Beyond Nighttime Lights: Multi-Sensor Satellite Fusion for Province-Constrained District-Level GDP Disaggregation in Data-Scarce Regions
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, cilt.19, ss.1-20, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 19
- Basım Tarihi: 2026
- Doi Numarası: 10.1109/jstars.2026.3711841
- Dergi Adı: IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
- Derginin Tarandığı İndeksler: Scopus, Materials Science & Engineering Collection (ProQuest), Technology Collection (ProQuest), Aerospace Database, Science Citation Index Expanded (SCI-EXPANDED), Compendex, Geobase, INSPEC, Directory of Open Access Journals
- Sayfa Sayıları: ss.1-20
- Dokuz Eylül Üniversitesi Adresli: Evet
Özet
Accurate subnational measures of economic activity are essential for spatial policy design, particularly in developing and middle-income countries where regional inequalities are pronounced. However, official economic statistics are often unavailable at fine administrative scales, leaving local economic dynamics insufficiently captured. This study develops a transparent and scalable framework for province-constrained small-area GDP disaggregation in data-scarce environments, using Türkiye as a case study. Rather than directly observing or supervising district-level GDP, which is not officially available in Türkiye, the framework uses the Socio-Economic Development Index (SEDI) as a supervisory benchmark to learn relative intra-provincial socioeconomic spatial structure. This learned structure is then combined with multi-source satellite and demographic indicators to allocate official provincial GDP totals to districts while preserving aggregate consistency. The framework integrates complementary remotely sensed indicators that capture different dimensions of production-related and socioeconomic activity: artificial illumination intensity from VIIRS Nighttime Lights (NTL), vegetation dynamics from MODIS NDVI, structural density from Sentinel-1 SAR backscatter, and atmospheric pollution from Sentinel-5P NO₂. These indicators are complemented by population size and surface area to account for demographic scale and spatial heterogeneity. A machine learning framework with province-grouped cross-validation is employed to compare alternative models, with Random Forest providing the most stable performance under spatially conservative validation. The resulting model-derived signals are transformed into province-specific allocation weights and incorporated into a mass-preserving dasymetric disaggregation procedure. Three main findings emerge. First, light-only indicators are insufficient for capturing fine-scale socioeconomic variation, particularly in heterogeneous or rural regions. Second, integrating multiple satellite-derived and demographic indicators improves the representation of intra provincial spatial heterogeneity relative to single-source approaches. Third, the resulting aggregate-consistent district-level estimates show meaningful alignment with external socioeconomic benchmarks, while these relationships are interpreted as convergent statistical associations rather than direct causal or fully independent validation of GDP accuracy. Overall, the study contributes a transparent and transferable framework for satellite-informed, SEDI-supervised, province-constrained GDP disaggregation, offering a practical approach for countries where official economic statistics remain spatially limited.