Integrating terrestrial and marine remote sensing indicators to estimate subnational economic output

Subnational economic data remain scarce across much of the Global South, particularly in fragmented island and coastal geographies, thereby constraining the capacity of researchers and policymakers to monitor local economic conditions. This study proposes and empirically evaluates a four-proxy remote sensing framework for subnational gross domestic product (GDP) estimation. The framework extends the established Night-time Lights (NTL)–Normalised Difference Vegetation Index (NDVI)–Forest Loss (FL) approach by incorporating sea surface Chlorophyll-a (CHLA) as a marine fisheries proxy. This achieves full proxy coverage across the non-agriculture, forestry, and fisheries sector through NTL, agriculture through NDVI, forestry through FL, and marine fisheries through CHLA. The empirical analysis employs Multiscale Geographically Weighted Regression (MGWR) and pooled ordinary least squares (OLS) across 102 regencies and cities in Sulawesi and Maluku, eastern Indonesia, over 2015–2024, with a province-level panel extending to 2002–2024. The stepwise MGWR R2 rises from 0.468 (NTL only) to 0.543 (NTL, NDVI, and FL) to 0.561 (full model), with a pooled Pearson correlation of 0.760. CHLA is positive and significant at both spatial scales after conditioning on all terrestrial proxies, confirmed by OLS coefficients of 0.029 at the regency/city level and 0.148 at the province level, significant at the 0.1% and 1% levels respectively. All proxy relationships replicate in direction and significance across scales and over the 23-year panel. The framework provides a freely replicable tool for subnational economic monitoring in data-scarce coastal and island settings, with direct applicability to marine-dependent primary sector economies globally.