?
Сеточные данные о плотности населения в России: валидация и коррекция глобальной модели GHS-POP
This paper presents the validation and correction of the global 100-m resolution GHS-POP population density dataset for Russia. Adapting global population density models to national contexts and improving their accuracy addresses the critical lack of high-resolution population distribution data required across geography and related disciplines. Modern population density models (e.g., LandScan, GHS-POP, WorldPop) employ dasymetric mapping principles, disaggregating statistical population data into regular grid cells using satellite-derived proxies. However, these models may yield divergent estimates due to conceptual differences in defining "population", auxiliary datasets, and algorithms. The GHS-POP model estimates resident population density based on built-up area volumes derived from Sentinel-2 imagery. We developed a two-stage data correction algorithm and applied it to three test sites with distinct urban morphologies: cities of Krasnodar, Saratov, and Naberezhnye Chelny, followed by validation of the resulting dataset. The correction 1) eliminated systematic discrepancies with the 2020(21) All-Russian Population Census results at municipal level and 2) resolved population misallocation artifacts in industrial zones using OpenStreetMap data on landuse. Validation against national housing registry data demonstrated the accuracy of the corrected grid of 47–56% at 100-m resolution and 75–83% at 1-km resolution. The scalable correction algorithm was applied to all of Russia, with corrected datasets openly available via the geoportal of the HSE University Faculty of Geography and Geoinformation Technology.