Abstract:Soil type mapping reveals the geographical distribution and characteristics of soils and provides a scientific basis for the use, protection and management of soil resources. The digital soil mapping method based on the soil-environment relationship is an important means for quickly acquiring high-precision and high-resolution soil spatial distribution information. However, its applicability and mapping accuracy for different topographical units require further investigation. Taking Pinggu District of Beijing, China as the research area, and dividing it into two terrain units: mountainous and hilly regions and plain regions. The random forest algorithm was used to establish a model linking soil types to environmental variables, upon which digital mapping of soil types was conducted for both terrain units, with mapping accuracy being compared subsequently. The results showed that the overall accuracy (OA) of digital mapping for soil group, subgroup, soil genus and soil species in mountainous and hilly regions was 100%, 93.1%, 89.7% and 75.9%, respectively, whereas the OA of soil group, subgroup, soil genus and soil species in plain regions was 73.7%, 55.3%, 52.6% and 23.7%, respectively. These results indicated that digital mapping of soil types using environmental variables performed well in mountainous and hilly regions, but its efficacy noticeably diminished in plain regions. As the classification units of soil types became more granular, transitioning from soil groups to soil species, the accuracy of soil type mapping supported by environmental variables was gradually decreased. It was recommended to make full use of readily available environmental variable data to enhance the accuracy of soil type mapping in mountainous and hilly regions. In plain regions, it was necessary to appropriately increase the number of soil type profiles to improve mapping accuracy. The research result can provide practical examples and technical support for digital soil type mapping in other regions.