Complex site topography can have a crucial bearing on a PV system’s energy yield. In a market characterised by tight margins and careful risk management, precision in terrain-aware modelling is an essential ingredient of smart engineering, writes Solargis CEO Marcel Suri.
In photovoltaics, terrain is not a cosmetic detail; it’s an important input into energy yield. The reality is that many utility-scale PV projects are not built on perfectly flat plots of land. On the contrary: they often sit on sloped, hilly or otherwise irregular grounds, with lots of variables that need to be taken into account.
Get Premium Subscription
Complex topography affects the fundamentals behind PV performance in two ways. Optically, it changes shading behaviour and view factors. Electrically, it affects operating points and losses that depend on irradiance variability, temperature and angle of incidence. If the geometry used in the simulation does not match the geometry used in the design, the uncertainty introduced is not random, it is systematic. Wrong geometry in, wrong yield out.
This is why PV layouts should reflect real terrain: not only to reduce construction risk, but also to avoid inaccuracies in yield estimates. In practice, however, the connection between design and simulation is often weaker than it appears.
When you import a detailed terrain model and design a PV layout in 3D, you expect that when you hit “simulate”, the design will reflect the terrain. But many solutions on the market simplify the terrain, smoothing out slopes and removing local variations until the whole power plant is treated as if it were built on flat ground. The terrain is simply lost in the process.
So how can PV designers and developers avoid this and simulate energy yield using terrain that actually reflects reality?
The first challenge is obtaining and importing terrain data into the PV design tool. The first prerequisite, of course, is using software that supports terrain adapted design.
The industry still uses a wide range of terrain sources. Many tools rely on global DEM (Digital Elevation Model) layers with 30-90-metre resolution, but this data is often several years to decades old – adequate for early screening, but it can be misleading in mountainous terrain, in valleys, on terraced land, or on sites shaped by earthworks, roads and drainage.
For detailed design, you typically need to import your own terrain. The most robust format in practice is often GeoTIFF, because it carries georeferencing information in the file header. When you load multiple tiles, they can align correctly without manual stitching.
This leads to a key point that is easy to overlook. “Terrain” is not always terrain. Remote sensing products such as LiDAR (Light Detection and Ranging) may include tree canopies, buildings or temporary structures. Sometimes that is exactly what you want, because it can capture near shading without manually modelling every obstacle.
Sometimes it is a problem, because artefacts can behave like solid barriers in a 3D mesh. Power lines, for example, can be reconstructed as a wall rather than cables in the air, which then produces unrealistic shading and false collisions. Terrain data needs scrutiny before it becomes design truth.
Even when you have good terrain, you face a different constraint: computation. A detailed LiDAR or drone scan produces a so-called ‘point cloud’ that can be many gigabytes large. It is often unrealistic to manipulate these raw datasets interactively in a browser-based environment, and you do not want to waste rendering capacity on parts of the scene that have no relevance to PV geometry.
A useful approach is to convert elevation data into an optimised triangular mesh that preserves the important ridgelines and breaks in slope, while reducing triangles in areas where the surface is nearly flat.
For example, algorithms such as Mapbox’s “Delatin” illustrate this principle well: you accept controlled, quantified loss in geometric fidelity in exchange for a mesh that can be rendered and edited smoothly. The goal is maximum relevance where design decisions are sensitive to slope, curvature and local shading.
Once the terrain is usable, the central question becomes: does the PV layout actually conform to it, and does the simulator honour that same conforming geometry?
On undulating ground, table placement changes continuously. Row-to-row clearance, pile heights and tracker rotations interact with slope and local curvature. In tracker systems, the situation is especially unforgiving because a table that clears the terrain at one rotation angle may intersect it at another. If you only check geometry at a neutral position, you can miss collisions that occur during morning or afternoon tracking, or during stow events.
A terrain-adapted layout therefore needs two qualities. It must place structures on the terrain with realistic constraints, and it must preserve the detailed geometry into the simulation step, where shading and irradiance distribution are computed. If the simulator replaces that with a flattened approximation, you lose the entire point of terrain aware design.
On complex terrain, collisions are not rare edge cases. They are a predictable consequence of steep slopes, short clearances, and non-uniform table heights. Collisions can occur between adjacent tables in a deep valley, between tables and the terrain when legs are too short and between tables and equipment such as inverters placed beneath structures.
Smart collision detection is about preventing late-stage redesign and reducing the risk that constructability problems will arise after procurement decisions are made. Robust software solutions detect collisions immediately as the designer edits the layout, because waiting until a final design review is often too late and too expensive.
Most engineers are familiar with slope limits. At some point, a surface is simply too steep for economically reasonable construction and maintenance. But slope magnitude alone does not describe terrain suitability.
In the northern hemisphere, a given slope angle can have very different implications depending on whether it faces south or north. The same grade can ither support favourable module orientation or create persistent shading and access problems.
That is why terrain azimuth, the directional orientation of slopes, deserves to be treated as a first-order design variable rather than an afterthought. In practice, the best terrain-aware decisions come from combining slope magnitude, slope direction and local shading context into a single constraint view, rather than relying on a single “maximum slope” number.
Terrain also changes the effective tilt and exposure of PV surfaces, which can influence soiling dynamics and snow behaviour. In mountainous regions, especially, differences in surface tilt and local wind patterns can affect how quickly modules shed snow, how rainfall cleans surfaces and how long soiling persists.
This does not mean a model can “solve” snow and soiling perfectly. It means that if you ignore terrain in the geometry, you can end up applying loss assumptions that are inconsistent with how the plant will actually behave. For banks and investors, these inconsistencies show up later as performance surprises. For engineers, they show up as overdesign margins and uncomfortable uncertainty.
If you work on projects in complex topography, there is a straightforward question worth asking any simulation provider: how does your engine represent undulating terrain in the actual yield calculation?
Does it use the same 3D geometry you designed, or does it use a flattened layout?
Terrain-adapted PV design is not about making a 3D image of a solar power plant that looks impressive. It is about ensuring that the geometry you model is the geometry you simulate, and that the physics of the site is preserved all the way into bankable energy yield assumptions.
When simulation is done at the cell level on a realistic terrain model, the loss breakdown becomes more honest: you can see which energy losses are driven by topography, weather conditions, and which by design choices such as row spacing, tilt, or tracker geometry.
In a market where margins are tighter and risks are scrutinised harder than ever, precision in terrain-aware modelling is no longer a nice to have. It’s part of a smart engineering practice.
Marcel Suri is an entrepreneur and cofounder of the solar data and software company Solargis. He is an expert in solar resource, photovoltaics and geoscience. Holding a PhD in geography and geoinformatics, Marcel has made significant contributions to solar energy through science and peer-reviewed research. Driven by a passion for innovation, he is dedicated to improving the efficiency of digital tools and data resources and analytics that mitigate weather-related risks and elevate industry standards.

source