Map scale effects of soil databases on modeling organic carbon dynamics for paddy soils of China

Paddy soils support important croplands in many parts of the world, especially in Asia, so an accurate estimation of the magnitude and dynamics of paddy soil organic carbon (SOC) stocks is critical to developing strategies for greenhouse gas mitigation. Soil databases of different map scales have been widely used in process-based models, but exactly how map scale affects the simulation of SOC dynamics on a national basis remains unclear. To improve the understanding of map scale associated uncertainty, the DeNitrification–DeComposition (DNDC) model were parameterized with three soil polygon-based databases at scales of 1:1,000,000, 1:4,000,000, and 1:14,000,000 to simulate SOC stocks (0–30 cm) for paddy soils of China and their changes from 1980 to 2008. Based on the relatively detailed 1:1,000,000 scale soil database, SOC stock for paddy soils of China ranged from 2.51 Pg C in 1980 to 2.65 Pg C in 2008. The estimates of SOC stock decreased as map scale decreased from 1:4,000,000 to 1:14,000,000, and differed by 27% and 40% from the 1:1,000,000 scale database. Map scale influences on simulation outcome are related to the total paddy soil area and initial SOC contents, of which the latter may account for 73.2%–79.5% of the variability of SOC density for paddy soils of China. The multi-scale simulation results indicate that the application of 1:1,000,000 scale soil polygon-based database could reduce the inaccuracy of national SOC stock estimates and contribute to improved assessment of continental and global carbon balance.

► The DNDC model was used for SOC dynamics estimation in China's paddy soils. ► 1:1,000,000, 1:4,000,000, and 1:14,000,000 scale soil polygon-based databases were built. ► We explored map scale effects of soil databases on modeling SOC dynamics. ► SOC stock estimates decreased as map scale decreased from 1:4,000,000 to 1:14,000,000. ► The 1:1,000,000 scale soil database reduced uncertainties in SOC simulation.