right and in the last column of the table. This is then our final time series of
effective precipitation, which can also be plotted, for example as a bar plot, together with the rainfall input time
series. In the example presented here, the effective precipitation is only a small proportion of the rainfall, as the
storage capacity of the catchment was relatively large. This is essentially the procedure to quantify a runoff
formation with the SCS curve number method. It allows the quantification of how much rainfall becomes direct
runoff. The timing of when the runoff arrives at the catchment outlet is dealt with in the next topic, which is
runoff concentration.

NEXT

 Flow routing is the third phase of the hydrological systems approach. It is required to merge the runoff
distributions from different subcatchments and to describe in-stream processes such as wave
propagation, attenuation and retention.

 Wave attenuation in streams can be best illustrated by a flat wave that propagates along the flow channel
downstream.
The flow experiences resistance due to friction at the channel bed, which in itself leads to an attenuation and
elongation of flood waves as they travel downstream. In a big flood, however, the river often exceeds the
channel bed. Flood planes and otherwise flooded areas have typically a higher surface roughness than the
channel bed and thus a higher resistance to flow. Enhancing flow resistance by vegetated flood planes as well as
the creation of additional flood areas are environmental friendly solutions for flood protection by enhanced
wave attenuation. From a technical point of view, bed resistance and retention capacity of a river reach are not
constant with the level of the flow and essentially depend on the river stage. This leads to unsteady flow
conditions. Wave attenuation can be visualized by plotting the flood hydrographs of a single event measured at
different locations along the river
A wide range of mathematical methods and concepts exist to describe flow routing. Empirical methods
comprise simple regression functions, statistical methods, or neural network approaches. Conceptual models
are widely used. Linear storage can describe floodplain retention and peak attenuation, and translation
functions at time delay. The Masttinger method is a popular approach for conceptual modelling and assumes a
linear relationship between the storage in a river section and the inflow and outflow of a particular river reach.
Flow routing concepts can also be applied to grid-based models, such as the mis-scale hydrological model MHM,
depicted schematically here. The left picture shows the model grid and overlays the digitalised river network,
including floodplain areas. The individual cells are then linked to a routing scheme depicted on the right, but the
true channel properties, like river length and floodplain area, are preserved as parameters for each cell in the
routing scheme. In addition to empirical and conceptual routing schemes, we can also use physical descriptions
of the flow, such as the shallow water equation. They consist of a continuity equation for flow, depicted on the
right, and the energy balance equation shown on the left. The first two terms of the energy balance equation are
the acceleration terms. They are often neglected, and the resulting form is called the diffusive wave equation. If
additionally the hydrostatic pressure is neglected, it is assumed that the friction slope equals the bed slope,
which simplifies the equations a lot and leads to the kinematic wave equation. In general, the mechanistic
approach to flow routing requires a lot of data, not only but also because it is spatially explicit. This is the reason
why these models are typically not used in conjunction with conceptual hydrological models. However, they are
advantageous for more detailed investigations, such as related to flood inundation, for example, or to inversely
estimate flood peak discharge from observed water levels. If such a detailed model is required, we can choose
from a range of software packages, such as the popular and freely available HECRA software, which is designed
to perform one and two-dimensional hydraulic calculations for a network of natural and constructed channels.
And with that, I like to conclude the topic of flow routing.

NEXT

However, the first choice of rainfall-runoff models are usually lump-conceptual models. They employ only a few
parameters, are computationally fast and require only rainfall and a few other methodological variables as
input. A potentially higher accuracy can be achieved with distributed conceptual rainfall-runoff models, which
account for different hydrological responses in different parts of the catchment. But the data requirements and
computational effort can quickly grow. Empirical rainfall-runoff models, such as neural network approaches,
gain again on popularity, but the major disadvantage is that they can't, at least so far, extrapolate beyond the
range of the training data, and that they require a lot of data to be trained. Mechanistic models are usually too
data-hungry and too computationally slow to be applied solely to describe the rainfall-runoff relationship,
although they have the best extrapolation ability when it comes to scenario simulation. Stochastic models are, in
a strict sense, not really a separate model type, but rather a modeling technique that is used in conjunction with
a model and allows the uncertainty quantification of model outputs by random variation of inputs or uncertain
parameters. In this respect, stochastic simulation with conceptual models can be quite useful. Here is an
example for the viral river catchment in New Zealand, which is reasonably large at 3430 square kilometers. The
lumped two-store four-parameter model G R4J was applied to predict river flow at the catchment outlet and
despite its simplicity it reproduces historic flow data relatively well. The model components are schematically
shown here on the right. There is the production store or soil moisture accounting store S and the routing store
R. The inputs of precipitation and potential avalputation are required to determine the net rainfall PN which is a
similar concept to the effective precipitation. There is also a linear routing scheme for the percolation with unit
hydrograph. This illustrates the similarity of the method and also the diversity as literally dozens of lumped
hydrological models have been developed and are freely available. The lumped models have their limitations
though. Looking closely the model simulations deviate from the observations for some situations particularly
for low flow periods. If these are the purpose of the model simulations the model would have to be adjusted
accordingly for example by a different target in the calibration or by separating the relatively large catchment
into a few smaller units of similar hydrological response. As a brief summary on the topic I would like to remind
you that models are never 100% accurate but that they should be useful. If it comes to rainfall runoff models
there are a few established ones that can be tested. The model choice and calibration should be aligned by the
model purpose and rainfall runoff models also don't serve a single purpose. It is a good practice to start simple
and add complexity as needed. However modeling is not like following a cooking recipe where the individual
steps are listed and you just have to follow through and the result will be more or less fine. Modeling is also a
kind of an art and sometimes it requires a few trials to get it right. With these remarks I like to conclude the very
brief section on rainfall runoff models and also the topic runoff as a whole. After the topics of precipitation and
runoff the next topic deals with the remaining component of the long term water balance which is
evapotranspiration.

NEXT

The second phase in the hydrological systems approach to quantify the processes that turn precipitation into
runoff is the phase of runoff concentration. Here we see again the schematic of these three phases. Whereas in
the first phase, the runoff formation is estimated how much effective precipitation is ending up as a direct runoff
in the stream.

The second phase aims at describing the timing of the runoff at the catchment outlet, the shape of the
hydrograph.
 Runoff concentration is the stream flow response that results from the effective precipitation, which
moves from the different contributing areas along quick surface and subsurface flow path to the stream
and forms a characteristic flow hydrograph at the catchment outlet.

 It should be mentioned here that the base flow component is also part of the hydrograph and must be
estimated by separate calculations.

1.
 Several factors influence the shape of the hydrograph and I will subsequently illustrate some of these
factors. The first is the non uniform temporal distribution of precipitation
Here we see two variants of the same total precipitation. In the case A, the majority of the rain falls in the first
time step followed by smaller amounts in the second time step. In case B, it is the other way around. Rainfall
goes into initial losses by wetting during the first time step and thus dampens the runoff response of case A and
delays the response in case B.
Correspondingly, the runoff peak of the dark blue hydrograph that belongs to case B arrives later at the
catchment outlet and is higher because the larger precipitation in the second time step hits a catchment that is
already wet. For completeness, the dashed line exhibits the uniform case when the precipitation in both time
steps is equal.
2.

 The second factor is the non uniform spatial distribution of the precipitation across the catchment. Here,
three cases are distinguished. In the cases A and B, it rains on the upper and lower half of the catchment
area respectively as illustrated by the colored area in the graphs on the left. In case C, the entire
catchment is covered by the rain which also means that the catchment receives twice as much rainfall as
in the other two cases. Cases B and C have the same onset of the runoff hydrograph as in both cases the
same flow path in the lower parts of the catchment are activated.
 But since the flow from the upper catchment is missing in case B, the corresponding dark blue
hydrograph peaks earlier and lower than the one of case C. The flow paths from the upper catchment to
the catchment outlet are longer and thus the response of runoff from case A is delayed and the peaks
slightly more damped compared to case B. This is shown by the light blue hydrograph.

3.

Precipitation events are often not stationary in space. They move with the fronts of cyclones for example or are
otherwise driven by winds.
The direction of these precipitation tracks is also influencing the timing of runoff at the catchment outlet. Here
we consider again three cases but the sum of the precipitation is the same for all three. In case A the rainfall
moves from the upper part of the catchment to what the catchment outlet. The rainfall takes the opposite
direction in case B and case C depicts a stationary rainfield. Again case B is similar to the stationary case
because the lower part of the catchment receives the rainfall first but B has a delayed peak and a longer tail of
the hydrograph because the rain arrives later at the upper part of the catchment. The discharge peak is again
more delayed in case A because of the travel time in the stream network from the upper parts of the catchment
that receives the rain first. However the peak is higher compared to the other two cases because the rain moves
in direction of the surface water flow and thus leads to a higher concentration of travel times along the various
flow paths in the catchment. T
4.
his also leads to a further impact. impact. impact. factor is the presence and location of urban areas in the
catchment. Runoff from urban areas is typically larger compared to runoff from natural catchments due to the
high percentage of sealed surfaces such as roads and pavements, roofs, parking lots etc. The natural condition is
schematically depicted in case D with no urban areas present. The hydrograph corresponding to a rainfall event
is shown by the green line. The three other hydrographs illustrate changes in the shape of the hydrograph if
urban areas are present in different parts of the catchment. If the urban area is located in the upper part of the
catchment, the flow peak gets more pronounced because the runoff from urban areas is larger, but the timing of
the peak remains largely unchanged. In contrast when the urban areas are located in the lower part of the
catchment, the flow peak arrives earlier because the lower part of the catchment drains faster. It can be noted
that generally the presence of urban areas or otherwise sealed areas enhance the flow peaks compared to
natural putty vegetated conditions.
5.
The shape and structure of the catchment is another factor that has a large influence on the shape of the
hydrograph. Narrow elongated catchments with very different flow paths and travel time lengths yield flat and
long flow hydrographs. In contrast, run catchments with uniform travel times from all locations of the
catchment yield spiky hydrographs with large peaks. The shape of the catchment can be characterized by
different form factors. Here a factor for the rundness, rb, is shown. It is simply the fraction of two numbers n1
and n2 that relate to the stream order system by Strala. And one and n2 donate the numbers of first and second
order streams in the catchment respectively. As a rule of thumb, the larger the fraction, the more first order
streams are present and the more elongated is the catchment. Round catchments exhibit a small rb number.
However, the structure of the stream network depends also on other factors than catchment shape, which also
have an influence on run-off concentration.

 the analysis of run-off times here shows typically some similarity between individual flood hydrographs
in a given catchment, which suggests that catchments have their own characteristic run-off response to
rainfall events.
 This characteristic response function is the link between effective precipitation and the run-off
hydrograph. Hence, it is also called the transfer function.
 A mathematical formulation of this function is needed to quantify run-off concentration.
 It could be any formal description like an empirical function or a mathematical model.
 In addition to the transfer function, a generalized mathematical concept is required that scales the run-
off response for arbitrary inputs of effective precipitation.
UNIT HYDROGRAPH
The underlying assumption of this concept is that a catchment exhibits a characteristic run-off response to a
unit of effective precipitation. Conceptually, the unit precipitation is applied instantaneously at time zero as
a delta impulse that integrates the unity. The impulse response H of t is the characteristic transfer function.
The law of mass conservation dictates that the integration of the response curve ht is also unity. The
response curve is therefore the hypothetical hydrograph that results from a unit of effective precipitation.
Correspondingly, the method is named the unit hydrograph method.

The unit hydrograph method is based on three fundamental concepts or assumptions. Stationarity,
proportionality and superposition. Stationarity means that the transfer function is time invariant. This is
illustrated by three unit impulses at t1, t2 and t3 in the graph to the left, and there are color-coded responses
in the graph to the right. The same impulse leads always to the same response function. It is only shifted by
the time between impulse and t3.
Proportionality denotes that the ordinate of the response function scale proportionally with the magnitude
of the input. The red curve is the response to a unit impulse that we have seen in the previous slide. If the
impulse is doubled, the response is doubled as depicted by the blue impulse and the blue response function.
If the impulse is halved, the response is halved too as shown in green. The linear scaling preserves the mass
balance between input and response.
The third principle is superposition, which means that the total runoff response of the responses. In the
example shown here, we have two identical unit inputs at time t1 and t2 and the corresponding individual
responses in red and blue. The total response is simply the sum of the two curves that are shown here in the
black line.
With these three principles, we can process arbitrary input functions of effective precipitation and get the
corresponding runoff response as illustrated here. I like to note here that proportionality and superposition
form a linear model, which is of advantage when it comes to implementing the method as a computer
program.
To express the unit hydrograph method mathematically, we assume that the input function, the effect of
precipitation, is a series of impulses, each creating its own response. This is schematically depicted in the top
graph. The blue bar is a single impulse, p tau, in the input times here, and tau is the time of the start of the
impulse. The increment of the impulse is denoted by delta tau. The second graph depicts the unit response
function, or unit hydrograph, which we assume known just now, but I will come back to this shortly. To
compute the runoff response to the blue input impulse, we first need to shift the unit hydrograph along the
time axis by tau, as shown in the third graph. This is the principle of statuarity. Secondly, we need to scale
the unit hydrograph according to the magnitude of the impulse by multiplication with p tau and d tau. In the
case depicted here, the impulse was larger than unity, and the response function, shown in the blue shaded
area in the bottom graph, is accordingly scaled up. This is the principle of proportionality. We can do this
procedure now for all increments of detail of the input function to get the total runoff response to the time
series of effective precipitation. Mathematically, this can be expressed by the convolution integral depicted
on the right. It sums up all response functions for the time increment d tau, which expresses the principle of
superposition. When it comes to the practical implementation of the unit hydrograph method, we have to
convert the convolution integral into a discrete form. This means we discretize the input times series with
constant increments of delta t, which can be minutes, hours, or even days, and replace the integral by a
summation of the ordinate of the individual responses. The discrete form of the convolution integral is
shown in the yellow box, where m donates the number of increments in the effective precipitation times
series, and n is the number of ordinate of the unit response function. I will demonstrate this here once more
on a brief example. Consider the first time step with effective precipitation P1 equals to 1. The unit
hydrograph has seven ordinate here, and the runoff response is exactly the unit hydrograph as depicted in
blue in the bottom graph. In the second time step, shown in green, the effective precipitation is twice as large
as in the first time step. The ordinate of the unit hydrograph are shifted by a time step to the right, and
scaled by a factor of 2 as depicted by the green line in the runoff response graph. The same procedure is
done in the third time step, where the effective precipitation is 1.5 times the effect of precipitation in the
first time step. Correspondingly, the response is shifted by two time steps to the right, and scaled by a factor
of 1.5. Finally, we sum up all the ordinate of the three response functions for each of the time steps as
defined by the discrete convolution integral, and we get the red curve, which is the composite runoff
response to the effective precipitation event. It should be clear that we need to evaluate the convolution
integral for m plus n time steps to get the full picture of the response.

With the unit hydrograph method, we have a quantitative method at hand to compute runoff concentration.
It can be used to transform other three times the use of effective precipitation into a runoff. However, one
piece of the puzzle is still missing. We have to specify the unit response function ht, which is specific for each
catchment.
One option is to derive the function graphically or mathematically from flow data of time series, and to
provide the ordinate of the unit hydrograph in tabular form. Another option is the application of
hydrological models. The most commonly used model class for that purpose are conceptual models. I have
introduced the linear storage model already in Lecture 3. Therefore, I will not repeat myself here. I only like
to reiterate that the linear storage model has one parameter, the storage coefficient k, which determines
how fast the water in the store is drained. And conveniently, there is an analytical solution to solve the linear
storage model. Now, let's have a closer look at the linear storage model. The first term of the analytical
 solution describes outflow from the reservoir without any inflows to the reservoir. The second term
describes the outflow considering the input function of effective precipitation p-tile. And if you look closely,
you can see here again the mathematical expression of the convolutional integral. To derive the response
function of the linear storage model to a unit impulse at t0, we first set the storage of the reservoir to unity.
From the storage equation depicted on the previous slide, it follows that the response of the reservoir at
time 0 qt0 equals to 1 over k. If we now insert 1 over k into the first term of the analytical solution at the top,
we obtain the runoff response from the linear storage model to a unit input, which is the unit response ht.
The response function only depends on the storage coefficient k and has the form of an impulse. The
function is plotted in the graph on the left. The exponential dk is well suited to describe the recession of
natural hydrographs, but the instantaneous jump at t0 is probably not very realistic and doesn't look like a
typical runoff response. To circumvent this problem, we can put several linear reservoirs in series. The
outflow from the first reservoir is the input to the second reservoir and so on. This method is called a linear
storage cascade and the unit response function of the storage cascade is shown by the second equation on
the right. Assuming that all reservoirs have the same storage coefficient k, the model has only one more
parameter, n, which is the number of linear reservoirs in the cascade. It should be noted that to be consistent
with the unit impulse concept, only the first store is full at t0 and all other reservoirs are empty. The two
parameters of the storage cascade make the model very flexible to cover a wide range of response function
shapes. The number of stores n influences the shape of the hydrograph as depicted in the graph to the left,
where k is held constant. The blue line shows the exponential output from a single reservoir. With increasing
number of stores, the response is increasingly delayed and the peak response reduced. The storage
coefficient k determines the retention in the reservoirs and influences the magnitude and position of the
peak flow as illustrated in the graph to the right for a constant number of stores, two, and different k values.
Smaller k values correspond to a smaller retention and yield larger peaks that arrive earlier as shown by the
blue line. The two parameters of the storage cascade can be estimated either directly from geomorphological
characteristics or indirectly from hydrograph analysis, for example by fitting it to historic flow data. The
storage cascade is just one example for the mathematical description of the transfer function ht. In principle,
any parametric function can be used in the unit hydrograph method to describe the characteristic response
of a particular attachment. This concludes the topic, runoff concentration. In summary, runoff concentration
aims at the mathematical transformation of the effective precipitation times years into a corresponding
times years of direct runoff. The transformation depends on many geomorphological and meteorological
factors and can be quantified with the unit hydrograph method and a specified transfer function, such as the
linear storage cascade or unit hydrograph ordinate directly derived from runoff times years. If these
concepts are applied to several subcatchments or hydrotobes, we have to merge the runoff contributions
from the individual areas and also consider potential in-channel processes. This is called flow writing and
the subject matter of the next topic.

NEXT