Spatiotemporal dynamics of solar photovoltaics development and ecological implications in Japan from cropland to forest

This study presents a nationwide assessment of solar photovoltaic (PV) mapping from 2008 to 2024 and long-term land-cover changes driven by PV installations in Japan from 2012 to 2024. A novel framework combining pixel-based spectral classification with time-series change detection was developed using the Google Earth Engine (GEE) platform. A Random Forest model trained on Sentinel-2 imagery and environmental features was used to generate a high-resolution (10 m) PV map for 2024, while installation years were estimated using the Continuous Change Detection and Classification (CCDC) algorithm applied to Landsat-derived spectral indices. Validation against governmental FIT records and OpenStreetMap data demonstrated high mapping accuracy and temporal agreement, with challenges noted in detecting installations smaller than 50 kW. The results reveal spatial shifts in PV development from cropland to forested areas, influenced by land availability, regulatory changes, and socio-economic factors. Regional differences in PV expansion were further examined through case studies in national parks, which represent Japan’s most ecologically significant landscapes. Analyses showed that most PV installations within national parks are concentrated in “Ordinary Areas,” the least regulated zones, with ecological impacts potentially extending beyond legal boundaries into adjacent areas. These findings contribute critical insights for harmonizing renewable energy expansion with land management and biodiversity conservation policies.