Algal Research 17 (2016) 67–78

Contents lists available at ScienceDirect

Algal Research

journal homepage: www.elsevier.com/locate/algal

Dynamic model of an industrial raceway reactor for

microalgae production
I. Fernández a, F.G. Acién b, J.L. Guzmán a,⁎, M. Berenguel a, J.L. Mendoza c
a
Department of Informatics, University of Almería, CIESOL, International Excellence Campus ceiA3, E04120 Almería, Spain
b
Department of Chemical Engineering, University of Almería, International Excellence Campus ceiA3, E04120 Almería, Spain
c
Faculty of Engineering and the Environment, University of Southampton, Southampton

a r t i c l e i n f o a b s t r a c t

Article history: A dynamic model for microalgae production in raceway reactors is developed in this work. The model takes into ac-
Received 14 January 2016 count fluid-dynamic, mass transfer, and biological phenomena taking place in microalgae bioreactors. The model has
Received in revised form 27 March 2016 been calibrated and validated using real data from a 100 m2 pilot-scale raceway reactor. Results demonstrate that in
Accepted 23 April 2016
raceway reactors large accumulation of oxygen takes place into the channel whereas carbon losses into this section
Available online xxxx
are scarce, these phenomena influencing the overall productivity of the reactor. The model can be used to determine
Keywords:
characteristic parameters (biological and engineering ones) of existing reactors. Moreover, it can be used to simulate
Microalgae the effect of different designs and/or operation conditions into the performance of the system. Simulations allow us
Modeling demonstrating that to increase the productivity of raceway reactors it is recommendable to reduce the water depth
Fluid dynamic into the cultures and to increase the mass transfer capacity of the entire reactor.
Photobioreactors © 2016 Elsevier B.V. All rights reserved.
Raceway

1. Introduction
Microalgae have a large biotechnological potential for producing valu-
able substances for feed, food, nutraceutical, and pharmaceutical indus-
tries [1,2]. Furthermore, other applications can be attributed to the
photosynthetic process performed by these microorganisms such as CO2
mitigation, wastewater treatment and biofuels production, thus allowing
researchers to develop new biotechnological processes [3–6]. By these
reasons, the use of microalgae is receiving a lot of interest recently. Inde-
pendently of the final application, two types of photobioreactors are
mainly used to produce microalgae: (i) closed photobioreactors as tubular
or flat panels reactors, in which high-value products are produced by
strains highly sensitive to contamination; and (ii) open reactors as open
ponds and raceway reactors, simpler and less expensive ones where
contamination-proof strains can be produced [7,8]. In addition to contam-
ination risk, raceway reactors have additional drawbacks such as low bio-
mass concentrations, poor gas-liquid mass transfer, and lack of
temperature control. In general, raceways have a scarce control of operat-
ing conditions that diminishes the final productivity achieved; however
at commercial scale N 95% of algae production worldwide is performed
⁎ Corresponding author at: Department of Informatics, University of Almería, Carretera
Sacramento s/n, E04120 Almería, Spain.
E-mail addresses: ifernandez@ual.es (I. Fernández), facien@ual.es (F.G. Acién),
joseluis.guzman@ual.es (J.L. Guzmán), beren@ual.es (M. Berenguel),
jlmm1f10@soton.ac.uk (J.L. Mendoza).
in raceway reactors due to their low constructions cost, easy scale-up,
and low energy requirement [9].
Engineering aspects of raceway reactors were previously studied [10],
but now they are being characterized to improve the efficiency of this
technology [11–15]. In this sense, the fluid-dynamic and mass transfer ca-
pacity of raceway reactors as a function of design and operational condi-
tions is being optimized. It is important to note that the fluid-dynamic and
mass transfer capacity determine the evolution of culture parameters as
dissolved oxygen, pH and CO2 availability, all of them influencing the per-
formance of the cells, and thus the final productivity achievable.
Models combining the fluid-dynamic and mass transfer capacity of
raceway reactors with the biological performance of the cells under dif-
ferent conditions are very scarce. Till now, few steady-state models has
been reported considering the entire reactor as a stirred tank reactor,
thus evaluating the performance of the cultures as a function of average
value of culture parameters as light availability, pH, and nutrients con-
centration [16–19]. However, raceway reactors are plug-flow reactors
exposed to changing solar light, thus culture conditions change on
time and space inside the reactor. Due to this reason, dynamic models
that take into account the temporal-spatial distribution of culture pa-
rameters are necessary to adequately simulate this type of reactors.
Moreover, these dynamic models are necessary to optimize the design
and operation of the systems, helping to understand the different dy-
namics and phenomena taking place. Furthermore, these models can
be used as predictive and simulation tools in order to properly design
and operate these systems, as well as to design control strategies for op-
timal biomass production [20,21].

http://dx.doi.org/10.1016/j.algal.2016.04.021
2211-9264/© 2016 Elsevier B.V. All rights reserved.
68 I. Fernández et al. / Algal Research 17 (2016) 67–78
In summary, in the present work, a dynamic model of microalgae
production in raceway reactors is developed. The model is based on a
previously reported model for tubular reactors [22]. The model is
based on mass balances, transport phenomena, thermodynamic rela-
tionships, and biological phenomena taking place into the reactor,
thus being based on fundamentals principles instead of empirical equa-
tions. It takes into account the kinetic of different phenomena inside the
reactor, thus a complete dynamic simulation model can be obtained.
The model allows predicting the evolution of the main variables of the
system such as biomass concentration, pH, dissolved oxygen, and total
inorganic carbon in the liquid phase, in addition to oxygen and carbon
dioxide exchange for the gas phase. The model has been calibrated
and validated using experimental data from a 100 m2 pilot-scale race-
way reactor, and it was used to determine the influence of design pa-
rameters in the performance of the system. This model is presented as
a powerful tool for the optimization of design and operation of this
type of photobioreactors.
2. Materials and methods
2.1. Microorganism and culture medium
The microalga strain used was Scenedesmus almeriensis (CCAP 276/
24). This microalgae is characterized by a high growth rate, withstand-
ing temperature up to 45 °C and pH values up to 10, although its opti-
mum conditions are 35 °C and pH 8 [23]. Experiments were
performed using Arnon medium prepared using fertilizers instead of
pure chemicals.
2.2. Raceway reactor and operation conditions
The raceway reactor used is located at 36° 48′N–2° 43′W in Research
Center “Las Palmerillas”, property of Cajamar Foundation (Almería,
Spain). The reactor consisted of two 50 m length channels (0.46 m
high × 1 m wide), both of them connected by 180° bends at each end,
with a 0.59 m3 sump (0.65 m long × 0.90 m wide × 1 m deep) located
1 m part of the way down one channel (Fig. 1). The entire reactor, in-
cluding the sump, was made of white 3 mm thick fiber glass. The liquid
was circulated by a marine plywood paddle-wheel with 8 paddle, with a
1.2 m diameter, which is driven by an electric motor (Ebarba, Barcelona,
Spain) with gear reduction and speed control using a frequency inverter
Fig. 1. Raceway reactor scheme showing dimensions, position of pH-T and DO probes used for the calibration and validation of the model, and geometric calculation of the shadow
projection on the perpendicular axis of the walls. Numbers indicates the position where the probes were situated. (1) before paddlewheel, (2) after paddlewheel-before sump,
(3) after sump-beginning of the channel, (4) end of the right channel.
(Ibérica, S.A. Barcelona, Spain). The reactor can be divided in three main
parts depending on its fluid-dynamic characteristics (channels, paddle-
wheel and sump), such as was observed in Fig. 1. For this reason, three
pH-T and dissolved oxygen probes were situated at the end of each of
these parts, (5083 T and 5120, Crison, Barcelona, Spain) connected to
transmitters (MM44, Crison, Barcelona, Spain) and data acquisition
software (Labview, National Instruments, USA). Air or CO2 gas was auto-
matically injected at the bottom of the sump through a diffuser to con-
trol the dissolved oxygen and pH of the culture. The gas flow rate
entering to the reactor was measured by a mass flow meter (PFM
725S-F01-F, SMC, Tokyo, Japan).
Experiments were performed in semicontinuous mode. For this pur-
pose, the reactor was filled with Arnon medium up to 15 cm water
depth (15 m3 volume), prepared from fertilizers instead of pure
chemicals, and it was inoculated with a 10% total volume of culture
from a 3.0 m3 tubular photobioreactor. Then, it was operated in batch
mode for one week. After that, the reactor was operated in
semicontinuous mode at 0.2 day−1 this being previously demonstrated
as optimal for this reactor [15]. To operate in semicontinuous mode, a
fixed culture volume of 3.0 m3 was harvested and replaced daily with
fresh medium over 6 h in the middle of the daylight period.
Semicontinuous operation was maintained till steady state was
achieved; only data around steady-state conditions being used. Evapo-
ration (6–10 l/m2 day− 1) inside the reactors was compensated by
adding fresh medium, in addition to the volume of fresh medium used
for the reactor’ semicontinuous operation. The culture medium was
not sterilized, simply filtered before entering the reactors using
200 μm pore-size filters to remove solids.
3. Model
To model a bioprocess, both biological aspects of the microorganism
and the engineering aspects of the reactor have to be considered. In this
case, a biological model for a microalgae previously reported in [24], and
an engineering characterization of the raceway reactor used in [14,15,
25] were used. This previous knowledge was integrated into the dy-
namic model of photobioreactors previously developed in [22] to obtain
a complete model of raceway reactors. Moreover, since the model is de-
veloped for outdoor conditions, the classical knowledge about solar ra-
diation availability (sun position and solar radiation as a function of
location, date, day of the year and solar hour) was also incorporated.
 I. Fernández et al. / Algal Research 17 (2016) 67–78 69

Table 1 Table 1 (continued)

Variables, constants and characteristic parameters used into the model.
Variable Definition Value (units)
Variable Definition Value (units)
KlaCO2p Volumetric gas-liquid mass transfer coefficient for s−1
Dynamic variables CO2 in the paddle-wheel
Cb Biomass concentration kg m−3 KlaCO2s Volumetric gas-liquid mass transfer coefficient for s−1
[CO2] Carbon dioxide concentration in the liquid phase mol m−3 CO2 in the sump
[CO⁎2 ] Equilibrium concentration with the gas phase for mol m−3 KO2 Oxygen inhibition constant 0.8373 mol m−3
dioxide carbon n form exponent 1.045
2−
[CO3 ] Bicarbonate specie mol m−3 PO2,max Maximum photosynthesis rate 2.06⋅10−5 kgO2
[CT] Total inorganic carbon concentration mol m−3 kg−1 s−1
[HCO3−] Carbonate specie mol m−3 RO2 Respiration coefficient for dissolved oxygen 9.58⋅10−7 kgCO2
PCO2 Carbon dioxide consumption rate kgCO2 kg−1 s−1 kg−1 s−1
PO2 Photosynthesis rate kgO2 kg−1 s−1 RCO2 Respiration coefficient for carbon dioxide 4.28⋅10−7 kgCO2
Iav Average solar irradiance μE m−2 s−1 kg−1 s−1
I0 Solar irradiance on an horizontal surface μE m−2 s−1 Yb/O2 Biomass yield coefficient 0.7273 kg
[O2] Dissolved oxygen concentration mol m−3 z form parameter 5.4333
[O⁎2] Equilibrium concentration with gas phase for oxygen mol m−3
yO2 Oxygen molar fraction
YO2 O2 to N2 molar ratio in gas phase mol O2 (mol N2)−1
Notice that Table 1 contains the definitions, units, and values of the dif-
yCO2 Carbon dioxide molar fraction
YCO2 CO2 to N2 molar ratio in gas phase mol CO2 (mol N2) −1 ferent variables and parameters appearing in the model equations pre-
sented in this section.
Constants
C0 Drift flux model parameter 0.996
[CT]m Total inorganic carbon in the medium 3 mol m−3 3.1. Solar radiation model
h Liquid height 0.17 m
hc Wall height 0.46 m Solar energy is the driving force for microalgae-based processes op-
hs Solar hour h erating under autotrophic conditions, thus the photosynthesis rate and
hss Subterranean height of the sump h
finally the biomass production are a function of light availability, which
HCO2 Henry's constants for carbon dioxide 38.36 mol atm−1
m−3 depends on location and design of the reactor, in addition to day of the
HO2 Henry's constants for oxygen 1.07 mol atm−1 year and solar hour. Light availability in an horizontal surface can be
m−3 easily estimated using classical solar radiation equations. However, the
ls Length of the sump 1m
net amount of light received in raceway reactors is a function of its de-
m Form parameter 0.0021
MCO2 Molecular weight of carbon dioxide 32 g mol−1
sign, especially the walls shadow having a large influence. In this sense,
MO2 Molecular weight of oxygen 44 g mol−1 shadow is generated by the channel walls in each cross-sectional area of
N Day of the year d the reactor, depending on the sun position and reactor geometry. There-
[O2]m Dissolved oxygen in the medium 0.2812 mol m−3 fore, shadow influences the photosynthesis rate, and it can be modelled
Qgas Volumetric flow rate of gas m3 s−1
as a distributed factor (αs) in each cross-sectional area. The shadow fac-
Qliq Volumetric flow rate of liquid m3 s−1
Qm Volumetric flow rate of culture medium m3 s−1 tor (αs) is calculated taking into account the length of the shadow pro-
Qh Volumetric flow rate during the harvesting process m3 s−1 jection on the perpendicular axis of the walls, such as shown in Fig. 1,
Sx Shadow onto the surface of the cross-sectional area m2 using Eqs. (1) and (2) [26]. According to these equations, azimuth (α)
U∞ Bubble accession rate 0.651 m s−1
and altitude angles (γ) are calculated as a function of the latitude (ϕ),
Ugas Superficial velocity of the gas 0.0186 m s−1
Uliq Superficial velocity of liquid 0.0441 m s−1
hour angle (ω) and sun declination (δ), these last two terms being a
v Velocity of the fluid 0.2 m s−1 function of the day of the year (N), and the solar hour (hs). The projec-
Vp Volume of the paddle-wheel 0.1651 m3 tion of the shadow generated by the channel walls onto the surface of
Vs Volume of the sump 0.8151 m3 the cross-sectional area (sx) can be described in terms of the wall height
Vmol Molar volume 20 L mol−1
(hc), the solar altitude angle, the Azimuth angle, and the angle measured
ws Width of the sump 1m
y N2 Nitrogen molar fraction from the North to the normal vector of the cross-sectional area of the re-
εs Gas hold-up in the sump actor (γ0), in this case 84° (Fig. 1). Finally, the distributed parameter is
αs Distributed factor calculated as the ratio to the total width of the channel,
α Solar altitude angle
δ Sun declination
ω Solar hour angle α ¼ sin−1 ð sinðδÞ sinðϕÞ þ cosðδÞ cosðϕÞ cosðωÞÞ ð1Þ
ϕ Latitude
 
γ Azimuth angle sinðδÞ cosðϕÞ− cosðδÞ cosðωÞ sinðϕÞ
γ0 Angle from North to the normal vector of the reactor γ ¼ cos−1 ð2Þ
cosðα Þ
Characteristic parameters
B1 Preexponential factor 2.4098 if sinðωÞN0 then γ ¼ 360−γ
ð3Þ
B2 Preexponential factor 533.009 if sinðωÞ≤0 then γ ¼ γ
C1 Activation energies 6.2684
C2 Activation energies 68.8062  
Ka Extinction coefficient 80 m2 kg−1 360ð284 þ NÞ
δ ¼ 23:45 sin ð4Þ
KCO2,c Transfer coefficient constants for CO2 in the column 0.91 365
KCO2,l Transfer coefficient constants for CO2 in the loop 0.91
Ki Form parameter 174 μE m−2 s−1
ω ¼ 15ð12−hs Þ ð5Þ
KlaO2c Volumetric gas-liquid mass transfer coefficient for 2.5000⋅10−6 s−1
CO2 in the Channels  
KlaO2p Volumetric gas-liquid mass transfer coefficient for 0.0219 s−1 ðhc −hÞ  sinðγ−ð180 þ γ 0 ÞÞ
sx ¼ abs ð6Þ
CO2 in the paddle-wheel tanðα Þ
KlaO2s Volumetric gas-liquid mass transfer coefficient for 0.0063 s−1
CO2 in the sump
KlaCO2c Volumetric gas-liquid mass transfer coefficient for s−1
sx
CO2 in the channels αs ¼ ð7Þ
w
70
3.2. Biological model
The growth rate of microalgae is a function of the light availability or
average irradiance (Iav) at which the cells are exposed to [27]. The aver-
age irradiance integrates the local irradiance values inside the culture
over the total culture volume, it being calculated as a function of the
total incident radiation on the photobioreactor surface (Io), the biomass
concentration (Cb), the light attenuation of the biomass (Ka), and the
light path or culture depth (h) [28]. Taking into account the variation
of biomass concentration with time, t, and position along the reactor,
x, the average irradiance in whatever section of the reactor can be calcu-
lated by Eq. (8).
I 0 ðt Þ
I av ðt; xÞ ¼ ð1− expð−K a C b ðt; xÞhÞ:
K a C b ðt; xÞh
The photosynthesis rate (PO2), defined as the oxygen production rate
per biomass mass unit, is correlated with the average irradiance by an
hyperbolic function, the response of photosynthesis rate to average irra-
diance being modulated by adequacy of culture conditions using nor-
malized factors [24]. A potential equation describes the influence of
dissolved oxygen concentration on the photosynthesis rate, whereas
for the pH a model based on the Arrhenius equation is used. Thus,
under nutrient sufficient conditions, Eq. (9) can be used to determine
the photosynthesis rate as a function of average irradiance, dissolved
oxygen concentration ([O2]) and pH into the culture [24]. In this equa-
tion, several biological parameters specific of microalga strain and
growth status of the cells are included, as the maximum photosynthesis
rate under the culture conditions (PO2,max), the form exponent (n), the ir-
radiance constant (as an exponential function of average irradiance, Ki
and m), the oxygen inhibition constant (KO2), a form parameter (z),
the pre-exponential factors (B1, B2), the activation energies of the Arrhe-
nius model (C1, C2), and the constant respiration rate (RO2).
n 
P O2; max Iav ðt; xÞ
P O2 ðt; xÞ ¼ ð1−α s Þ 1−
   K i expðI ðt; xÞmÞ 
av þI av ðt; xÞn
−C 1 −C 2
B1 exp −B2 exp −α s RO2
pHðt; xÞ pHðt; xÞ
Once the photosynthesis rate is modelled, Eq. (10) allows determin-
ing the biological carbon dioxide uptake (PCO2) considering a one-to-one
molar ratio between oxygen and carbon dioxide (from basic equation of
photosynthesis). Moreover, considering a mean value of oxygen coeffi-
cient yield (Yb/O2) the net production of biomass can be determined by
Eq. (11).
P CO2 ðt; xÞ ¼ −P O2 ðt; xÞ
P b ðt; xÞ ¼ Y b=O2 ðt; xÞP O2 ðt; xÞ
3.2.1. Engineering model of the reactor
The raceway reactor used in this work has been previously charac-
terized in both fluid-dynamic and mass transfer capacity [14,15,25]. Ac-
cording to this previous knowledge, the reactor can be divided into
three main zones: channel, paddle-wheel, and sump. The channel per-
forms as a plug-flow reactor, thus perfect mixing exits into the cross sec-
tion of the channel, and axial gradients are considered due to biological
and mass transfer phenomena. Biological phenomena (production of
oxygen, consumption of inorganic carbon, production of biomass, etc.)
take place into the channel, in addition to mass transfer between the
culture and the atmosphere (oxygen and carbon dioxide exchange).
Paddle-wheel performs as stirred tank, thus perfect mixing existing in
the liquid phase with no gradients taking place. In this section of the re-
actor, biological phenomena also take place in addition to mass transfer
I. Fernández et al. / Algal Research 17 (2016) 67–78
between liquid and atmosphere. The sump performs also as stirred tank
for the liquid, and thus no gradients exist. However, as plug-flow for the
injected gas is considered, gradients of oxygen and carbon dioxide into
the gas phase appear. Inside the sump, the same biological phenomena
take place, but the mass transfer is a function of gas phase composition
along the sump. Assuming constant velocity (v) and liquid height (h) in-
side the channel, the volumetric flow rate of liquid (Qliq) is defined as
the multiplication of velocity and cross-sectional area of the channel
(calculated using the liquid height and the width of the channel, w).
This flow rate is constant for the three sections of the reactor. Regarding
the mass transfer, it is a function of mass transfer coefficient (Klal) and
driving force in each position, the driving force being a function of the
component concentration into the liquid phase and that in equilibrium
with the gas phase in contact.
ð8Þ
3.2.2. Equations of the model
To model the raceway reactor, mass balances have been applied to
each reactor sections. A model using partial differential equations
(PDEs) has been used to cope with the existence of plug-flow behavior
in some parts of the reactor. PDEs are used in many physical problems,
such as fluid flow, heat transfer, solid mechanics, and biological process-
es. Only ordinary differential equations (ODE) has been applied to
stirred tank sections of the reactor as sump and paddle to reduce the
computational effort.
3.2.2.1. Mass balances in the liquid phase. The three main components
considered into the liquid phase are biomass concentration (Cb), dis-
solved oxygen concentration ([O2]), and total inorganic carbon concen-
tration ([CT]). A mass balance is defined for each one of the three
components in each section of the reactor. Thus, the proposed balances
for each one of the three components are shown in Eqs. (12) to (14).
∂Cbðt; xÞ ∂Cbðt; xÞ
wh ¼ −whv þ whP O2 ðt; xÞCbðt; xÞY b=O2 ð12Þ
  ∂t ∂x
½O2 ðt; xÞ z
K O2 : ∂½O2 ðt; xÞ ∂½O2 ðt; xÞ
wh ¼ −whv þ
∂t ∂x ð13Þ
P O ðt; xÞCbðt; xÞ  

þwh 2 þ whK laO2c O2 ðt; xÞ−½O2 ðt; xÞ
ð9Þ M O2
∂½C T ðt; xÞ ∂½C T ðt; xÞ
wh ¼ −whv þ
∂t ∂x
P CO2 ðt; xÞCbðt; xÞ  
: ð14Þ
þwh þ whK laCO2c CO2 ðt; xÞ−½CO2 ðt; xÞ
M CO2
The oxygen mass transfer is a function of the volumetric coefficient
for oxygen into the channel (KlalO2c) and the logarithmic driving force
ð10Þ ([O2⁎]−[O2]) [29]. The dissolved oxygen concentration in equilibrium
with air surrounding the channel ([O2⁎]) is calculated as a function of
ð11Þ the oxygen molar fraction into the air (0.21) based on Henry's law.
The carbon dioxide mass transfer is calculated in the same way as a
function of volumetric mass transfer for carbon dioxide into the channel
(KlalCO2c), that could be directly related to KlalO2c by a factor of 0.93, which
takes into account the difference in aqueous diffusivity of the two gases.
Regarding the carbon dioxide concentration ([CO2]), it is a function of
total inorganic carbon ([CT]) and pH, due to the existence of bicarbonate
buffer [29]. The carbon dioxide concentration in equilibrium with the
gas phase ([CO2⁎]) is also calculated as a function of the carbon dioxide
molar fraction into the air (0.0003) based on Henry's law.
Analogous mass balances are applied to the paddle-wheel, consider-
ing that this section can be represented by ODEs. For the paddle-wheel,
the concentration of the major components at inlet is that calculated as
outlet from the channel. Eqs. (15) to (17) allow us calculating the bio-
mass, dissolved oxygen, and inorganic carbon concentration at the out-
let of the paddle-wheel taking into account the specific dimensions and
mass transfer coefficients in this section. It is important to note that in
 I. Fernández et al. / Algal Research 17 (2016) 67–78 71

paddle-wheel air is also the gas in contact with the liquid phase, its con- 3.2.2.2. Mass balances to the gas phase. Air (21% O2, 0.03% CO2) or flue gas
centration being constant in spite of oxygen and carbon dioxide ex- (6% O2, 10% CO2) is injected into the sump to control the dissolved oxy-
change. gen concentration and the pH of the culture. Thus, when the pH is
higher than setpoint (pH = 8) flue gas is injected to reduce the pH
dCbout ðt Þ Q liq and supply inorganic carbon, otherwise air being injected to minimize
¼− ðCbout ðt Þ−Cbin ðt ÞÞþ
dt Vp ð15Þ the accumulation of dissolved oxygen and to avoid achieving toxic dis-
þP O2 ðt ÞCbin ðt ÞY b=O2 solved oxygen concentrations (N 250%Sat) [24]. The injected gas mod-
ifies its composition along the sump due to mass transfer, thus mass
d½O2 out ðt Þ Q liq
balances are also applied to the gas phase to determine its variation
¼− ½O2 out ðt Þ−½O2 in ðt Þ þ
dt Vp along the sump. Since the nitrogen molar flow can be considered con-
ð16Þ
P O ðt ÞCbout ðt Þ  
stant, because its solubility is low, the balances are formulated by rela-
þ 2 þ K laO2p O2 ðt Þ−½O2 ðt Þ lm
M O2 tions from the rest of gases to nitrogen molar ratio. According to
Eq. (23), the variation of the oxygen to nitrogen molar ratio into the
d½C T out ðt Þ Q liq
gas phase (YO2) is a function of the nitrogen molar fraction into the
¼− ½C T out ðt Þ−½C T in ðt Þ þ
dt Vp gas phase (yN2). For the carbon dioxide, an analogous mass balance
: ð17Þ
P CO2 ðt ÞCbout ðt Þ  
can be defined to determine the variation of the carbon dioxide to nitro-
þ þ K laCO2p CO2 ðt Þ−½CO2 ðt Þ lm
MCO2 gen molar ratio in the gas phase (YCO2).

Similar mass balances are applied to the sump also considering that dYO2;out ðt Þ Q gas

¼− YO2;out ðt Þ−YO2;in ðt Þ
this section can be represented by ODEs. For the sump, the concentra- dt V s ð1−ε s ðt ÞÞ
ð23Þ
tion of major components at inlet is that calculated as outlet from the V ð1−ε s ðt ÞÞ   

−K laO2s mol O2 ðt Þ−½O2 ðt Þ lm
paddle-wheel. Eqs. (18) to (20) allow us calculating the biomass, dis- yN2 ε s ðt Þ
solved oxygen and inorganic carbon concentration at the outlet of the
sump, respectively. dYCO2;out ðt Þ Q gas

¼− YCO2;out ðt Þ−YCO2;in ðt Þ
dt V s ð1−εs ðt ÞÞ
: ð24Þ
dCbout ðt Þ Q liq V ð1−ε s ðt ÞÞ   

¼− ðCbout ðt Þ−Cbin ðt ÞÞþ −K laCO2s mol CO2 ðt Þ−½CO2 ðt Þ lm
dt V s ð1−ε s ðt ÞÞ yN2 ε s ðt Þ
ð18Þ
Qm
þP O2 ðt ÞCbin ðt ÞY b=O2 − Cbout ðt Þ
V s ð1−ε s ðt ÞÞ
3.2.2.3. Solvers and software. PDEs and ODEs balances, which establish
d½O2 out ðt Þ Q liq
the base of the model, have been implemented by the software Matlab
¼− ½O2 out ðt Þ−½O2 in ðt Þ þ
dt V s ð1−ε s ðt ÞÞ 8.3 (MathWorks, Massachusetts, USA). Since the computational cost re-
P O ðt ÞCbout ðt Þ  
quired to solve these kinds of equations is high, above all concerning
þ 2 þ K laO2s O2 ðt Þ−½O2 ðt Þ lm þ ð19Þ
M O2 PDEs equations, a general procedure for first order hyperbolic equations
Qm

þ ½O2 m −½O2 out ðt Þ has been used by means of the well-known method of lines. On the
V s ð1−ε s ðt ÞÞ other hand, ODEs balances have been calculated by a forward first
order finite difference approximation method. Note that for the calibra-
d½C T out ðt Þ Q liq
tion procedure a multidimensional nonlinear minimization process has
¼− ½C T out ðt Þ−½C T in ðt Þ þ
dt V s ð1−ε s ðt ÞÞ been formulated, using Sequential Quadratic Programming (SQP)
P CO2 ðt ÞCbout ðt Þ  

þ þ K laCO2s CO2 ðt Þ−½CO2 ðt Þ lm þ : ð20Þ methods since a Quadratic Programming (QP) sub-problem was consid-
MCO2 ered. For that reason, an appropriated simulation time of the model is
Qm

þ ½C T m −½C T out ðt Þ required in order to solve the optimization problem suggested in a rea-
V s ð1−ε s ðt ÞÞ
sonable time. Model calibration has been performed comparing real
data of pH and dissolved oxygen concentration at the end of the three
Since air or carbon dioxide is injected into the sump, and the exchange sections of the reactor (paddlewheel, sump and channel); with the sim-
of oxygen and carbon dioxide from the gas phase to the liquid modifies ulation response obtained using estimated values of characteristic
the oxygen and carbon dioxide molar fraction into the gas phase, the ox- parameters.
ygen and carbon dioxide in the liquid equilibrium with the gas phase
must be dynamically calculated. The outlet of biomass, dissolved oxygen 4. Results
and inorganic carbon from the system due to harvesting is included by
considering the volumetric flow of medium (Qm). In these equations, 4.1. Calibration of the model
the volume of each section is corrected by gas hold-up (εs) to determine
the right liquid volume in each section. The gas hold-up can be approxi- The proposed model is based on first-principles and thus it takes into
mated by Eq. (21) taking into account the difference between the volu- account the major phenomena taking place in the system, both biolog-
metric flow rate of gas introduced in the sump, Qgas, and the total ical and engineering phenomena. Biological aspects of the model in-
volumetric flow rate of the system (Qgas +Qliq). clude the response of the cells to prevailing culture conditions, where
a certain number of characteristic parameters are used. Although the
Q gas ðt Þ
ε s ðt Þ ¼ : ð21Þ value of these characteristic parameters can be determined at laborato-
Q gas ðt Þ þ Q liq ðt Þ ry scale, the model allows estimating them by calibration versus exper-
imental data. In this sense, the determined values are more realistic,
To apply these equations, the different volume of each section is con- representing the performance of cells into the culture at real outdoor
sidered, thus the volume of the sump is calculated by Eq. (22) consider- conditions. Regarding the engineering aspects, a certain number of
ing the height (hs), wide (ws) and length (hs) of the sump in addition to characteristic parameters are also included into the model, such as
the volume of channel comprises over the sump. mass transfer coefficients. These parameters can be determined by engi-
neering characterization of the reactor, but it implies to perform addi-
V s ¼ hss ws ls þ hws ls : ð22Þ tional experiments with an ‘empty’ reactor requiring long time. In the
72 I. Fernández et al. / Algal Research 17 (2016) 67–78
same way than biological characteristic parameters, the engineering
characteristic parameters can also be obtained from calibration process.
Thus, a calibration procedure was performed using experimental data of
outdoor cultures in order to determine the value of characteristic pa-
rameters of the model. To this end, a nonlinear minimization process
is used to fit experimental data into the proposed model, where an
error metric, like integrated absolute error (IAE) or integrated squared
error (ISE), can be chosen as loss function. In this case, the characteristic
parameters are the decision variables of the optimization problem, es-
tablishing as initial point for the optimization the experimental values
previously measured, and constraining their maximum and minimum
values by linear constraints in order to preserve physical and chemical
senses.
The experimental data were obtained under normal operating con-
ditions of the reactor (liquid velocity at 0.2 m s−1, pH = 8 by injection
of flue gas, semicontinuous operation at 0.2 day−1 dilution rate with
volumetric flow rate of medium at 30 l min−1) for different dates, cov-
ering a wide-range of solar radiation conditions. Measurements of dis-
solved oxygen and pH at the end of each part of the reactor (channel,
paddle-wheel and sump) were registered, in addition to environmental
conditions (solar radiation, temperature), and operation parameters
(CO2 injection, dilution of the culture). Initial values of characteristic pa-
rameters were obtained from previous knowledge. Thus, value of bio-
logical parameters such as the extinction coefficient (Ka), maximum
photosynthesis rate (PO2 ,max), form parameters (m, n, z), oxygen inhibi-
tion constants (KO2), pre-exponential factors (B1, B2), activation ener-
gies (C1, C2) and the respiration constant (RO2) were taken from [24].
On the other hand, value of mass transfer coefficients for each part of
the reactor were taken from [15,25]. To determine the real values of
these characteristic parameters, data of dissolved oxygen and pH from
Fig. 2. Experimental and simulated data of dissolved oxygen concentration and pH at the end of the channels, paddle-wheel and sump. All of them as a function of CO2 injection and solar
radiation of two representative days used for model calibration purposes.
the real reactor were compared with the output of the model using an
optimization problem on the IAE error, an optimum value of these pa-
rameters being determined for each day used in the calibration proce-
dure. Fig. 2 allows to compare the real and simulated data obtained
using the developed model. As observed, the model response fits to
the experimental data for both dissolved oxygen and pH. The model
captures smooth variations of the photosynthetic rate produced by the
bell-shaped form of the solar radiation. Increase of dissolved oxygen
and pH due to photosynthesis by solar radiation availability is simulat-
ed. Changes produced through CO2 injections, as pH and dissolved oxy-
gen decrease, are also represented by the proposed model, taking into
account the different dynamics taking place with and without presence
of solar radiation. Moreover, the model allows us observing how the
characteristic delay of the system produces an increase of pH and dis-
solved oxygen in the channels and paddle-wheel, both variables de-
creasing after the sump. The mean errors between the simulated and
experimental data were of 8.2%, 8.6% and 10.0% for the dissolved oxygen
at the end of the channel, paddle-wheel and sump respectively, and of
1.5%, 1.8% and 1.6% for the pH for the same sections.
As result of the calibration procedure the average values of character-
istic parameters were determined (Table 1). Regarding biological param-
eters the maximum photosynthesis rate was 2.06·10−5 kgO2 kg−1 s −1,
whereas the extinction coefficient was 80 m2 kg −1, and the coefficients
respiration rate (RO2 and RCO2) were 9.58·10−7 kgO2 kgg−7 s−1 and
4.28·10−6 kgCO2 kg−1 s−1 for oxygen and carbon dioxide respectively.
The rest of biological parameters being n=1.045, Ki =174 μ E m−2 s−1
and m=0.0021. Regarding the volumetric mass transfer coefficients for
oxygen in the different sections of the reactor, the values obtained were
2.5·10−6 s−1, 2.2·10−3 s−1 and 6.3·10−3 s−1, for channel, paddle-
wheel and sump, respectively.
 I. Fernández et al. / Algal Research 17 (2016) 67–78
4.2. Validation of the model
Validation of the model was performed to check the model quality,
using a different set of data in different dates (validation data), in
order to avoid possible bias and variance errors. For this validation pro-
cess, the reactor was operated under the same conditions that in the cal-
ibration stage, where model includes the characteristic parameters
obtained from calibration process being compared with experimental
data. Fig. 3 shows how the simulation fits to experimental results, the
adequacy of the model to simulate experimental data being confirmed.
The mean error for the dissolved oxygen was 4.9%, 1.6% and 1.5% at the
end of the channel, paddle-wheel and sump. On the other hand, the dif-
ference between the experimental pH of the culture and the simulated
pH was 1.1%, 1.0% and 1.2% for the same sections, proving an accurate
response of the model for real conditions of operation. It can be conclud-
ed that the proposed model captures the variations of the dissolved ox-
ygen caused by the daily variation of solar radiation, and consequently
of the photosynthesis rate. Furthermore, the model represents the
transfer and consumption of carbon dioxide, thus allowing to simulate
the variation of pH as a function of photosynthesis rate and CO2 injec-
tions performed to feed the system and control the pH of the culture.
4.3. Estimation of additional variables
Once the model is validated, it can be used to determine additional
variables that cannot be measured directly such as the average irradi-
ance, the photosynthesis rate or the CO2 losses, considering the time
and the different sections of the reactor. In this sense Fig. 4 shows the
variation of these additional variables for one of the days used for vali-
dation purposes. This figure shows how, in spite that solar irradiance
Fig. 3. Experimental and simulated data of dissolved oxygen concentration and pH as a function of CO2 injection and solar radiation of two representative days used for validation purposes.
73
achieves values up to 1200 μE m−2 s−1, the maximal value of average
irradiance was only 180 μE m−2 −1. In these conditions, the maximal
photosynthesis rate was 8·10−5 mol m−3−1, clearly fitting the varia-
tion of average irradiance inside the culture, thus no photo-inhibition
being observed. The dissolved oxygen at noon reaches values of
0.40 mol m−3 in the channels, below of the inhibition constant deter-
mined for dissolved oxygen of 0.72 mol m−3, and being reduced 25%
by the air injections into the sump, up to 0.3 mol m−3, showing that
the negative effect of the accumulation of dissolved oxygen concentra-
tion takes place in the mid hours of the daylight period when the dis-
solved oxygen concentration is higher. Regarding CO2, data show as
CO2 stripping into the sump is mainly related with the injection of CO2
to control the pH, anyway it being really low, with maximal values of
0.010 molCO−1 2 , thus confirming the adequacy of injection system
used. Moreover, the CO2 stripping into the channels is ten times lower
than into the sump, thus confirming that optimal design and operation
of the sump is a critical issue in raceway reactors. Regarding oxygen de-
sorption into the sump, it is much higher than CO2 stripping, maximal
values up to 0.025 molO−1 2 being measured, but oxygen desorption tak-
ing place during the entire daylight period due to the injection of air into
the sump. In any case, the injection of CO2 and air allows us to control
the pH and dissolved oxygen concentration of the culture in the ade-
quate ranges to maximize the performance of the cultures (pH = 8,
[O2]b 250%Sat-0.7 mol m−3).
4.4. Local gradients
In addition to local or overall values, the developed dynamic model
allows us to determine the value of any variable as a function of time
and position along the reactor. Therefore, the model helps to analyze
74 I. Fernández et al. / Algal Research 17 (2016) 67–78

Fig. 4. Daily variation of measured culture parameters (pH, dissolved oxygen concentration, CO2 inlet, solar radiation) and simulated ones (average irradiance, photosynthesis rate, CO2 and
air stripping). Simulation data by the proposed model using characteristic parameters determined experimentally.

the local values of culture parameters, thus identifying gradients at
which the cells are exposed inside the reactor and that potentially can
reduce the performance of the cells. Fig. 5 shows the gradients of dis-
solved oxygen, carbon dioxide and pH in liquid phase at noon, for
each one of sections in which the reactor was divided (channels,
paddle-wheel and sump). These profiles have been determined for the
same day and condition that Fig. 4. It can be observed how the dissolved
oxygen concentration increases into the channel due to the photosyn-
thesis performed by the cells, whereas it decreases both in the paddle-
wheel, caused by stripping to the atmosphere, and into the sump, due
to the injection of flue gas with a low molar fraction of oxygen (6%).
On the contrary, carbon dioxide diminishes along the channel mainly
due to consumption by photosynthesis, greatly increasing into the
sump due to flue gas injection containing CO2 (10% CO2). Regarding
pH, it increases linearly along the channel as a direct function of photo-
synthesis rate, reducing when the CO2 is injected into the sump. These
results demonstrate the existence of gradients along the channel that
may lead to adverse culture conditions in some parts of the reactor, al-
though the culture conditions are adequate in the rest of the reactor.
4.5. Tool for the design of raceway reactors
Once the model has been calibrated and validated, it can be also used
to simulate different designs, conditions or scenarios. In this case, only dif-
ferent designs or configurations are simulated. Thus, due to the influence
of the height of the reactor wall on the light availability, and finally the
biomass productivity of the reactor, this issue was first studied. Fig. 6
shows the influence of normal wall height of the reactor (h/hliq) demon-
strating that using walls with heights double than liquid reduces the pro-
ductivity a 30%. As was expected, the optimal point was found in the unit
of this relation, since on the contrary the difference between these heights
generates shadows when the solar radiation is projected in the surface of
the liquid, producing a lower level of production. Once the wall height is
optimized, the other relevant parameter is the water depth because it
 I. Fernández et al. / Algal Research 17 (2016) 67–78 75

Fig. 5. Profiles of dissolved oxygen, carbon dioxide concentration and pH in the liquid phase throughout the reactor at noon. Data from simulations performed using the proposed model
and values of characteristic parameters determined experimentally.

determines the light availability inside the culture through the average ir-
radiance. Fig. 7 shows the results obtained modifying the liquid height
into the reactor but maintaining the rest of operational and design condi-
tions (length and wide) of the reactor in any case. Note that in this case,
the wall height and the liquid height have been established to equal
values in order to avoid the negative effects produced by shadows in
the system. Data shows how the biomass productivity of the system ex-
ponentially increases when reducing the water depth, increasing 72%
when the water depth reduces from 30 to 5 cm. It is important to note
that these results do not consider the difficulty of operating large race-
ways at low water depth, as losses of efficiency on the paddle-wheel or
water depth variations along the reactor. Finally, the last parameter
studied was the channel length by its influence into the accumulation of
dissolved oxygen and inadequate pH. In this case, simulations were per-
formed maintaining the overall configuration of the reactor (waster
depth, wide of the channel), except the length of channel that was varied
from 10 to 300 m. Fig. 8 shows how the biomass productivity was similar
when using lengths channels lower than 100 m, but over this value rele-
vant dissolved oxygen and pH gradients exist into the cultures, that re-
duce the performance of the reactor. The biomass productivity reduces
by 50% when the channel length increases from 10 to 300 m. To solve
this problem, several alternatives exist, but all of them impose higher
costs related to raceway reactor building or operation, thus being manda-
tory to accurately consider these phenomena.

Fig. 6. Influence of normalized wall height on normalized biomass productivity of Scenedesmus almeriensis semicontinuous cultures in outdoor raceway reactor under standard conditions.
76 I. Fernández et al. / Algal Research 17 (2016) 67–78
Fig. 7. Influence of water depth on normalized biomass productivity of Scenedesmus almeriensis semicontinuous cultures in outdoor raceway reactor under standard conditions.
5. Discussion
The developed model is a powerful tool for the study of existing
raceway reactors producing microalgae biomass. Thus, from direct mea-
surements of dissolved oxygen, pH and CO2 injection, it is possible to
obtain the values of characteristic parameters of the system, both bio-
logical and engineering ones. Data obtained from calibration of the
model versus experimental data agree with previously reported for
this reactor [14,15,25]. The extinction coefficient of the biomass is in
the range of 50 to 200 m2 kg−1 reported for microalgae, whereas the
photosynthesis rate is more than one order of magnitude higher than
the respiration rate, of 2.06·10−5 and 9.58·10− 7 kgO2 kg−1 s− 1, re-
spectively. The model also allows us to determine characteristic param-
eters of the growth model of the strain used (n, Ki, m) that are usually
determined at laboratory conditions, requiring long time and numerous
experiments [24]. Regarding the mass transfer coefficients determined
from the calibration procedure, results agree with previously reported
for the same reactor [14], confirming that mass transfer mainly takes
place in sump and paddle-wheel. The model confirms that channels
perform as a ‘tubular reactor’, where no relevant exchange of oxygen
and CO2 takes place between the culture and the atmosphere, besides
the general knowledge that in raceway reactors oxygen accumulation
does not take place and large CO2 losses take place into the channel.
Thus, the model allows us to demonstrate that oxygen is accumulated
into the culture up to values of 0.4 mol/m3 because the photosynthesis
rate is higher than oxygen desorption capacity (Fig. 4). On the opposite,
the injected CO2 is higher than CO2 consumption by the cells, being
Fig. 8. Influence of channel length on normalized biomass productivity of Scenedesmus almeriensis semicontinuous cultures in outdoor raceway reactor under standard conditions.
mainly lost to the atmosphere into the sump and paddle-wheel. Howev-
er, CO2 losses into the channels are three orders of magnitude lower
(Fig. 4). This behavior is analogous to that reported in tubular
photobioreactors [29]. Then, it can be concluded that the design of
both type of reactors, raceway and tubular, is not so different as usually
reported.
Results here reported concerning the spatial-temporal variation of
culture parameters demonstrate that static models cannot be used to
adequately represent this type of reactors. Thus, microalgae have differ-
ent responses (photosynthesis rate, etc.) to changes in the culture con-
ditions not only along the day but also at the different positions inside
the reactor [22]. Several studies reported dissolved oxygen concentra-
tions in raceways as high as 500%Sat., causing inhibition of photosyn-
thesis and growth, and eventually leading to culture death [15,30–32].
Otherwise, it has also been reported that algal cultures in raceways
can become carbon limited if only CO2 from the air is available [33], to
maximize the productivity being necessary to maintain a CO2 concen-
tration in the bulk liquid of at least 65 μmol/L and a pH of 8.5 for high
productivity of some microalgae [10]. To provide CO2 and maintain
the pH, it is possible to supply flue gases, but it is necessary to adequate-
ly design and operate the CO2 supply unit [25]. Moreover, the existence
of pH gradients into the reactor reduces the performance of microalgae
cultures [34,35]. Experimental data here reported confirm the existence
of relevant variations of culture conditions with time (Fig. 4) and posi-
tion inside the reactor (Fig. 5). Variation of culture parameters along
the day is a consequence of solar daily variation, whereas variations
along the different sections of the raceway is a consequence of the
 I. Fernández et al. / Algal Research 17 (2016) 67–78
different rates of phenomena taking place (biological, physic and
chemical).
To represent the behavior of microalgae cultures usually steady-
state models are formulated taking into account mean daily values of
culture conditions, disregarding the characteristic times of different
phenomena taking place. For instance, microalgae growth has a charac-
teristic time of hours, being mainly a function of photosynthesis rate
with characteristic time of minutes, and depending on the solar radia-
tion at a certain time instant. Furthermore, other factors such as the
mixing or mass transfer phenomena have characteristic times of
seconds-minutes. In this scenario, dynamic models, as that developed
in this work, are required to simulate the behavior of microalgae culture
systems, taking into account both the spatial and temporal characteris-
tics of the system. The developed model allows us analyzing the condi-
tions at which the cells are exposed in each part of the reactor, helping
to design an optimal structure of the reactor, bearing in mind variations
of culture conditions with time and space. Moreover, this type of models
are extremely useful from a control point of view, being possible to de-
velop advanced control strategies such as based-model predictive con-
trollers or optimal and hierarchical structures widely used in other
industrial process such as solar energy, manufacturing, robotics or bio-
logical process related to greenhouses [36,37]. Implementation of ad-
vanced control strategies based on simple dynamical models allows to
enhance the yield of microalgae cultures by diminishing the gradients
of certain variables such as pH or temperature at which the cells are ex-
posed to [22,34,38,39].
The developed model can also be used to simulate the performance
of the raceway reactor at different conditions. The simplest analysis is to
study the influence of the reactor design (configuration) into the perfor-
mance of the system. In this sense, the influence of the wall height,
water depth and channel length into the performance of the culture
has been performed. Results demonstrate the large influence of wall
height into the performance of the reactor due to the shadow that
large walls create into the water surface, reducing the solar light avail-
ability (Fig. 6). As expected, the larger the wall height the lower was
the biomass productivity, but what is relevant in the quantification of
the phenomenon, up to 30% of productivity being lost if using walls larg-
er than double of water depth. Most interesting is the analysis of the
water depth. This analysis demonstrates that reducing the water
depth the biomass productivity increases, up to 72% when reducing
the water depth from 30 to 5 cm (Fig. 7). These results agree with the
actual trend to reduce the water depth into the reactors in order to in-
crease the biomass concentration by improvement of light availability,
as a strategy to operate more robust systems, with lower risk of contam-
ination or failure. Moreover, the use of thin-layer reactors with very low
water depth was reported time ago [40], but now they are expanding
his application at pilot and commercial scale [41,42]. Regarding the
channel length, results show how the biomass productivity is indepen-
dent of these parameters at small scale (short channels) but upper than
100 m the biomass productivity reduces notably (Fig. 8). In addition to
power consumption, the increase of the channel length enhances the
adverse effects derived from inadequate gradients of culture conditions
that reduce the performance of the overall system. It is important to re-
mark that this issue does not mean that raceway reactors cannot be
larger than 100 m, but to use channels larger than this length the oxy-
gen and CO2 mass transfer in each one of the section of the reactors
must be accurately designed and performed.
Summarizing, the developed model has demonstrated to reproduce
the spatial and temporal variation of main variables (light, biomass, dis-
solved oxygen, carbon dioxide, pH) in raceway reactors. This model is
an useful tool to design and operate photobioreactors in a conservative
way, for the boundary conditions of maximum solar radiation availabil-
ity or whatever other situation. However, as the model also allows us to
include the variation with time of culture parameters, it can be used for
the implementation of advanced control strategies, and to refine the de-
sign and operation of open reactors taking into account the dynamic
77
accumulation or uptake of compounds, thus optimizing it. Dynamic
models based on first principles as that here reported, are a necessary
and powerful tool for the improvement of raceway reactors. These reac-
tors represent N95% of the worldwide production of microalgae, thus
the improvement of this technology will have a large impact on
microalgae industry.
6. Conclusions
A dynamic model based on first principles of the production of
microalgae in raceway reactors has been developed, calibrated, and val-
idated. The model integrates the biological and engineering aspects of
this type of systems, considering both spatial and temporal variations
inside the reactor. The developed model allows to determine character-
istic parameters (biological and engineering ones) of existing systems,
but also to simulate the effect of different design and/or operational
conditions in the performance of the process. Data here reported dem-
onstrate that, in opposite to what is usually accepted, in raceway reac-
tors large accumulation of oxygen takes place into the channel
whereas carbon losses into this section are scarce. This conclusion is re-
ally remarkable, since these phenomena have a relevant impact into the
biomass production of the system. Simulations performed modifying
the geometry of the reactor demonstrate that it is recommendable to re-
duce the water depth into the culture but more important, the total
length of the channels is limited by the mass transfer capacity of the en-
tire reactor. Future works will be focused on improving the model to in-
clude the variation of temperature and evaporation losses as a function
of location, day of the year, and weather conditions will be considered.
Acknowledgements
This work has been partially funded by the following projects:
DPI2014-55932-C2-1-R (Spanish Ministry of Economy and Competi-
tiveness and FEDER funds); Controlcrop PIO-TEP-6174 (Junta de
Andalucía); EDARSOL CTQ2014-57293-C3 (Spanish Ministry of Science
and Innovation); PURALGA RTA2013-0056-C03 (INIA), and supported
by Cajamar Foundation.
References
[1] P. Spolaore, C. Joannis-Cassan, E. Duran, A. Isambert, Commercial applications of
microalgae, J. Biosci. Bioeng. 101 (2006) 87–96.
[2] M. Koller, A. Muhr, G. Braunegg, Microalgae as versatile cellular factories for valued
products, Algal Res. 6 (2014) 52–63.
[3] W.J. Oswald, Introduction to advanced integrated wastewater ponding systems,
Water Sci. Technol. 24 (1991).
[4] R. Muñoz, B. Guieysse, Algal-bacterial processes for the treatment of hazardous con-
taminants: a review, Water Res. 40 (2006) 2799–2815.
[5] Y. Chisti, Biodiesel from microalgae beats bioethanol, Trends Biotechnol. 26 (2008)
126–131.
[6] F.G. Acién, C.V. González-López, J.M. Fernández-Sevilla, E. Molina-Grima, Conversion
of CO2 into biomass by microalgae: how realistic a contribution may it be to signif-
icant CO2 removal? Appl. Microbiol. Biotechnol. 96 (2012) 5577–5586.
[7] C. Posten, Design principles of photo-bioreactors for cultivation of microalgae, Eng.
Life Sci. 9 (2009) 165–177.
[8] F.G. Acién, J.M. Fernández, E. Molina-Grima, Photobioreactors for the production of
microalgae, Rev. Environ. Sci. Biotechnol. 12 (2013) 1–21.
[9] J. Benemann, Microalgae for biofuels and animal feeds, Energies 6 (2013)
5869–5886.
[10] J.C. Weissman, R. Goebel, J.R. Benemann, Photobioreactor design: mixing, carbon
utilization, and oxygen accumulation, Biotechnol. Bioeng. 31 (1988) 336–344.
[11] K. Sompech, Y. Chisti, T. Srinophakun, Design of raceway ponds for producing
microalgae, Biofuels 3 (2012) 387–397.
[12] D. Chiaramonti, M. Prussi, D. Casini, M.R. Tredici, L. Rodolfi, N. Bassi, G.C. Zittelli, P.
Bondioli, Review of energy balance in raceway ponds for microalgae cultivation:
re-thinking a traditional system is possible, Appl. Energy 102 (2013) 101–111.
[13] B. Ketheesan, N. Nirmalakhandan, Modeling microalgal growth in an airlift-driven
raceway reactor, Bioresour. Technol. 136 (2013) 689–696.
[14] J.L. Mendoza, M.R. Granados, I. de Godos, F.G. Acién, E. Molina, C. Banks, S. Heaven,
Fluid-dynamic characterization of real-scale raceway reactors for microalgae pro-
duction, Biomass Bioenergy 54 (2013) 267–275.
[15] J.L. Mendoza, M.R. Granados, I. de Godos, F.G. Acién, E. Molina, S. Heaven, C.J. Banks,
Oxygen transfer and evolution in microalgal culture in open raceways, Bioresour.
Technol. 137 (2013) 188–195.
78 I. Fernández et al. / Algal Research 17 (2016) 67–78
[16] H. Guterman, A. Vonshak, S. Ben-Yaakov, A macromodel for outdoor algal mass pro-
duction, Biotechnol. Bioeng. 35 (1990) 809–819.
[17] H. Jupsin, E. Praet, J. Vasel, Dynamic mathematical model of high rate algal ponds
(HRAP), Water Sci. Technol. 48 (2003) 197–204.
[18] S.C. James, V. Boriah, Modeling algae growth in an open-channel raceway, J. Comput.
Biol. 17 (2010) 895–906.
[19] L. Xin, H. Hong-Ying, G. Ke, S. Ying-Xue, Effects of different nitrogen and phosphorus
concentrations on the growth, nutrient uptake, and lipid accumulation of a freshwa-
ter microalga Scenedesmus sp, Bioresour. Technol. 101 (2010) 5494–5500.
[20] N. Norsker, M.J. Barbosa, M.H. Vermuë, R.H. Wijffels, Microalgal production - a close
look at the economics, Biotechnol. Adv. 29 (2011).
[21] F. Acién, J.M. Fernández, E. Molina-Grima, Economics of microalgae biomass produc-
tion, Biofuels Algae 14 (2013) 313–325.
[22] I. Fernández, F.G. Acién, J.M. Fernández, J.L. Guzmán, J.J. Magán, M. Berenguel, Dy-
namic model of microalgal production in tubular photobioreactors, Bioresour.
Technol. 126 (2012) 172–181.
[23] J.F. Sánchez, J.M. Fernández, F.G. Acién, A. Rueda, J. Pérez-Parra, E. Molina, Influence
of culture conditions on the productivity and lutein content of the new strain
Scenedesmus almeriensis, Process Biochem. 43 (2008) 398–405.
[24] T.A. Costache, F.G. Acién, M.M. Morales, J.M. Fernández-Sevilla, I. Stamatin, E. Molina,
Comprehensive model of microalgae photosynthesis rate as a function of culture
conditions in photobioreactors, Appl. Microbiol. Biotechnol. 97 (2013) 7627–7637.
[25] I. de Godos, J.L. Mendoza, F.G. Acién, E. Molina, C.J. Banks, S. Heaven, F. Rogalla, Eval-
uation of carbon dioxide mass transfer in raceway reactors for microalgae culture
using flue gases, Bioresour. Technol. 153 (2014) 307–314.
[26] R. Kittler, S. Darula, Determination of time and sun position system, Sol. Energy 93
(2013) 72–79.
[27] F.G. Acién, F. García, J.A. Sánchez, J.M. Fernández, E. Molina, A model for light distri-
bution and average solar irradiance inside outdoor tubular photobioreactors for the
microalgal mass culture, Biotechnol. Bioeng. 55 (1997) 701–714.
[28] E. Molina, J.M. Fernández Sevilla, J.A. Sánchez Pérez, F. Garca Camacho, A study on
simultaneous photolimitation and photoinhibition in dense microalgal cultures tak-
ing into account incident and averaged irradiances, J. Biotechnol. 45 (1996) 59–69.
[29] F. Camacho, F.G. Acién, J.A. Sánchez, F. García, E. Molina, Prediction of dissolved ox-
ygen and carbon dioxide concentration profiles in tubular photobioreactors for
microalgal culture, Biotechnol. Bioeng. 62 (1999) 71–86.
[30] F.J. Marquez, K. Sasaki, N. Nishio, S. Nagai, Inhibitory effect of oxygen accumulation
on the growth of spirulina platensis, Biotechnol. Lett. 17 (1995) 225–228.
[31] D.P. Singh, N. Singh, K. Verma, Photooxidative damage to the cyanobacterium spiru-
lina platensis mediated by singlet oxygen, Curr. Microbiol. 31 (1995).
[32] A. Vonshak, Spirulina: growth, physiology and biochemistry, in: A. Vonhask (Ed.),
Spirulina platensis (arthrospira): Physiology, Cell-Biology and Biotechnology, Taylor
and Francis, London, 1997.
[33] D. Stepan, R. Shockey, T. Moe, R. Dorn, Carbon Dioxide Sequestering Using
Microalgae Systems, US Department of Energy, Pittsburgh, PA, US FC26 - 98FT
40320, 2002.
[34] M. Berenguel, F. Rodríguez, F.G. Acién, J.L. Garca, Model predictive control of pH in
tubular photobioreactors, J. Process Control 14 (2004) 377–387.
[35] J.L. García, M. Berenguel, F. Rodríguez, J.M. Fernández Sevilla, C. Brindley Alias, F.G.
Acién, Minimization of carbon losses in pilot-scale outdoor photobioreactors by
model-based predictive control, Biotechnol. Bioeng. 84 (2003) 533–543.
[36] F. Rodrguez, J.L. Guzmán, M. Berenguel, M.R. Arahal, Adaptive hierarchical control of
greenhouse crop production, Int. J. Adapt. Control Signal Process. 22 (2008)
180–197.
[37] C.M. Cirre, M. Berenguel, L. Valenzuela, R. Klempous, Reference governor optimiza-
tion and control of a distributed solar collector field, Eur. J. Oper. Res. 193 (2009)
709–717.
[38] I. Fernández, J. Peña, J.L. Guzman, M. Berenguel, F.G. Acién, Modelling and control is-
sues of pH in tubular photobioreactors, IFAC Proc Vol (IFAC-PapersOnline), 11 2010,
pp. 186–191.
[39] A. Pawlowski, J.L. Mendoza, J.L. Guzmán, M. Berenguel, F.G. Acién, S. Dormido, Effec-
tive utilization of flue gases in raceway reactor with event-based ph control for
microalgae culture, Bioresour. Technol. 170 (2014).
[40] K. Lívanský, J. Doucha, CO2 and O2 gas exchange in outdoor thin-layer high density
microalgal cultures, J. Appl. Phycol. 8 (1996) 353–358.
[41] J. Masojídek, J. Kopecký, L. Giannelli, G. Torzillo, Productivity correlated to
photobiochemical performance of chlorella mass cultures grown outdoors in thin-
layer cascades, J. Ind. Microbiol. Biotechnol. 38 (2011) 307–317.
[42] M.M. Morales-Amaral, C. Gómez-Serrano, F.G. Acién, J.M. Fernández-Sevilla, E.
Molina-Grima, Production of microalgae using centrate from anaerobic digestion
as the nutrients source, Algal Res. 9 (2015) 297–305.