Dynamics and control of a periodically forced microalgae

culture
Carlos Martínez, Francis Mairet, Pierre Martinon, Olivier Bernard

To cite this version:

Carlos Martínez, Francis Mairet, Pierre Martinon, Olivier Bernard. Dynamics and control of a peri-
odically forced microalgae culture. 2018. �hal-01891648�

HAL Id: hal-01891648

[Link]
Submitted on 9 Oct 2018

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
 Dynamics and control of a periodically
forced microalgae culture ?
∗,∗∗
Carlos Martı́nez Francis Mairet ∗∗∗∗ Pierre Martinon ∗∗∗
Olivier Bernard ∗,∗∗
∗
Université Côte d’Azur, Inria, INRA, CNRS, UPMC Univ Paris 06,
BIOCORE team, France 1
∗∗
LOV-UPMC Sorbonne-CNRS, UMR 7093, Station Zoologique, B.P.
28, 06234 Villefranche-sur-mer, France
∗∗∗
INRIA Saclay , École Polytechnique, CMAP 91128 Palaiseau,
France. 2
∗∗∗∗
IFREMER PBA, Nantes, France 3

Abstract: Microalgae cultivation with wastewater is a promising way of reducing the energetic
needs for wastewater treatment and the costs of biofuel production. However, the very turbid
medium is not favorable for the development of microalgae. Indeed, light, the key element for
photosynthesis, rapidly vanishes along depth due to absorption and scattering. Therefore it
is crucial to understand the effects of the depth on turbid cultures. In this work, we study
theoretically the long-term behavior of a continuous culture of microalgae exposed to a periodic
source of light. By allowing periodic variations of the depth and the hydraulic retention time,
we show that the microalgae population is forced to a periodic regime. Finally, we address
numerically the problem of determining the optimal variations of the depth and the hydraulic
retention time for maximizing the productivity of the culture in the periodic regime.

Keywords: Dynamics and control; Industrial biotechnology; Wastewater; Biomass productivity;

Microalgae

1. INTRODUCTION 50 cm are generally recommended.

Microalgae grown as a by-product of wastewater treatment The objective of this paper is to give some insights in the
in high rate algal ponds (HRAPs) represents a great oppor- optimization of biomass productivity by playing with the
tunity for reducing costs of biofuel production (Park et al. depth of the culture and considering a natural light source.
(2011); Christenson and Sims (2011)). Moreover, HRAPs In our approach, we combine the modelling approaches of
show a great potential for nutrient removal in wastewa- Droop (1968) and Huisman et al. (2002) to build a model
ter treatment (Abdel-Raouf et al. (2012)). One problem accounting for light and substrate limitation. Our model is
with the cultivation of algae in wastewater systems is periodically forced by the incident light with period of one
that light, one of the main factors affecting microalgae day. We argue that by providing periodic input and output
growth, rapidly becomes limiting after its decrease due to flows to the system, any solution of the model reaches
absorption and scattering by algal cells and the high con- a 1-day periodic regime i.e. approaches asymptotically a
tent of particulate matter and colored substances (Morel periodic solution. Thus, we study the productivity of the
and Bricaud (1981); Borowitzka (1998)). We refer to the system in a periodic regime by controlling the input and
light extinction due to all non-microalgae components as output flow rates, or equivalently, the depth of the system
background turbidity. In the standard analyses of microal- and the hydraulic retention time.
gae cultures, background turbidity is neglected. However,
Martı́nez et al. (2018) showed theoretically that back-
ground turbidity results in a reduction of the productivity This article is organized as follows. In section 2, we intro-
when increasing the depth of the system. Indeed, according duce the model of a microalgae culture given by a system
to Larsdotter (2006), depths of the HRAP between 15 and of nonautonomous differential equations. In section 3, we
study the dynamics of the model. The main result of this
section (Theorem 2) gives sufficient conditions to ensure
? This work was supported by the CONICYT doctoral grant (Carlos that any solution approaches asymptotically to a periodic
Martı́nez), and by the Phycover (ANR-14-CE04-0011) and IPL- solution characterized by the presence of microalgae (i.e.
Algae in silico projects. different from the washout). This result follows from reduc-
1 INRIA Sophia Antipolis, 2004, route des Lucioles ing the model to a cooperative system and following some
BP 93, 06902 Sophia Antipolis Cedex, France (e-mail
ideas of Smith (1997). In section 4, we state an optimal
{[Link],[Link]}@[Link])
2 [Link]@[Link] control problem for maximizing the microalgae produc-
3 [Link]@[Link] tivity per day in the periodic regime. Periodic optimal
problems have already been studied by Grognard et al. We note that d is the unique cubic function satisfying
(2014); Bayen et al. (2015). Here, we present some numer- d(q0 ) = 1 , d(qL ) = 0 , and d0 (q0 ) = d0 (qL ) = 0. (7)
ical solutions obtained with the software BOCOP. Finally,
we illustrate the asymptotic behavior of the model when The volume (V ) of the culture varies with the supply flow
repeating each day the 1-day periodic optimal control. rate (Qin ) and the withdrawal rate (Qout ) according to
V̇ = Qin − Qout . (8)
2. MODEL DESCRIPTION
The transversal area (A) of the culture is constant in time
Let us consider a perfectly mixed microalgae culture of and along the depth of the culture. After dividing both
depth L illuminated from above with an incident light sides in Eq.(8) by A, we obtain the following equation for
Iin . We assume that light is attenuated exponentially the depth of the culture
according to the Lambert-Beer law i.e. at a distance z ∈
[0, L] from the illuminated surface, the corresponding light L̇ = Fin − Fout , (9)
intensity I(x, z) satisfies
with Fin := Qin /A and Fout := Qout /A linear flows. Mass
I(x, z) = Iin e−(kx+Kbg )z , (1) balances in the culture give the following equations (see
Chapter 2 in Dochain (2013) and Bougaran et al. (2010))
with x the microalgae concentration, k > 0 the specific Fin
light attenuation coefficient of microalgae, and Kbg ≥ 0 the ẋ = [µ̄(t, x, q, L) − ]x
background turbidity. We assume that microalgae growth L
is limited by light and a substrate s. Based on the model q̇ = ρ(q, s) − µ̄(t, x, q, L)q
of Droop (1968), we assume that the specific growth rate µ (10)
depends on an internal cell quota q. The latter corresponds Fin
to an internal pool of nutrient per unit of biomass. The ṡ = (sin − s) − ρ(q, s)x,
L
light intensity effects are described by a Monod type
function. Thus, the specific growth rate is taken to be

q0

I where sin is the (constant) substrate influent concentra-
µ(I, q) = µmax 1 − − r, (2) tion.
q KI + I

where r > 0 is the respiration rate, q0 > 0 represents 3. PERIODIC SYSTEM AND ASYMPTOTIC
the value of q at which growth ceases, KI > 0 is a half- BEHAVIOR.
saturation constant, and µmax > 0 is the maximal specific
growth rate. Following Huisman et al. (2002), we compute We assume that the incident light is variable in time
the average growth rate in the reactor, denoted µ̄, by according to
integrating the local growth rates over all the reactor Iin (t) = Imax max{0, sin(2πt/ω)}2 , (11)
ZL
1 with ω > 0 the length of a day. Thus, by assuming that
µ̄(·) := µ(I(z), q)dz. (3) Fin and Fout vary periodically with a period ω, the system
L
0 (9)-(10) becomes an ω-periodic system.
A simple computation leads to
   We use the following notation for the time-average of an
µmax KI + Iin q0 ω-periodic function 4
µ̄(·) := ln 1− − r,(4)
(kx + Kbg )L KI + Iout (x) q Zω
1
< f >:= f (t)dt. (12)
with Iout (x) = I(x, L) the light intensity at the bottom of ω
the culture. 0

By integrating Eq.(9) we obtain that

Cell quota q decreases with cell growth and increases with
Zt
nutrient uptake, whose rate function is taken to be
( s L(t) = L(0) + [Fin (s) − Fout (s)]ds. (13)
ρmax d(q) if q ≤ qL , 0
ρ(q, s) = Ks + s (5)
0 if q > qL ,
We assume that
where ρmax is the maximal uptake rate of nitrogen, qL < Fin (t) − Fout (t) >= 0.
is the hypothetical maximal quota, and Ks is a half- Then L is an ω-periodic function.
saturation constant. d(q) ∈ [0, 1] is a down regulation
term. While d(q) is usually taken as a linear function
(Bougaran et al. (2010); Morel (1987)), here we take d(q) We define the total amount of limiting element both in the
as substrate and in the biomass by means of

qL − q
2 
qL − q q − q0
 S = s + xq. (14)
d(q) = +3 (6) 4
qL − q0 qL − q0 qL − q0 A function f is ω-periodic if f (t + ω) = f (t) for all t
A simple calculation shows that S satisfies the differential Since xs satisfies the differential equation
equation dxs

ρ(q, sin − xs )

= xs −D , (21)
Ṡ = D(t)(sin − S), (15) dt q

with D(t) : FL(t)

in (t)
known as the dilution rate. We can we have that xs (t) = x(t)q(t) ≤ sin for all t ≥ 0.
rapidly verify that S(t) → sin as t → ∞ for any value of Consequently, P is positively invariant with respect to Eq.
S(0). Thus, the solutions of Eq.(10) approach the surface (16). Thus, if x(0) > 0, then x(t) ≤ sin /q0 for all t ≥ 0.
This implies that the set K := {(x, xs ) ; x ∈ (0, ∞), q0 <
sin = s + xq. xs /x, xs ≤ sin } is positively invariant with respect to
We take advantage of this by dropping the equation for s (20). Consequently, any solution of (20) starting on K is
and replacing s in Eq.(10) by s = sin − xq in the second bounded. Thus, any solution of Eq.(20) starting on K,
equation. This results in the two-dimensional system or any solution of Eq. (16) starting on P , approaches a
ẋ = [µ̄(t, x, q) − D(t)]x periodic solution. The following theorem gives conditions
. (16) to avoid the washout (0, q ∗ ) as an attracting periodic
q̇ = ρ(q, sin − xq) − µ̄(t, x, q)q solution.
Theorem 2. If < µ̄(t, 0, q ∗ (t)) − D(t) > is greater than 0,
In the following we study the asymptotic behavior of then any solution of Eq. (16) starting in P is asymptotic
Eq.(16). Care must be exercised in extrapolating results to an ω-periodic solution of Eq.(16) with positive x-
obtained from the analysis of Eq.(16) to analogous results component.
for Eq.(9)-(10). A discussion about this can be found in
Smith (1997). Proof. We use the Floquet theory (see Chapter 7 in the
book of Sideris (2013)) to determine the stability of the
washout. The characteristic multipliers of the washout
As s must be non-negative, initial conditions for Eq. (16) (0, q ∗ ) are given by ρ1 = exp(ω < a(t) >) and ρ2 =
must satisfy x(0)q(0) ≤ sin . We note that if x(0) = 0 (resp. exp(−ω < c(t) >) with
x(0) > 0), then x(t) = 0 (resp. x(t) > 0) for all t ≥ 0 (see
Zanolin (1992) to find rigorous arguments). We also note a(t) = µ̄(t, 0, q ∗ (t)) − D(t), (22)
that q(0) ≥ q0 implies q(t) ≥ q0 for all t ≥ 0. Therefore we
are only interested in solutions starting on the set and
P := {(x, q) ; x > 0, q ≥ q0 , and xq ≤ sin }. ∂ρ ∗
(17) c(t) = − (q (t), sin ) + µ̄(t, 0, q ∗ (t))
∂q (23)
A solution (x, q) of Eq.(16) will be called an ω-periodic ∂ µ̄
+q ∗ (t) + (t, 0, q ∗ (t)) > 0.
solution provided each component is ω-periodic. There is ∂q
a trivial periodic solution corresponding to the absence of
microalgae in the chemostat. Putting x = 0 in the second Thus, ρ1 > 1 and ρ2 < 1, and the washout is not stable.
equation of (16) results in This implies that not all the solutions of Eq.(16) starting
on P are asymptotic to the washout. Since any solution
q̇ = ρ(q, sin ) − µ̄(t, 0, q)q. (18) converges to an ω-periodic solution, there exists an ω-
periodic solution different from the washout that attracts
The following proposition guarantees the uniqueness of the some solutions. Following the same arguments as in the
washout. proof of Proposition 1.3 in Smith (1997), we can conclude
Proposition 1. Eq. (18) has a unique periodic solution q ∗ that this is true for any solution starting on P . 
to which all solutions are attracted.
Note that Theorem 2 does not say anything about the
Proof. Since q̇ is strictly decreasing with q, we can use number of periodic solutions with positive x-component.
the same arguments as in the proof of Proposition 1.1 in
Smith (1997). 
4. OPTIMAL CONTROL PROBLEM
Thus, the trivial periodic solution is given by
We define the areal biomass harvest on the interval of time
x = 0, q = q ∗ (19)
[ti , tf ] as
The change of variables xs := qx leads the system (16) to Ztf
ẋ = [µ̄(t, x, xs /x) − D(t)]x H= Fout (t)x(t)dt. (24)
(20)
ẋs = ρ(xs /x, sin − xs )x − D(t)xs . ti

Eq. (20) is a cooperative system i.e. H represents the quantity of biomass that is harvested on
∂ ẋ ∂ ẋs the interval of time [ti , tf ] per unit of area.
, ≥ 0.
∂xs ∂x
The importance of this system lies on the fact that Here we consider the case where the chemostat is operated
any bounded solution of a planar, cooperative, periodic in long term, with a periodic daily biomass production
system of differential equations is asymptotic to a periodic from the reactor outlet. The problem that we consider is
solution ( see Chapter 7 in Smith and Waltman (1995)). the maximization of the biomass production over a single
day. The control variables are Fin (u1 ) and Fout (u2 ). The Table 1. Parameters
problem can be stated as Parameter Value Unit Reference
Zω µmax 1.72 d−1 Huisman et al. (2002)
max H(u1 , u2 ) = u2 (t)x(t)dt, KI 107 µmol m−2 s−1 Huisman et al. (2002)
k 0.2 m2 gC −1 Béchet et al. (2013)
0 ρmax 11.2 gN gC −1 d−1 Bougaran et al. (2010)
s.t. ẋ = [µ̄(t, x, q, L) − u1 /L]x, gN gC −1
(25) qL 0.28 Bougaran et al. (2010)
q̇ = ρ(q, sin − xq) − µ̄(t, x, q, L)x, q0 0.07 gN gC −1 Bougaran et al. (2010)
L̇ = u1 − u2 Ks 0.0007 gN m−3 Bougaran et al. (2010)
x(0) = x(ω) > 0, q(0) = q(ω), L(0) = L(ω) r 0.1 d−1 -
u1 (t) ∈ [0, Fin,max ] , u2 (t) ∈ [0, Fout,max ] sin 10 gN m−3 -
Imax 2000 µmol m−2 s−1 -
Kbg 0 or 10 m−1 -
In the formulation of our problem we use Eq. (16) together
Fin,max 1000 L m−2 d−1 -
with Eq. (9).
Fout,max 1000 L m−2 d−1 -

4.1 Numerical results

We solve numerically the problem (25), with a direct 40

Microalgae (gC/m3 )
method implemented in the sofware BOCOP (Bonnans
et al. (2011)) (version 2.10). The problem is discretized
by a two-stage Gauss-Legendre method of order 4 with 35
200 time steps. We consider a constant initialization, and
the tolerance for IPOPT NLP solver is set at 10−12 .
30
We consider two microalgae cultures; culture 0 without
background turbidity i.e. Kbg = 0, and culture 1 with a A
background turbidity Kbg = 10 m−1 . For both cultures 25
the rest of parameters are given in Table 1. The value of 0 0.2 0.4 0.6 0.8 1
Kbg for the culture 1 was set as high as the numerical
method allowed. We were not able to obtain convergence 0.29
for values of Kbg equal to or higher than 11 m−1 . Huisman
Quota (gN/gC)

et al. (2002) reported a background turbidity of 7.2 m−1

for a nutrient-rich medium used in a laboratory culture. 0.28

Figures 1 and 2 show the results. In both cultures, algae 0.27

concentration increases during half or almost all the day
(first half of the period), then decreases at night (see B
Figure 1A). We note that in culture 1 the microalgae 0.26
concentration is higher, and consequently the nutrient in 0 0.2 0.4 0.6 0.8 1
the medium is lower (see Figure 1C). However, the daily
microalgae productivity is 1.7 gC m−2 in contrast to the 2
culture 0 that reaches a daily microalgae productivity of
Substrate (gN/m3 )

4.0 gC m−2 (i.e. 8.0 gDW m−2 ). 1.5

The control structures are Bang-Singular (u1 ) and Bang- Culture 0
Bang (u2 ) for both cultures (see Figure 2A). These controls 1 Culture 1
result in a different depth for each culture. Figure 2B shows
that there is a difference of almost 40 cm during all the 0.5
period of 24 hours with a lower depth for the culture that C
accounts for background turbidity. 0
0 0.2 0.4 0.6 0.8 1
Consider the controls (u1 , u2 ) (see Fig. 2A) obtained from
Time (d)
the optimization of one periodic day-night cycle. We apply
the respective control to each culture (i.e. to solve Eq.(9)-
(16)) repeatedly for a long interval of time. The initial
depth for each culture is equal to the initial depth of the Fig. 1. Optimal solutions for problem (25) with two differ-
respective optimal solution (see Figure 2B). The initial mi- ent background turbidities. A. Microalgae concentra-
croalgae concentration and the initial quota are 5 gC/m3 tion. B. Cell quota. C. Substrate concentration in the
and 0.28 gN/gC respectively for both cultures. Figures 3 medium.
and 4 show that the microalgae concentration approaches culture 0.
the optimal periodic microalgae concentration (Fig. 1A)
found previously. For the culture 1 the microalgae concen-
tration approaches more slowly to the periodic solution. By using the solver bvp4c (Shampine et al. (2000)) in
Indeed, it needs 50 days in contrast to 30 days for the Matlab we can determine numerically q ∗ and evaluate the
 40
1000
35
Control (L/m2 d)

500
u1 30
0
25
-u 2
-500 20
A
15
-1000
0 0.2 0.4 0.6 0.8 1 Evolution of x
10
Optimal periodic solution

0.6 5
0 10 20 30 40 50 60 70
Time (d)
0.5
Depth (m)

Fig. 4. Comparison of the optimal periodic microalgae

0.4 Culture 0 concentration and the evolution of the microalgae
Culture 1 with a low initial concentration when the periodic op-
timal control is applied all the time. With background
0.3
turbidity.
B
5. CONCLUSIONS AND FUTURE WORK
0.2
0 0.2 0.4 0.6 0.8 1
A first series of numerical simulations using a direct
Time (d) method illustrates the impact of the background turbidity
on the choice of the depth of the culture for maximiz-
Fig. 2. Optimal solutions for problem (25) with two ing the microalgae productivity. We have shown that a
different background turbidities. A. Control variables. background turbidity of 10 m−1 can result in a reduction
B. Depth of the culture. of 40 cm of the optimal depth and a productivity loss
of 55 %. Thus, the background turbdity should not be
neglected. The choice of the depth is not trivial and should
40 be adapted to the overall turbidity, in particular when
both light and nutrient limitations can occur. This choice
35 should also include thermal considerations, since low depth
induces lower thermal inertia. Processes with low depth
30 are likely to overwarm very rapidly in hot seasons (De-
Luca et al. (2017)) and a trade-off between light and
25 temperature performances should be found.

20
We also studied theoretically the long-term behavior of
15 the microalgae population when the system is operated
periodically in agreement with the period of the light
10 Evolution of x source. We established sufficient conditions to ensure that
Optimal periodic solution any solution of the model, independently of the initial
5 conditions, is attracted to an oscillatory solution with a
0 10 20 30 40 50 positive microalgae concentration.
Time (d)
Photoinhibition (i.e. reduction of the photosynthetic rate
Fig. 3. Comparison of the optimal 1-day periodic microal- due to high light intensities) should be included in a future
gae concentration and the evolution of the microalgae work. Low depths of the culture could enhance the pho-
with a low initial concentration when the periodic toinhibtory process resulting in a lower productivity, and
1-day optimal periodic control is applied. Culture for low initial microalgae concentrations the population
without background turbidity. could go extinct (see Gerla et al. (2011); Martı́nez et al.
(2017)).
sufficient condition given in Theorem 2. For both cases,
we have that < µ̄(0, q ∗ (t), t) − D(t) >≥ 0.3, thus, any
solution to Eq. (16) starting in P should converge to an ω- In this work, we emphasized the productivity of microalgae
periodic solution of the system with a positive microalgae using wastewater streams without paying attention to the
concentration. aspect of wastewater treatment. In a future work, other
optimization problems will be studied. For example, the application to specific absorption of phytoplankton.
maximization of the removal of nitrogen with restrictions Deep-Sea Res, 28(11), 1375–1393.
over the volume of wastewater to be treated. Morel, F.M. (1987). Kinetics of nutrient uptake and
growth in phytoplankton 1. Journal of phycology, 23(1),
137–150.
REFERENCES Park, J., Craggs, R., and Shilton, A. (2011). Wastewater
treatment high rate algal ponds for biofuel production.
Abdel-Raouf, N., Al-Homaidan, A., and Ibraheem, I. Bioresource technology, 102(1), 35–42.
(2012). Microalgae and wastewater treatment. Saudi Shampine, L.F., Kierzenka, J., and Reichelt, M.W. (2000).
journal of biological sciences, 19(3), 257–275. Solving boundary value problems for ordinary differen-
Bayen, T., Mairet, F., Martinon, P., and Sebbah, M. tial equations in matlab with bvp4c. Tutorial notes,
(2015). Analysis of a periodic optimal control problem 2000, 1–27.
connected to microalgae anaerobic digestion. Optimal Sideris, T.C. (2013). Ordinary differential equations and
Control Applications and Methods, 36(6), 750–773. dynamical systems.
Béchet, Q., Shilton, A., and Guieysse, B. (2013). Modeling Smith, H.L. and Waltman, P. (1995). The theory of
the effects of light and temperature on algae growth: the chemostat: dynamics of microbial competition, vol-
state of the art and critical assessment for productivity ume 13. Cambridge university press.
prediction during outdoor cultivation. Biotechnology Smith, H. (1997). The periodically forced droop model
advances, 31(8), 1648–1663. for phytoplankton growth in a chemostat. Journal of
Bonnans, J., Grelard, V., and Martinon, P. (2011). Bocop, Mathematical Biology, 35(5), 545–556.
the optimal control solver, open source toolbox for Zanolin, F. (1992). Permanence and positive periodic
optimal control problems. URL [Link] org. solutions for kolmogorov competing species systems.
Borowitzka, M.A. (1998). Limits to growth. In Wastewater Results in Mathematics, 21(1-2), 224–250.
treatment with algae, 203–226. Springer.
Bougaran, G., Bernard, O., and Sciandra, A. (2010).
Modeling continuous cultures of microalgae colimited by
nitrogen and phosphorus. Journal of theoretical biology,
265(3), 443–454.
Christenson, L. and Sims, R. (2011). Production and har-
vesting of microalgae for wastewater treatment, biofuels,
and bioproducts. Biotechnology advances, 29(6), 686–
702.
De-Luca, R., Bezzo, F., Béchet, Q., and Bernard, O.
(2017). Exploiting meteorological forecasts for the
optimal operation of algal ponds. Journal of Process
Control, 55, 55–65.
Dochain, D. (2013). Automatic control of bioprocesses.
John Wiley & Sons.
Droop, M.R. (1968). Vitamin b 12 and marine ecology.
iv. the kinetics of uptake, growth and inhibition in
monochrysis lutheri. Journal of the Marine Biological
Association of the United Kingdom, 48(3), 689–733.
Gerla, D.J., Mooij, W.M., and Huisman, J. (2011). Pho-
toinhibition and the assembly of light-limited phyto-
plankton communities. Oikos, 120(3), 359–368.
Grognard, F., Akhmetzhanov, A.R., and Bernard, O.
(2014). Optimal strategies for biomass productivity
maximization in a photobioreactor using natural light.
Automatica, 50(2), 359–368.
Huisman, J., Matthijs, H.C., Visser, P.M., Balke, H.,
Sigon, C.A., Passarge, J., Weissing, F.J., and Mur, L.R.
(2002). Principles of the light-limited chemostat: theory
and ecological applications. Antonie van Leeuwenhoek,
81(1-4), 117–133.
Larsdotter, K. (2006). Wastewater treatment with
microalgae-a literature review. Vatten, 62(1), 31.
Martı́nez, C., Bernard, O., and Mairet, F. (2017). Maxi-
mizing microalgae productivity by shading outdoor cul-
tures. IFAC-PapersOnLine, 50(1), 8734–8739.
Martı́nez, C., Mairet, F., and Bernard, O. (2018). Theory
of turbid microalgae cultures. Journal of theoretical
biology, 456, 190–200.
Morel, A. and Bricaud, A. (1981). Theoretical results
concerning light absorption in a discrete medium, and