How to map soil organic carbon stocks in highly urbanized regions?

•Including urban areas into the regional SOC assessment increases total and specific stock on 10-15%.•Urban soils are responsible for the major part of SOC spatial variability in the region.•Ignoring urban-specific factors result in considerable overestimation of regional carbon stocks.•“Urban-specific” model has shown the best performance and explained 30% of total SOC variability.

Urbanization is among the most impetuous current land-use change trends, resulting in a permanently increasing role of urban ecosystems in regional and global environments. Urban soil organic carbon (SOC) is probably the least understood stocks because of the lack of appropriate methodology to analyze and map it. Cities represent a small-scale patchwork of very contrasting soil features. This creates high short-term spatial variability. Urban-specific factors including size and age of the city, soil sealing and cut-off profiles dominate the anthropogenic soil forming factors. Considering these specific urban environments, our study aimed to adapt the digital soil mapping (DSM) approach to map topsoil and subsoil SOC stocks in a highly urbanized region. Field SOC data collected for different environmental conditions in the Moscow region (five soil types and five land-use types of 244 mixed samples for topsoil and subsoil) were linked to available auxiliary data, including both traditional (relief, climate, vegetation etc.) and urban-specific (functional zoning, size and history of the settlements) factors. Separate general linear models (GLM) were developed for the three different cases: i) excluding urban areas from the analysis (non-urban model); ii) including urban areas but only considering traditional soil forming factors (conventional model); and iii) including urban factors (urban-specific model). Total and specific carbon stocks, spatial variability represented by coefficient of variance (CV %) and the determination coefficient with a validation dataset were compared for the three models. The conventional model dramatically overestimated carbon stocks and underestimated of SOC's spatial variability. Total and specific carbon stocks estimated by non-urban model were 10-15% less than ones given by urban-specific model. The urban-specific performed best and explained more than 30% of total variability. Urban areas showed the highest spatial variability and specific carbon stocks, 90% of which was stored in subsoils. Even when the high uncertainty of the absolute values is considered, urban areas contributed to regional carbon stocks. Considering urban-specific factors to estimate carbon stocks and their spatial variability is thus necessary.