Mapping Lebanon’s Rooftop Solar Revolution

Lebanon’s ongoing energy crisis has driven a rapid expansion of decentralized rooftop solar photovoltaic (PV) systems, creating an urgent need for up-to-date and spatially explicit information on solar deployment. However, comprehensive ground-based inventories remain unavailable, limiting the ability of policymakers and utilities to assess installed capacity and support energy planning.

This work presents the first large-scale map of rooftop solar installations in Lebanon, covering 5,850 km² (approximately 57% of Lebanon’s territory), selected to represent the country's major urban and peri-urban settlement patterns. While most rooftop PV detection studies rely on very high-resolution aerial or satellite imagery (typically 10–30 cm), this study utilizes Airbus Vision-1 imagery with an 87 cm spatial resolution acquired between June and July 2023. At this resolution, photovoltaic installations are represented by only a few pixels and exhibit limited textural information, significantly increasing the difficulty of automated detection. To address this challenge, we developed a custom Analysis Ready Data (ARD) processing pipeline incorporating geometric correction, radiometric calibration, and pansharpening. A U-Net++ segmentation model was fine-tuned using a locally annotated dataset containing 4,080 verified photovoltaic installations. To overcome domain shift and severe class imbalance, we introduced a smart patch-based training strategy, targeted data augmentation, and composite loss functions.

The proposed approach achieved an Intersection-over-Union (IoU) of 0.70 and an F1-score of 0.73 on validation data. Detected solar panel areas were subsequently converted into estimates of installed capacity and annual energy production. Across the mapped area, rooftop solar installations are estimated to generate an annual energy production of 778 GWh/year. Results reveal a highly uneven spatial distribution of solar deployment, with districts such as Minieh–Dannieh and Saida exhibiting the highest concentrations of installations and energy yield. Given the rapid pace of rooftop PV deployment in Lebanon, these estimates should be interpreted as a snapshot of conditions during the June–July 2023 acquisition period rather than as a current inventory.

The study demonstrates that accurate rooftop PV detection is achievable from sub-meter satellite imagery that is substantially coarser than the datasets commonly used in the literature. The resulting map provides a valuable evidence base for energy planning, infrastructure assessment, and policy development in data-scarce environments, while the proposed processing pipeline is readily adaptable for nationwide mapping in similar contexts.