Background Soil erodibility(named K factor) is one of the key parameters of the soil erosion equation. The K is the basic data for soil erosion monitoring, reflecting the difficulty of soil dispersion and transportation under the action of rainfall erosion, and its size is related to the characteristics of the soil. However, the existing national K map was made based on soil species data, and the survey was conducted nearly 40 years again. Besides, the soil-polygon linked method was used to produce the legacy map, which cannot reflect the K variability existing in the same soil polygon.

Methods This article updated the national K-value map based on the soil series survey(completed from 2008 to 2018) and the random forest regression model. Firstly, the soil texture and organic matter content of 4 327 sample points were collected, and the K was calculated using the nomograph equation; when the soil organic matter content was >12%(mass fraction), the corrected EPIC formula was used to calculate the K. Secondly, the random forest regression model was used to train the K of the sample points, and a variety of environmental factors were used as prediction variables, including climate, surface temperature, vegetation index, terrain and parent rock type, and then remote sensing images were used to carry out spatial mapping.

Results The cubic spline function combined with natural logarithm interpolated the content of very-fine sand content(≥0.050-0.100 mm) with high R² and reasonable value. An exponential equation was built between the Nomo-K and EPIC-K values(R²=0.807 1). The updated map showed that the range of national K values was 0.005 1-0.074 5 t·hm²·h/(MJ·mm·hm²), with an average value of 0.029 8 t·hm²·h/(MJ·mm·hm²). The map of soil erodibility in China showed the macro rule that the K of the Loess Plateau and North China Plain was the largest, that of the southern and northeastern regions was the middle, and that of the Qinghai Tibet Plateau was the lowest. The spatial difference of the K factor was related to the distribution of main soil types in China. Besides, the divisions of soil and water conservation regions had various K values due to the comprehensive effect of natural factors and human activities. However, the updated K values still need to be corrected based on the measured data of runoff plots when applied in practice.

Conclusions Our study improved the calculation method of the K, which is more accurate and objective. The updated K map reflects the recent soil condition and expresses the spatial variation of K in more detail with a 30 m resolution grid on a nationwide scale. This study will provide methods and data support for soil erosion investigation and monitoring.