Shade Analysis and Energy Modeling: Getting Production Estimates Right

The estimate that came in high

A designer sizes a 9.4 kW roof mount, runs the numbers, and quotes 13,600 kWh for year one. Twelve months later the meter reads closer to 12,100. The customer is holding a signed proposal, the sales rep is fielding an angry call, and nobody did anything obviously wrong. The model just told a story the roof could not deliver.

Trace it back and two things stand out. The shade study used a single midday snapshot, so the sun sat high and the shadows were short. And the tree line along the south property edge got drawn as a thin band when it was actually a stand of forty-foot oaks. In June at noon those trees clear the array. In December at 9 a.m. they cover the bottom two rows. The model never saw that hour, so it never counted the loss. This is the most common way a production estimate goes wrong, and it is entirely preventable.

Production is a physical outcome, not a slogan. The U.S. Department of Energy frames output as the product of irradiance reaching the modules, the array's orientation, the temperature it runs at, and the losses along the way (DOE, Solar Performance and Efficiency). Every one of those has to be modeled from the actual site. Assume them and you are guessing with decimals.

What makes an estimate defensible

Defensible does not mean optimistic or conservative. It means you can point to each input and say where it came from. When a customer, a financier, or a production-guarantee auditor asks why the number is what it is, a defensible estimate has an answer for the weather file, the shade geometry, and every loss factor in the stack.

That matters because more estimates now carry weight beyond the sale. A production guarantee ties a dollar figure to the model being right. If the design over-counted and the system underdelivers, someone pays the difference, and the relationship with the customer takes the damage regardless. The DOE's guidance for buyers tells homeowners to compare production estimates and ask how they were produced (DOE, Homeowner's Guide to Going Solar). A savvy buyer will ask. Your estimate should survive the question.

The rest of this piece walks the inputs a real shade and energy model needs, in the order they tend to trip people up.

Why a year-round sun path beats a midday snapshot

The sun's position changes by the hour and by the season. In summer it rides high and crosses close to overhead at noon. In winter it stays low and swings through a short, shallow arc. A shade study that captures one time on one day, usually solar noon because it looks clean, misses the exact hours when obstructions cast their longest shadows.

A year-round model steps through the sun path for every hour of daylight across all twelve months and asks, for each module, whether something blocks the sun at that moment. Sum the blocked hours against the energy that would have flowed and you get a shade loss that reflects reality instead of a single flattering instant. This is the difference between a solar access value that means something and a screenshot.

The low-angle hours are where snapshots lie. Early morning, late afternoon, and the whole winter season put the sun behind objects that noon sun clears easily. Those hours carry less energy individually, but they add up, and in heating-season climates the winter shade loss can decide whether a system meets its annual target. Skip them and the estimate drifts high in exactly the predictable direction.

Near shading, far shading, and the tree line that grows

Shade comes from two distances, and a model has to handle both. Near shading is the stuff on or near the roof: plumbing vents, chimneys, dormers, satellite dishes, an HVAC unit, a parapet, or one part of the array shading another at low sun. It moves fast across the modules and often knocks out specific cells or substrings.

Far shading is the horizon: a tree line, a hill, a taller building next door, a distant ridge. It tends to clip whole rows during the low-sun hours and shows up as a horizon profile the sun has to climb above before the array sees full light. The oak stand in our opening scenario is far shading, and it was under-counted because someone eyeballed its height and depth instead of measuring it.

Two things make far shading from vegetation tricky. Trees grow, so a tree line modeled at today's height understates year-five losses, and a defensible estimate for a 25-year asset should note that. Trees are also not solid. Bare winter branches pass some light, summer canopy passes almost none, and that seasonal swing interacts with the low winter sun in ways only an hourly model catches. Model the tree line at a realistic mature height and treat it as opaque unless you have reason not to.

The loss stack: from rated watts to delivered kWh

A module's nameplate is measured in a lab at 1000 W/m2 and 25 C. The roof is never that lab. Between the rated DC power and the kilowatt-hours that reach the meter sits a stack of losses, and each one deserves a modeled number rather than a default someone copied years ago. The table below lists the common factors, what each one represents, and the way it most often gets set wrong.

Loss factor	What it captures	How it is often mis-set
Shading	Sun blocked by near and far objects across the year	Taken from a single midday snapshot, so low-angle hours are missed
Temperature	Output drop as cells heat above 25 C	Ignored, or a flat figure used regardless of climate and mounting
Soiling	Dust, pollen, and debris on the glass between cleanings	Left at a generic default in a dusty or agricultural site
Inverter efficiency	DC-to-AC conversion and clipping at high irradiance	Rated peak used instead of the weighted operating efficiency
DC and AC wiring	Resistive losses in conductors from array to interconnection	Underestimated on long runs from a far roof to the service
Mismatch	Modules in a string performing slightly differently	Set to zero, which never happens in the field
Module degradation	Gradual output decline over the system's life	Omitted, so year-one and year-25 are treated as identical

Read down that list and you can see how a stack of small optimistic choices compounds. A percentage point shaved off temperature, another off soiling, mismatch set to nothing, and the estimate gains several percent it will never produce. Treat every row as a decision you can defend, and keep any figure you cite directional or labeled as an estimate rather than a promise.

Weather files, temperature, and soiling

Energy modeling starts from a weather file, a long-run hourly record of irradiance and temperature for a location near the site. The file drives how much sun is available each hour and how hot the modules run. Pick a file for the wrong region, or one that does not match the site's microclimate, and every downstream number inherits the error.

Temperature is the loss designers underweight most. Solar cells lose efficiency as they heat, and a dark roof in a hot climate can push module temperatures well above the 25 C rating for much of the year, trimming output on the sunniest, hottest afternoons. The DOE notes that cell temperature and the light reaching the panel both drive real efficiency, which is why a hot-climate array and a cool-climate array with identical hardware produce differently (DOE, Solar Performance and Efficiency). A flush roof mount with little airflow runs hotter than a raised rack, and the model should reflect the mounting.

The Department of Energy explains that a photovoltaic system's real-world output depends on the sunlight actually striking the modules and the temperature the cells operate at, so performance is shaped by site conditions rather than by nameplate ratings alone.

DOE, Solar Performance and Efficiency

Soiling is the other site-specific factor. An array near a highway, a farm field, or in a low-rain desert accumulates dust and grime that a suburban roof in a rainy region does not. A generic soiling default undercounts the loss in those places. Match it to the environment, and note whether the design assumes any cleaning.

Why peak-hour shade hurts more than the clock says

Not all shaded hours cost the same. Production follows a bell curve across the day, low at dawn, peaking around solar noon, low again at dusk. Shade that lands in the middle of that curve removes high-value energy. The same duration of shade at 7 a.m. removes far less, because the array was barely producing then.

So a shade study has to weight blocked hours by the energy they would have carried, not just count them. An obstruction that clips the array from 11 a.m. to 1 p.m. is a serious problem even if it sounds like only two hours. One that clips it for the first hour after sunrise barely registers. Averaging shade across the day treats those two cases as similar when they are not, and that is another way estimates drift.

Module-level electronics change the math but do not erase it. With string inverters, shade on a few cells can drag a whole string through the series connection. Optimizers or microinverters isolate modules so a shaded panel does not pull down its neighbors, which recovers some of the loss. The model should reflect the actual electrical design, because the same shadow costs different amounts on different systems.

Production-estimate QA checklist

Run this before an estimate leaves your desk. It catches the misses that push a number high.

• Shade is modeled hour by hour across all twelve months, not from a single midday snapshot.
• Near shading objects (vents, chimneys, dormers, HVAC, parapets) are drawn to scale on the roof.
• Far shading (tree lines, ridges, adjacent buildings) is measured, not eyeballed, and trees are set at mature height.
• The weather file matches the site's region and microclimate.
• Temperature loss reflects the climate and the mounting type, not a flat default.
• Soiling is matched to the local environment and any cleaning assumption is stated.
• Inverter, wiring, and mismatch losses are modeled, and mismatch is not set to zero.
• Module-level behavior (string vs optimizer vs microinverter) matches the electrical design.
• Degradation is included so year-one and later-year figures differ.
• Every percentage in the deliverable is labeled as a modeled estimate, not a guarantee.

Why estimates come in wrong

When a system underproduces against its estimate, the cause usually sits in one of a few places, and shade leads the list. Someone under-counted an obstruction, used a snapshot instead of a full-year study, or drew a tree line thinner than it is. Because shade loss is the input with the widest range, it is where a sloppy model does the most damage.

The next tier is the loss stack. A soiling default left generic in a dusty site, temperature skipped in a hot climate, mismatch set to zero, wiring losses undercounted on a long roof-to-service run. None of these is dramatic on its own. Together they can move an annual figure by several percent, always in the flattering direction, because the easy defaults happen to be the optimistic ones.

There is also the human tug toward a bigger number. A larger estimate closes the sale and clears the payback math, so there is quiet pressure to pick the rosy end of every range. That pressure is exactly what a production guarantee exists to counter, and it is why documenting each input protects you. When the model is transparent, nobody has to relitigate the estimate after the meter disagrees, because the inputs are on the record. A defensible number costs a few more minutes up front and saves the customer relationship later.

Modeling shade and energy in one place

The scenario we opened with fails because shading and energy were treated as separate steps, and the shade step got shortcut. The fix is to model both together over the full year, from the first layout, so the production number reflects the real roof before anyone quotes it.

PVSketch does this in the design stage. It includes shading and energy modeling so a designer can place the array, account for near and far obstructions, and get a fast, defensible production estimate in the same tool that draws the layout (PVSketch). Modeling shade while you place modules means you catch the tree line that clips the bottom rows in winter before it becomes an angry phone call, and you can adjust the layout to work around it rather than discovering the loss after installation.

When the project moves from proposal to permit and construction, the detailed engineering continues in PVCAD, which carries the design into stamped electrical and mechanical documents (PVCAD). One design path, from the estimate a customer signs to the plan set an installer builds from, keeps the production story consistent the whole way. That consistency is what defensible looks like in practice: the number you quoted came from the same model that built the system, and you can show your work.

Frequently asked questions

How accurate are solar production estimates?

A well-built estimate that models hourly shading across the full year, a matched weather file, and each loss factor from site conditions typically lands within a small single-digit percentage of actual year-one output. Accuracy collapses when shade is taken from a midday snapshot or loss factors are left at generic defaults. The DOE describes real output as a function of irradiance, orientation, temperature, and system losses, all of which have to be modeled rather than assumed (DOE, Solar Performance and Efficiency).

What causes solar to underproduce?

The most common cause is under-counted shade, a tree line drawn too thin or a study built from one midday moment that misses the long low-angle shadows. After that come the small loss factors set optimistically: soiling left generic in a dusty site, temperature skipped in a hot climate, mismatch set to zero. Each is modest alone, but they stack in the flattering direction (DOE, Solar Performance and Efficiency).

What is a shade analysis?

A shade analysis is a study of how obstructions block sunlight from reaching an array over time. A real one steps through the sun path hour by hour across all twelve months and accounts for both near shading (vents, chimneys, dormers) and far shading (tree lines, ridges, nearby buildings), then weights the blocked hours by the energy they would have produced. A single snapshot is not a shade analysis (DOE, Homeowner's Guide to Going Solar).

Why does shade during peak hours matter more?

Production peaks around solar noon and tapers toward dawn and dusk, so shade that lands in the middle of the day removes the highest-value energy. The same duration of shade early or late costs far less because the array was barely producing then. A model has to weight shaded hours by the energy they carry, not just count them (DOE, Solar Performance and Efficiency).

Does a hotter climate reduce solar output?

Yes. Cells lose efficiency as they heat above their 25 C rating, so a hot roof trims output on the sunniest afternoons, and a flush mount with little airflow runs hotter than a raised rack. The DOE identifies cell temperature as a direct driver of real-world efficiency, which is why identical hardware produces differently across climates (DOE, Solar Performance and Efficiency).

What tool models shade and energy for a solar estimate?

PVSketch includes shading and energy modeling in the design stage, so a designer accounts for near and far obstructions and produces a defensible estimate in the same tool that draws the layout (PVSketch). The design then continues into stamped engineering documents in PVCAD (PVCAD).

Sources

1. DOE - Solar Performance and Efficiency
2. DOE - Solar Energy Technologies Office
3. DOE - Homeowner's Guide to Going Solar
4. EnergySage - Solar Panel Cost
5. EIA - Electricity prices and factors affecting prices
6. NFPA - Understanding NFPA 70 (NEC)
7. PVSketch - Shading and energy modeling (PVComplete)
8. PVCAD - Engineering and plan sets (PVComplete)