LETTER • OPEN ACCESS You may also like

- Crop harvested area, not yield, drives

Agrivoltaic system design tools for managing variability in crop production in Iran
Ehsan Eyshi Rezaei, Gohar Ghazaryan,
trade-offs between energy production, crop Rooholla Moradi et al.

- Evaluation of the Agricultural Reality of the

productivity and water consumption Industrial Yellow Corn Crop in the Iraqi
Province of Kirkuk
Dawood Fahad Abdullah
To cite this article: Emily Warmann et al 2024 Environ. Res. Lett. 19 034046
- The productivity of five-field crop rotations
and their influence on humus content in
southern Russia
V S Pashtetskiy, L A Radchenko, E N
Turin et al.
View the article online for updates and enhancements.

This content was downloaded from IP address 161.18.37.114 on 07/04/2024 at 23:02

 Environ. Res. Lett. 19 (2024) 034046 https://doi.org/10.1088/1748-9326/ad2ab8

LETTER

Agrivoltaic system design tools for managing trade-offs between

OPEN ACCESS
energy production, crop productivity and water consumption
RECEIVED
27 April 2023 Emily Warmann1,2,∗, G Darrel Jenerette3 and Greg A Barron-Gafford1,2
REVISED 1
26 January 2024 School of Geography, Development and Environment, University of Arizona, Tucson, AZ, United States of America
2
Office of Research, Development, and Innovation, Biosphere 2, University of Arizona, Tucson, AZ, United States of America
ACCEPTED FOR PUBLICATION 3
19 February 2024
Department of Botany and Plant Sciences, University of California, Riverside, CA 92521, United States of America
∗
Author to whom any correspondence should be addressed.
PUBLISHED
5 March 2024 E-mail: [email protected]

Keywords: agrivoltaic, design, energy, production, crop

Original content from
this work may be used Supplementary material for this article is available online
under the terms of the
Creative Commons
Attribution 4.0 licence.
Any further distribution Abstract
of this work must
maintain attribution to
Agrivoltaic systems that locate crop production and photovoltaic energy generation on the same
the author(s) and the title land have the potential to aid the transition to renewable energy by reducing the competition
of the work, journal
citation and DOI. between food, habitat, and energy needs for land while reducing irrigation requirements.
Experimental efforts to date have not adequately developed an understanding of the interaction
among local climate, array design and crop selection sufficient to manage trade-offs in system
design. This study simulates the energy production, crop productivity and water consumption
impacts of agrivoltaic array design choices in arid and semi-arid environments in the Southwestern
region of the United States. Using the Penman–Monteith evapotranspiration model, we predict
agrivoltaics can reduce crop water consumption by 30%–40% of the array coverage level,
depending on local climate. A crop model simulating productivity based on both light level and
temperature identifies afternoon shading provided by agrivoltaic arrays as potentially beneficial for
shade tolerant plants in hot, dry settings. At the locations considered, several designs and crop
combinations exceed land equivalence ratio values of 2, indicating a doubling of the output per
acre for the land resource. These results highlight key design axes for agrivoltaic systems and point
to a decision support tool for their development.

1. Introduction can make agrivoltaics competitive relative to conven-

tional solar arrays and, in some cases, conventional
Agrivoltaics, the practice of co-locating photovol- agriculture (Amaducci et al 2018, Barron-Gafford
taic arrays and agricultural production on the same et al 2019, Agostini et al 2021)
land, is of growing interest as the world trans- The central challenge of agrivoltaics is appor-
itions to renewable energy (Macknick et al 2022). tioning sunlight among the crop and the photovol-
Historically, more than half of the land used for taic panels to support adequate crop and electricity
photovoltaic production is converted from crop pro- production. To date, the agrivoltaics field has not
duction (Kruitwagen et al 2021), and photovolta- systematically evaluated design approaches—largely
ics could occupy 10% or more of the land currently because the optimal partition of sunlight depends
used for crops in some regions by 2050 (OECD/IEA on geography and crop selection. Implementations
2021, van de Ven et al 2021). Agrivoltaics presents the range from low density, semi-transparent panels at
opportunity to reduce the competition between food high coverage levels to vertically oriented, widely
and energy needs for scarce land resources (Dupraz spaced bifacial panels that provide negligible shad-
et al 2011, Barron-Gafford et al 2019). Results from ing and obtain substantial illumination from reflec-
initial agrivoltaic pilot projects suggest system bene- ted light (Elamri et al 2018, Othman et al 2020,
fits from interactions between the crop and solar array Ott et al 2020, Riaz et al 2020, Abidin et al

© 2024 The Author(s). Published by IOP Publishing Ltd

Environ. Res. Lett. 19 (2024) 034046 E Warmann et al

2021, Agostini et al 2021, Trommsdorff et al 2021). gap in ability to evaluate and predict whole system
The diversity of agrivoltaic pilot projects reflects performance.
agricultural diversity, but translating the knowledge Here we evaluate the effects of array design and
between settings is challenging without understand- crop selection on agrivoltaic energy production, crop
ing the interplay between system components and production, and water use in three case study loca-
the effect of local climate conditions on those inter- tions in the southwestern United States. We present
actions. Current models neglect key drivers of plant a new intermediate modeling approach that adds a
function and thereby overlook critical features of temperature response to the light response compon-
agrivoltaic system interactions. ent and estimates water use. The goals of the ana-
Agrivoltaic researchers have adopted the concept lysis are to identify array design characteristics that
of the land equivalence ratio (LER) from intercrop- minimize impacts to crop productivity and maximize
ping systems (Mead and Willey 1980, Trommsdorff savings of irrigation water. We anticipate identifying
et al 2021). This nondimensional metric is equal to some system configurations that increase crop pro-
the amount of land required to replicate the agrivol- ductivity in tandem with significant water savings in
taic system’s yield of electricity (in kW-hr yr−1 ) arid environments. Finally, we combine the analysis
and crops (in the metric appropriate for the spe- of energy production from the arrays to determine a
cific crop considered) using conventional solar array trade-off with electricity generation for array designs
and agricultural practices on separate land plots that maximize crop productivity or water savings.
(equation (1)),
2. Methods
YieldCrop,AV YieldElec, AV
LER = + . (1)
YieldCrop,Conv YieldElec, Conv 2.1. Modeling approach
We approach the question of how an agrivoltaic array
While this metric is succinct, it reduces all per- will affect water consumption and plant growth by
formance comparison to land savings. Important determining how the array changes the illumination
for dryland regions, the LER approach neglects the conditions in the field and then applying those illu-
impact on water consumption—the primary limit- mination conditions to (1) a model of ET and (2) a
ing factor in food production (Jaeger et al 2017). plant productivity model (figure 1). The same cal-
The addition of a photovoltaic array directly affects culation captures the illumination of the photovol-
soil evaporation and crop transpiration, collect- taic panels and hence the electricity production. The
ively known as evapotranspiration (ET), by alter- model is implemented in MATLAB (MathWorks).
ing the radiation environment. Because freshwater For simplicity, the model assumes conditions under
withdrawals already exceed 100% of renewable water the array are the same as the control setting in all
resources in some regions, the impact of agrivoltaics aspects other than illumination. While multiple stud-
on water consumption is a crucial metric. ies have reported differences in air temperature in
The agrivoltaics field has begun building on exist- the presence of photovoltaic arrays (Barron-Gafford
ing modeling frameworks developed for photovolta- et al 2019, Broadbent et al 2019, Jiang et al 2021),
ics (Usama Siddiqui et al 2012, Liu et al 2016, Jain the effects are not consistent in magnitude or sign
et al 2017) and agronomy (Valiantzas 2013, Jones across or within studies. To examine the sensitiv-
et al 2017) to predict crop and energy production ity of the results to this assumption, we also con-
for integrated systems. To date, these efforts have sidered scenarios where the air temperature under the
fallen into two categories. In one, modelers apply a array differed from the ambient by up to ±5 ◦ C (see
light response curve combined with the reduction supplemental).
in photosynthetically active radiation (PAR) caused
by shading (Campana et al 2021, Trommsdorff et al 2.2. Array geometry and shading
2021, Riaz et al 2022), thereby neglecting vital plant The first modeling step generates an array geometry
physiology determinants of performance and simply based on the unit panel dimensions, the panel ori-
predicting a decline in crop productivity with increas- entation (azimuth and tilt angle) and the panel
ing array coverage. These simulations predict a dir- height (figure 2(a)). The array unit cell includes the
ect tradeoff between food and energy production in unit panel (2 m wide and 1.3 m in height, elev-
the settings considered but have not explored climate ated 2.5 m) and variable space, tiled two directions
contributions to agrivoltaic performance. The second (figure 2(b)). The analysis considered array rows of
modeling approach uses parameterized crop models either North-South (NS) or East-West (EW) axes.
(Amaducci et al 2018, Elamri et al 2018) tying these We also examined three panel tilt angles, 0, 20 and
models to specific input and location cases. These 30 degrees to the South (for EW rows) or West (for
more physiologically detailed efforts have considered NS rows). Together, these design permutations span
a fixed array architecture and emphasized finding the density and orientation of most fixed-panel solar
crop partners for agrivoltaics accompanied by exper- installations and includes average tilt angles for latit-
iment. These endpoints of model complexity leave a udes up to approximately 45 degrees.

2
Environ. Res. Lett. 19 (2024) 034046 E Warmann et al

Figure 1. Schematic representation of modeling scheme for this study. User-selected inputs of array geometry, time and location
are used to calculate the shade pattern at six minute intervals. The agrivoltaic shading is in turn used to determine electricity and
crop production and water consumption in combination with data on local solar and climate conditions. The model combines
electricity and crop production to determine the land equivalence ratio (LER), which, along with the water consumption, is the
model output.

Figure 2. Schematic view of agrivoltaic array parameters that can be varied within the model. In panel (a), dimension of panel
width, W and height (H); mounting height (Z); and mounting angle (α). As the image shows, the panel may be composed of
multiple photovoltaic devices in one unit, depending on the design. In panel (b), the array parameters: angle of row (β) relative to
ordinal directions, spacing between rows (d1) and spacing between panels within rows (d2). The array unit cell illustrates the
element tiled in two directions to create the full array.

Array coverage (projected area of the panels from ∆ (Rn − G) + ρa cp es −ea

ET = ( ) ra . (2)
nadir divided by the field area) ranges from 100% to ∆ + γ 1 + rras
1% and is the primary basis of comparison between
the designs in the subsequent analysis. To generate
the shade pattern, we first calculated the solar angle This model estimates the ET of a plant layer based
at six-minute intervals for each day of the year at each on the energy balance between incoming solar radi-
location (Blanco-Muriel et al 2001). We determined ation (G), outgoing thermal radiation (Rn ), and latent
the shade fraction by comparing the area of the panel heat from ET, as governed by the local air conditions:
shadow projected onto the ground plane to the area of vapor pressure deficit (es –ea ), the air density (ρa ), spe-
the array unit cell. Any self-shading by adjacent panels cific heat (cp ), psychrometric constant (γ) and slope
was corrected in total shading and electricity produc- of the vapor saturation pressure curve (∆), and bulk
tion calculations. and aerodynamic resistances (rs and ra ) that represent
the restrictions to water transport through the soil,
plant tissues and canopy.
2.3. Modeling of water consumption For full sun ET estimates, G is the global hori-
We model the array impact on ET using the FAO zontal irradiance, and for full shade conditions, G is
formulation of the Penman–Monteith equation the diffuse irradiance, both taken from the National
(equation (2)) (Allen et al 1998). Solar Radiation Database (Sengupta et al 2018). We

3
Environ. Res. Lett. 19 (2024) 034046 E Warmann et al

assume that the leaf surface temperature is equal to Temperature productivity response
( )( )
the air temperature, the sky temperature is 273 K Ha/
RT 1−T0/ T
and the albedo is 0.7, which together determine e 0
= C∗ (4)
the outgoing radiation (Rn ) (Evangelisti et al 2019). (Sv T − Hd )/
1+e (RT)
Temperature, ambient water vapor pressure and wind
speed are hourly climate normals for the location,
Ha is the activation energy and Hd the deactiva-
interpolated for model timesteps (Arguez et al 2012).
tion energy, R is the gas constant, T 0 is a reference
The surface roughness/diffusivity parameters of the
temperature (298 K), T is the leaf temperature, and
PM model, ra and rs , are the baseline values for grass
Sv is an entropic term. For this model we use the ref-
(Allen et al 1998).
erenced tabulated values for Ha and Sv corresponding
To analyze water savings for each array case, we
to Brassica rapa, while we adjust the value of Hd to set
weighted the baseline full sun and full shade ET rates
the desired peak temperature and C is a scaling con-
at each timestep by the sun and shade fractions and
stant that adjusts the peak value to one. The full plant
calculated an average ET rate for the field as a whole.
model productivity (equation (5)) is the product of
This instantaneous ET rate was then integrated over
the light response (LPR) and temperature response
each day or the total year to quantify seasonal and
(TPR) for a given set of conditions:
cumulative trends.
Instantaneous Crop Prod (T, I) = LPR (I) ∗ TPR (T) .
2.4. Modeling of crop productivity
(5)
As in many current ecosystem modeling efforts,
our model avoids distinguishing among plant spe- The model output value range lies between zero
cies, but rather collapses plant differences into two and one, where one represents the maximum possible
broad types: shade-tolerant plants and sun-tolerant instantaneous productivity.
plants. Productivity in our model refers to photosyn- Sun tolerant plants use the same model paramet-
thetic activity, which we treat as a proxy for all crop ers for both full sun and full shade conditions, spe-
products. We assume that the plants are managed to cifically I sat is 300 W m−2 and the peak response tem-
avoid water or nutrient restriction without specify- perature is 26 ◦ C. For the shade-tolerant plants, the
ing an irrigation method, and our model generates an model uses the same parameters in the shade, but in
instantaneous crop productivity value based on illu- full sun the peak response temperature is 22 ◦ C, simu-
mination and temperature. We include temperature lating plants that are more sensitive to high temperat-
in our model, because photosynthesis has a strong ures in full sun. The values for the peak response tem-
dependence on temperature and because temperat- peratures and I sat parameters are chosen to match the
ure has diurnal, seasonal and location variations that aggregate photosynthetic performance of a variety of
interact with the diurnal and seasonal illumination plants in agrivoltaic experiments under full sun and
changes caused by the agrivoltaic array (Farquhar et al full shade conditions (Barron-Gafford et al n.d.).
1980, Leuning 2002, Sage and Kubien 2007). We calculate the plant response for each six-
Our plant model calculates productivity as the minute time step under full sun and full shade con-
product of two responses. First, plant productivity ditions. Agrivoltaic productivity for the field for each
increases to an asymptote with irradiance (Pickett and time step is the average of the full sun and full shade
Myers 1966) (equation (3)): productivity values, weighted by the instantaneous
( ) shade fraction of the field. The instantaneous pro-
I
Light productivity response = tanh 1.5 ∗ . ductivity values are integrated over each day and over
Isat
(3) a growing season from day 120 to day 275 of the year.
We also include an integration over the whole year
Here I is the cumulative irradiance on the plants to obtain cumulative values of the crop productivity
in W m−2 and I sat is the saturation irradiance. We potential for settings that may grow crops over mul-
use the cumulative irradiance rather than the PAR, tiple seasons. All productivity values are compared to
because PAR is well-correlated with cumulative irra- an unshaded baseline case.
diance at the level of precision required by this model
(Yu et al 2015). 2.5. Model application
The second component of the plant model is a The analysis examines the impact of agrivoltaic
temperature response that rises with temperature to a designs on water consumption and crop and electri-
peak value and then declines. The function is based on city productivity for an agrivoltaic system in Tucson,
the maximum catalytic rate as a function of leaf tem- AZ (mean temperature 21 ◦ C, mean annual precip-
perature, as reported in (Leuning 2002), which cap- itation 269 mm) (Arguez et al 2012). The analysis is
tures the asymmetric rise and fall of activity of the further extended to Stockton, CA (mean temperat-
enzyme Rubisco with temperature (equation (4)): ure 17 ◦ C, mean annual precipitation 342 mm) and

4
Environ. Res. Lett. 19 (2024) 034046 E Warmann et al

Aurora, CO (mean temperature 11 ◦ C, mean annual baseline (figure 3(b)). The NS array has a higher water
precipitation 367 mm) to explore how water con- consumption in the winter months, while the EW
sumption, crop productivity and LER depend on sys- array is higher in the early summer, suggesting sea-
tem design, location and crop selection. The three loc- sonal variations in the shading pattern for the two
ations span the climatic space for the majority of the classes of array.
Southwestern United States, a region that has large The daily and seasonal trends in shading fraction
agricultural production, arid or semi-arid climate, for these two arrays (see supplemental) illuminate
and water resources under increasing stress due to these seasonal differences in water consumption. For
climate change and demand growth (Overpeck and south-tilted EW array, the shade fraction is higher in
Udall 2020). winter months when the sun’s average zenith angle is
For each location we calculated the shade pattern, larger, but the shade fraction does not vary over the
total water consumption and crop productivity for course of the day. By contrast, the NS array has a con-
arrays with both EW and NS orientation, panel cov- sistent shade fraction over the course of the year, but
erage ranging from 1% to 100% and tilt angles of 0, the tilt to the west causes a higher shade fraction in the
20 and 30 degrees to the South (for EW rows) or West afternoon, with most light reaching the ground in the
(for NS rows). We simultaneously calculated the elec- morning. The shade fraction averaged over the day for
tricity production, with panel illumination calculated a NS array is independent of the panel tilt angle, only
as the sum of the direct irradiance projected onto the the distribution of light within the day is affected.
panel area and the diffuse irradiance. Electricity pro-
duction of select array configurations at each loca- 3.2. Crop productivity results
tion were compared to projections from the PVWatts For the sun-tolerant crop (figure 4(a)), productivity
(Dobos 2014) model for validation. Portions of the in the shade is lower than productivity in the sun
panel shaded by adjacent panels are assumed to pro- for the entirety of both days considered. By contrast,
duce zero electricity. For all scenarios, we compare the the shade-tolerant model (figure 4(b)) shows lower
production of crops and electricity for the agrivoltaic in-shade productivity on day 50, but higher in-shade
system to a baseline case consisting of a full sun con- productivity the majority of day 200. On day 200, the
ventional agriculture field and a second photovoltaic- temperature rapidly rises to a level above the peak
only field with a baseline photovoltaic array consist- activity temperature for the model in full-sun, and
ing of EW rows of panels tilted 30 degrees to the south productivity is suppressed. In the shade, the model
and 50% area coverage. For both conventional and has a higher peak activity temperature and suffers a
agrivoltaic crop productivity we consider a summer less severe mid-day depression in productivity. Over
growing season from day 120 to day 275. By com- the course of the year, the shade-tolerant daily plant
paring the agrivoltaic crop and energy productivity to productivity in the agrivoltaic setting shows a large
the baseline, we obtain an LER value for each design variation, ranging from 40% of the control productiv-
permutation. ity in the winter to over 110% of the control set-
ting productivity for days in the mid to late sum-
3. Results mer. The sun-tolerant plant model also exhibits lower
productivity in winter, particularly for EW arrays at
3.1. Water consumption results roughly 50%–60% of the control setting, however in
There is a clear correlation between array cover- summer the productivity is only 80% of the control
age and water savings, due to shading reducing setting value for both EW and NS arrays.
the incoming irradiation and therefore the ET rate Comparing the performance of both crop mod-
(figure 3(a)). The NS arrays show a linear trend until els over the full year (figure 5(a)), as the field cov-
roughly 70% coverage for all tilt orientations. The EW erage fraction increases, plant productivity suffers a
arrays follow three slightly different linear trends cor- larger penalty relative to the control case. The shade-
responding to different panel tilts, with larger tilt val- tolerant crop model shows a slightly higher crop pro-
ues experiencing larger water savings. At very high ductivity compared to the sun-tolerant model for
coverage fractions, the linear trend for all array types arrays with NS rows. For field coverage fractions of
shifts to a lower rate of increase with area, due to 40% or higher, plant productivity varies by roughly
increasing self-shading of the array. If the air tem- 10 to 20 percentage points, depending on the com-
perature were warmer under the array, water savings bination of crop and array design.
decrease by roughly 3% per degree of temperature dif- Reanalysis for the summer growing season alone
ference. Conversely, savings increase by 3% per degree (figure 5(b)) shows a large separation in the beha-
if the air is cooler (see supplemental). vior of the two plant models. The sun tolerant crop
Agrivoltaic arrays with different azimuth orienta- productivity still declines with array coverage, but
tion but equivalent coverage both decrease water con- the shade-tolerant crop under NS arrays with a west-
sumption by a similar amount relative to the full sun ern tilt has a productivity up to 8% higher than the

5
Environ. Res. Lett. 19 (2024) 034046 E Warmann et al

Figure 3. In panel (a), the cumulative annual water consumption predicted for agrivoltaic arrays in Tucson, AZ as a percentage of
the full-sun annual evapotranspiration (ET) demand, versus the fraction of the field covered by the arrays in a top-down view. In
panel (b), the daily evapotranspiration rate over the course of the year for agrivoltaic arrays with ∼50% coverage in Tucson, AZ.
The North-South (NS) array has panels tilted 20 degrees to the west and the East-West (EW) array has panels tilted 30 degrees to
the south.

Figure 4. In panel (a), the plant productivity over the course of days 50 and 200 for plants in full sun and full shade based on a
sun-tolerant plant model. In panel (b), the plant productivity over those days in full sun and full shade for a shade-tolerant plant
model. In both (b) and (c), solid traces indicate full-sun conditions and dotted traces indicate full shade. The color of the trace
indicates the day of year which supplied the temperature and illumination values for the plant model inputs. In panel (c), the total
daily plant productivity as a percentage of the full sun productivity for both plant models under agrivoltaic arrays with 50%
coverage in Tucson, AZ. The east-west oriented array traces are blue and the north-south traces in red. Solid curves indicate the
sun-tolerant plant model and dotted curves correspond to the shade-tolerant model.

full-sun baseline. This effect comes from the uneven full-sun control depending on the coverage level, with
distribution of shading over the course of the day productivity enhanced above the baseline for shade-
for west-tilted NS arrays, which allows the shade- tolerant crops (figure 5(b)).
tolerant crop to take advantage of morning sun while
temperatures are lower. The crop productivity curves 3.3. Location comparison
(figure 4(b)) illustrate this behavior in detail. The For Tucson (figure 6(a)) the shade-tolerant crop
shade tolerant crop under EW arrays shows a pro- model productivity and water saving trade-off is small
ductivity penalty of 10% or less compared to the to positive. However, for the sun-tolerant model, the
baseline, but it receives no net benefit from the water savings are close to the loss in crop perform-
shading. ance regardless of design. In the results for Stockton
The water consumption over the summer grow- (figure 6(b)) the shade-tolerant model performs well
ing season (day 120–day 275) versus the field cover- but does not show productivity benefits. Tilted NS
age fraction (figure 5(c)) is almost identical to the arrays lose less than 5% relative to the baseline. This
full-year trend (figure 3(a)). For both NS and EW performance difference compared to Tucson is due to
arrays, water consumption declines by roughly 45% the cooler temperatures in Stockton, which also result
of the array coverage fraction. This indicates that for in lower water savings. For Aurora, CO (figure 6(c)),
NS arrays with 40% or higher coverage, the agrivol- the behavior of both crop models and both array ori-
taic field would have 20%–45% water savings over the entations collapse into a single behavior trend. Based

6
Environ. Res. Lett. 19 (2024) 034046 E Warmann et al

Figure 5. In panel (a), the cumulative annual plant productivity as a percentage of productivity in full sun for plants grown under
agrivoltaic arrays relative to the coverage fraction of the array. In panel (b), the plant penalty vs the coverage fraction over the
summer growing season (day 120 to day 275). In both panels, east-west array points are shaded blue, while north-south points are
shaded red. Filled markers indicate the sun-tolerant plant model, while open markers indicate the shade-tolerant plant model.
The dashed line at y = 100% is included to guide the eye. In panel (c), the water consumption over the summer growing season,
as a percentage of the full sun water consumption, versus the field coverage. Again, East-West array points are drawn as blue
circles and north-south points as red triangles and these relationships reflect Tucson, AZ.

on these simulations, an agrivoltaic system in Aurora up to 0.6. In all cases, the range of potential LER val-
would suffer a plant productivity penalty of roughly ues increases with the coverage, however the Aurora
two thirds of the water savings over the summer grow- location LER values are much more closely bounded
ing season regardless of the coverage level when com- for a given coverage level. As the narrow green region
pared to a full-sun field. The water savings are low- indicates, this is due to the Aurora location having
est for this location among the three, at roughly 1/3 productivity independent of crop type (figure 7(c)).
of the coverage level. The effect of warmer or cooler
temperatures under the array also depends on loc-
ation, crop type and season considered. In Tucson 4. Discussion
warmer temperatures reduce summer shade tolerant
crop productivity to a level comparable to that in Growing land-use conflicts between food and energy
Stockton, while crops benefit from warmer temper- production have increased interest in agrivoltaics as
atures in Aurora (see supplemental). a dual-use approach that could benefit both sectors.
The design challenge of managing the tradeoffs asso-
3.4. Land equivalent ratio analysis ciated with overstory PV panel light capture versus
All agrivoltaic scenarios considered had LER values light transmission to food production has become
greater than one, indicating that even low-coverage a barrier to wider-scale adoption. Our modeling
designs increase the land use efficiency. While LER approach illustrates that agrivoltaics have potentially
increases with area coverage, there is a large degree large benefits across the food, energy, and water sec-
of variation in the LER at comparable coverage levels tors within drylands, such as the Southwestern US,
(figure 7(a)). At coverage of 60% and higher, the but the benefits vary depending on the design of the
choice of array orientation and tilt and the shade tol- agrivoltaic system. All locations examined showed a
erant vs sun tolerant crop can change the LER value by reduction in water consumption proportional to the

7
Environ. Res. Lett. 19 (2024) 034046 E Warmann et al

Figure 6. Plant productivity vs water consumption over the summer growing season (day 120 to day 275) for both shade-tolerant
and sun-tolerant plant models at three locations. In panel (a), Tucson, AZ. In panel (b), Stockton, CA. In panel (c), Aurora, CO.
In all panels, blue points correspond to East-West arrays, red to North-south, while filled markers indicate the sun-tolerant plant
model and open markers indicate the shade-tolerant model. The dashed line at y = 100% in panel (a) is included to guide the eyes.

array coverage level. This would represent a direct trade-offs to coupled food, energy, and water systems
savings of irrigation inputs in a region with more than in their climate setting.
8 million irrigated acres (Cody and Johnson 2015, This new model identifies the role of array design
Lahmers and Eden 2018) where the future water sup- in mediating the impact of the agrivoltaic array on
ply is projected to decline by 10%–20% on average crop productivity. Prior simulations of agrivoltaics
(Pathak et al 2018, Sheikh and Stern 2019, Ray et al that did not account for diurnal temperature vari-
2020, Miller et al 2021). We also illustrated multiple ations have predicted that crop productivity would
combinations of crop and array design minimally det- decline with array coverage (Dupraz et al 2011, Riaz
rimental or even beneficial to crop productivity— et al 2020, Trommsdorff et al 2021), in contrast to
even with array coverage exceeding 50%. These res- experimental results (Barron-Gafford et al 2019). By
ults indicate that tradeoffs between energy produc- including effects of temperature, the model can sim-
tion and crop productivity are sensitive to local con- ulate crop productivity increases with coverage for
ditions. At the hottest, driest location, array design specific array designs that create an uneven distribu-
and crop selection combine to have large impacts on tion of shading over the course of the day, with shad-
crop productivity. By contrast, at the coolest, wet- ing coinciding with high temperatures. This design
test location, crop productivity varied only a small approach results in shade-tolerant crop productiv-
amount with crop type and array design. Finally, all ity up to 8% higher over the summer growing sea-
locations showed the potential to have LER values of son than arrays with comparable coverage where the
2 or higher. PV developers have long been able to util- shade fraction does not vary over the course of the day
ize a range of tools, such as the System Advisor Model, and suggests that tracking arrays (ITRPV 2022) may
to anticipate electricity production within traditional further allow tailoring of the shade profile for max-
PV systems (Blair et al 2018). This model provides imum plant growth.
a similar tool for predicting plant productivity in an The second important insight from the model
agrivoltaic system, allowing stakeholders to quantify is the significance of crop selection. At the Tucson

8
Environ. Res. Lett. 19 (2024) 034046 E Warmann et al

however this model gives insight into particular

plant physiological characteristics that will perform
well in agrivoltaic settings. Plants that are sensit-
ive to high heat under high illumination condi-
tions may benefit from agrivoltaic shading, partic-
ularly at hot locations. Our results are consistent
with crop-specific models such as GECROS, which
have predicted that agrivoltaics can increase unirrig-
ated crop yields relative to full-sun controls in years
with low precipitation (Amaducci et al 2018). Future
research could modify these model functions with
field-derived data on plant performance from spe-
cific crop types and any degree to which these crops
can adapt to a shade. Photosynthetic acclimation—
whereby leaves alter their morphology and/or bio-
chemistry to adapt photosynthetic efficiency subject
to long-term changes in the light environment—is
well documented (Townsend et al 2018); however,
no studies on plant acclimation within agrivoltaics
exist to-date. Photosynthetic acclimation of crops in
an agrivoltaics setting would likely increase the cal-
culated benefits of agrivoltaics, and this represents a
potentially valuable subfield of study.
The third fundamental characteristic considered
in this study, geography, also governs the trade-offs
between crop and energy in agrivoltaic systems. While
all three locations saw a reduction in water con-
sumption, the two hottest, driest locations showed
a bifurcation in performance between shade and
sun-tolerant crop types. At cooler, wetter Aurora,
CO the model predicts no beneficial shading effects.
Intriguingly, recent results from a nearby site showed
that grasses growing beneath a photovoltaic array
exhibited higher growth in regions receiving morning
sun and afternoon shade despite having lower aver-
age soil moisture content than regions with morning
shade and afternoon sun (Sturchio et al 2023). These
experimental results suggest that (1) plants may accli-
Figure 7. In panel (a) the land equivalence ration (LER) vs matize to agrivoltaic shading or (2) that the bene-
area coverage for agrivoltaic arrays in Tucson, AZ including fits of afternoon agrivoltaic shading may be more
crop productivity over the summer growing season (day
120–day 275). In panels (b)–(d), the upper and lower
pronounced when crops experience water restriction.
bounds of LER vs area coverage for agrivoltaic systems in Both possibilities may be valid simultaneously.
Tucson, AZ, Stockton, CA, and Aurora, CO. Portions of the The sensitivity of both water savings and crop
plot shaded in green represent the amount of variation in
LER controlled by crop selection. Portions shaded in gray productivity to the air temperature under the array
represent the LER variation controlled by array design. The (see supplemental) signals a need for more detailed
heavy black curve in the central region is the LER trend for
arrays with sun-tolerant crops and EW arrays at 50%
modeling capturing the impact of the array on the
coverage and a 30 degree southern tilt, the baseline local microclimate. Existing observations of array
configuration. temperature impacts have focused on unirrigated
dryland settings and a narrow range of array designs.
A model that accounts for array design and density
and predicts temperature effects on diurnal and sea-
and Stockton locations, agrivoltaic shading is either sonal time scales will be valuable for predicting array
minimally detrimental or positive for shade-tolerant impacts on crop productivity.
plants. By contrast, sun-tolerant crops in this model The array, crop, and geographic drivers all affect
respond to any shading with a loss of productiv- LER. At all locations, the highest LER values at every
ity proportional to the coverage level. The need coverage level were those that maximized electricity
to pair solar arrays with crop partners that toler- production rather than crop productivity. The west-
ate shading has been well established (Dupraz et al tilted NS arrays that maximize crop production suffer
2011, Trommsdorff et al 2021, Riaz et al 2022), a larger reduction in electricity production (∼10%)

9
Environ. Res. Lett. 19 (2024) 034046 E Warmann et al

than the increase in crop productivity (at most 8%) This array orientation reduces electricity production
relative to the comparable coverage EW array with a 10%, but crop production remains high under very
30% tilt to the south. Within that context, the agrivol- high coverage levels. In meeting the land needs for
taic system designer is presented with a trade-off: photovoltaic energy generation and ongoing invest-
maximize crop production through afternoon shad- ments in agriculture, agrivoltaics may provide bene-
ing at the expense of total electricity production. fits whose performance reflects geographically vari-
Similarly, a farmer seeking to reduce the irrigation ability, array design choices, and crop performance.
demand of a shade-tolerant crop may choose to part-
ner with an agrivoltaic array, thereby increasing the Data availability statement
total system productivity with little or no crop pro-
ductivity loss. All data that support the findings of this study are
The results of this analysis suggest several avenues included within the article (and any supplementary
for further investigation to improve the fidelity and files).
scope of subsequent crop models. Extending the plant
model in this analysis to include performance under Acknowledgments
water stress would allow extension to settings depend-
ent entirely on precipitation. Experiments aimed at This research received funding support from (1) the
identifying within-crop variability in traits linked National Science Foundation Geography and Spatial
with responses to environmental variability would Sciences through Award 2025727 and (2) the National
facilitate tailored crop selection. Renewable Energy Laboratory’s InSPIRE project,
through the U.S. Department of Energy Office of
5. Conclusion Energy Efficiency and Renewable Energy (EERE)
Solar Energy Technologies Office under award DE-
This analysis shows that integrated agrivoltaic sys- EE00034165
tems have potential to benefit coupled energy, food
and water systems by reducing land costs of photovol- ORCID iD
taic production while maintaining agricultural pro-
duction with lower water use. While the photovoltaic Emily Warmann  https://orcid.org/0000-0002-
array can have a negative impact on crop productiv- 2810-4608
ity, all agrivoltaic systems considered had higher land
use efficiency LER than equivalent conventional sys- References
tems and reduced water consumption proportional to
Abidin M A Z, Mahyuddin M N and Zainuri M A A M 2021 Solar
the coverage level. Agrivoltaic systems with roughly
photovoltaic architecture and agronomic management in
50% coverage will reduce crop water consumption agrivoltaic system: a review Sustainability 13 18–21
by 15%–20% or more depending on the location. Agostini A, Colauzzi M and Amaducci S 2021 Innovative
Ultimately location, crop selection and array design agrivoltaic systems to produce sustainable energy: an
economic and environmental assessment Appl. Energy
combine to determine the agrivoltaics optimization
281 116102
landscape (Macknick et al 2022), and these results Allen R G, Pereira L S, Raes D and Smith M 1998 Crop
demonstrate the value of a modeling approach that evapotraspiration guidelines for computing crop water
captures agrivoltaic sensitivity to geography, crops, requirements FAO Irrigation & Drainage Paper 56 (FAO,
Food and Agriculture Organization of the United Nations)
and array design.
Amaducci S, Yin X and Colauzzi M 2018 Agrivoltaic systems to
This model allows system designers to quantify optimise land use for electric energy production Appl.
crop and energy tradeoffs for individual systems. Energy 220 545–61
For a given location and coverage the variation in Arguez A, Durre I, Applequist S, Vose R S, Squires M F, Yin X,
Heim R R and Owen T W 2012 Noaa’s 1981–2010 U.S.
LER depended on array design and crop selection
climate normals Bull. Am. Meteorol. Soc. 93 1687–97
choices. A system combining photovoltaics with a Barron-Gafford G A et al 2019 Food—energy—water nexus in
sun-tolerant crop or a system located in a climate sim- drylands Nat. Sustain. 2 848–55
ilar to Aurora, CO, will face a direct tradeoff between Barron-Gafford G A, Lepley K, Rouini N, Barnett-Moreno I,
Salazar A, Murphy P and Macknick J E n.d. Agrivoltaics as a
energy production and crop productivity. If there is
climate-smart and resilient solution for midday depression
some maximum acceptable reduction in crop pro- in photosynthesis across dryland regions Nat. Sustain. Agric.
ductivity for the farmer, that threshold will determ- Blair N et al 2018 System advisor model (SAM) general
ine the maximum array density tolerable and con- description system advisor model (SAM) general description
(version 2017.9.5) (National Renewable Energy Laboratory)
sequently the maximum LER achievable by those sys-
(available at: www.nrel.gov/docs/fy18osti/70414.pdf)
tems. By contrast, in hot, dry locations with a shade- Blanco-Muriel M, Alarcón-Padilla D C, López-Moratalla T and
tolerant crop, productivity may in some cases increase Lara-Coira M 2001 Computing the solar vector Sol. Energy
with array coverage, thereby achieving a high LER 70 431–41
Broadbent A M, Krayenhoff E S, Georgescu M and Sailor D J 2019
with high density arrays. Maximizing crop produc-
The observed effects of utility-scale photovoltaics on
tion requires an array design that provides predom- near-surface air temperature and energy balance J. Appl.
inantly afternoon shading over a shade tolerant crop. Meteorol. Climatol. 58 989–1006

10
Environ. Res. Lett. 19 (2024) 034046 E Warmann et al

Campana P E, Stridh B, Amaducci S and Colauzzi M 2021 directly underneath agrivoltaic conditions with orthosiphon
Optimisation of vertically mounted agrivoltaic systems J. stamineus crop cultivation Agronomy 10 1472
Cleaner Prod. 325 129091 Ott E M, Kabus C A, Baxter B D, Hannon B and Celik I 2020
Cody B A and Johnson R 2015 California Agricultural Production Environmental analysis of agrivoltaic systems Reference
and Irrigated Water Use Renée Johnson Specialist in Module in Earth Systems and Environmental Sciences
Agricultural Policy Specialist in Natural Resources Policy pp (https://doi.org/10.1016/b978-0-12-819727-1.00012-1)
1–28 (available at: https://sgp.fas.org/crs/misc/R44093.pdf) Overpeck J T and Udall B 2020 Climate change and the
Dobos A P 2014 PVWatts version 5 manual aridification of North America Proc. Natl Acad. Sci. USA
(NREL/TP-6A20-62641) National Renewable Energy 117 11856–8
Laboratory (NREL) p 20 (available at: www.nrel.gov/docs/ Pathak T B, Maskey M L, Dahlberg J A, Kearns F, Bali K M and
fy14osti/62641.pdf) Zaccaria D 2018 Climate change trends and impacts on
Dupraz C, Marrou H, Talbot G, Dufour L, Nogier A and Ferard Y California Agriculture: a detailed review Agronomy
2011 Combining solar photovoltaic panels and food crops 8 1–27
for optimising land use: towards new agrivoltaic schemes Pickett J M and Myers J 1966 Monochromatic light saturation
Renew. Energy 36 2725–32 Plant Physiol. 41 90–98
Elamri Y, Cheviron B, Lopez J M, Dejean C and Belaud G 2018 Ray P, Wi S, Schwartz A, Correa M, He M and Brown C 2020
Water budget and crop modelling for agrivoltaic systems: Vulnerability and risk: climate change and water supply
application to irrigated lettuces Agric. Water Manage. from California’s central valley water system Clim. Change
208 440–53 161 177–99
Evangelisti L, Guattari C and Asdrubali F 2019 On the sky Riaz M H, Imran H, Alam H, Alam M A and Butt N Z 2022
temperature models and their influence on buildings energy Crop-specific optimization of bifacial PV arrays for
performance: a critical review Energy Build. 183 607–25 agrivoltaic food-energy production: the
Farquhar G D, von Caemmerer S and Berry J A 1980 A light-productivity-factor approach IEEE J. Photovolt.
biochemical model of photosynthetic CO2 assimilation in 12 572–80
leaves of C3 species Planta 149 78–90 Riaz M H, Imran H and Butt N Z 2020 Optimization of PV array
ITRPV 2022 International Technology Roadmap for Photovoltaic, density for fixed tilt bifacial solar panels for efficient
2021 Results agrivoltaic systems Conf. Record of the IEEE Photovoltaic
Jaeger W K et al 2017 Finding water scarcity amid abundance Specialists Conf. (June 2020) pp 1349–52
using human–natural system models Proc. Natl Acad. Sci. Sage R F and Kubien D S 2007 The temperature response of C3
USA 114 11884–9 and C4 photosynthesis Plant Cell Environ. 30 1086–106
Jain R K, Qin J and Rajagopal R 2017 Data-driven planning of Sengupta M, Xie Y, Lopez A, Habte A, Maclaurin G and Shelby J
distributed energy resources amidst socio-technical 2018 The national solar radiation data base (NSRDB)
complexities Nat. Energy 2 17112 Renew. Sustain. Energy Rev. 89 51–60
Jiang J, Gao X, Lv Q, Li Z and Li P 2021 Observed impacts of Sheikh P A and Stern C V 2019 Management of the Colorado
utility-scale photovoltaic plant on local air temperature and River: water allocation, drought, and the federal role Renew.
energy partitioning in the barren areas Renew. Energy Resour. J. 33 8–15
174 157–69 Sturchio M A et al 2023 Grassland productivity responds
Jones J W et al 2017 Brief history of agricultural systems modeling unexpectedly to dynamic light and soil water environments
Agric. Syst. 155 240–54 induced by photovoltaic array Ecosphere 13 e4334
Kruitwagen L, Story K T, Friedrich J, Byers L, Skillman S and Townsend A J, Retkute R, Chinnathambi K, Randall J W P,
Hepburn C 2021 A global inventory of photovoltaic solar Foulkes J, Carmo-Silva E and Murchie E H 2018 Suboptimal
energy generating units Nature 598 604–10 acclimation of photosynthesis to light in wheat Canopies
Lahmers T and Eden S 2018 Water and irrigated agriculture in Plant Physiol. 176 1233–46
Arizona Arroyo 2 16 Trommsdorff M, Kang J, Reise C, Schindele S, Bopp G,
Leuning R 2002 Temperature dependence of two parameters in a Ehmann A, Weselek A, Högy P and Obergfell T 2021
photosynthesis model Plant Cell Environ. 25 1205–10 Combining food and energy production: design of an
Liu H, Aberle A G, Buonassisi T and Peters I M 2016 On the agrivoltaic system applied in arable and vegetable
methodology of energy yield assessment for one-Sun farming in Germany Renew. Sustain. Energy Rev.
tandem solar cells Sol. Energy 135 598–604 140 110694
Macknick J et al 2022 The 5 Cs of agrivoltaic success factors in the Usama Siddiqui M, Arif A F M, Kelley L and Dubowsky S 2012
United States: lessons from the InSPIRE research study p 80 Three-dimensional thermal modeling of a photovoltaic
(available at: www.nrel.gov/docs/fy22osti/83566.pdf) module under varying conditions Sol. Energy
Mead R and Willey R W 1980 The concept of a ‘land equivalent 86 2620–31
ratio’ and advantages in yields from intercropping Exp. Valiantzas J D 2013 Simplified forms for the standardized FAO-56
Agric. 16 217–28 Penman–Monteith reference evapotranspiration using
Miller O L, Putman A L, Alder J, Miller M, Jones D K and limited weather data J. Hydrol. 505 13–23
Wise D R 2021 Changing climate drives future streamflow van de Ven D J, Capellan-Peréz I, Arto I, Cazcarro I, de Castro C,
declines and challenges in meeting water demand across the Patel P and Gonzalez-Eguino M 2021 The potential land
southwestern United States J. Hydrol. X 11 100074 requirements and related land use change emissions of solar
OECD/IEA 2021 Key World Energy Statistics 2021 (IEA) (available energy Sci. Rep. 11 1–12
at: www.iea.org/reports/key-world-energy-statistics-2021) Yu X, Wu Z, Jiang W and Guo X 2015 Predicting daily
Othman N F, Yaacob M E, Su A S M, Jaafar J N, Hizam H, photosynthetically active radiation from global solar
Shahidan M F, Jamaluddin A H, Chen G and Jalaludin A radiation in the Contiguous United States Energy Convers.
2020 Modeling of stochastic temperature and heat stress Manage. 89 71–82

11