Soil organic carbon response deficits reveal constrained vegetation-soil carbon coupling in Tibet.

Vegetation productivity enhancement is often treated as evidence of ecosystem recovery, but its conversion into soil organic carbon (SOC) gain remains uncertain in alpine systems. We developed an expected-response framework for Tibet using the 2001-2020 NPP trend and the change in modelled SOC density between the 2000-2005 and 2015-2020 endpoints. Expected SOC response was estimated among pixels matched by ecological zone, productivity-enhancement decile and long-term grazing-intensity class. Under this grazing-adjusted median benchmark, 44.74% of valid 0-20 cm pixels had a positive SOC response deficit (CSD-SOC > 0). Monte Carlo propagation of NPP, SOC-endpoint and benchmark uncertainty showed that 2.17% of pixels had P(CSD-SOC > 0) ≥ 0.95 and 1.55% retained a positive 95% Monte Carlo lower bound. Five-fold spatial block cross-validation gave R2 = 0.232 ± 0.019; removing all LHGI variables reduced R2 to 0.199 ± 0.018. Elevation and ecological zone remained the leading predictors, while LHGI class, mean LHGI and LHGI trend added explanatory value. Results were stable across 5, 15 and 20 bins and a continuous matching model. The weaker 0-100 cm signal coincided with higher SOC-product uncertainty. We use the resulting maps to identify locations for field validation and process investigation rather than direct management prescriptions.