Use case: energy and utilities

Finding rooftops suited to solar power

A roof suits solar panels when its area, tilt, orientation and shading fit and enough solar radiation reaches it. Databases such as PVGIS derive that radiation from satellite images. AI on satellite imagery can detect existing panels, and machine learning can estimate rooftop area at scale. In solar cadastres such as Styria's, the roof geometry comes from airborne laser scanning.

Reference
UC-08
Technologies
  • Earth observation
  • Artificial intelligence
Industry
Challenge
Site selection and planning
Updated
Published by
LETO SPACE

01Technology

How a rooftop solar analysis works

A roof suits solar panels when it has enough usable area, a favourable tilt and orientation, little shading from neighbouring buildings and trees, and when enough sunlight reaches the site over the year. An analysis answers these points roof by roof.

The solar cadastre of the Austrian region of Styria shows how this works in practice. A high-resolution surface model from airborne laser scanning describes the roofs. GIS methods then calculate the annual direct and diffuse radiation on a 1 by 1 metre grid and sort the roofs into three classes: very well suited, well suited and suited. For photovoltaics, the result includes the expected annual yield and the peak power in kWp Source: Land Steiermark, accessed 6 October 2026Land SteiermarkSolarpotenzial Steiermarkaccessed 6 October 2026Open source (opens technik.steiermark.at).

Satellites contribute in two ways. Their images are the basis of solar radiation databases such as PVGIS Source: European Commission, Joint Research Centre (JRC), accessed 8 October 2026European Commission, Joint Research Centre (JRC)PVGIS: Data sources and calculation methodsaccessed 8 October 2026Open source (opens joint-research-centre.ec.europa.eu). And deep learning on satellite imagery can find existing solar panels and estimate their size Source: Joule, 19 December 2018JouleDeepSolar: A Machine Learning Framework to Efficiently Construct a Solar Deployment Database in the United States (Yu et al., Joule 2, 2605 to 2617)19 December 2018Open source (opens doi.org). Machine learning and geospatial analysis have also been used to estimate rooftop area and solar potential for the whole globe Source: Nature Communications, 5 October 2021Nature CommunicationsHigh resolution global spatiotemporal assessment of rooftop solar photovoltaics potential for renewable electricity generation (Joshi et al., Nature Communications 12, 5738)5 October 2021Open source (opens doi.org).

  • very well suited
  • well suited
  • suited
  • no class
  • Existing panelsfound by AI
  • Per roofyield and kWp
  • Shadinglowers the class
The radiation is computed on the roof model itself. Satellite radiation data, with a resolution of a few kilometres, is too coarse for single roofs. Detecting existing panels by AI is a separate layer. Illustrative street, not real data. Source: Land Steiermark

The European Solar Rooftops Initiative aims to unlock the large and underused solar potential of roofs in the EU Source: European Commission, EUR-Lex, 18 May 2022European Commission, EUR-LexEU Solar Energy Strategy, COM(2022) 221 final18 May 2022Open source (opens eur-lex.europa.eu).

02Data

Data and satellites

Solar radiation
Satellite-based, METEOSAT, about 4 km resolution, long time series (PVGIS)
Roof shape
Airborne laser scanning, results on a 1 m grid (Styria)
Very high resolution satellite
30 cm per pixel (Pléiades Neo)
Free optical satellite
10 m per pixel (Sentinel-2)

Solar radiation. The European Commission's PVGIS tool estimates solar radiation from METEOSAT satellite data, at about 4 km resolution below the satellite, and offers long time series Source: European Commission, Joint Research Centre (JRC), accessed 8 October 2026European Commission, Joint Research Centre (JRC)PVGIS: Data sources and calculation methodsaccessed 8 October 2026Open source (opens joint-research-centre.ec.europa.eu). The Copernicus Atmosphere Monitoring Service provides time series of solar radiation for a chosen point, which helps to plan both large solar farms and rooftop installations Source: Copernicus Atmosphere Monitoring Service (CAMS), accessed 6 October 2026Copernicus Atmosphere Monitoring Service (CAMS)Solar energyaccessed 6 October 2026Open source (opens atmosphere.copernicus.eu).

Roofs and existing panels. Commercial satellites such as Pléiades Neo record images at 30 cm resolution Source: European Space Agency (ESA), Earth Online, accessed 6 October 2026European Space Agency (ESA), Earth OnlinePléiades Neoaccessed 6 October 2026Open source (opens earth.esa.int). The DeepSolar framework analysed satellite imagery with deep learning to locate solar panels and estimate their size across the contiguous United States Source: Joule, 19 December 2018JouleDeepSolar: A Machine Learning Framework to Efficiently Construct a Solar Deployment Database in the United States (Yu et al., Joule 2, 2605 to 2617)19 December 2018Open source (opens doi.org).

03Deliverables

What you get

Energy companies can use such results to decide where to offer systems, which automating the sales of renewable energy systems describes. Municipalities can use them to inform citizens and to plan their own buildings.

04Limits

Limits

05Next steps

How we help

Regional cadastres, satellite archives, commercial imagery and AI providers all offer parts of this analysis. In strategy and use cases we work out what the result must decide and which data source gives the accuracy you need at a price that fits.

Spacegraph, our discovery platform, has use-case categories such as "Object detection & counting" and "Site selection & spatial planning".

06Use cases

Frequently asked questions

Q1Which data does a rooftop solar analysis need, and how current is it?

It needs a model of the roofs with their tilt, orientation and shading, and solar radiation data for the site. In Styria, for example, the model comes from airborne laser scanning, and the radiation is calculated on a grid of 1 by 1 metre Source: Land Steiermark, accessed 6 October 2026Land SteiermarkSolarpotenzial Steiermarkaccessed 6 October 2026Open source (opens technik.steiermark.at). Satellite-based radiation data covers long time series, up to 30 years or more Source: European Commission, Joint Research Centre (JRC), accessed 8 October 2026European Commission, Joint Research Centre (JRC)PVGIS: Data sources and calculation methodsaccessed 8 October 2026Open source (opens joint-research-centre.ec.europa.eu). The roof model is only as current as the survey or imagery it is built from, so check its date.

Q2Can free Sentinel-2 images find suitable roofs?

Hardly. Sentinel-2 records its finest bands at 10 m per pixel Source: European Space Agency (ESA), accessed 6 October 2026European Space Agency (ESA)Sentinel-2: Facts and figuresaccessed 6 October 2026Open source (opens esa.int), too coarse to show the shape of a single roof. Commercial satellites such as Pléiades Neo reach 30 cm Source: European Space Agency (ESA), Earth Online, accessed 6 October 2026European Space Agency (ESA), Earth OnlinePléiades Neoaccessed 6 October 2026Open source (opens earth.esa.int), and the aerial imagery used by the Copernicus Emergency Management Service 10 to 20 cm Source: Copernicus EMS, Copernicus Emergency Management Service, accessed 6 October 2026Copernicus Emergency Management ServiceUsing Earth Observation Dataaccessed 6 October 2026Open source (opens mapping.emergency.copernicus.eu).

Q3How do we start?

Check first whether your region already publishes a solar cadastre, as Styria does Source: Land Steiermark, accessed 6 October 2026Land SteiermarkSolarpotenzial Steiermarkaccessed 6 October 2026Open source (opens technik.steiermark.at). If it does, the task is often to combine it with your own data, not to build a new map. If it does not, a pilot over one district shows which data source gives the accuracy you need at a price that fits. Our consulting overview shows the path from the problem to operation.

Sources (9)

  1. Land Steiermark, accessed 6 October 2026Solarpotenzial Steiermark (opens technik.steiermark.at)
  2. European Commission, Joint Research Centre (JRC), accessed 8 October 2026PVGIS: Data sources and calculation methods (opens joint-research-centre.ec.europa.eu)
  3. Copernicus Atmosphere Monitoring Service (CAMS), accessed 6 October 2026Solar energy (opens atmosphere.copernicus.eu)
  4. European Space Agency (ESA), accessed 6 October 2026Sentinel-2: Facts and figures (opens esa.int)
  5. European Space Agency (ESA), Earth Online, accessed 6 October 2026Pléiades Neo (opens earth.esa.int)
  6. Copernicus Emergency Management Service, accessed 6 October 2026Using Earth Observation Data (opens mapping.emergency.copernicus.eu)
  7. Nature Communications, 5 October 2021High resolution global spatiotemporal assessment of rooftop solar photovoltaics potential for renewable electricity generation (Joshi et al., Nature Communications 12, 5738) (opens doi.org)
  8. Joule, 19 December 2018DeepSolar: A Machine Learning Framework to Efficiently Construct a Solar Deployment Database in the United States (Yu et al., Joule 2, 2605 to 2617) (opens doi.org)
  9. European Commission, EUR-Lex, 18 May 2022EU Solar Energy Strategy, COM(2022) 221 final (opens eur-lex.europa.eu)

Contact

Which roofs do you want to assess, and what should the result decide?

Prefer to talk?

Tell one of our experts about your project. Together we work out where you stand, whether satellite data and AI can help and what a sensible next step would be.

Both calls are free of charge and take place online. You choose a date in our booking calendar.