【Objective】Soil maps are essential for land resource management, agricultural planning, and ecological conservation. Their accuracy directly affects applications such as arable land quality assessment, crop suitability zoning, and soil carbon stock estimation. However, limited by the high cost and low frequency of traditional surveys, historical soil maps often suffer from outdated classifications, incomplete taxonomy, and low precision, making them inadequate for current monitoring needs. This study explores how to extract effective soil–environment relationships from historical maps and combine them with new sample data to support efficient updates. 【Method】Using the soil map from the second national soil survey of Jingshan City (Hubei Province) as a baseline, this study integrates multi-source data—including land use, parent material, remote sensing, and topography—and applies pedogenic theory to guide a hierarchical soil classification updating strategy. Key classification indicators were extracted via machine learning algorithms to iteratively update soil types from soil order to soil species. 【Result】(1) The proposed strategy effectively updated the historical soil map. Validation using 30 field samples showed classification accuracies of 90% for soil order and suborder, and 83% and 73% for soil group and species. (2) Classification structure was adjusted: soil orders remained at five; suborders decreased from 13 to 12; soil groups from 43 to 38 (14 removed, 9 added); and soil species increased from 283 to 323. (3) Spatially, about 24% of the area underwent soil order changes, with fluvisols increasing by 172% and paddy soils decreasing by 32%. Suborder and group-level changes covered 32% and 36% of the area, respectively. Most soil species categories were revised due to naming logic reconstruction. 【Conclusion】This study verified the effectiveness of the classification hierarchical update framework based on soil formation theory and machine learning in improving the accuracy of historical soil maps, improving the classification system, and revealing the evolution of spatial patterns. It also provided feasibility and reference for the efficient updating of historical soil maps.