Agrivoltaics Design: Spacing, Height and Crop-Light Calculations

Agrivoltaic arrays (solar over working farmland) need taller panels and wider rows than standard ground-mount PV, so crops get light and machines can pass. Fraunhofer ISE reports modules 3 to 5 m above arable fields. Dupraz finds most crop yields stay above 80% of the no-panel yield when ground coverage ratio (GCR, the share of land the panels cover) is below about 25%.

SurgePV publishes this guide and sells solar design software. The worked examples are teaching calculations, not customer projects or measured crop results. For the wider planning process, see the agrivoltaics design guide. For a deeper look at row geometry, see agrivoltaics row spacing.

Agrivoltaic system types and when each fits

Fraunhofer ISE, a German solar research institute, splits open agrivoltaics into two groups. Interspace PV sits low, and you farm between the rows. Overhead PV sits high on stilts, and you farm under the modules. The German standard DIN SPEC 91434 draws the line at 2.1 m of clearance. Category I is overhead PV above 2.1 m. Category II is interspace PV below 2.1 m (Fraunhofer ISE guideline, February 2024).

The table shows the designs that NREL, the US national renewable energy laboratory, costed in 2020 next to the Heggelbach research plant in Germany. NREL’s baseline is a standard ground-mount on bare ground.

NREL’s inter-row spacing assumes a 44 degree tilt for every design except the vertical mount. Metric values are converted from NREL’s feet. The Heggelbach ratio is 3.4 divided by 9.5, which gives 0.36. The source does not state a GCR for that plant.

Elevated fixed or overhead. Fraunhofer ISE reports modules typically 3 to 5 m above the field for arable farming. Hops need more than 7 m. Vineyards need 2 to 3 m. Wider rows typically cut ground coverage to about one third.

Vertical bifacial. These panels stand upright and collect light on both faces (bifacial). Fraunhofer ISE says they cost less because the mounting structure is lower. They also give you fewer ways to manage light. NREL’s 2023 overview calls vertical bifacial panels the most farm-equipment friendly, with the largest trade-off in energy production.

Single-axis or dual-axis tracking. Trackers tilt the modules to follow the sun. They give flexible light control but raise cost and control work. Fraunhofer ISE says dual-axis systems with large module tables tend to leave a deep shadow behind each table. Other parts of the field then stay in full sun. That suits crops less well.

Narrow modules and tubular modules. Japan’s “solar sharing” keeps panels facing south and uses narrower modules to spread light. Tubular thin-film modules aim for an even pattern of light and water.

Define six inputs before comparing layouts

Record the actual farming operation and the model’s conventions. These six inputs are the minimum set:

Add crop rotation, future equipment, headlands, irrigation, terrain, water, animals and maintenance to the brief. A tractor’s straight-line width does not give its turning envelope. A layout built for today’s hand harvest may not suit the farm’s equipment in five years.

What GCR research says about crop yield

Christian Dupraz’s 2023 synthesis of GCR and crop productivity defines GCR as PV panel area divided by land area. He used only studies that report both GCR and relative crop yield (yield under panels compared with a plot without panels). He left out some studies because of edge effects or unreliable control plots. His abstract finds that a GCR below 25% is needed to keep most relative crop yields above 80%. In the text, a GCR of 25% gives an average relative plot yield of 80%.

Two yield rules frame the target. The same abstract notes minimum relative yields of 80% in Japan and 66% in Germany. DIN SPEC 91434 asks for at least 66% of the reference yield. It also limits land loss to 10% of the project area for category I and 15% for category II (Fraunhofer ISE, 2024).

NREL’s 2020 crop benchmarks use GCRs of 20% to 28%. NREL also reports a design from at least one company at a GCR of 44%. NREL expects it to cut installed cost. Its effect on lifetime cost and revenue still needs study.

Calculate row pitch from GCR

For GCR defined as row width divided by row pitch:

row pitch = row width / GCR

These numbers are geometry only. Take a 2 m footprint on an 8 m pitch. The strip between rows is 6 m, before you allow for supports, terrain, cables or buffers. If 2 m is the width of a sloping table, subtract its horizontal footprint instead.

Heggelbach shows a real case. Its rows sit 9.5 m apart and each row is 3.4 m wide. If you read 3.4 m as the footprint, the strip between rows is 9.5 − 3.4 = 6.1 m. Fraunhofer ISE reports that this spacing cut installed capacity per hectare by about 25% compared with conventional ground-mount PV.

GCR is not the share of crop light lost each day. Light depends on sun position, weather, tilt, terrain, structures and where the crop sits. Ask how your model turns geometry into a light map.

Set clearance for the largest machine you expect

Set the usable envelope with the farmer and qualified reviewers. Include raised implements, moving parts, crop height, access routes and turning. For trackers, separate positions used for generation, stow and maintenance.

Clearance above ground differs from clearance above crop canopy. Check which reference your model and your structural drawings use.

Fraunhofer ISE gives a rule of thumb for east-west fixed modules. Clearance should be at least 1.5 times the row width, or 2 times for tracked modules. This avoids a deep shadow under the row. For a 2 m row width, that means 3 m fixed and 4 m tracked.

Crop rotation drives the decision. NREL notes that developers often design for several crops and harvest methods, because farming practice and markets change. That can mean higher panels or more spacing. Fraunhofer ISE adds that in arable farming the design must suit every crop in the rotation. With permanent crops, the design can fit one crop.

DOE’s Farmer’s Guide to Going Solar describes wider spacing between panels and rows as a common adaptation. It also notes that more spacing reduces the electricity produced on a given piece of land. Your structural and geotechnical engineers set loads, foundations and investigations. Fraunhofer ISE does not recommend concrete foundations on farmland and names piled or screw foundations as alternatives.

Calculate daily light integral from photon flux

Daily light integral (DLI) is the total light plants can use over one day, in mol/m²/day. Purdue Extension separates this daily total from an instantaneous light reading and describes logging with a quantum sensor.

Light plants can use at one moment is called photosynthetic photon flux density (PPFD), in µmol/m²/s. For interval-average PPFD:

DLI = sum(PPFD for each interval × interval duration in seconds) / 1,000,000

Use valid intervals across the whole day and record gaps, sensor location and the integration method. A shaded-area fraction from a photo does not replace photon-flux data.

Unit check. Say PPFD stays at 500 µmol/m²/s for 10 hours. That is 500 × 36,000 / 1,000,000 = 18 mol/m² for that period. Real PPFD changes through the day, so add up the real curve.

Reduction example. Assume an open-field DLI of 40 mol/m²/day and a 30% daily light reduction. Then DLI under the panels is 40 × (1 − 0.30) = 28 mol/m²/day. The 30% must come from a light model or sensors, not from GCR alone.

For a real reference point, Fraunhofer ISE reports that lettuce under an agrivoltaic system grew more leaf area when light fell by about 30%. Crop light needs come from farm research on your variety, growth stage, climate and management.

Keep bifacial performance separate from crop light

The hard part of modeling bifacial panels is the light that hits the back. Sandia’s PVPMC overview says this rear light changes with position on the array, array design and nearby objects such as combiner boxes and racking. It also depends strongly on albedo, the share of light the ground reflects. Sandia names two modeling methods: ray tracing and view factor.

Use the real module data, front and rear light, structures, vegetation and the electrical model. Ground cover changes with the crop calendar, so make your albedo assumptions follow it. Transparency and bifacial response are separate properties, so check the datasheet. Do not apply the bifaciality factor to a rear power rating twice.

Worked examples: pitch, energy and light

Each example reuses the formulas above. All inputs are assumptions for teaching, not site data.

Notes on each:

• Leafy greens. The 6 kWp is an assumption, not derived from an acre or a GCR. Capacity comes from module count and rating. This also means no crop yield is claimed.
• Berries. Multiplying $21,600 by 25 years gives $540,000, an undiscounted constant-price total. It is not a lifetime forecast or a return, because prices, costs and generation change.
• Sheep pasture. NREL’s Table 2 gives 7.0 acres per MW for the vertical mount and 4.5 for conventional fixed tilt. Define row direction, azimuth of both faces and spacing, then model seasonal generation and pasture light. Get a grazing plan from an animal-welfare and site reviewer. The French sheep-grazing research review shows what that evidence looks like and where it ends.

Costs and land equivalent ratio

NREL’s 2020 cost model puts the extra installed cost at $0.07 to $0.80 per watt DC. That is for ground-mounted dual-use PV compared with a standard ground-mount on bare ground. The highest extra costs are for PV plus crops, because the support structures change (NREL TP-6A20-77811).

Fraunhofer ISE gives 2020 German estimates and says they do not reflect current prices. For stilted overhead systems above 4 m, mounting costs 372 euros per kWp on average, within a range of 243 to 500. For ground-mounted PV it is 76 euros per kWp. The NLR agrivoltaics overview says agrivoltaics is not always more expensive than traditional solar development. Get quotes for your own configuration.

Land equivalent ratio (LER) adds two ratios: crop yield and electricity yield on shared land, each compared with single-use land. An LER of 1.3 means 10 ha of agrivoltaics gives as much crop and electricity as 13 ha of separate farms and solar plants. For Heggelbach, Trommsdorff (Fraunhofer ISE, 2020) subtracts land losses of 8.3%:

LER 2018 = 255.26 / 230.02 + 249,857 / 301,032 − 0.083 = 1.110 + 0.830 − 0.083 = 1.86

The inputs are potatoes at 255.26 dt/ha under panels against 230.02 dt/ha on the reference plot. Electricity is 249,857 kWh/year against 301,032 kWh/year for a conventional plant. The same source gives about 1.5 for 2017 wheat. That is roughly 80% of wheat yield and 80% of electricity, with about 10% land loss. Potato yield under the array was 18% lower than the reference in 2017 and 11% higher in 2018.

The Fraunhofer ISE 2024 guideline reports a land-use rate of 160% in the first project year (2017). It says crop yields stayed above 80% of reference. Treat these as results from one site in one region.

For illustrative crop and electricity ratios of 0.80 and 0.75, LER is 0.80 + 0.75 = 1.55. LER is a land-use ratio. It does not measure profit or approval. Use per-area quantities, comparable periods and a defensible reference.

Common design mistakes

1. Treating GCR as a direct percentage of daily crop-light loss.
2. Confusing row pitch with the clear machine aisle or turning area.
3. Using one ground-referenced clearance for every crop, tracker position and future rotation.
4. Substituting midday shadow coverage for integrated photon data.
5. Predicting annual bifacial gain from albedo alone.
6. Presenting illustrative energy or DLI arithmetic as measured crop or financial results.
7. Reading one study’s GCR recommendation as every site’s specification.

When crop rotation, equipment, geometry or contracts change, redo the agricultural, technical and financial work that depends on them.

Model your layout in SurgePV

SurgePV covers solar array layout, shading analysis and energy-yield and financial modeling. Crop light needs and crop yield come from agricultural data and field trials, not from PV design software. To size a plot first, see the solar land area calculator and the inter-row spacing guide. For mounting trade-offs, read tracker vs fixed tilt.

Frequently Asked Questions

What GCR should an agrivoltaic design use?

Dupraz’s 2023 review of published crop data finds that a GCR below about 25% is needed for most crop yields to stay above 80% of the no-panel yield. NREL’s 2020 crop benchmarks use 20% to 28%. Fix how you define GCR first. Then test it against your crop, climate and local yield rules.

How do I calculate row pitch from GCR?

Define GCR as row width divided by row pitch. Then pitch equals width divided by GCR. A 2 m row width at GCR 0.25 gives an 8 m pitch. That leaves a 6 m strip between rows, before supports and obstacles.

How high should agrivoltaic panels be?

Fraunhofer ISE reports modules 3 to 5 m above the field for arable farming. Hops need more than 7 m and vineyards 2 to 3 m. DIN SPEC 91434 calls systems with clearance above 2.1 m overhead (category I). The Heggelbach research plant uses 5 m. Size clearance to the largest machine you expect over the project life.

How should daily light integral be calculated under panels?

Multiply each light reading (PPFD, in micromoles per square meter per second) by the length of its interval in seconds. Add the results and divide by one million. That gives moles per square meter for the day. Midday shadow or GCR alone does not give DLI.

Does panel height determine bifacial gain?

No. Light on the back of the panel depends on the array design, nearby objects and how much the ground reflects. Use a ray-tracing or view-factor model with your real module and site. Height is one input to that model. It is not a gain figure.

What are examples of agrivoltaic systems?

Fraunhofer ISE lists research sites in Germany. They include a 194 kWp fixed system over arable crops at Heggelbach (2016). They also include a 134 kWp vertical system at Dresden-Pillnitz (2022) and a 300 kWp east-west fixed system over vines at Tuniberg (2023).