Tomato Yield Under Different Shading Levels in an Agrivoltaic Greenhouse in Southern Spain

Abstract

Agrivoltaic greenhouses in southern Spain offer a sustainable way to manage excessive irradiance levels by generating renewable energy. This study presents a shading experiment on tomato cultivation in a raspa-y-amagado greenhouse in Almeria, southern Spain, during the 2023–2024 growing season. Photovoltaic modules were mimicked by opaque plastic sheets that were arranged in a checkerboard pattern on the roof of the greenhouse. Two shading zones (30% and 50% roof cover ratio) were compared against an unshaded control zone. Microclimate, plant physiology, yield and quality were monitored during the study. The results show that shading influenced the microclimate, which directly impacted crop yield. The 30% and 50% shading zones resulted in 15% and 26% crop yield reductions, respectively. A preliminary, theoretical analysis of potential revenues of the photovoltaic yield showed that reductions in crop yield can be overcompensated by the energy generated by the PV system. For the summer crop cycle, a higher PV production and lower crop yield reductions can be expected. The economic advantage demonstrates the potential of agrivoltaic greenhouses in southern Spain.

1. Introduction

During the last few decades exponential growth in installed photovoltaic (PV) capacity has been observed worldwide. In order to reach the targets of the Paris Agreement and combat climate change, many countries are progressively increasing their share of renewable energies. Since large-scale PV systems have proven to be a green, cheap and reliable energy source, increasing PV energy production is a suitable approach for this task. The rapid growth of PV systems has thereby triggered direct land use changes and ground-mounted PV systems have been installed on arable land formerly used for food production [1]. However, renewable energy does not threaten agricultural production; indeed, they can coexist in harmony as illustrated in the concept of agrivoltaics (APVs), describing the combination of agriculture and PV energy production on the same plot of land [2]. APVs are a promising solution when it comes to meeting the growing demand for both food and energy while also addressing land use conflicts caused by expanding PV installations on farmland [3].

One way to realize APV systems are PV greenhouses (GHs), where PV panels are installed on the roof or inside the GH [4,5]. In southern Spain, especially in the region of Almeria, over 30,000 ha of land are occupied by GHs [6]. According to the European Commission’s PVGIS database, typical values of annual global horizontal irradiation in the Almeria region range between 1900 kW h m⁻² and 2100 kW h m⁻² per year [7,8]. These local environmental conditions make year-round crop production in GHs possible.

As photosynthesis is the key process in the production of biomass, sunlight is one of the most important environmental factors influencing plant growth and eventually yield in GHs [9]. For photosynthesis, plants capture a fraction of the total radiation spectrum, the so-called photosynthetic active radiation (PAR) spectrum. The PAR spectrum includes wavelengths in the range of 400 nm to 700 nm and is quantified by the photosynthetic photon flux density (PPFD [μmol m⁻² s⁻¹]). For tomatoes, light not only fuels photosynthesis but also affects plant morphology, flowering, fruit set and ripening. Depending on the stage of plant development, tomatoes typically require 400 μmol m⁻² s⁻¹ to 700 μmol m⁻² s⁻¹ of PAR for efficient photosynthesis [10]. Besides light intensity, the directional light distribution in the GH throughout the canopy also plays an important role [11]. Several studies showed that plants use diffuse light, which leads to a more homogeneous light profile in the plant canopy, more efficiently than direct light [12,13,14]. Furthermore, the photosynthetic rate of leaves shows a nonlinear response to the light flux density, converging towards a saturation point in which a further increase in PAR radiation does not translate into a higher value of photosynthetic rate [15]. The light saturation point for tomato photosynthesis is around 700 μmol m⁻² s⁻¹ to 1000 μmol m⁻² s⁻¹, above which photosynthesis plateaus or decreases due to stress or photoinhibition. In addition, the exposure of tomatoes to intense light for long periods affects the accumulation of antioxidants in the fruits, like lycopene and Vitamin C, especially during the ripening stage, as these elements are sensitive to oxidation. It has been shown that excessive light may lead to oxidative degradation, while moderate shading can enhance or preserve these compounds in tomato fruit by reducing stress and photoinhibition [16,17].

As high light levels usually lead to photosynthetic saturation and a decrease in light use efficiency, farmers actively manage light and heat inside GHs [6]. Typically, plastic covers for the GHs are used for this purpose [6]; however, excessive sunlight requires farmers to further reduce the transmission of GH plastics throughout the crop cycles, often by applying white chalk paint [18]. In addition, there are other techniques to reduce light transmission and manage the GH microclimate, such as the use of shading nets (applied internally or externally) [19], thermal screens [20], or even wavelength-selective plastic films, that block out a fraction of the infrared (IR) spectrum [21,22]. APV concepts offer another option to actively support this microclimate management of the growers. Shading techniques, both traditional shading nets or APV structures, can modify light intensity, spectrum and duration, affecting both yield and quality of the GH crops. To maintain successful GH cultivation, the balance between sufficient light for photosynthesis and avoiding stress-induced degradation is crucial. Traditional GH plastics, such as low-density polyethylene films, are often multi-layered with an ultraviolet (UV)-protection film toward the outside and agrochemical-resistant films toward the inside of the GH [23]. Depending on their material composition, they mostly block out UV radiation and transmit important PAR. Traditional shading methods, such as shading nets or whitening agents, primarily reduce light intensity and slightly alter the spectrum by scattering more UV and blue light and blocking out a fraction of IR light. In contrast, traditional PV modules absorb a portion of PAR, especially in the blue and red regions, to generate electricity. As this may negatively affect the availability of PAR to the crops, more recent approaches for GH-implemented APV solutions focus on semi-transparent or spectrally selective PV technologies in order to allow more PAR transmission while still producing energy. It is therefore crucial to study the effect of shading caused by PV panels in APV GHs and the availability and distribution of PAR inside APV GHs.

In APV systems, irradiance could also be used to produce energy and the grower’s dependence on rising energy prices could be reduced [24]. According to Dupraz et al. [2], the productivity of APV GHs can increase up to 73% for crops benefiting from low-light conditions. A study by Hanrieder et al. demonstrated that a theoretical maximum PV coverage of approximately 44% in Almeria is feasible while ensuring that irradiance levels inside the GH remain above the typical irradiance thresholds necessary for healthy plant growth [25].

Nevertheless, the integration of PV panels into the GH roof can create a significant reduction in solar radiation inside the GH with possible effects on microclimate and subsequently crop production [26,27,28]. Several APV GH experiments and numerical simulations of APV GHs have been performed in recent years, and, depending on the local climate, these experiments reported different observations based on GH type and investigated crop. Marrou et al. observed that when mean daily air temperature and relative humidity (RH) in a GH were similar to full-sun conditions, the growth rate of some crops, such as lettuce and cucumber in summer, was not significantly different [29]. Moreover, no reduction in lettuce yield was observed under a 50% and 70% reduction in incoming radiation in summer [30]. Cossu et al. [9] found that each 1% of roof coverage by PV modules reduces transmitted irradiance by approximately 0.8%. A checkerboard PV arrangement and a north–south orientation of the GH enhance light uniformity beneath the panels, potentially reducing shading stress on crops [9]. López-Díaz et al. [23] reported that shading levels above 30% delayed yield and reduced fruit quality in a Venlo GH in the Almeria region. The studies of Ezzari et al. [31] and Sánchez et al. [32] found that a 10% checkerboard PV cover did not significantly affect tomato yield in a Mediterranean GH. Beyond yield effects, shading can modify plant morphology, flowering time and fruit quality [26]. In Mediterranean regions, where solar radiation typically satisfies crop requirements of high-light demanding crops, such as tomatoes, quality parameters such as Soluble Solids Content (SSC), color, firmness and pH remain largely unaffected by rooftop PV shading [32,33]. However, shading levels exceeding 40% have been shown to increase fruit firmness while reducing SSC in southeastern Spanish GH production systems [19].

The results of the experimental studies show that there is no simple solution for a perfect APV GH and it is difficult to predict crop responses to artificial shading. In order to improve the understanding of APV systems and their influence on crop yield, modeling approaches are increasingly being employed in APV. Crop growth models enable simulating plant biomass accumulation, fruit yield and other physiological processes over time. However, for each application the model should be calibrated to local conditions, taking into account GH structure, climate variations and crop-specific physiological parameters. In the case of tomatoes, the Tomato Growth (TOMGRO) model of Jones et al. and its adaptations (e.g., the reduced TOMGRO model) have been widely validated for GH cultivation and tested under different environmental scenarios [34]. Only recently have these modeling frameworks been applied in the context of APV systems, often focusing on the interplay between shading, temperature and solar irradiance on crop growth [35,36].

The present study addresses the research gap in experimental studies on APV GH with high shading ratios and provides a comprehensive analysis of partial shading using opaque dummy PV modules arranged in a checkerboard pattern on a raspa-y-amagado GH in Almeria. The objective of the study is to enhance the understanding of changes in plant development and impacts on crop yield in order to support the implementation of PV modules on GH. The radiation distribution, microclimate and the yield and quality of tomato plants in three different zones of the GH, with 50%, 30% and 0% (control) roof cover ratios, are experimentally evaluated. The experimentally measured data helps to improve our understanding of shading-related changes in tomato crops early on in the growing cycle. The study results can help to reduce shading-related yield reductions in future APV GH installations and motivate further research on APV GHs. Furthermore, the experimentally collected data was used to apply the reduced TOMGRO model under the specific shading conditions. The simulated dry weight of mature tomato fruits is compared with experimentally measured data. The application of the reduced TOMGRO model to APV GHs enables the simulation of detailed crop development and adaptations in physiological processes without the explicit need for experimental studies. The model can be used to simulate theoretical crop yield of future APV GHs and help further improve our understanding of crop responses to APV installations.

Section 2 describes the experimental setup, including the GH layout and the PV module pattern, as well as the data acquisition and measurement protocols. The modeling approach using the reduced TOMGRO model is presented in Section 2.4. Section 3 reports the measured and simulated results, and Section 4 discusses the results in the context of different studies. Section 5 summarizes the main conclusions and gives directions for future research.

2. Materials and Methods

2.1. Experimental Greenhouse

The experiment was conducted during the 2023–2024 tomato growing season in a raspa-y-amagado GH in Almeria at Fundación Finca Experimental UAL-Anecoop (36.864° N, −2.282° E, 180 m above sea level), covering a total area of 1765 m². Due to their low construction and maintenance costs, raspa-y-amagado GHs are the most common GH type in the Almeria region [6]. Figure 1 shows the location of the experimental test GH within the Almeria region.

The cultivation cycle started with the transplantation of ‘Lygalan’ tomatoes, a novel variety developed by ANECOOP S.COOP [37], into the GH on 16 September 2023. Pruning was carried out on 9 January 2024 (115 days after transplantation). The last harvest took place on 15 March 2024 (181 days after transplantation).

The GH structure consisted of galvanized steel tubes, featuring a central corridor and two parcels 8 m in width expanding towards each side of the corridor. The roof’s lowest points (the corridor) were at a height of 3.4 m, while the highest points at the gables reached 4.5 m. The roof and walls were covered with a multilayered polyethylene film of 200 μm thickness with an 89% light transmission in the 400 nm to 700 nm range.

Passive ventilation was provided by three zenithal windows, each 0.8 m wide and protected with insect-proof nets. These windows were opened when the indoor to outdoor temperature difference exceeded 10 °C. A central irrigation system served the entire GH without separation of the test zones. The overall GH and crop maintenance, i.e., ventilation, irrigation, fertilization and pest control, was carried out by Fundacion UAL-Anecoop (Appendix Figure A1 and Figure A2).

To mimic the shading effect of PV modules, opaque plastic panels (1.0 m × 1.7 m) were mounted on the roof. Two shaded test zones with 30% and 50% roof cover ratios were compared against a control zone without any shading, as shown in Figure 2 and Figure 3. Based on the GH simulation framework of Kujawa et al. [38,39], the panels were arranged in a checkerboard pattern to improve light homogeneity and to avoid boundary effects and shade overlapping issues throughout the year. In the 50% shading zone, 140 panels were installed (14 per row with 0.56 m gaps between modules), whereas the 30% zone contained 90 panels (9 per row with 1.44 m gaps). Figure 2 shows the GH and the layout of the ‘PV’ modules.

2.2. Microclimate Monitoring System

Throughout the experiment, microclimate parameters, such as global horizontal irradiance (GHI), RH and temperature, were measured at one-minute intervals. Figure 2 presents the location of the different sensors inside the GH.

In each test zone, three pyranometers were installed at the center of the zone and one additional pyranometer was placed near the edge of the zones to monitor boundary effects caused by radiation entering from the side walls. Temperature and RH were recorded at two distinct locations in the center of each zone. All microclimate measurement sensors, i.e., four pyranometers, two temperature and RH sensors per zone, were mounted at a height of 2.8 m above the GH ground, corresponding to the maximum canopy height of the tomato plants during cultivation. The two shading zones and the control zone were not physically separated from one another.

Table 1. Measurement devices used in the greenhouse (GH).

Parameter	Unit	Device	Manufacturer
GHI (center)	W m−2	CMP10	Kipp & Zonen, OTT HydroMet B.V., Delft, Netherlands
GHI (center)	W m−2	SP110	Apogee Instruments, Logan, United States of America
GHI (boundary)	W m−2	SPlite2	Kipp & Zonen, OTT HydroMet B.V., Delft, Netherlands
Temperature	°C	RC-51H	Elitech, Greenwich, United Kingdom
Relative Humidity	%	RC-51H	Elitech, Greenwich, United Kingdom

summarizes all microclimate measurement devices.

2.3. Plant Physiology and Fruit Quality Parameters

Two categories of measurements were conducted on the plants during the experiment: plant physiological and fruit quality assessments. For each treatment zone, 20 plants were selected within a central, representative area of each GH zone, as indicated in Figure 3. Physiological measurements were taken weekly between 16 October 2023 and 9 January 2024, when the plants were pruned.

Table 2. Measured plant physiological parameters.

Parameter	Unit	Device/Method
Height of stem 1	cm	Measuring tape
Height of stem 2	cm	Measuring tape
Number of branches	–	Visual count
Number of leaves for stem 1	–	Visual count
Number of leaves for stem 2	–	Visual count
Trunk diameter of stem 1	mm	Digital caliper
Trunk diameter of stem 2	mm	Digital caliper
Distance between leaves in apical part	cm	Measuring tape
Distance between developed clusters in apical part	cm	Measuring tape
Number of fruits	–	Visual count

summarizes all measured plant physiological parameters.

In addition, fruit quality measurements were performed at each harvest date, beginning on 18 January 2024 (125 days after transplantation) and ending on 4 March 2024 (171 days after transplantation). Immediately after harvesting, fruits from each repetition were washed and examined. Fruit firmness was measured using a digital fruit penetrometer. Then, fruits were cut in half and total width and length and flesh wall thickness were determined. The solid matter content, expressed as °Brix, was determined by expressing some drops of liquid from the halves and adding them to a refractometer according to the AOAC method no. 932.12 [40]. In addition, for six selected plants per treatment zone, fresh and dry weight of the harvested fruits were determined. For dry weight, the cut-in-halves tomatoes were dried in a drying oven at 70 °C for five days. If the samples still contained liquid after five days, they were kept an additional day in the oven.

Table 3. Measured fruit quality parameters.

Parameter	Unit	Device	Manufacturer
Fruit width and length	mm	Digital caliper	-
Flesh wall thickness 1	mm	Digital caliper	-
Fruit firmness 1	kg/cm3	Digital fruit penetrometer (model 4105)	STEP Systems, Nürnberg, Germany
Soluble Solids Content	°Brix	Pocket Refractometer PAL-1	ATAGO, Saitama, Japan
Fresh weight 2	g	NewClassic ML scale	Mettler Toledo, Gießen, Germany
Dry weight 2,3	g	Universal drying oven	Memmert, Schwabach, Germany

¹ Measured at three individual points per fruit. ² Measured in six selected plants per treatment zone for each date of harvest. ³ Fruits were dried at 70 °C for 5–6 days and weighed with the same scale as for fresh weight.

summarizes the measured parameters.

2.4. Simulation of Tomato Yield Using the Reduced TOMGRO Model

The crop modeling in this study is based on the reduced version of the State-Variable Tomato Growth Model (TOMGRO), which simulates tomato growth and yield using hourly inputs of temperature and PPFD. The TOMGRO model, developed by Jones et al. [34] in 1991, consists of 574 state variables to capture processes such as photosynthesis, transpiration and nutrient transport. In its reduced version, developed by Jones et al. [41] in 1999, this model was simplified to five state variables, i.e., number of nodes, leaf area index, aboveground dry weight, fruit dry weight, and mature fruit dry weight. Figure 4 shows the workflow of the reduced TOMGRO model.

In general, the reduced TOMGRO model works with hourly time resolution for the two input variables PPFD and temperature. The plant development is calculated based on these hourly time steps. The yield of the plant, described by the five output variables, is then calculated on a daily time scale.

To apply the reduced TOMGRO model to the presented APV GH experiment, the experimentally measured microclimate data was used. For each shading zone of the GH, the hourly PPFD, calculated based on GHI according to the method presented in Section 3.1, and the hourly temperature, measured by the two temperature sensors in the central area, were averaged and used as input data. In this study, mature fruit dry weight was the only output variable of the reduced TOMGRO model that was measured. To apply the model to local conditions of the experimental GH, calibration of the model parameters was necessary. First, all model parameters and their value ranges were summarized from the literature.

Table A1. Model parameters, their original values, and range of variation.

Parameter	Description	Original Value	Range of Variation	Unit
Nm	Node appearance rate per day	0.5 [34]	[0.1–0.9] [74]	node d−1
$\delta$	Maximum leaf area expansion per node	0.038 [41]	[0.01–0.1] [74]	m2node−1
$\beta$	Coefficient in expolinear equation	0.169 [41]	[0.06–0.5] [74]	node−1
Nb	Projection of linear segment of LAI vs. N to horizontal axis	16.0 [41]	[8.0–25.0] [74]	node
LAImax	Maximum leaf area index	4.0 [41]	[±10% of 4.0] [34]	unitless
p1	Loss of leaf dry weight per node after LAImax is reached	2.0 [41]	[±10% of 2.0] [34]	g node−1
Vmax	Maximum increase in vegetative tissue dry weight growth per node	8.0 [41]	[2.0–12.0] [74]	g node−1
NFF	Nodes per plant when first fruit appears	22.0 [41]	[4.0 [75]–22.0 [41]]	node
$k_F$	Development time from first fruit to first ripe fruit	5.0 [41]	[0.5 [76]–5.0 [34]]	node
$\tau$	Carbon dioxide use efficiency	0.0693 [34]	[0.01–0.5] [74]	μmol m−2 s−1
D	Maximum leaf conductance to water vapor	2.593 [34]	[0.1 [76]–2.6 [41]]	g m−2 h−1
K	Light extinction coefficient	0.58 [34]	[0.3–0.9] [74]	unitless
m	Leaf light transmission coefficient	0.1 [34]	[±10% of 0.1] [34]	unitless
Qe	Quantum efficiency of photosynthesis	0.0645 [34]	[0.01–0.5] [74]	unitless
Q10	Coefficient in maintenance respiration equation	1.4 [34]	[1.4 [41]–1.6 [76]]	unitless
rm	Maintenance respiration coefficient	0.016 [41]	[0.005 [76]–0.016 [34]]	g[CH2O] g−1[d.w.] d−1
E	Growth efficiency	0.7 [41]	[0.5–0.9] [74]	g[d.w.] g−1[CH2O]
$T_{\mathrm{CRIT}}$	Critical temperature for photosynthesis decline	24.4 [34]	[17.0–29.0] [74]	°C
${\alpha }_F$	Assimilate fraction allocated to fruit growth per day	0.8 [34]	[0.1–0.8] [74]	d−1
v ($\vartheta$)	Transition coefficient between vegetative and full fruit growth	0.135 [34]	[0.05–0.9] [74]	node−1
$fF\_$	Fraction of dry matter allocated to fruits	not specified	[0–1] [41]	unitless

in Appendix B summarizes all model parameters and their initial values. A sensitivity analysis of these parameters, using the Extended Fourier Amplitude Sensitivity Test (eFAST), was performed to identify the parameters most sensitive to calibration [42]. The corresponding value ranges for all parameters are also provided in Table A1. Following the sensitivity analysis, the Particle Swarm Optimization (PSO) algorithm [43] was then employed to find the set of calibrated model parameters by minimizing the root mean square error (RMSE) between simulated and observed mature fruit dry weights. The RMSE is defined as in Equation (1):

$$\mathrm{RMSE}=\sqrt{{\displaystyle \frac{1}{n}}\sum _{i=1}^n{(\widehat{y_i}-y_i)}^2}$$

(1)

where:

• n is the number of days.
• $\widehat{y_i}$ is the simulated dry weight for day i (kg m⁻²).
• $y_i$ is the experimentally measured dry weight for day i (kg m⁻²).

To achieve an (as far as possible) independent validation from the calibration process with data for only one crop cycle available, the model was calibrated using measurements from the 50% shading zone. The resulting set of parameters was then tested against data from the 30% shading and control zones. In order to quantify the deviation between the simulated dry weight curves and the experimentally measured dry weight for the two remaining shading zones, the Mean Absolute Error (MAE) and the Coefficient of Determination $R^2$ were calculated in addition to the RMSE. MAE and $R^2$ are defined according to Equations (2) and (3).

$$\mathrm{MAE}={\displaystyle \frac{1}{n}}\sum _{i=1}^n\left|\widehat{y_i}-y_i\right|$$

(2)

and

$$R^2=1-{\displaystyle \frac{{\sum }_{i=1}^n{(y_i-\widehat{y_i})}^2}{{\sum }_{i=1}^n{(y_i-\overline{y})}^2}}$$

(3)

with the same parameter definition as in Equation (1). $\overline{y}$ describes the mean of the experimentally measured values, calculated as $\overline{y}={\displaystyle \frac{1}{n}}{\sum }_{i=1}^n{y}_i$.

3. Results

This section presents the microclimate measurements, the plant physiological and fruit quality measurements and the results of the crop yield simulation using the reduced TOMGRO model.

3.1. Shading Effect on Microclimate

The pyranometers in the GH monitored GHI throughout the study period. As photosynthetic processes for crop growth are mainly dependent on the incoming PAR spectrum, the GHI measurements were converted into PPFD according to the linear regression model of Vindel et al. [44]. While there are more complex approaches, such as radiative transfer and spectral decomposition models, which provide more accurate PAR estimations, these methods often require detailed spectral data. The method of Vindel et al. was developed using satellite-derived broadband GHI and PAR estimations for mainland Spain and has been validated against ground measurements at different sites. For the validation of the linear regression model, the site of CIEMATs’ Plataforma Solar de Almeria (PSA) was chosen, which is located at a distance of about 50 km from the location of the experimental APV GH. For this location, a correlation coefficient higher than 0.99, a mean bias error of −2.36% and an RMSE of 2.8 W m⁻² were found. Equation (4) describes how the model converts GHI [W m⁻²] into PAR [μmol m⁻² s⁻¹].

$$PAR\left(t\right)=\left[a\right(month)·GHI(t)+b(month\left)\right]·conv.$$

(4)

The conversion factor $conv.$ has a constant value of 4.6 μmol J⁻¹ [45]. For the region of southern Spain, the coefficients of [44], shown in

Table 4. Coefficients a and b for the region of southern Spain taken from [44].

	Jan	Feb	Mar	Apr	May	Jun	Jul	Aug	Sep	Oct	Nov	Dec
a [-]	0.42	0.42	0.41	0.41	0.38	0.39	0.36	0.39	0.39	0.41	0.41	0.41
b [W m−2]	0.35	0.49	1.37	2.15	10.94	9.37	18.05	8.63	7.12	2.25	1.45	1.25

, are applied.

Figure 5 shows the daily mean PPFD as well as the relative fraction of the 30 and 50% zones with respect to the control zone.

The daily average PPFD showed that irradiance decreased in the winter months and increased again towards the end of the crop cycle in March. The lowest consecutive irradiance values were monitored at the beginning of January 2024 with values below 150 μmol m⁻² s⁻¹ for the control zone, below 100 μmol m⁻² s⁻¹ for the 30% shading zone and below 60 μmol m⁻² s⁻¹ for the 50% shading zone. The highest PPFD values were reached in March 2024 with values of up to 600 μmol m⁻² s⁻¹ for the control zone, 490 μmol m⁻² s⁻¹ for the 30% shading zone and 300 μmol m⁻² s⁻¹ for the 50% shading zone. Low PPFD values in all three zones on the same day could be identified throughout the entire crop cycle and were mostly caused by cloudy and rainy conditions at the test site.

From the PPFD, the daily light integral (DLI) can be calculated [45]. Tomatoes are classified as high-light demanding crops, requiring an optimal DLI between 30 mol m⁻² d⁻¹ and 12 mol m⁻² d⁻¹ for sufficient crop growth according to Cossu et al. [46]. Figure 6 shows the evolution of the average DLI in the three zones during the study period.

The overall trend of decreasing DLI towards the winter months and increasing DLI towards the ripening stage of the fruits in March is visible in all zones. The 0% control zone exceeded 12 mol m⁻² d⁻¹ on clear sky days until beginning of December 2023 when a few days fell below the threshold. After the middle of December 2023, the 0% zone shows an overall increasing trend. In the months January to March 2024, a few days fell below the threshold again due to cloudy weather at the test site. The DLI values of the 30% zone fell below 12 mol m⁻² d⁻¹ three weeks earlier in November 2023. Starting from the end of January 2024, the overall trend in the DLI values of the 30% zone is above the threshold. The values of the 50% zone remained constantly below the threshold for the entire study period apart from a few days before the final harvest in March 2024.

Throughout the study period, radiation levels were reduced on average by 48% in the 50% zone and 25% in the 30% zone. This slight decrease with respect to the theoretical shading reduction of the 30% zone was caused by the orientation of the GH and the positioning of the zones. For comparison, the study of Cossu et al. [9] indicated that the annual global radiation decreases by 0.8% for each additional 1.0% of PV panels located in a roof-top PV GH. As we only compare two shading zones against the control zone, we do not present a general linear regression for the radiation reduction. In the presented study, each 1% of additional shadowed area caused a 0.83% (30% zone) or 0.96% (50% zone) decrease in PAR according to the measurements, which aligns well with the observations of Cossu et al. [9].

In addition to GHI, temperature and RH were also monitored and measured at two distinct points in the center of each zone. In general, the best growth of tomatoes occurs with sunny weather, with steady temperatures between 20 °C and 25 °C [47]. Fruit quality is strongly influenced by the temperature, whereas high temperatures together with a high RH promote diseases [47]. Climate change will cause an increase in temperature, which can affect crop yield due to a reduced fruit set and changes in fruit quality [48]. Hence, APV systems might be beneficial for GH temperature control in future scenarios. Figure 7 shows the diurnal mean temperature and RH, calculated between 9:00 and 18:00 o’clock of each day, and the temperature and RH difference with respect to the control zone.

The evolution of temperature in the GH shows a decreasing trend among all zones from the start of the study period towards the winter months followed by a temperature increase towards the end of the study period. At the beginning of the study period, the temperature values exceeded 25 °C in all three zones. The lowest temperature values were recorded at the end of December 2023, with values below 12 °C in the 50% shading zone. Starting from January 2024 the mean diurnal temperature values increased again throughout all zones and the highest measured values were obtained at the beginning of March 2024 with above 35° in both the 0% and 30% zone. On average for the entire study period, the mean temperature values were 24.2 °C for the control zone, 23.8 °C for 30% shading and 23.0 °C for 50% shading.

The differences in temperature with respect to the control zone show that temperature values in the 50% zone were consistently lower than in the unshaded control zone with an increasing trend towards the end of the study period. On average, the diurnal mean temperature in the 50% zone was 2.2 °C lower than in the control zone throughout the study period. The temperature measured in the 30% shading zone behaves similarly to the measurements in the control zone. On average, the diurnal mean temperature in the 30% zone was 0.4 °C lower than in the control zone. These differences were observed with the use of a shared central irrigation system and without physical separation between the zone.

RH generally increased during the colder months and declined as temperatures rose towards spring. On average for the entire study period, 58.5% RH was measured in the control zone, 60.5% in the 30% shading zone and 64.5% in the 50% shading zone. Looking at the difference in RH with respect to the control zone, the data shows that in both shading zones an increase in diurnal mean RH values was observed except from few days at the beginning of the study period. The measured RH in the 30% zone was on average 2.0 percentage points (p.p.) higher than in the control zone. For the 50% zone, the measured RH was on average 6.0 p.p. higher than in the control zone.

3.2. Shading Effect on the Crops

3.2.1. Crop Yield—Experimental Results

This section presents the fresh and dry weight measurements of the investigated fruits. Fresh weight and dry weight are mathematically related through the moisture content of the tomato samples. In the context of this study, we chose to present both quantities as they serve distinct purposes. Fresh yield represents the marketable yield, which will be critical to evaluate the economic viability of the APV GH. On the other hand, dry weight offers a measure of total biomass without variability due to water content. Dry weight of the fruits can be linked to photosynthetic efficiency, nutrient uptake, environmental conditions and overall plant growth. Additionally, dry weight data was required for parameterization and validation of the reduced TOMGRO model used in our study.

The tomato plants were distributed with a density of 2 plants per m² in the GH; hence, absolute weight measurements of the single fruits were normalized to g m⁻². Figure 8 presents the accumulated fresh weight measurements and the measurements for the different days of the harvest as well as the relative and the absolute deviation from the control zone.

The data shows that the highest accumulated fresh weight of mature tomato fruit was consistently measured in the control zone. Over the harvesting period, the difference in fresh weight between the control zone and both shaded zones progressively decreased, thus indicating a potential delay in ripening due to shading. By the end of the harvest, the control zone accumulated the highest fresh weight at 3298 g m⁻², followed by the 30% shading zone with 2791 g m⁻² and 2446 g m⁻² measured in the 50% shading zone. Compared to the control zone, a 15% yield reduction was measured in the 30% shading zone with a difference of 507 g m⁻². For the 50% shading zone, a 26% yield reduction (851 g m⁻²) was recorded with respect to the control zone on the final day of harvest. The highest relative deviation between fresh weight in the shaded zones and fresh weight in the control zone was already measured on the first day of harvest in January 2024. Overall there is a decreasing trend in relative deviation for the 50% zone.

The numbers align with the study of Callejón-Ferre et al. [19], who also investigated the effect of high shading ratios (above 30%) on tomato crops in a raspa-y-amagado GH in Almeria and reported a proportional decrease in yield with an increase in shading ratio. Riga et al. [49] also observed lower values of tomato yield in GHs under lower PAR levels compared to those receiving natural sunlight. They found that a 30% and 50% reduction in incoming PAR led to a 14.3% and 38.1% reduction in tomato yield, respectively.

In addition to fresh weight of the mature fruits, the dry weight was also monitored to trace back the shading effect on the quality of the fruits. Figure 9 presents the dry weight of mature tomato fruit across the three different shading zones.

At the end of the growing season, the highest dry weight was measured in the control zone with an accumulated total value of 330 g m⁻². In the 30% shading zone 282 g m⁻² were measured and in the 50% shading zone 267 g m⁻² were measured. The 30% zone had therefore an overall yield reduction in dry weight of 46 g m⁻² with respect to the control zone, which is equivalent to a 14% reduction. For the 50% shading zone, the overall reduction in dry weight with respect to the control zone was determined to be 61 g m⁻² at the end of the crop cycle, which is approximately 19% less yield for that day. Similar to the observations in the fresh weight measurements in Figure 8, the largest relative deviation in dry yield was already observed at the first day of weight measurements. On the first measurement day, the control zone exhibited the highest yield with 81 g m⁻² in dry weight. On that day, 12% less dry weight was measured in the 30% zone and 38% less in the 50% zone compared to the control zone. Towards the end of the study period, this relative difference decreases in both shading zones.

3.2.2. Crop Yield—Simulation Results with the Reduced TOMGRO Model

The model calibration was performed using measured irradiance and temperature data from the 50% shading zone. Seven model parameters were selected for calibration based on the eFAST sensitivity analysis. The PSO algorithm was applied to minimize the RMSE between simulated and experimentally measured dry weight data.

Table 5. Calibrated model parameters.

Parameter	Description	Calibrated Value
Nm	Node appearance rate per day	0.38 d−1
D	Maximum leaf conductance to water vapor	2.33 g m−2 h−1
Qe	Quantum efficiency of photosynthesis	0.44 mol(CO2) fixed/mol(photons)
T_CRIT	Critical temperature for photosynthesis decline	19.76 °C
${\alpha }_F$	Assimilate fraction allocated to fruit growth per day	0.86 d−1
fF_	Fraction of dry matter allocated to fruits	0.90

summarizes the seven parameters and their calibrated values. The whole parameter set and the initial values are summarized in Appendix B. The minimum RMSE value found in the calibration process was 8.54 g m⁻².

The calibrated model was then tested against the experimentally measured data of the 0% and 30% shading zone of the experiment. Figure 10 presents the simulated and observed dry weight data for all three shading zones.

The deviation between the simulated dry weight curves and the experimentally measured dry weight for the two remaining shading zones is evaluated based on the RMSE, MAE and $R^2$.

Table 6. Error metrics for the control zone and the 30% shading zone for the validation of the calibrated reduced TOMGRO model. The values of the 50% shading zone were obtained in the calibration process and are given for comparison.

	0% Shading	30% Shading	50% Shading
RMSE	7.76 g m−2	11.56 g m−2	8.54 g m−2
MAE	4.14 g m−2	7.11 g m−2	4.34 g m−2
$R^2$	0.99	0.98	0.99

summarizes the RMSE, MAE and $R^2$ for the control zone and the 30% and 50% shading zone.

In general, both datasets of the experimentally measured values are described realistically by the simulation with an RMSE below 12 g m⁻². The lower RMSE and MAE values of the 0% zone indicate better agreement of the simulation with the experimentally measured values than for the 30% zone. The MAE remains at 7.11 g m⁻² for the 30% shading zone and 4.14 g m⁻² for the control zone. At the end of the crop cycle, the deviation of the model from the experimentally measured values of the 30% shading zone was 4%. For the control zone, the simulations resulted in a deviation of 3% from the experimentally measured values at the end of the crop cycle.

The results point out the importance of a careful parameter calibration and also the need for independent validation datasets. The results presented here would potentially look different if the calibration had been performed with the dataset from a different zone than the 50% zone. Furthermore, the results highlight the importance of using real experimentally measured input data that correctly maps the microclimate in the APV GH. A common approach in recent studies using TOMGRO for yield estimation in APV GHs is yield simulation without real experimental input data under the assumption of constant and optimal temperature conditions [36,50]. The presented study showed that for shading levels and PV patterns of the experiment, a change in microclimate and non-optimal temperature conditions should be expected. Hence, adapted development of the tomatoes to the shading conditions should be observed subsequently. These changes can only be modeled accurately if the input data is adapted and a calibration of the model parameters to the site’s conditions is performed. The main objective of the simulation was the first application of a biophysical modeling approach to estimate crop yield under real APV conditions. The results of the simulation show that the model can be adapted in general to simulate crop yield under reduced irradiance and temperature conditions. For future studies, the work can be extended to test the parameter calibration for additional crop cycles and crop types. An extension of the model to different GH types or climatic scenarios can also be studied in the future. The goal for the future and a possible application of the model would be one in which growers who are potentially interested in an APV setup for their GH could estimate the evolution of crop yield for their specific use case. Thus, an optimization related to ripening delays and market entry could be possible.

3.3. Plant Physiology and Crop Quality

In the presented experiment, differences in the development of the tomato plants were visible early in the growing stage. This aligns with the findings of Heuvelink et al. [51] who showed that plant growth and productivity are directly correlated with the light intensity and environment temperature in the GH. The figures of the plant physiological measurements listed in Table 2 are collectively summarized in Appendix C. On average for the entire growing period, the plants in the 50% shading zone were 38 cm higher with a 0.6 mm smaller diameter of the trunk (see Figure A3, Figure A4, Figure A5 and Figure A6). The plants had on average 5.1 more leaves than the plants in the control zone (Figure A9 and Figure A10). For 30% shading, the plants were on average 9 cm higher with a 0.2 mm smaller trunk diameter and had 5.0 more leaves for the whole growing period (see also Figure A3, Figure A4, Figure A5, Figure A6, Figure A9 and Figure A10).

In addition to plant physiology, fruit quality was determined by measuring fruit width and length, flesh wall thickness, fruit firmness and SSC [°Brix] (as described in Table 3) of the single harvested fruits for each date of measurement during the harvest period. Since the SSC is a particularly interesting variable to assess the quality of tomatoes, it will be investigated in more detail in what follows. The mean SSC values alongside the sample standard deviations are depicted in Figure 11 for all harvested fruits among the zones. The figures of the other fruit quality parameters are collectively summarized in Appendix C.2.

In general, all zones show an increase in SSC towards the end of the harvest period, as later-harvested fruits grow higher on the plant and receive more irradiance throughout the growing period. The fruits taken from the 50% shading zone show on average an absolute reduction of 0.6 °Brix compared to the fruits of the control zone throughout the entire harvest period, which corresponds to an average reduction of 11%. On the last measurement day, the fruits of the 50% zone show an absolute reduction of 1.0 °Brix compared to the control zone. The fruits of the 30% shading zone follow a similar trend as the 50% zone; however, an increase in SSC towards the end of the harvest period is notable. On average, a reduction of 3% in SSC was observed throughout the harvest period. On the last day of measurement, the fruits of the 30% zone had a reduction of 0.4 °Brix compared to the control zone. These observations align well with the study of López-Díaz et al. [23], who found that shading levels over 30% decreased the SSC in the investigated tomato crop in comparison with the control treatment for a Venlo GH in Almeria. In the study of Callejón-Ferre et al. [19], the measured SSC of tomato crops was decreased for shading ratios above 40%. Furthermore, other studies showed that the SSC of tomatoes can increase slightly with increasing temperature, which might have a positive influence on the flavor of the crops [51,52].

Similar to the observed trend of an increase in SSC for the fruits of the 30% shading zone, the fruits also showed an increase in firmness towards the end of the harvest period and showed on average a similar behavior compared to the fruits of the control zone (see Figure A15). These observations are also in agreement with the previous study of Callejón-Ferre et al. [19] who showed that shading over 40% increases the firmness of the fruits. In the study of López-Díaz et al. [23], it was also observed that with shading levels over 30% an increase in firmness compared to the control treatment can be expected. In addition, different studies observed a decrease in the proportion of larger fruits and an increase in the proportion of smaller fruits at 50% shading [49,53].

Since both the plant physiology as well as the fruit quality parameters were measured from a set of representative plants in each zone, a statistical compatibility analysis was performed to complement the above comparison of the mean parameter values in the different zones. The measurement samples in the different zones were compared on the last day of measurements with a Welch’s t-test (allowing for unequal variances of the two compared populations) with a standard significance threshold value of $\alpha =0.05$.

Table 7. Overview of mean and standard deviation values of the measured plant physiological and fruit quality parameters on the last date of measurement. The last two columns represent the p-values for the null hypothesis that the measured values in the 30 and 50% zones are compatible with the measurements in the control zone ($\alpha =0.05$).

		0% Shading	30% Shading	50% Shading	p-Value 30	p-Value 50
Plant physiology
Height stem 1	(cm)	270.8 ± 14.4	281.1 ± 17.2	296.1 ± 12.4	0.052	1.1 × 10−6
Height stem 2	(cm)	251.2 ± 22.2	284.4 ± 26.0	269.3 ± 30.6	1.5 × 10−4	0.045
Trunk diameter stem 1	(mm)	15.4 ± 1.3	15.2 ± 1.3	13.7 ± 1.3	0.62	3.1 × 10−4
Trunk diameter stem 2	(mm)	15.4 ± 1.1	15.0 ± 0.8	14.0 ± 1.1	0.22	4.7 × 10−4
Number of branches		21.8 ± 1.6	21.6 ± 1.9	18.5 ± 1.4	0.73	5.1 × 10−8
Distance branches	(cm)	15.6 ± 5.0	15.8 ± 3.9	14.8 ± 2.8	0.93	0.59
Number of leaves stem 1		36.1 ± 3.3	37.5 ± 3.1	38.6 ± 2.6	0.17	0.01
Number of leaves stem 2		32.3 ± 4.7	38.3 ± 3.1	36.4 ± 3.3	4.8 × 10−5	0.003
Distance leaves	(cm)	5.8 ± 1.7	6.8 ± 1.8	8.5 ± 1.7	0.12	5.9 × 10−5
Number of fruits		34.8 ± 4.5	36.1 ± 3.1	32.9 ± 4.3	0.30	0.20
Fruit quality
Fruit diameter	(mm)	52.6 ± 2.5	51.8 ± 3.6	48.1 ± 5.0	0.58	0.01
Flesh wall thickness	(mm)	5.8 ± 0.5	5.5 ± 0.7	5.0 ± 0.7	0.29	0.004
Fruit firmness	(kg cm−3)	3.4 ± 0.4	3.9 ± 0.5	3.3 ± 0.7	0.02	0.54
Soluble Solids Content	(°Brix)	7.8 ± 0.3	7.4 ± 0.7	6.8 ± 1.2	0.14	0.02

summarizes the means and standard deviations for all plant physiology and fruit quality parameters on the last day of measurement as well as the p-values for the compatibility of 30% shading to the reference zone and 50% shading to the reference zone. The 50% shading zone shows significant deviations from the reference zone for almost all parameters, except for the distance between branches, the number of fruits and the fruit firmness. The only significant deviations in the 30% shading zone are determined to be the height of stem 2, the number of leaves in stem 2 and the fruit firmness.

4. Discussion

4.1. Experimental Results

The artificial shading in the presented experiment not only caused a radiation reduction but directly impacted the whole microclimate in the GH. Therefore, the development of the cultivated tomato plants was affected early on from the beginning of the crop cycle. Over the entire study period, there was an average reduction of 48% in radiation in the 50% shading zone, and a 25% reduction in the 30% shading zone in comparison to the control zone. These findings align well with the radiation reductions determined by Cossu et al. [9], who indicated a radiation reduction of 0.8% for each additional 1.0% of PV shading.

The provided shading in the experiment resulted in 15% yield reduction for the 30% shading zone and 26% yield reduction for the 50% shading zone in measured fresh weight at the end of the crop cycle. These numbers align well with the findings of Callejón-Ferre et al. [19] and Riga et al. [49], who both reported a similar decrease in yield with an increase in shading ratio. Several other experimental studies also confirm that high shading levels lead to reduced tomato yield in GHs due to radiation reduction [23,53,54,55,56,57]. Nevertheless, it is important to point out that tomato yield is not an isolated characteristic and depends on the growth of the whole plant [51]. If the tomato plant does not grow well then it will never give a high yield. Therefore yield is determined by the interaction between plant morphology, physiology and growth conditions.

In the presented experiment, the two shading zones and the control zone were not physically separated from one another. Hence, there was an exchange in air flow between the zones and no separation of temperature or RH within the different areas in the GH. Nevertheless, changes in temperature and RH were measured at the sensor locations at the center of the zones. The combination of changes in temperature and RH and reduction in irradiance directly affected the tomato plants at each stage of their development. In the presented experiment, differences in the development of the tomato plants were visible early in the growing stage, as the plant physiological measurements confirmed. These observations align well with the findings of Heuvelink et al. [51], who showed that plant growth and productivity are directly correlated with the light intensity and environment temperature in the GH, which affect two important physiological processes: photosynthesis and respiration. Photosynthesis depends on both light intensity and temperature, while respiration only on temperature. Therefore, final yield and sugar content depend on the trend of these two physiological processes [57].

In the presented experiment, plants from both shading zones developed on average 20% less fruits than the plants in the control zone throughout the entire crop cycle. Furthermore, a delay in ripening of the fruits was determined for both shading treatments, i.e., 6 days for 30% shading and 9 days for 50% shading after the first harvest in the control zone. Previous studies also showed that shading in GHs caused by PV panels reduces the number of produced fruits and causes a delay in ripening [53,58,59]. A delay in ripening and a delayed tomato harvest can actually have strategic market advantages, especially if it shifts production into a period of lower supply and higher prices. García-Martínez et al. [60] analyzed price trends in GH tomatoes across several Spanish regions and found that the highest seasonal price indices occurred in March and April. By strategically timing the delay in yield in an APV GH to coincide with periods of reduced supply and higher selling prices, growers can potentially compensate yield reductions resulting from the shading.

Apart from their influence on plant physiological development and yield, the effect of shading treatments on tomato quality, fruit size and dry matter content was also determined. At the end of the crop cycle a reduction in dry weight of 14% and 19% was observed for the 30% and 50% shading zones, respectively. In addition, for the fruits of the 50% shading zone, a reduction of 0.9 °Brix was measured at the end of the crop cycle. For the fruits of the 30% shading zone, a reduction of 0.2 °Brix was observed. SSC is an important quality parameter considering that the amount of sugar and acid and their interactions are related to flavor quality in tomatoes [61]. According to Dorais et al. [10], SSC is highly affected by the GH microclimate and environment including solar radiation, temperature, day-length, water availability, soil mineral content, irrigation and fertilization.

4.2. Theoretical PV Yield and Revenues

4.2.1. Electricity Self-Consumption and Feed-In

As APV concepts describe a trade-off between crop cultivation and energy production, it is important to emphasize the potential advantages of installing PV modules on GHs. One key benefit is that the electricity generated by the PV system can first be used to cover the GH’s own energy consumption and, secondly, to potentially overcompensate reductions in crop yield by selling surplus energy to the grid. This section provides an estimation of the potential economic benefits from self-consumption and energy sales, aimed at compensating crop yield reductions. The estimations are based on simplified assumptions and are not intended as a detailed economic analysis. Rather, they offer an initial indication of how APV systems can help to stabilize income and reduce dependence on fluctuating market prices.

For the economic evaluation, we consider three hypothetical scenarios for the entire GH: (1) no PV modules installed, (2) 30% of the roof covered with PV modules, and (3) 50% of the roof covered with PV modules. Crop and PV yield values are scaled to the 1765 m² GH area for each of the three scenarios.

Based on the crop yield measurements of the 0% shading zone, the tomato yield scaled to the entire experimental GH under full-sun conditions without PV installation is 5820 kg. According to the measurements of the two shading zones, the applied shading treatments resulted in a 15% crop yield reduction for the 30% shading zone and 26% reduction for the 50% shading zone. This corresponds to approximately 895 kg less tomatoes for the 30% APV GH and 1504 kg less tomatoes for the 50% shading scenario. At a market price for tomatoes of 0.58 € kg⁻¹ for the 2023–2024 campaign [62], the GH without PV would have lead to an income of EUR 3376 by solely selling the tomatoes. The reduction in crop yield in the two APV scenarios would have resulted in a loss of EUR 519 and EUR 872, respectively.

To estimate the PV yield of the two APV systems (30% and 50% roof coverage of the entire GH roof), the PVGIS SARAH radiation database was used [8]. The breakdown of the monthly PV yield, as well as monthly prices for electricity are given in

Table A2. Monthly breakdown of prices, costs and savings for the economic evaluation presented in Section 4. The prices for electricity consumption (con.), electricity feed-in tariffs and the levelized prices corrected with the levelized costs of electricity are presented. The PV yields for the 30% shading scenario and the 50% shading scenario are given as well as the consumption of the GH. The savings due to self-consumption and the revenues (rev.) from the two PV systems are then calculated.

	Days	Electricity	Feed-In	Levelized	PV Yield	PV Yield	Cons.	Saving	PV Rev.	PV Rev.
		Price	Price	Price	30%	50%		Self-Cons.	30%	50%
		[63,64]	[66,67]	[65]	[7]	[7]
		[ct/kWh]	[ct/kWh]	[ct/kWh]	[MWh]	[MWh]	[kWh]	[EUR]	[EUR]	[EUR]
09/2023	14	11.2	10.3	5.1	6.08	10.13	160	18	304	512
10/2023	31	10.4	9.0	3.8	10.93	18.22	354	37	403	681
11/2023	30	8.0	6.3	1.1	8.38	13.96	342	27	92	156
12/2023	31	8.5	7.2	2.0	7.86	13.10	354	37	151	257
01/2024	31	8.6	7.4	2.2	8.46	14.10	354	37	179	304
02/2024	29	5.0	4.0	−1.2	9.45	15.75	331	17	−109	−185
03/2024	15	3.4	2.0	−3.1	6.11	10.18	171	6	−188	−318
Sum	181	–	–	–	57.26	95.44	2065	165	832	1407

in Appendix D. For the period from September to March, the estimated energy production was 57.26 MWh for the 30% shading scenario and 95.44 MWh for the 50% shading scenario, assuming standard silicon modules and default system losses. The total energy consumption of the experimental test GH during this period was approximately 2065 kWh. For the economic analysis, we assume the same energy consumption for all three scenarios. Based on average Spanish electricity prices from September 2023 to March 2024 [63,64], self-consumption of PV electricity to cover the GH’s energy demand during the tomato cycle could therefore save around EUR 165 in both APV scenarios. The levelized cost of electricity for the PV systems was estimated as 5.2 ctkWh⁻¹, based on the findings for rooftop PV from [65] for the irradiation level of the site of this study. To calculate potential revenues from selling excess electricity, average feed-in tariffs during the study period [66,67] were considered. For the 30% shading scenario, the overall energy sales would therefore yield approximately EUR 832. For the 50% shading scenario, the selling of the excess PV energy would yield EUR 1407.

Table 8. Overview for the economic analysis.

		0% PV	30% PV	50% PV
Crop yield	(kg)	5820	4925	4316
Crop revenue	(EUR)	3376	2857	2504
PV yield	(MWh)	0	57.26	95.44
GH energy consumption	(kWh)	2065	2065	2065
Costs energy consumption	(EUR)	−165	0	0
PV revenue	(EUR)	0	832	1407
Total sum	(EUR)	3211	3686	3911
Profit increase in comparison to 0% PV			+ EUR 478	+ EUR 700
			+14.8%	+21.8%

summarizes the comparison of the calculated PV and crop yield revenues and estimates the overall profit increase for the two APV scenarios.

The hypothetical 30% shading system integrated into our 1765 m² GH generated an estimated 32.44 kWh m⁻² per crop cycle (September to March), which corresponds to a direct revenue of approximately 50 ct per m². For the hypothetical 50% shading system, the PV yield can be estimated with 54.07 kWh m⁻² for September to March, which corresponds to a revenue of 80 ct per m². In APV GH systems, the revenue generated from the PV installation can serve as a supplementary income and help offset operational costs of the GH such as electricity for climate control, irrigation or lighting. Over time, the APV GH dual-use approach can improve the economic viability and resilience of horticulture, particularly in regions with high solar irradiance [68,69].

In our two theoretical APV GH scenarios, both cases would overcompensate the reduction in crop yield and lead to a profit increase at the end of the crop cycle. For the 30% shading scenario, the additional income can be estimated as EUR 478, which corresponds to a profit increase of +14.8% with respect to the traditional GH without PV. The 50% cover ratio APV GH would have led to EUR 700 of additional income, which corresponds to an increase of +21.8%.

These results already demonstrate the economic potential of APV GH systems. Even during the winter months of the study period, when irradiance and PV output are naturally lower, a positive net income could be observed. During summer, PV yields are higher and radiation levels can even be harmful to crops. It can therefore be expected that the benefits of APV systems considering the full year are even greater, leading to a higher overall annual advantage.

4.2.2. Powering of Desalination Plants

The generated electricity of the PV system could be used to power desalination plants for water production, which could subsequently be used for GH irrigation [70]. Such estimations become particularly important in years with low tomato market prices or under extreme climatic conditions. The desalination plant of Carboneras in the Almeria region, for example, has an energy requirement of 4.25 kW h per 1 m³ produced freshwater [71]. It should be noted that this value refers to one specific, exemplary plant and more modern and energy-efficient desalination technologies may require less energy. The PV yield of the 50% cover ratio system could lead to a production of 22,455 m³ of desalinated water during the tomato cycle. The presented 30% cover ratio GH could supply up to 13,473 m³ of freshwater during the same period.

The experimental GH required 271 m³ of water during the tomato cultivation period. The PV yield from the 50%-cover-ratio APV system could hence supply the experimental GH 82 times with freshwater. For the 30%-cover-ratio APV system, the water consumption would be covered 49 times. As mentioned in the previous subsection, PV yields are even higher in summer. The overall annual advantage can therefore be even higher considering the full year.

4.3. Discussion of Simulation with the Reduced TOMGRO Model

In addition to the presented experimental measurements, we also used the collected data to model crop growth and estimate yields with the reduced TOMGRO model. The modeling results presented in the previous chapter show a reasonable agreement with the experimentally measured dry weight data. The RMSE values remained below 12 g m⁻² for the two presented validation tests of the 30% shading zone and 0% control zone. At the end of the crop cycle, the model overestimated the experimentally measured values of the 30% shading zone by 4%. For the control zone, the model overestimated the experimentally measured values by 3% at the end of the crop cycle. In comparison to other studies that also used the TOMGRO model or its adaptions to simulate APV scenarios, the presented results point out the advantages of a detailed calibration and validation campaign of the model. The study of Ghaffarpour et al. [35] integrated the reduced TOMGRO model into their APV GH; however, they validated their simulations against the experimental measurements of Hemming et al. [72] and the simulation values from the work of Gong et al. [73]. The modeling results of Ghaffarpour et al. underestimated both the experimental and the simulated values, with deviations of approximately 14% at the end of the crop cycle. Other studies on tomato yield in APV GHs neglect the calibration to site-dependent conditions and do not validate their application of the reduced TOMGRO model against experimental measurements [36,50]. Furthermore, these models assume optimal and stable temperature conditions and do not take into account the potential effect of shading on temperature inside the GH. The presented experiment showed that shading may alter the microclimate inside the GH such that the cultivated plants adapt to these conditions. Hence, the experimental measurement of these changes in microclimate is very important for future research. In addition, the yield simulations of crops under APV conditions would also be more accurate due to the correct mapping of the input microclimate data.

Nevertheless, it is important to mention possible limitations of the presented modeling approach and discuss potential improvements for future studies. First, the presented experiment covered the time period of one growing cycle. To utilize the data from one growing cycle to test the TOMGRO model, the data from the 50% shading zone was used for model calibration. Subsequently, the model was applied to the 30% and 0% shading zones. In order to fully generate an independent validation, the experimentally measured data of a second crop cycle would be necessary. In future studies, the model could also be tested against measurements from different GH types or sites.

In the presented study, the model was calibrated and tested using experimentally measured mature fruit dry weight data. The reduced TOMGRO model also simulates four other state variables, i.e., number of nodes, leaf area index, aboveground dry weight and fruit dry weight. These datasets were not recorded in the presented study as the main goal was to monitor experimentally the changes in microclimate and successively changes in plant physiology and yield of the tomato crops. In order to perform comprehensive model calibration and validation, additional datasets of these four parameters would be beneficial and could be monitored during a second growing cycle.

In addition to the available data, there are several other factors influencing the growth process of the tomato plants, which were considered in the current modeling approach. These include information about soil, irrigation and diseases or pests that is not directly included in the reduced TOMGRO model, as well as the assumption of a constant CO₂ content in the GH. In general, soil properties, irrigation and fluctuations in CO₂ concentration in the GH can play a non-negligible role in how plants develop and might not be mapped fully correctly in the presented modeling approach. Consequently, performing simulations with a new independent dataset could lead to diverging results as certain stress factors or growth conditions may not be correctly mapped right now. In future studies, a comprehensive measurement campaign could focus on monitoring such additional GH environment and tomato growth variables.

In general, the results show that a biophysical model-based approach with the reduced TOMGRO model can be used to perform yield estimations for tomatoes grown under APV conditions. In order to generate a fully independent and accurate validation, the data of another growing cycle would be needed. The model could also be applied to different types of crops, such as bell pepper or zucchini. Considering further calibration of the model parameters or an extension to cover different local environmental conditions, such as different climates or sites, could also be implemented.

4.4. Uncertainties and Limitations of the Study

Both the experimental measurements and the simulations of the presented study have sources of uncertainty that should be taken into account in the interpretation of the data.

First of all, the measurements of the sensors introduce a potential source of uncertainty to the microclimate measurements and subsequently the simulation with the reduced TOMGRO model. Sensor soiling, environmental influences or technical limitations might affect the accuracy of the input data of the model and thus influence the outputs. The sensors deployed in the experiment were calibrated in advance and constantly cleaned during the measurement period in order to minimize uncertainties due to soiling or tilting. Furthermore, daily quality control was performed with the measured microclimate data. In addition, PAR data, which is one of the input variables of the reduced TOMGRO model, was estimated by applying the linear regression model developed by Vindel et al. [44] to the experimentally measured GHI data. Although deviations are minimal for validation locations, it should be kept in mind that the method may introduce additional uncertainties to the model, as the relationship between GHI and PAR is not always linear and can be influenced by local environmental conditions. Furthermore, the method of Vindel et al. [44] has been developed for outside conditions. The irradiance spectrum within the GH can be different from the outside due to the transmission through the GH plastic, which introduces another source of uncertainty in the presented study. In a future measurement campaign, specifically chosen PAR radiation sensors, only monitoring the PAR spectrum, could be deployed in the GH to avoid the additional conversion step in the data evaluation.

In the presented experimental study all three zones were located within the same GH without any physical separation. Hence, a constant exchange of air flow between the zones was not actively prevented. In addition, all plants were irrigated equally and a potential change in water needs of the plants due to shading was not observed. Ideally, a separate GH would be needed for each shading treatment to ensure that the specific conditions of each zone are studied without interference from neighboring zones. Nevertheless, clear differences in the microclimate were observed with the presented experimental setup. These deviations were then also mapped realistically in the crop yield simulations of the three different zones.

4.5. Practical Applications

This work contributes to the development and refinement of APV systems for GHs by expanding the available experimental datasets. The experimental results improve our understanding of how PV modules installed on GH roofs influence the microclimate, crop development and expected yields under APV conditions. Future research could focus on optimizing PV module layouts to maintain light homogeneity inside the GH and minimize yield losses due to shading. Another promising direction is the development of smart shading strategies that dynamically adapt to the crop’s light requirements.

This study highlights the potential of APV GHs to serve as an additional and independent source of income for growers. The study included a preliminary economic analysis for the study period, which showed that reductions in crop yield may be overcompensated by the energy generated from PV modules. These benefits are likely to be even greater over the full year, particularly during summer months when solar irradiance is higher and GHs are not cultivated. In addition, the study incorporated tomato yield modeling using the reduced TOMGRO model. The results demonstrate that the model can be successfully adapted to APV conditions to simulate crop yield. The model can therefore be a valuable tool for estimating tomato yields in APV GHs without requiring extensive new experimental data.

5. Conclusions

The presented experiment investigated the impact of shading caused by PV modules on tomato crops in a commercial scale raspa-y-amagado GH. The experiment was conducted in the 2023–2024 growing season in Almeria, southern Spain. Two distinct shading zones with roof cover ratios of 50% and 30% were compared against a control zone without artificial shading. The experiment aimed to fill a research gap regarding experimental studies on APV GHs (type raspa-y-amagado) with high shading levels (above 30% roof cover ratio). The study contributes to a broader understanding of APV GHs and creates a baseline for future studies. The results of the presented study show that high shading levels lead to changes in microclimate, which then impact yield and fruit quality and cause a delay in harvest. The 30% and 50% shading zones resulted in 15% and 26% yield reductions, respectively, in measured fresh weight at the end of the crop cycle in comparison to the control zone. The study results showed that the 30% shading zone had a lower impact on fruit quality than the 50% shading zone with an increasing trend in SSC towards the end of the study. In contrast, 50% shading resulted in significant reductions in yield and quality of the fruits. The study showed that experimental data on these microclimate changes and impacts on yield is important to fully understand and also optimize APV scenarios.

Furthermore, the presented study included crop yield simulations using experimentally measured microclimate data. The simulation results highlighted the importance of using real-world climatic and radiation data to better understand crop responses to PV shading. The crop yield modeling was limited to the dataset of the presented study of one crop cycle, which introduced uncertainties to the presented results. Future studies could therefore include the validation of the yield simulations with a totally independent dataset from a different crop cycle. Additionally, adapting the model to other sites and crops can also be tested in the future.

A theoretical, preliminary evaluation of the expected PV yield and economic viability of the APV GH system was performed. The calculations were based on simplified assumptions, such as a constant selling price for the harvested crops and constant GH energy demand that is fully covered by PV energy and monthly average electricity prices. It should be kept in mind that this economic analysis is not a profound investigation of the specific case of the APV system, but rather aims to show the potential and benefits of APV GHs in general. As discussed previously, the revenues of the investigated PV systems are expected to compensate for the observed reductions in crop yield and indicate that APV GHs can create an additional source of income for growers, support self-sufficiency and improve independency from fluctuation market prices. Delayed harvests of crops in APV GHs can also present strategic market opportunities, as late-season tomato prices tend to rise when early supply declines. In order to achieve high prices for the APV crops, fruit quality should be comparable to the quality of unshaded crops.

The present study already showed the high potential of APV GHs in southern Spain. To minimize reductions in crop yield and create higher revenues, future studies could also include semitransparent PV modules or dynamic PV tracking systems.