France's PV Atlas

Best estimate: 13,447 MWp, with a 90% range of 11,501–17,836 MWp. The range covers two measured sources of uncertainty, not everything that could move this number. Below: what is inside the range, what is not, what calibration areas it rests on, and how it compares with other sources.

Inside the range:

• The two area-to-power constants. Panel area on a roof converts at about 0.18 kWp per m², the rating of a typical module; the outline of a ground-mounted park converts at about 0.05 kWp per m², measured on parks whose capacity is known. Each is varied over its own range (90%: 0.15–0.21 and 0.035–0.075 kWp per m²).
• How reliable the segmentation model's detections are. How often its detections turn out to be real, and how many mapped arrays it misses, were measured size class by size class against mapped installations. Those measurements have their own sampling error, which is drawn into the range.

Outside the range, and not something more arithmetic can add back:

• The calibration areas were chosen by hand, not drawn at random, so the range is not a statistical margin of error for the country. Re-drawing them shows how much the areas disagree with each other, not how far the places the estimate is applied to differ from all of them. Only a random sample of areas can measure that.
• "Every panel mapped" holds as of the mapping imagery, which is older than the satellite imagery. Panels installed in between are missing from the reference: the model's detections of them count as false alarms and its misses of them go uncounted. Each of these makes this figure more likely too low than too high.
• This page is a precision floor, not a recall-corrected estimate. France has no glint sample, so the chance that an unmapped detection is real is set to zero and the model's credit collapses to the share of its detections that OpenStreetMap already confirms. The recall correction that roughly doubles comparable Pakistani figures was deliberately not applied: the only French ground truth available to measure it against is a sub-400 m² population, and a correction fitted there would have been manufactured rather than measured.

Treat this as an early estimate from an active research pipeline, not a census. What it offers is a reproducible way to estimate distributed solar from free satellite imagery and open-source models where official statistics are thin or missing. Expect the figures to move as the evidence behind them grows.

Fourteen French communes were mapped panel by panel, and what they measured is the limit of the instrument rather than a correction to it. Across 3,335 hand-drawn features on sub-metre imagery, the median installation is 20 m² against a Sentinel-2 pixel of 100 m². Detection recall climbs with array size across those communes (Spearman +0.83) while an independent model reading the same installations from sub-metre imagery is flat, which places the gradient in the sensor and not in the annotator. That is why this page has no small-panel half at all: the per-building classifier that supplies most of Pakistan's estimate scores 0.710 here against 0.857 there, and at a usable precision it flags 16 buildings out of 44,314. The communes are drawn on the map above; they set the floor, they do not raise the total.

France's own register says this page is missing most of the fleet, which is the expected result. The ODRE register puts French PV at roughly 34.6 GWp, several times the figure above, and 92.0% of French installations by count sit below this pipeline's 400 m² detection floor. Per commune, the detection estimators correlate with the register at Spearman 0.45 after retraining in domain, against 0.29 zero-shot -- the retrained model got the geography substantially better while recovering about a third of the capacity. Read the map for where French PV is, not for how much there is.

Layers. "Show" picks the quantity the cells are coloured by: the Best estimate, the OSM-mapped capacity alone, the satellite detections above 400 m², rooftop PV below it, farm ground PV below it (arrays beside fields, where the farm detector has run), or the gap between the Best estimate and what is already mapped, which is where mapping effort pays off most. Each cell shows its total in MWp. The colour scale is logarithmic and is refitted to whichever quantity is shown; the legend always states its range.

Your own data. "Layers" → "Add a layer" overlays a GeoJSON file, a CSV with latitude/longitude columns, a GeoJSON URL, an XYZ raster tile URL or a WMS (use {bbox-epsg-3857} in its GetMap URL). Files are read in your browser and are not uploaded anywhere. The live OpenStreetMap overlay queries the public Overpass API for the area in view when you zoom in, so it shows mapping done after this page was built.

Exports. Every figure downloads as PNG (for slides), SVG (vector, for papers) and CSV (the numbers behind it); the map as PNG with its legend and attribution, and the grid as GeoJSON or CSV. Basemap imagery and tiles keep their own providers' terms: OpenStreetMap data is © OpenStreetMap contributors (ODbL), the grey canvas and the satellite imagery © Esri and its partners.