Field-observed soil–regolith depth covaries with rainfall regime and landscape position along a semi-arid tropical mountain transect

Abstract Purpose Field-observed soil–regolith depth records the upper profile-to-rock transition and can reflect weathering, pedogenesis, preservation and erosion, but need not equal the thickness of the entire weathered-bedrock zone. Climatic and geomorphic associations are difficult to disentangle in semi-arid crystalline mountains. We examined how recorded depth covaries with rainfall regime and landscape position. Methods We analyzed 14 georeferenced profiles from an observational, non-probability mountain-to-pediment transect, including two purposively selected environmental end members. Field descriptions were combined with AESA rainfall, CHIRPS and IMERG rainfall-regime metrics, TerraClimate, Copernicus GLO-30, MERIT Hydro and SGB geology. An exploratory six-model comparison used AICc and cross-validation, supplemented by robust and spatial-confounding diagnostics. Results Within the selected profiles, depth was positively associated with AESA precipitation (Spearman ρ = 0.851) and negatively with CHIRPS PCI (ρ = −0.701), IMERG interannual rainfall variability (ρ = −0.711) and altitude (ρ = −0.657). Depth and coarse fragments shared a P7–P8 change point (permutation p ≈ 0.006 for each). Although three climate-plus-altitude models had ΔAICc < 1, these predictors represented correlated environmental gradients, not independent causal mechanisms. PCI was strongly related to transect position (ρ = −0.952); its partial depth association was approximately zero after controlling for position (partial ρ = −0.002, p = 0.995). Adding PCI to a position-plus-altitude sensitivity model increased R² by only 0.00014 ( p = 0.920). Conclusion The selected transect documents a coupled hydroclimatic–geomorphic association and a distinct P7–P8 shift in recorded profile depth and coarse-fragment abundance. Rainfall concentration is a spatially confounded co-gradient, not an independently identified driver of erosion or regolith thickness. Purposive sampling and the exploratory candidate set restrict generalization beyond the studied profiles; the production–preservation–removal framework remains a testable interpretation rather than a measured process balance.