1 s2.0 S2772801322000100 Main
1 s2.0 S2772801322000100 Main
1 s2.0 S2772801322000100 Main
a r t i c l e i n f o a b s t r a c t
Keywords: Efficient energy and resource management are essential for a more sustainable food production system. On-
Aquaponics demand coupled multi-loop aquaponics, containing at least a fish, plant and anaerobic digester unit, has shown
horticulture great potential for optimizing nutrient use. However, balancing the nutrients for both fish and crop production
sustainable agriculture
becomes increasingly complex at higher latitudes. In these areas also significant reductions in energy use are
dynamic model
required. Hitherto, the effects of energy-saving strategies on the dynamics of an aquaponics system have not
design scenarios
been explored.
The objective of the study was to investigate and mathematically model an aquaculture-hydroponic (aquapon-
ics) system with Nile tilapia (Oreochromis niloticus L.) and lettuce (Lactuca sativa L.) or tomato (Solanum lycoper-
sicum L.), and to verify it for the Netherlands.
In the mathematical model several physical and operational parameters were varied to find their influence
on water and nutrient use efficiency, energy use and the growing environment, in particular in the greenhouse.
The variation in crop transpiration was found to be a simple, yet effective, indicator for the influence of
greenhouse settings on the performance of on-demand coupled multi-loop aquaponics systems. The correlation
between water or nutrient use efficiencies and transpiration variation (0.98) is stronger than the correlation with
energy use (0.78). The maximum allowed humidity (70% to 90%) and use of artificial lighting (0 – 200 W/m2 )
has the largest effect on both performance and energy use. Other settings, such as inclusion of thermal insulation
of the cover, minimum temperature, addition of heat storage, and use of screens, have effect on the energy
efficiency, but have little effect on the nutrient dynamics. To counter the negative effects of variation caused by
other energy-saving strategies or climatic variations, the use of a buffer tank, placed between anaerobic digester
and hydroponic system, was found a feasible option.
Aquaponics is an emerging technology in sustainable food production and this research supports the intro-
duction of on-demand coupled aquaponics systems in northern latitudes.
List of abbreviations: DWC, deep water culture; HPS, hydroponic system; RAS, recirculating aquaculture system; RH, relative humidity; UASB, upflow anaerobic
sludge blanket reactor; AP, aquaponics; NFT, nutrient film technique; AD, anaerobic digester; KPI, key performance indicator; RO, reverse osmosis; ETc, crop
transpiration; CV, coefficient of variation; COD, Chemical oxygen demand.
∗
Corresponding author.
E-mail address: karel.keesman@wur.nl (K.J. Keesman).
https://doi.org/10.1016/j.clcb.2022.100012
Received 25 March 2022; Received in revised form 20 May 2022; Accepted 23 May 2022
2772-8013/© 2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/)
L. Jansen and K.J. Keesman Cleaner and Circular Bioeconomy 2 (2022) 100012
2
L. Jansen and K.J. Keesman Cleaner and Circular Bioeconomy 2 (2022) 100012
Figure 2. The placement of the two possible buffers in the AP system. Buffer 1
is for the RAS to HPS flow and buffer 2 is for the AD to HPS flow. The direct flow
between the RAS and HPS carries most of the nitrogen and its flow rate is linked
to the transpiration rate (Goddek & Körner, 2019). Most of the phosphorus enters
the HPS as digester effluent, at a relatively constant rate. Figure 3. The four sub-models of the aquaponics model and the dynamic vari-
ables that are exchanged between them. As the nutrient solution is kept within
suitable bounds for the crop, it is assumed that the Water & nutrients sub-model
with DWC (200 L/m2 ) having a far greater volume than NFT (3 L/m2 ) has no further influence on the Plants sub-model.
(Stanghellini et al., 2019). For both technologies the nutrients not taken
up by the plant is returned to the sump for recirculation, forming closed
loop systems. The balance equations of the four sub-models are solved numerically
In this study, the preparation of the nutrient solution is different using the Euler forward method. A time-step of one day was found to be
from standard hydroponics, as it is largely based on a mix of RAS water sufficiently accurate for the Fish production and Water & Nutrients sub-
and digestor effluent, topped up with synthetic fertilizers. By managing models as variation in the nutrient concentration throughout the day
the fertilizer supplementation and adjustable combination of fresh water was not considered in this study. For the Greenhouse climate and Plants
intake and discharge of HPS water the solution can be kept within plant- sub-models, five minutes was chosen as the smallest acceptable time-
specific ranges (Bittsánszky et al., 2016). Transpiration of the crops is step in terms of computational requirements.
the main driver of water transfer between RAS and HPS, as the water The growth of fish in the rearing tanks is calculated using
level in the HPS is kept constant. Given the fact that also the amount the equations from Timmons & Ebeling (2010), as implemented by
of water in the RAS is kept constant, the variation in transpiration (i.e. Dijkgraaf et al. (2019). Assuming the growing environment is optimal,
radiation) results in a varying freshwater intake in the RAS. in terms of e.g. pH, dissolved oxygen and fish feed, the weight gain of the
In addition to the three basic elements, as shown in Figure 1, a re- fish is calculated using only water temperature and several fish-specific
verse osmosis (RO) filtration system can be implemented to further in- parameters.
crease the difference in nutrient concentration between the RAS and Besides being used to calculate the fish yield, the sub-model is used
HPS (Goddek & Keesman, 2018). As a result of the filtration process, a to determine the feed requirements throughout the production cycle. For
nutrient-rich brine is added to the HPS, while the remaining water with this, a feed conversion rate is used, which describes the ratio between
largely reduced nutrient concentrations is fed back to the RAS. In this feed uptake and weight gain, as explained by Dijkgraaf et al. (2019).
study, the system has a fixed hourly flow but is deactivated when either Staggered production of fish provides a smoother nutrient supply
the nitrate or phosphorus concentration in the HPS nears its upper limit, (Goddek et al., 2016). As this study specifically estimates the perfor-
to prevent the need for dilution. The RO filtration system was not used mance over one year of operation, the start-up phase of the RAS is omit-
in most of the scenarios in this study and it is specifically stated in which ted.
cases it was used. The Water & Nutrients sub-model calculates the flows of nutri-
Limited seasonal variation solar radiation climate was found to be ents and water throughout the system. It is based on the work of
beneficial for a balanced and efficient aquaponics system by Goddek & Dijkgraaf et al. (2019), with several corrections (Supplementary Ma-
Körner (2019). In seasonally varied climates, a mismatch can arise be- terials SM 2). The model considers nitrogen and phosphorus, for their
tween the stable nutrient flow from the RAS to the HPS and the varied importance in crop growth and dynamics of RAS and AD. Potassium (K)
nutrient demand of the HPS (between summer and winter). Two addi- was not considered as it makes up an insufficient fraction of the fish feed
tional strategies were tested to achieve the benefits of stable climates, (Shiau & Hsieh, 2001) to influence the system (<0.3% K, compared to
using water-buffering tanks as shown in Figure 2. The RAS to HPS buffer 1% P and 7% N).
and the AD to HPS buffer store nutrient rich water in the winter to be In the HPS, represented by the sub-models Greenhouse climate and
released in summer. Both options aim to better match the supply and Plants, the concentrations of N and P were determined by the incom-
demand of nutrients in the HPS throughout the year. ing flow and the nutrient uptake of the crops. Given the very complex
process of nutrient uptake by plants (Roose et al., 2001; Picart et al.,
2015) and given the overall aim of the study to study the operation
2.2. Aquaponics model and design of a full aquaponics system, the basic assumption that the
nutrient (N and P) uptake by the crops is equal to the product of
For ease of access, the model created in this study was imple- the crop transpiration and nutrient concentration, as in the studies of
mented in Microsoft ExcelTM and consists of various sub-models, as (Dijkgraaf et al, 2019; Goddek & Körner, 2019), was made. Nutrient
shown in Figure 3. While discussed only briefly in the following sec- supplementation or dilution of the HPS water was used to keep the con-
tions, each is described in detail in Supplementary Materials SM 1. centrations within the suitable range for the crop. The ranges used in
For more advanced aquaponics model implementations we refer to this study are 140 – 180 mg/L nitrate (NO3 ) and 40 – 60 mg/L phosphate
Karimanzira et al (2016) and Reyes Lastiri et al (2018). For clarity in (PO4 ), following Resh (2013).
categorizing parameters and performance indicators, a distinction was The main function of the Plants sub-model is to use data from the
made between the aquaponics system (Fish production + Water & Nutri- greenhouse climate (temperature, humidity and radiation) to deter-
ents) and the greenhouse system (Plants + Greenhouse climate). Unless mine the transpiration of the crop, according to Stanghellini and de
stated otherwise, the parameter values used in this study are presented Jong (1995) and (Graamans et al., 2017). Crop transpiration (ETc) de-
in Appendix 1. pends on plant-specific parameters and environment variables. In this
3
L. Jansen and K.J. Keesman Cleaner and Circular Bioeconomy 2 (2022) 100012
Table 1
Parameter description and values for the aquaponics system, based on Timmons & Ebeling (2010), Stanghellini et al. (2019) and Goddek
& Keesman (2018).
HPS area Area of the greenhouse planted with crops. 2000 1000 – 4000 m2
P content feed Mass fraction of phosphorus in the fish feed. 1% 0.8%, 1%, 1.4% −
System volume Water volume of the HPS, relative to HPS area. 200 3, 200 L/m2 HPS
Desalination flow Input flow for reverse osmosis desalination, relative to HPS area. 0 0 – 12 L/m2 HPS/d
Table 2
Parameter description and values for the greenhouse system, based on Stanghellini et al. (2019), Seginer et al. (2020), Kempkes & Janse (2016) and Van Beveren
et al. (2015).
study, the growth of the crops was not affected by the Water & Nutri- the performance of the system. Table 3 lists the KPIs that were used to
ents or Fish production sub-model, as nutrient concentrations were kept evaluate the different parameter settings of the aquaponics system.
within optimal bounds. The unweighted average of the WUE, NUE and NSS is used for a quick
In the Greenhouse climate sub-model, the temperature, humidity and overview and referred to as the ‘average KPI’. As the RAS volume was
light inside the greenhouse are simulated, representing the growing en- kept constant and the fish growing conditions were kept within desired
vironment of the crops. The greenhouse climate is described by physics- bounds, the feed rate and yield of the fish remain unchanged and are
based, dynamic balance equations for uniform temperature (energy) and not displayed.
absolute humidity (mass) of the greenhouse, as done by Van Beveren The KPIs used to evaluate the greenhouse system are shown in
et al. (2015), see Supplementary Materials SM 1. All variables and pa- Table 4 and give information on the transpiration rate, growing envi-
rameters are also explained in SM 1. While the model created for this ronment and energy use.
study is of relatively low complexity, the outcomes fit within those found Because of the explorative character of the study, we used param-
in the literature, as shown in Supplementary Materials SM 1.5. To en- eter sensitivity and scenario analyses to have an indication about the
sure a suitable temperature and humidity for the crops, management influence of individual design and control parameters (Tables 1 and 2)
and control strategies have been modelled, as well. For instance, the on the KPIs.
energy screen is deployed when solar radiation is below a threshold,
as described in Kempkes & Janse (2016). The shade screen is deployed 2.5. Location
when solar radiation exceeds a maximum or when outside temperatures
are close to the maximum allowed greenhouse temperature. Artificial The model was designed to be applicable for different locations with
lighting, if used, is activated between 8 am and 8 pm if solar radiation corresponding climate data (temperature, relative humidity and solar
is below a threshold and is disabled from May to October to save energy. radiation). For this study, the operational strategies and simulated tech-
Temperature and humidity are controlled based on the energy and va- nical equipment have been designed to meet the demands of moderate
por fluxes of the greenhouse in four steps, as displayed in Supplementary and sub cold climates, such as in Canada, Northern Europe and North-
Materials SM 1. ern Asia, which share strong seasonal differences. This study used hourly
weather data from 2017 of de Bilt in the Netherlands (KNMI 2021).
2.3. Design and operation parameters
3. Results
In this study, four aquaponics parameters were varied to investi-
3.1. Fish production
gate their influence on the nutrient dynamics and system performance
(Table 1). Additionally, seven greenhouse parameters were varied for
For all the scenarios of this study, the parameters of the RAS, as
their effect on the energy use and crop transpiration (Table 2).
shown in Supplementary Materials SM1, were left unchanged. As the
For the greenhouse parameters, also two combination scenarios were
growing environment for the fish is kept within optimal ranges, it is
tested. In both scenarios we assumed a well-insulated cover (𝛼 glazing:
assumed that the production is also the same for each scenario. For a
1.7 W/m2 /K) and passive heat storage (0.3m). Combination 1 also had
total tank volume of 100 m3 , the calculated feed rate varied between 34
a lower minimum temperature (10°C) and combination 2 had a higher
and 42 kg per day (14 tons per year) and a yearly fish yield of 11640
maximum humidity (90%).
kg.
For tomato production, the parameters different from lettuce pro-
duction are presented in Table A5.1.
3.2. Aquaponics parameter sensitivity analysis
2.4. Model data analysis First, the general impact that individual parameters (Table 1) have
on the nutrient balances, and by extension, on the system performance
From the large amount of data generated by the model, the most is presented. Figure A2.1 in Appendix 2 provides an additional visual
important outcomes are summarized in a set of key performance indica- representation of the effect of each parameter on the nutrient dynamics
tors (KPIs). Together, the KPIs cover the relevant factors for determining throughout the year.
4
L. Jansen and K.J. Keesman Cleaner and Circular Bioeconomy 2 (2022) 100012
Table 3
KPIs for the aquaponics system.
𝑡𝑟𝑎𝑛𝑠𝑝𝑖𝑟𝑎𝑡𝑖𝑜𝑛
WUE (%) Water Use Efficiency. 𝑤𝑎𝑡𝑒𝑟 𝑖𝑛𝑝𝑢𝑡
𝑢𝑝𝑡𝑎𝑘𝑒
NUE N & P (%) Nutrient Use Efficiency for both nitrogen and phosphorus. 𝑡𝑜𝑡𝑎𝑙 𝑖𝑛𝑝𝑢𝑡
𝑖𝑛𝑝𝑢𝑡 𝑓 𝑟𝑜𝑚 𝑓 𝑖𝑠ℎ
NSS N & P (%) Nutrient Self Sufficiency: fraction of nutrients entering the HPS originating from the fish (as opposed to supplementation). 𝑡𝑜𝑡𝑎𝑙 𝑖𝑛𝑝𝑢𝑡
Table 4
KPIs for the crop and greenhouse.
5
L. Jansen and K.J. Keesman Cleaner and Circular Bioeconomy 2 (2022) 100012
Figure 5. KPIs for different HPS volumes. Left: an NFT system with 3 L/m2 . Right: a DWC system with 200 L/m2 .
Figure 6. Total energy use (heating, dehumidification and lighting) against transpiration variation (CV) of 37 scenarios. Each of the parameters is varied along the
range shown in Table 2, with one or two parameter values present to indicate the direction. The lower graph displays a zoomed-in fraction of the upper graph, as
marked by the dashed rectangle.
early with the power of lighting deployed, until the summer and winter the energy screen is deployed, lowering the heat transfer coefficient of
DLI are almost equal at 200 W/m2 . the greenhouse by approximately 40%. Extending the closed period be-
yond a minimum radiation of 25 W/m2 does not provide additional
benefits.
3.3.3. Energy-saving screen
As energy-saving screens (bottom graph of Figure 6) have become
3.3.4. Diameter of passive heat storage
common in new Dutch greenhouses (Kempkes et al., 2014; Van Beveren
For passive heat storage (PHS) (bottom graph of Figure 6), long
et al., 2015), it is used in the reference scenario. From sunset to dawn,
water-filled cylinders are installed in parallel under the grow beds. As
6
L. Jansen and K.J. Keesman Cleaner and Circular Bioeconomy 2 (2022) 100012
Figure 7. KPI values against the corresponding variation of the transpiration for 37 different scenarios. Total yearly transpiration has been kept equal by resizing
the HPS area. Initial HPS areas: 2000 and 3000 m2 , 1% P content in the feed, DWC.
the volume of the PHS is relatively small and without active control, and NUE, at the expense of NSS, as also shown in Figure 5. For an NFT
it only functions as a short-term energy buffer. Additional storage vol- system, the drop in performance is roughly double.
ume decreases the rate at which the temperature and humidity in the To allow for better comparison between greenhouse parameter set-
greenhouse change, which slightly lowered the energy requirement in tings, the HPS area was resized slightly for most scenarios to maintain
the spring and fall. equal average transpiration, which leaves only the impact of the varia-
tion. This is explained in more detail in Appendix 3.
3.3.5. Minimum daytime temperature
Changing the minimum temperature (bottom graph of Figure 6) has 3.5. Experimental buffering strategies
a minor effect on the transpiration rate and its variation. Lowering the
minimum leads to a substantial decrease in heating demand while the 3.5.1. Buffering RAS to HPS flow
energy required for dehumidification increases slightly. An actively managed buffer between the RAS and HPS stabilizes the
direct flow between the RAS and HPS, effectively lowering the variation
3.3.6. Average heat transfer coefficient of the glazing throughout the year. RAS effluent is added to the buffer when transpi-
An increase in the insulation of the greenhouse (decrease of trans- ration is low and released to the HPS when transpiration is high, as is
fer coefficient) results in a minor decrease in variation as the average shown in Appendix 4. For this study, the average transpiration was used
temperature in winter is higher, paired with large energy savings. to set an upper and lower bound on the flow from the RAS to the HPS.
Increasing the buffer size allows for tighter bounds, resulting in a lower
variation.
3.3.7. Shade screen
In Figure 8 (left), the difference in total phosphate demand of the
In the reference situation, the shade screen was required for the
AP system, i.e. phosphate in buffer 1 and HPS (Figure 2), for the case
warmest days as ventilation cooling could no longer prevent overheat-
without and with a buffer is shown. The total phosphate (P) demand
ing of the crop. When activation of the shade screen is also based on
was calculated such that the phosphate level in the HPS remained be-
the solar radiation, the transpiration and variation decrease, with a ne-
tween 40 – 60 mg/L phosphate (PO4 ), the required range for the crop
glectable change in energy use.
(Resh, 2013). Figure 8 (right) shows the combined KPIs for five scenarios
and different buffer volumes. For the RAS, the average NO3 concentra-
3.3.8. Combination scenarios
tion is slightly decreased with this buffering system, that is 3% for a
Both scenarios, as described in Section 2.3 and shown in top graph of
2000 m2 HPS and 50L/m2 buffer capacity. More importantly, adding
Figure 6, combine insulation with passive heat storage. Also decreasing
buffer 1 decreases the highest peaks in the nutrients concentrations that
the minimum temperature by two degrees further lowers the energy
could result in dilution.
demand, with little effect on variation. Some more savings are achieved
when increasing the maximum humidity, as done in combination 2. Due 3.5.2. Buffering the AD to HPS flow
to the higher maximum humidity, transpiration is reduced mainly in As an alternative to the first case, a buffer between the digester and
winter. Allowing a higher humidity also lowers the dehumidification HPS (buffer 2, Figure 2) can be used to match supply to AP system de-
demand, but it increases the variation. A lower transpiration rate does, mand. This buffer ensures a good match between phosphorus supply
however, increase the risk of tipburn due to insufficient transport of and total (buffer 2 and HPS) demand throughout the year, as shown in
calcium to the leaves (Frantz et al., 2004). Figure 9 (left). For nitrogen, the supply is already to some extent linked
to the transpiration rate, but there is still an improvement with the AD
3.4. Impact of transpiration variation on AP performance buffer. As the nutrients in the AD effluent are far more concentrated, a
relatively small-sized buffer can be used to realize an increase in perfor-
Using the crop transpiration data from the 37 unique greenhouse mance of 4% – 10%. Also, as shown in Figure 9 (right), similar combined
scenarios, the AP model was run to calculate the KPIs related to water KPIs as in Figure 8, were found, but with a much smaller buffer capacity.
and nutrients. In Figure 7 the CV of the transpiration rate was used as
an indicator for the influence of each greenhouse scenario on the AP 3.5.3. Cost estimation
system for HPS areas of 2000 and 3000 m2 , respectively. Both graphs An estimation of the costs to construct both buffer types is made us-
show either no effect or a decrease in the performance indicators with ing interpolated data for the Netherlands (Raaphorst & Benninga, 2019,
an increasing variation. Increasing the HPS area results in a higher WUE p. 75) and is shown in Table 5.
7
L. Jansen and K.J. Keesman Cleaner and Circular Bioeconomy 2 (2022) 100012
Figure 8. Buffering of RAS to HPS flow. Left, the supply and demand for phosphorus are displayed for a 2000 m2 HPS and the reference scenario. For the upper
graph, no buffer is used, resulting in a stable supply and variable demand. For the lower graph, a buffer of 50 L per m2 HPS buffer is used (100 m3 for 2000 m2 )
which stabilizes the flow from the RAS to the HPS, decreasing the difference between supply and demand. In the right-hand graph, the combined KPI is shown for
five GH scenarios with NFT and various buffer volumes.
Figure 9. Buffering of the AD to HPS flow. Left, the supply and demand for phosphorus are displayed for a 2000 m2 HPS. For the upper graph, no buffer is used,
resulting in a stable supply and variable demand. For the lower graph, an 8 L/m2 buffer is used (16 m3 for 2000 m2 ) of which the output is linked to the crop
transpiration rate, and thus, to nutrient demand. In the right graph, the combined KPI is shown for five GH scenarios with NFT and various buffer volumes.
Table 5
Estimated cost ranges for construction of buffers, scaled for various HPS areas. A buffer volume of 50 L/m2 was used
for the RAS to HPS buffer and 8 L/m2 for the AD buffer.
2500 m2
125 m 3
€ 4,000 - 5,100 20 m3
€ 2,150 - 2,750
20000 m2 1000 m3 € 17,000 - 21,000 160 m3 € 4,750 - 6,000
50000 m2 2500 m3 € 23,500 - 30,000 400 m3 € 8,900 - 11,100
Cost per m2 HPS Between €0.50 and €2.05 per m2 Between €0.20 and €1.10 per m2
3.6. Combining greenhouse and aquaponics scenarios an on-demand coupled aquaponics system with anaerobic digester and
potential use of RO filtration system. The research focused on changes
Using the CV of transpiration, all 37 greenhouse scenarios were that could result in energy savings, as it is one of the crucial aspects
combined with the 9 settings of the aquaponics system, as shown in of sustainable food production in northern latitudes. While there are
Figure 10. For visualization, the KPIs of the AP system are combined studies that compare the system performance at different locations, the
as an unweighted average. A similar analysis was made for the tomato specific influence of physical and operational parameters was not yet
crop, which is shown in Appendix 5.2. investigated for aquaponics systems.
The objective of this research was to investigate and demonstrate Goddek & Körner (2019) suggested that a plant production system
the effect of changes in various physical and operational parameters on with a constant nutrient demand (which is more or less equivalent to a
8
L. Jansen and K.J. Keesman Cleaner and Circular Bioeconomy 2 (2022) 100012
constant transpiration), would result in a balanced system with better creased relatively more than the decrease in transpiration rate between
performance. This observation resulted in our hypothesized correlation the different scenarios.
between the (seasonal) variation in transpiration and the performance of
the AP system. It has been found that the coefficient of variation (CV) 4.2. Buffers and their potential for low-volume systems
of the transpiration rate, as representative for the different scenarios
(Figure 6 and 7), was an appropriate factor that highly correlates with The addition of a buffer for the direct flow from the RAS to the HPS
the chosen KPIs of the AP system (correlation of 0.98 to 1 for WUE and reduces the variation in this flow, which leads to a significant increase in
NUE), given that the HPS area is resized as shown in Appendix 3. This performance compared to the reference (Figure 10). However, the vol-
strong correlation also holds for the average KPI, as was demonstrated ume required to have this effect is substantial, with more than 10L/m2
in Figure 10. required for each percentage point of performance increase. The buffer
While increasing variation generally results in decreasing efficiency, for the AD output has shown similar benefits (Figure 9) with a much
the sensitivity to it was not the same for different settings of the smaller volume of 0.6 to 0.75 L/m2 for an increase of one percentage
aquaponics system. Starting from the lowest line in Figures 10, it is point, which greatly reduces the estimated construction costs.
shown that a poor P:N ratio (surplus of P) results in significantly lower The AD to HPS buffer increases performance by linking the (normally
efficiencies than the other combinations, for all levels of variation. The constant) nutrient supply to the seasonally variable demand. While this
NFT system, with its lower system volume, shows the highest sensitiv- mainly influenced phosphorus, the NUE of N was also improved in some
ity to variation (steepest slope). The 3000 m2 HPS scores lower than scenarios due to lower dilution requirements from phosphorus peaks.
the reference due to its relatively low fraction of nutrients supplied by The RAS to HPS buffer flattens the peaks of NO3 in the RAS, which is
the fish (NSS of N & P). The addition of the RO system for a 2500 m2 especially beneficial for sensitive fish.
HPS resulted in only a small performance difference with the reference It is not yet studied how the water quality is affected by the long-
(<1.2%), due to the low availability of nutrients in the RAS when de- term storage suggested here. This concerns both food safety aspects, as
mand is high in summer. Therefore, RO has limited added value in well as chemical reactions that could change the nutrient composition
northern latitudes. The scenarios with buffer show very low sensitiv- during storage. It is expected that both chemical and microbial reactions
ity to transpiration variation, of which an AD buffer for the NFT system continue in the storage tank, which could decrease the water quality and
shows the greatest performance increase (up to 8.5%). For the DWC sys- safety (M. van Eekert, personal communication, March 3rd , 2021). If re-
tem, the AD buffer outperforms the RAS to HPS buffer, despite requiring quired, several adjustments can be made to lower the risk of pathogens
less than a sixth of the volume. in the storage tanks, such as temperature and pH control and disinfec-
The production of tomatoes (Solanum lycopersicum L.) was also simu- tion.
lated and briefly explored to support the theories brought forward in this Buffering the brine flow of the RO system was not explored in this
study. The corresponding methods, results and discussion can be found study, but we expect a similar increase in performance as for the RAS to
in Appendix 5. For a tomato crop, a similar decline in AP efficiency is HPS buffer. While not only resulting in a more concentrated inflow than
found with increasing variation in the seasonal transpiration. However, buffering RAS water, and thus a smaller required buffer, it would allow
the changes in the greenhouse parameters resulted in a smaller differ- the RO to be activated more during the winter. This is when the nutrient
ence in variation than with lettuce. concentrations in the RAS are generally the highest, while RO cannot be
Most of the GH parameters tested in this study had only a small effect used as it would result in a nutrient surplus and subsequent dilution in
on the transpiration variation and could, therefore, be readily changed the HPS. Buffering the brine flow could likely flatten the nitrate peaks
to save energy. This includes the thermal insulation of the cover, min- displayed by Goddek & Körner (2019, figs. 3 & 4).
imum temperature, heat storage, and screens, as shown by the smaller
differences in transpiration in figure 6. Further increasing the bounds 5. Conclusion
of maximum humidity lowered the performance of the aquaponics sys-
tem the most and is, therefore, not advised for low volume systems. In This study showed an elaborate sensitivity and scenario analysis with
agreement with Dijkgraaf et al. (2019), it is in most cases better to have respect to many parameters in an aquaponics greenhouse concerning the
a system that requires supplementation rather than dilution. For scenar- efficient use of resources and energy. The study provides a methodology
ios with a high CV as a result of energy savings, an increase in the HPS to gain insight into some of the many settings and to support a basis for
area is a suitable measure to decrease dilution requirements, with an constrained optimization in design studies.
acceptable loss in nutrient self-sufficiency (NSS). The same can be seen The coefficient of variation (CV) of the daily transpiration values
in the outcomes of Goddek & Körner (2019), where the HPS area is in- was shown to be a good indicator for the impact of different greenhouse
9
L. Jansen and K.J. Keesman Cleaner and Circular Bioeconomy 2 (2022) 100012
settings on an aquaponics system, with the performance of the AP system CRediT authorship contribution statement
generally decreasing with a higher variation. For the tested scenarios of
lettuce production in the Netherlands, the CV ranged from 42% to 75%, Luuk Jansen: Conceptualization, Methodology, Software, Writing –
with a decrease of up to 6.5% and 10% for the nutrient use efficiency original draft, Visualization, Investigation. Karel J. Keesman: Supervi-
(NUE) and nutrient self-sufficiency (NSS) when comparing the lowest and sion, Writing – review & editing.
highest variation. For tomato production, the CV ranged from 40% to
61%, which also resulted in a smaller difference in the NUE and NSS at Acknowledgments
1.4% and 3%, respectively, for the highest and lowest variation.
For an aquaponics system in the Netherland, the benefits of energy- This work is part of the (i) research program SUGI/Food-Water-
saving techniques are likely to outcompete the decreased efficiency of Energy-Nexus “CITYFOOD”, which is (partly) financed by the Nether-
the aquaponics system. This study shows that several parameters, such lands Organization for Scientific Research (NWO) with project number
as thermal insulation of the cover, minimum indoor temperature, the 438-17-402, and (ii) BlueCycling project that has received funding from
addition of heat storage and use of energy/ shade screens, could be the European Union’s Horizon 2020 research and innovation program
changed to increase the energy efficiency of the greenhouse with little under grant agreement No 862555 within the 2019 Joint Call of the
effect on the nutrient dynamics. ERA-NET Cofund on Food Systems and Climate. We also like to thank
The use of buffers could greatly improve the water and nutrient use Dr Feije Zwart (Wageningen UR Greenhouse Horticulture) for his sug-
efficiencies of the aquaponics system, especially in high CV scenarios. gestions on greenhouse modelling.
Buffering the effluent of the anaerobic digester raised the overall effi-
ciency of the aquaponics system from 86% to 94%, at an estimated cost Supplementary materials
between 0.20 – 1.10 euro/m2 of cultivated area.
Supplementary material associated with this article can be found, in
the online version, at doi:10.1016/j.clcb.2022.100012.
Declaration of Competing Interest
Appendix 1. Initial model settings
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influence Initial parameter values are shown in Table A1.1. These are the val-
the work reported in this paper. ues used in the study for each parameter, unless specifically stated oth-
erwise.
Table A1.1
The initial settings for the parameters used in the model.
10
L. Jansen and K.J. Keesman Cleaner and Circular Bioeconomy 2 (2022) 100012
Table A1.2
Initial values for several variables.
Startup-time 252 days Time until all tanks are in use, with an 18-day interval
RAS TAN 0.5 mg/L
RAS NO3 50 mg/L
RAS P 0.7 mg/L
HPS N 160 mg/L Middle of optimum range
HPS P 50 mg/L Middle of optimum range
Fish biomass 1870 kg Biomass at start-up, determined
As the start-up phase is excluded in this study, the following values shown in Figure A3.3. If the reference transpiration is 33% lower than
are used at the start of the modeled year (Table A1.2). in the selected scenario, decreasing the HPS area of that scenario by
33% will result in the same total transpiration. Other parameters (LAI,
Appendix 2. Nutrient overviews for different aquaponics settings cropping density) remain unchanged and transpiration per square meter
is not affected significantly within the range of 1000 to 3000 m2.
Figure A2.1 shows the dynamic effect of several of the aquaponics Using this method to maintain equal total transpiration, the effect
settings, as described in Section 3.1 Aquaponics parameters on the nu- of variation shown clearly in the right graph of Figure A3.2. With only
trient concentrations in the HPS throughout the year. Notice that the the variation causing the difference in performance, an average KPI of
reference trajectories are shown in Figure A2.1-5 (central subfigure). 94.0% was estimated for the lowest and 90.0% at the highest variation.
Each of the graphs in this figure represents a different setting of the For an NFT system, with a lower buffering capacity, the decrease in
aquaponics system: performance is roughly doubled, with the average KPI decreasing from
92.3% at the lowest variation to 84.3% at the highest variation.
1) Decreasing the phosphorus concentration from 1% to 0.8%.
2) Increased transpiration variation, which could be caused by changes
in the greenhouse system (Section 3.3 Greenhouse parameter sensitivity Appendix 4. Buffer operation and sensitivity analysis
analysis).
3) Lower water volume in the HPS, which increases the sensitivity to tran- 4.1. RAS to HPS buffering
spiration variation.
4) Decreased HPS area from 2000 m2 to 1500 m2 , changing the RAS: HPS In Figure A4.1 the operation and effect of a RAS to HPS buffer is
ratio to 15 m2 /m3 . shown. This buffer creates an upper and lower limit for the direct flow
5) Reference scenario: 2000 m2 HPS, DWC, CV of 66%, P content of 1%, from the RAS to the HPS. If this flow is too high, water is diverted to
no desalination. the buffer. Conversely, if the flow is too low, water is released from
6) Increased HPS area from 2000 m2 to 3000 m2 , changing the RAS: HPS the buffer. The result is a more constant flow of RAS water to the HPS,
ratio to 30 m2 /m3 . reducing volatility in the nutrient concentrations of both the RAS and
7) Desalination module added to the RAS, with a maximum flow of 3m3 /d. the HPS.
This equates to 1.5 L/m2 HPS/d. While Figure A4.1 shows the optimal operation with the chosen
8) Decreased transpiration variation (CV of 40%), either through buffer- buffer volume, this is based on information that would normally not
ing (Section 3.5 Experimental buffering strategies) or adjustments of the be available. While a poor estimate of the average transpiration rate re-
greenhouse system (Section 3.3 Greenhouse parameter sensitivity analy- sults in a lesser performance improvement, the day-to-day timing and
sis). buffered volume are of lesser importance. Basing the flow to and from
9) Increased phosphorus concentration from 1% to 1.4%. the buffer on the previous five days did not decrease performance sig-
nificantly and a large difference between the start and end could still be
Appendix 3. Resizing the HPS area to maintain equal prevented in the final months of the year.
transpiration
Using the crop transpiration data from the 37 unique greenhouse 4.2. AD to HPS buffering
scenarios, the AP model was run to estimate the performance indica-
tors related to the nutrients and water. The coefficient of variation (CV) Optimal control of the incoming and outgoing flow of the buffer,
of the daily transpiration values was used as to quickly compare the again, requires some predictions of the coming year. An estimate of the
different GH scenarios. However, these scenarios also had different av- total yearly transpiration is required to prevent a surplus or deficit of
erage transpiration rates, e.g. because of extra transpiration in winter buffer water at the end of the year. While the decision-making for filling
with artificial lighting or reduced transpiration with additional shading. and discharging the buffer is currently made only based on the transpira-
Figure A3.1 displays the strong correlation (-0.93) between the average tion rate, a more advanced approach could consider the concentrations
transpiration and the variation for the calculated scenarios. in the HPS, lowering the reliance on estimations. The results of a sensi-
A change in the average transpiration rate has a significant effect on tivity analysis (Table A4.1) show how well a poor guess for the first half
the nutrient dynamics. In practice, it has a similar effect to a change in of the year can be recovered from.
the HPS area, where WUE and NUE increase with high transpiration and Overall, the performance difference between the worst and optimal
NSS increases at low transpiration rates. As this effect is stronger than case was found to be small, with the worst-case performing better than
the negative effect of variation, Figure A3.2 (left) shows increasing NSS the case without a buffer for all parameters. This would suggest that the
N & P with higher variation. control of the buffer does not necessarily have to be as precise as done
To distill the effect of transpiration variation on performance, the in this study.
changes in the average transpiration were canceled out. To maintain As the HPS area has been adjusted for the total transpiration of each
equal average transpiration for the different scenarios, the HPS area was greenhouse scenario, the initial HPS size (2000 m2 ) was used to calcu-
resized inversely proportional to the change in average transpiration, as late the total buffer volume (8 L/m2 ).
11
L. Jansen and K.J. Keesman Cleaner and Circular Bioeconomy 2 (2022) 100012
Figure A2.1. Nutrient concentrations, dilution and supplementation throughout the year for different aquaponics settings. For each of the nine graphs the yellow
line on top displays the nitrogen concentration in the HPS throughout the year, while the green line at the bottom displays the phosphorus concentration (left
axis). The concentrations are kept within their respective bounds (dotted lines). In the middle, the nitrogen and phosphorus added through supplementation and
removed through dilution are shown (right axis). In general, a higher sum of the dilution and supplementation results in lower performance (of WUE/ NUE and NSS
respectively).
12
L. Jansen and K.J. Keesman Cleaner and Circular Bioeconomy 2 (2022) 100012
Table A4.1
Performance differences for the AD buffer with a poor estimate of the average transpiration.
Reference - No
buffer Difference between guessed average and the actual average for the first half of the year difference optimum - difference no
worst case buffer - worst case
-20% -15% -10% 0 +10% +15% +20%
WUE 89.5% 95.0% 96.1% 96.1% 96.1% 96.1% 96.1% 95.2% -1.2% 6.1%
NUE N 90.3% 92.5% 92.9% 92.9% 92.9% 92.9% 92.9% 92.5% -0.4% 2.4%
NSS N 89.1% 94.8% 95.2% 95.1% 95.0% 95.1% 95.2% 95.5% -0.1% 6.4%
NUE P 89.2% 93.7% 94.6% 94.6% 94.6% 94.6% 94.6% 93.9% -1.0% 5.0%
NSS P 96.3% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 0.0% 3.9%
KPI avg 90.9% 95.2% 95.8% 95.7% 95.7% 95.7% 95.8% 95.4% -0.5% 4.7%
Table A5.1
Crop specific parameters
Crop rs (s/m) rb (s/m) LAI start LAI max Time to LAI max Growing season (days)
Lettuce 100 – 450, following eq. SM1.4 200 0.2 1.6 40 45, staggered
Tomato 200-450 following eq. A5.1 200 0.2 3.5 80 360
almost constant LAI due to the staggered production, the tomato crops
modeled in this study do not. The plants start with a small LAI, after
which a linear increase is assumed until it reaches its maximum, as
shown in Table A5. At the end of the year, the plants are removed to
make space for the plants of the following year, resulting in several days
without transpiration. The ranges for the nutrients remained the same
as for lettuce, as this roughly covers the nutrient demand of the different
growth phases (Resh, 2013).
As not all of the parameters for equations SM1.4 and SM1.5 were
available for tomato production, different equations were used for the
Figure A3.3. The concept of resizing the HPS area to maintain equal total tran-
stomatal resistance and net radiation on the crop. The stomatal resis-
spiration. If transpiration increases from 2 to 3 L/m2 /d, the HPS area is de-
tance (rs ) was calculated based on the temperature, radiation and the
creased from 300 to 200 m2 to maintain the total transpiration of 600 L/d.
leaf area index (LAI), as in Van Beveren et al. (2015) and the net ra-
diation on the crop (Rnet ) was calculated with an estimation of the soil
Appendix 5. Tomato production parameters, results and cover, as done by Stanghellini et al. (2019).
discussion ( ) ( ( )2 )
𝑟𝑠 = 82∗ 1 + 6.95∗𝑒−0.4∗𝐼𝑠𝑢𝑛 ∕𝐿𝐴𝐼 ∗ 1 + 0.23∗ 𝑇𝑎𝑖𝑟 − 20 (A5.1)
( )
The same model and procedures were used to assess the performance 𝑅𝑛𝑒𝑡 = 0.86∗ 1 − 𝑒−0.7∗𝐿𝐴𝐼 ∗𝐼𝑠𝑢𝑛 (A5.2)
of an aquaponics system with a tomato crop (Solanum lycopersicum L.).
The differences in the methodology compared to lettuce, as well as the 5.2. Results
results and discussion are presented in this Appendix.
Compared to lettuce, the transpiration relative to the leaf area is
5.1. Methodology higher for low radiation (mainly in winter and at night) and lower for
high radiation, as shown in Figure A5.1. In total, the transpiration is
In terms of operational parameters, the minimum temperature was higher and less varied for tomato plants than for lettuce.
changed from 12 to 18°C and the substrate on which the tomatoes Figure A5.2 displays the average AP KPI for all 37 greenhouse sce-
(Solanum lycopersicum L.) are grown (rockwool) has different water hold- narios (represented by their corresponding CV, x-axis) and 7 settings for
ing capacity at 10 L/m2 (Stanghellini et al., 2019). While lettuce has an the aquaponics system (lines).
13
L. Jansen and K.J. Keesman Cleaner and Circular Bioeconomy 2 (2022) 100012
Figure A5.1. The estimated LAI development of a tomato crop (right) as well as daily transpiration values of both tomato and lettuce throughout the year. The
tomato crop has a small CV of 48%, compared to 66% for lettuce.
Figure A5.2. The average KPI for 7 AP settings (lines) against the CV of 37 GH scenarios with tomato production. With the plants grown in rockwool, the water
volume in the system is similarly small to NFT at 10 L/m2. The HPS sizes of 1350, 1650 (reference) and 1950 m2 result in roughly the same total transpiration as
the lettuce crop with HPS areas of 2000, 2500 and 3000 m2 respectively. The legend is sorted to match the order in which the lines appear. The full dataset is shown
in Supplementary materials SM 3.2.
Like the lettuce crop, the average KPI decreases with variation, as For tomato production, the CV ranges from 40% to 61%, which also
shown in Figure. The transpiration variation is generally lower than for results in a smaller difference in the NUE and NSS at 1.4% and 3%,
lettuce, as was also seen in Goddek & Körner (2019). As all AP settings respectively, for the highest and lowest variation. The difference caused
had the same system volume of 10 L/m2 , there was little difference in by the variation in transpiration is smaller for tomato than for lettuce.
sensitivity to variation, except for the AD buffer. Like with lettuce, the use of an AD buffer greatly decreased the influence
of transpiration variation.
14
L. Jansen and K.J. Keesman Cleaner and Circular Bioeconomy 2 (2022) 100012
References Kempkes, Frank, & Janse, J. (2016). Teelt en energie 2SaveEnergy kas (GTB-1402).
https://edepot.wur.nl/385096
Baganz, G.F.M., Junge, R., Portella, M.C., Goddek, S., Keesman, K.J., Baganz, D., Kloas, W., Groß, R., Baganz, D., Graupner, J., Monsees, H., Schmidt, U., Staaks, G., Suhl, J.,
Staaks, G., Shaw, C., Lohrberg, F., Kloas, W., 2021. The aquaponic principle—It is Tschirner, M., Wittstock, B., Wuertz, S., Zikova, A., Rennert, B., 2015. A new con-
all about coupling. Rev. Aquac. raq. 12596. doi:10.1111/raq.12596. cept for aquaponic systems to improve sustainability, increase productivity, and re-
Bakker, J.C., 2009. Energy Saving Greenhouses. Chronica Horticulturae 49 (2), 19–23. duce environmental impacts. Aquaculture Environment Interactions 7 (2), 179–192.
https://research.wur.nl/en/publications/energy-saving-greenhouses. doi:10.3354/aei00146.
Bittsánszky, A., Uzinger, N., Gyulai, G., Mathis, A., Junge, R., Villarroel, M., Kotzen, B., KNMI. (n.d.). KNMI - Uurgegevens van het weer in Nederland. Retrieved February 27, 2021,
Kőmíves, T., 2016. Nutrient supply of plants in aquaponic systems. Ecocycles 2 (2), from https://www.knmi.nl/nederland-nu/klimatologie/uurgegevens
17–20. doi:10.19040/ecocycles.v2i2.57. Lennard, W., Goddek, S., 2019. Aquaponics: The Basics. In: Goddek, S., Joyce, A.,
Craig, S., Helfrich, L.A., Kuhn, D., Schwarz, M.H.others, 2017. Understand- Kotzen, B., Burnell, G.M. (Eds.), Aquaponics Food Production Systems: Combined
ing fish nutrition, feeds, and feeding. Virginia Cooperative Extension. Aquaculture and Hydroponic Production Technologies for the Future. Springer In-
https://fisheries.tamu.edu/files/2019/01/FST-269.pdf. ternational Publishing, pp. 113–143. doi:10.1007/978-3-030-15943-6_5.
David, L.H., Pinho, S.M., Agostinho, F., Costa, J.I., Portella, M.C., Keesman, K.J., Garcia, F., Maucieri, C., Nicoletto, C., Os, E.van, Anseeuw, D., Havermaet, R.Van, Junge, R, 2019.
2022. Sustainability of urban aquaponics farms: An emergy point of view. Journal of Hydroponic Technologies. In: Goddek, S., Joyce, A., Kotzen, B., Burnell, G.M. (Eds.),
Cleaner Production 331, 129896. Aquaponics Food Production Systems: Combined Aquaculture and Hydroponic Pro-
De Zwart, H.F., 1996. Analyzing energy-saving options in greenhouse cultivation using a duction Technologies for the Future. Springer International Publishing, pp. 77–110.
simulation model. Wageningen University PhD Thesis. doi:10.1007/978-3-030-15943-6_4.
Delaide, B., Monsees, H., Gross, A., Goddek, S., 2019. Aerobic and Anaerobic Treat- Nedelciu, C.E., Ragnarsdottir, K.V., Schlyter, P., Stjernquist, I., 2020. Global phosphorus
ments for Aquaponic Sludge Reduction and Mineralisation. In: Goddek, S., Joyce, A., supply chain dynamics: Assessing regional impact to 2050. Global Food Security 26,
Kotzen, B., Burnell, G.M. (Eds.), Aquaponics Food Production Systems: Combined 100426. doi:10.1016/j.gfs.2020.100426.
Aquaculture and Hydroponic Production Technologies for the Future. Springer In- Neto, R.M., Ostrensky, A., 2015. Nutrient load estimation in the waste of Nile tilapia
ternational Publishing, pp. 247–266. doi:10.1007/978-3-030-15943-6_10. Oreochromis niloticus (L.) reared in cages in tropical climate conditions. Aquaculture
Dijkgraaf, K.H., Goddek, S., Keesman, K.J., 2019. Modeling innovative aquapon- Research 46 (6), 1309–1322. doi:10.1111/are.12280.
ics farming in Kenya. Aquaculture International 27 (5), 1395–1422. Picart, D., Ozier-Lafontaine, H., Omrane, A., Louison, L., 2015. Modeling plant nutrient
doi:10.1007/s10499-019-00397-z. uptake: Mathematical analysis and optimal control. Evolution Equations and Control
Eding, E.H., Janssen, K., Heinsbroek, L.T.N., Verreth, J.A.J., Schrama, J.W., 2012. Can Theory 4 (2), 193–203. doi:10.3934/eect.2015.4.193.
water phosphorus level in recirculating aquaculture systems (RAS) compensate for Raaphorst, M. G. M., & Benninga, J. (2019). Kwantitatieve Informatie voor de Glastuinbouw
low dietary phosphorus level in nile tilapia (Oreochromis niloticus)? In: Proceedings 2019.
of the Ninth International Conference on Recirculating Aquaculture, pp. 34–35. Resh, H.M., 2013. Hydroponic Food Production. CRC Press.
Elings, A., Kempkes, F.L.K., Kaarsemaker, R.C., Ruijs, M.N.A., Van De Braak, N.J., Reyes Lastiri, D., Geelen, C., Cappon, H.J., Rijnaarts, H.H.M., Baganz, D., Kloas, W., Ka-
Dueck, T.A., 2005. The energy balance and energy-saving measures in green- rimanzira, D., Keesman, K.J, 2018. Model-based management strategy for resource
house tomato cultivation. Acta Horticulturae 691, 67–74. doi:10.17660/ActaHor- efficient design and operation of an aquaponic system. Aquacultural Engineering 83,
tic.2005.691.5. 27–39. doi:10.1016/j.aquaeng.2018.07.001.
FAO. (n.d.). Nile tilapia - Oreochromis niloticus. Retrieved November 10, 2020, from Roose, T., Fowler, A., Darrah, P., 2001. A mathematical model of plant nutrient uptake. J
http://www.fao.org/fishery/affris/species-profiles/nile-tilapia/nile-tilapia-home/en/ Math Biol 42, 347–360. doi:10.1007/s002850000075.
Frantz, J.M., Ritchie, G., Cometti, N.N., Robinson, J., Bugbee, B., 2004. Explor- Searchinger, T., Waite, R., Hanson, C., Ranganathan, J., Dumas, P., Matthews, E., & Klirs,
ing the limits of crop productivity: Beyond the limits of tipburn in lettuce. C. (2019). Creating a sustainable food future: A menu of solutions to feed nearly 10
Journal of the American Society for Horticultural Science 129 (3), 331–338. billion people by 2050. Final report.
doi:10.21273/jashs.129.3.0331. Seginer, I., van Straten, G., van Beveren, P.J.M., 2020. Day-to-night heat storage in green-
Goddek, S., Delaide, B.P.L., Joyce, A., Wuertz, S., Jijakli, M.H., Gross, A., Eding, E.H., houses: 4. Changing the environmental bounds. Biosystems Engineering 192, 90–107.
Bläser, I., Reuter, M., Keizer, L.C.P., Morgenstern, R., Körner, O., Verreth, J., doi:10.1016/j.biosystemseng.2020.01.005.
Keesman, K.J., 2018. Nutrient mineralization and organic matter reduction perfor- Semple, L., Carriveau, R., Ting, D.S.-K., 2017. Assessing heating and cooling demands of
mance of RAS-based sludge in sequential UASB-EGSB reactors. Aquacultural Engi- closed greenhouse systems in a cold climate. International Journal of Energy Research
neering 83, 10–19. doi:10.1016/j.aquaeng.2018.07.003. 41 (13), 1903–1913. doi:10.1002/er.3752.
Goddek, S., Espinal, C.A., Delaide, B., Jijakli, M.H., Schmautz, Z., Wuertz, S., Shiau, S.-Y., Hsieh, J.-F., 2001. Quantifying the dietary potassium requirement of juvenile
Keesman, K.J., 2016. Navigating towards decoupled aquaponic systems: A system dy- hybrid tilapia (Oreochromis niloticus×O. aureus). British Journal of Nutrition 85 (2),
namics design approach. Water (Switzerland) 8 (7), 1–29. doi:10.3390/W8070303. 213–218. doi:10.1079/bjn2000245.
Goddek, S., Keesman, K.J., 2018. The necessity of desalination technology for design- Stanghellini, C., de Jong, T., 1995. A model of humidity and its applica-
ing and sizing multi-loop aquaponic systems. Desalination 428 (May 2017), 76–85. tions in a greenhouse. Agricultural and Forest Meteorology 76 (2), 129–148.
doi:10.1016/j.desal.2017.11.024. doi:10.1016/0168-1923(95)02220-R.
Goddek, S., Körner, O., 2019. A fully integrated simulation model of multi-loop aquapon- Stanghellini, C., Van ’t Ooster, B., Heuvelink, E., 2019. Greenhouse hor-
ics: A case study for system sizing in different environments. Agricultural Systems ticulture. Greenhouse horticulture. Wageningen Academic Publishers
171, 143–154. doi:10.1016/j.agsy.2019.01.010. doi:10.3920/978-90-8686-879-7.
Graamans, L., van den Dobbelsteen, A., Meinen, E., Stanghellini, C., 2017. Plant fac- Sugiura, S.H., Marchant, D.D., Kelsey, K., Wiggins, T., Ferraris, R.P., 2006. Effluent profile
tories; crop transpiration and energy balance. Agricultural Systems 153, 138–147. of commercially used low-phosphorus fish feeds. Environmental Pollution 140 (1),
doi:10.1016/j.agsy.2017.01.003. 95–101. doi:10.1016/j.envpol.2005.06.020.
Karimanzira, D., J.Keesman, K, Kloas, W, Baganz, D, Rauschenbach, T, 2016. Dynamic Timmons, M., Ebeling, J., 2010. Recirculating Aquaculture Systems. In: Tidwell, J.H. (Ed.),
modeling of the INAPRO aquaponic system. Aquacultural Engineering 75, 29–45. Aquaculture Production Systems. Ithaca Publishing Company, Ithaca, pp. 245–277.
doi:10.1016/j.aquaeng.2016.10.004. doi:10.1002/9781118250105.ch11.
Kempkes, F., De Zwart, H.F., Munoz, P., Montero, J.I., Baptista, F.J., Giuffrida, F., Gilli, C., Van Beveren, P.J.M., Bontsema, J., Van Straten, G., Van Henten, E.J., 2015. Minimal
Stepowska, A., Stanghellini, C., 2017. Heating and dehumidification in production heating and cooling in a modern rose greenhouse. Applied Energy 137, 97–109.
greenhouses at northern latitudes: Energy use. Acta Horticulturae 1164, 445–452. doi:10.1016/j.apenergy.2014.09.083.
doi:10.17660/ActaHortic.2017.1164.58. Zandvakili, O.R., Barker, A.V., Hashemi, M., Etemadi, F., Autio, W.R., 2019. Com-
Kempkes, F.L.K., Janse, J., Hemming, S., 2014. Greenhouse concept with high insulating parisons of commercial organic and chemical fertilizer solutions on growth
double glass with coatings and new climate control strategies; From design to results and composition of lettuce. Journal of Plant Nutrition 42 (9), 990–1000.
from tomato experiments. Acta Horticulturae 1037, 83–92. doi:10.17660/ActaHor- doi:10.1080/01904167.2019.1589505.
tic.2014.1037.6.
15