International audience ; Aim: Although soil biodiversity is extremely rich and spatially variable, both in terms of species and trophic groups, we still know little about its main drivers. Here, we contrast four long-standing hypotheses to explain the spatial variation of soil multitrophic diversity: energy, physiological tolerance, habitat heterogeneity and resource heterogeneity.Location: French Alps.Methods: We built on a large-scale observatory across the French Alps (Orchamp) made of seventeen elevational gradients (-90 plots) ranging from low to very high altitude (280-3,160 m), and encompassing large variations in climate, vegetation and pedological conditions. Biodiversity measurements of 36 soil trophic groups were obtained through environmental DNA metabarcoding. Using a machine learning approach, we assessed (1) the relative importance of predictors linked to different ecological hypotheses in explaining overall multi-trophic soil biodiversity and (2) the consistency of the response curves across trophic groups.Results: We showed that predictors associated with the four hypotheses had a statistically significant influence on soil multi-trophic diversity, with the strongest support for the energy and physiological tolerance hypotheses. Physiological tolerance explained spatial variation in soil diversity consistently across trophic groups, and was an especially strong predictor for bacteria, protists and microfauna. The effect of energy was more group-specific, with energy input through soil organic matter strongly affecting groups related to the detritus channel. Habitat and resource heterogeneity had overall weaker and more specific impacts on biodiversity with habitat heterogeneity affecting mostly autotrophs, and resource heterogeneity affecting bacterivores, phytophagous insects, enchytraeids and saprotrophic fungi.Main Conclusions: Despite the variability of responses to the environmental drivers found across soil trophic groups, major commonalities on the ecological processes structuring soil ...