SolarMapped › Sources
Sources and method
SolarMapped turns public solar-resource data into something you can read at a glance. This page lists what we use, how we use it, and what the numbers do and don’t mean.
Solar estimates: PVGIS
All solar figures come from PVGIS (Photovoltaic Geographical Information System, version 5.3), provided by the European Commission, Joint Research Centre (JRC). PVGIS combines satellite-derived and reanalysis weather data with a PV performance model. The report uses PVGIS’s PVcalc tool. PVGIS picks the most suitable radiation database for each location, and the report shows which one was used (for example PVGIS-SARAH3 in Europe, Africa and Asia, PVGIS-NSRDB in the Americas or PVGIS-ERA5 elsewhere) along with its year range.
Requests go from your browser to SolarMapped’s own API, which calls PVGIS, converts the answer to our format and caches it. Coordinates are rounded to 0.01° (about 1 km) before calculating and caching.
Map layer
The colours on the map show annual PV output in kWh per kWp per year for fixed panels at their optimal tilt. The layer was generated by running the same PVGIS calculation on a 2° global grid (4,316 land points, values from 557 to 2,373), then smoothing between points and projecting to the web map. It is a regional overview: it will not show local effects such as valleys, shading or coastal fog, which the location report can. The layer was last generated on 21 September 2026.
Weekly radiation layer
The timeline under the map switches it to solar radiation per week, in kWh/m². This is the sunlight energy arriving on a panel at the optimal tilt during that week — not electricity, and not the annual PV figure the default layer shows. The two layers therefore use different colour scales, and the legend changes with them.
It comes from PVGIS hourly time series (2021–2023) at the same 2° land grid points (4,316 of them). Every hour is added to its week and the years are averaged, giving a typical year rather than any real one — no week shown has actually happened in that exact form. The year is split into 52 weeks of seven days, with the last week absorbing the leftover days so the weeks line up across years. The colour scale is fixed from 0 to 70 kWh/m² worldwide, so a week in June and a week in December are directly comparable; the sunniest week in the data set reaches 68.1 kWh/m². The report's weekly chart runs the same calculation for your exact point. Note that this series averages a shorter run of years than the annual PVcalc figure, so an individual week is a noisier estimate than the annual total: one unusually cloudy or clear year carries more weight in it. Generated on 21 September 2026.
Calculation assumptions
| PV technology | Crystalline silicon |
|---|---|
| System size | Calculated for 1 kWp and scaled linearly to your chosen size |
| System losses | 14% (cables, inverter, dirt, ageing and similar). Temperature and reflection losses are modelled separately by PVGIS. |
| Mounting | Fixed, free-standing |
| Orientation | Facing the equator: south in the northern hemisphere, north in the southern |
| Tilt | Optimised by PVGIS for the location |
| Shading | Terrain horizon from a digital elevation model, where available. Buildings and trees are not modelled. |
| Cloud and haze | Included. Taken from satellite observations of the actual sky over each location, not from a clear-sky model. |
| Air temperature | Included. Drives the efficiency the panels lose when hot, from reanalysis weather data. |
| Snow | Not modelled, in either direction — see below. |
| Ground reflectance | A fixed value all year. Seasonal changes, including snow cover, are not modelled. |
| Weather | Long-term multi-year averages, not a forecast |
These are model estimates, not an installer quote or a roof-specific simulation. Real output depends on the actual roof, nearby shading, equipment, and the weather in a given year.
How weather is handled
Weather is not a correction applied on top of a latitude calculation; it is the primary input. The radiation databases PVGIS draws on are built from geostationary weather satellite imagery, so the cloud cover, haze and aerosol over each point are observed rather than assumed. Air temperature and wind come from reanalysis data and feed the thermal model, because panels lose efficiency as they heat up.
This is why two locations at the same latitude can return very different figures, and why dry regions outrank cloudier ones at the same sun angle. The figures are averaged over many years, so they describe a typical year rather than a forecast or any particular year. Each report shows the database and year range that applied to that location.
Snow
Snow is not modelled, and it is the most significant known gap in these estimates. Two separate effects are missing, and they pull in opposite directions:
- Snow lying on the panels blocks production until it clears. Nothing in the model represents this. The 14% system-loss figure above is a flat, year-round allowance for cables, inverter, mismatch and dirt; it is not a snow term and does not vary by season or latitude.
- Snow on the ground raises surface reflectance sharply, and a tilted panel gains from light bounced off it. Because ground reflectance is held at a single fixed value all year, this winter gain is absent as well.
The two effects partly offset each other. The residual error is also smallest where snow is most frequent: at high latitudes the midwinter weeks contribute only a small fraction of the annual total, so even a complete loss of those weeks shifts the yearly figure by a few per cent at most. The larger risk is late winter and early spring, when irradiation has climbed steeply but snow can still cover a roof for days at a time. In snowy climates, read those months as an optimistic bound.
Qualitative labels
The label next to a result (for example “Good”) is a fixed, worldwide banding of kWh/kWp/year. It is not relative to the country or region, and the map legend uses the same bands.
| Label | kWh/kWp/year |
|---|---|
| Low | below 700 |
| Moderate | 700 to 1000 |
| Good | 1000 to 1300 |
| Very good | 1300 to 1600 |
| Excellent | 1600 to 1900 |
| Exceptional | 1900 and above |
“Optimal orientation” is reported as the direction facing the equator and “optimal tilt” as the fixed angle from horizontal that maximises annual output in the model.
Place search
Place names come from a city index built into the site from GeoNames — every city above 15,000 people, with its country and region. The GeoNames data is used under the Creative Commons Attribution 4.0 licence. The index downloads once, the first time you search or click.
Naming a point you click never leaves your browser. The report shows the city the point falls in, or, out in open country, how far it is from the nearest one — and over water, ice or empty desert it simply shows the coordinates, because there is no named place there to report.
Searching is different: a search that finds nothing is a dead end, so anything the index can’t answer — a street, a landmark, a village below that size, or a spelling it files differently — falls back to Nominatim, the search engine for OpenStreetMap data, operated by the OpenStreetMap Foundation under its usage policy. Searches run when you submit the form (not as you type) and results are cached. Data © OpenStreetMap contributors, available under the Open Database Licence (ODbL).
Basemap
Map tiles are served by OpenFreeMap using the OpenMapTiles schema, with data © OpenStreetMap contributors. The map is drawn with MapLibre GL JS.
Updates
Solar-resource averages change slowly. The map layer was generated on 21 September 2026 using PVGIS 5.3; the per-location report uses the same calculation, cached for up to 30 days and versioned so improvements to the method never mix with older results.