A 1 km soil organic carbon density dataset with depth of 20 and 100 cm from 1985 to 2020 in China

Abstract

Soil organic carbon (SOC) is an important component of the global carbon cycle and a vital indicator of ecosystem health, playing key roles in agricultural productivity and climate change mitigation. To trace the spatiotemporal dynamics of SOC in China, a high-resolution (1 km) Soil Organic Carbon Density (SOCD) dataset for the 0–20 and 0–100 cm depths spanning the period from 1985 to 2020 is produced in this study. By integrating Landsat archives, topographic and meteorological data, and 11 743 soil profile measurements, we produced the SOCD dataset from 1985–2020 in China using the Random Forest ensemble learning approach. Specially, a climate zoning strategy was developed to account for the significant environmental heterogeneity across China. The validation of our SOCD estimated results with 0–20 cm depth with independent testing samples showed strong agreement with R2=0.63 and RMSE=2.03 (kgCm-2) for 0–20 cm SOCD estimation and R2=0.62 and RMSE=6.16 (kgCm-2) for 0–100 cm. Moreover, our SOCD estimated results with 0–20 cm depth are aligned well with independent samples (R2=0.76, RMSE=1.75 kgCm-2) and Xu's dataset (R2=0.68, RMSE=1.70 kgCm-2). Furthermore, the validation of our SOCD estimated results with 0–100 cm depth with independent measurements from Dong et al. (2024a) showed strong agreement (R2=0.50, RMSE=4.93 kgCm-2). Furthermore, our SOCD product exhibits high consistency with existing global datasets (HWSD, SoilGrids250 m, and GSOCmap), showing the best fit with SoilGrids250 m (R2=0.74, RMSE=1.03 kgCm-2). Comparisons of model predictions to independent datasets from the 1980s, 2000s, and 2010s in China reveal substantial connections and demonstrate strong performance over time.