From National to Sub-National: Estimating a Proxy Nature Relationship Index from Satellite Imagery and Local Statistics for Regional Policy

The Nature Relationship Index (NRI) evaluates human–nature relationships through three dimensions (nature thriving, used with care, and safeguarded), but it exists only at the national level, whereas many policies are implemented regionally. Regional statistics describe resource use and protection, but harmonized information on ecological condition is rarely available for regions. This study examines whether satellite imagery can fill this gap. Transformer-based models were trained to estimate the Thriving (ecological condition) dimension from Sentinel-2 and Landsat-8 imagery of 136 countries, with labels constructed from forest cover and two satellite-derived land-condition indicators, while Care and Safeguarded were computed directly from emission, energy, and protected-area statistics. In repeated country-level cross-validation, imagery was essential for Thriving: a statistics-only model had no predictive skill (R2 = −0.04), whereas the image model reached R2 = 0.36 (p < 0.001), and transformer backbones performed similarly, with no gain from ensembling. The image-based Thriving estimates agreed with regional reference values across 42 Japanese prefectures (Spearman ρ = 0.63) and nine Korean provinces (ρ = 0.85) but not for small metropolitan cities. Combining image-based Thriving with statistics-based Care and Safeguarded yields a sub-national NRI whose profiles show whether a region’s standing is driven by ecological condition, resource use, or protection.