Soil maintenance depends on erosion not exceeding soil replacement, yet tolerable soil loss rates (defined as the maximum erosion rate possible before soil profiles begin to thin over time) remain poorly constrained and are often represented by broad global defaults that do not adequately reflect environmental variability. This study updates the global evidence base for soil production from the weathering of parent material and explores how these rates can be used to derive more spatially explicit estimates of tolerable soil loss. We compiled a global database of 528 soil production rate (SPR) observations with soil depth information from 29 publications covering approximately the last 30 years. Across the main dataset, SPRs span approximately six orders of magnitude, ranging from 6 x 10-5 to 2.47 mm yr⁻¹, with a mean of 0.0869 mm yr⁻¹ and a median of 0.033 mm yr⁻¹. Strong differences were evident among environmental settings, with especially low rates in dry climates, and the highest rates associated with steep slopes and metamorphic lithologies.
To move beyond global averages, we used an ensemble of 10,000 regression trees to identify robust combinations of environmental controls on SPRs, resulting in the delineation of seven globally mappable soil production zones. The selected tree structure indicates that climate, lithology, slope, and sampling depth (as a proxy for overlying soil profile thickness) interact to produce distinct regional rates of soil production. Our results are mapped at full global scale at 1 km resolution to allow for comparison with global model estimates of soil erosion rates. Mapping these zones for upland-hillslope soils (where explicit soil profile thickness estimates are available and where soils are most vulnerable to loss through erosion) shows that SPRs are lowest across hot arid regions, including much of Australia, central Asia, southern Africa, and the southern USA–Mexico border region, while higher SPRs occur widely in tropical regions and in some higher latitude regions with metamorphic lithology. These results illustrate that tolerable soil loss thresholds should be spatially differentiated, and that commonly cited universal values (e.g. 2 t ha-1 yr-1) likely overestimate tolerable soil loss rates in many landscapes.