GeoNatureAgent Benchmark - Dataset (tasks, results, derived environmental indicators)

GeoNatureAgent Benchmark — Dataset (v1.2). Complete benchmark data: 93 tasks; per-case scored results for the seven evaluated LLMs; compiled leaderboard / per-category / per-case tables; the two indicator COGs (CO2 suitability, gully erosion); Spanish administrative boundaries (provinces, autonomous communities, municipalities) and Portuguese districts; BigEarthNet V2 land-cover statistics; and the full raw agent conversation traces. Reproduces the leaderboard in the accompanying paper. Third-party source datasets (BigEarthNet V2, LUCAS 2022, MITECO, Sentinel-2) are cited, not redistributed. Code is in the companion software record.