Mapping global human presence for nature conservation using geotagged social media data

Abstract

Quantifying human presence in natural areas is crucial for understanding anthropogenic pressures on biodiversity and informing conservation strategies, yet monitoring at global scales remains challenging due to limited data and spatial sampling bias. This study presents a data-driven approach to mapping global human presence at a 0.01-degree resolution using geotagged social media data. We developed a human presence indicator (HPI) that categorizes locations into four intensity levels: no presence, occasional presence, frequent presence, and sustained presence. Using over 195 million geotagged microblogs from China and 76 covariate layers representing natural and social factors, we trained a random forest model to predict human presence worldwide. The model's effectiveness was validated through comprehensive cross-validation with external datasets, including manually labeled global samples, data from X (formerly Twitter), and global human settlement and population distributions. The inferred HPI map showed detectable human presence through social media covering at least 13.41 % of Earth's terrestrial surface, with substantial regional variations across continents and biodiversity hotspots. Analysis of 1995 strictly protected areas showed that while 67 % had minimal human presence (<1 % of their area), 163 protected areas exhibited human presence in over 10 % of their domain, indicating potential conservation challenges. Despite limitations in data quality and sampling rates, this dataset provides valuable estimates of global human presence, particularly for remote or poorly monitored protected areas. The trained model and dataset, which we make freely available, can support consistent cross-regional comparisons and evidence-based conservation planning globally.