Time-Window-Dependent Nonlinear Decline and Asymptotic Convergence of Temporal Stability of Soil Carbon, Nitrogen, Phosphorus, and pH in Tibetan Grasslands

Observation-window length can shape estimates of ecosystem temporal stability, yet its influence on belowground soil properties remains less explored. Using annual gridded soil datasets reconstructed from random forest models for alpine grasslands of the Qinghai–Xizang Plateau (2000–2022), we quantified the temporal stability of soil organic carbon, total nitrogen, total phosphorus, their stoichiometric ratios (C:N, C:P, and N:P), and pH across 3–23-year sliding windows. Analyses covered a climate-only scenario, a combined climate–NDVI scenario, and their ratio-based contrast. Estimated stability declined nonlinearly with increasing window length, with progressively smaller changes at longer windows. Exponential models described these relationships well (R2 ≥ 0.96), although decay rates and fitted asymptotic values differed among soil variables and scenario representations. Time-window length and climatic temporal stability together accounted for at least 87% of the variation in soil stability estimates within the common window-based statistical framework. The strong sensitivity of short-window estimates and gradual asymptotic behavior at longer windows highlight how the mean-to-standard-deviation metric captures temporal variation across scales. These results underline the importance of explicitly considering observation-window length when comparing soil stability estimates and provide a basis for examining the scale dependence of soil properties under changing environmental conditions.