Scale-aware emergent constraints on Tibetan Plateau warming: contrasting global and regional predictors

Emergent constraints are increasingly used to reduce uncertainty in regional climate projections by exploiting relationships across models between observable historical predictors and projected changes. However, the choice of historical window and predictor scale can strongly affect constraint performance. Using the Tibetan Plateau (TP), a region of complex terrain, as a testbed, we compare two widely used predictors for constraining future TP warming: large-scale global mean surface air temperature (GSAT) and regional Tibetan Plateau surface air temperature (TPSAT). We show that GSAT-based constraints are sensitive to the temporal evolution of external forcing and are strongest for windows beginning around 1980, reflecting the differing historical effects of greenhouse gases and aerosols. In contrast, TPSATbased constraints generally become stronger with longer windows, indicating that extended records help suppress internal variability and better reveal the regional forced signal. At the local scale, GSAT performs better in lower-elevation regions, whereas TPSAT is more skillful over high-elevation and topographically complex areas where local feedbacks play a larger role. Motivated by this spatial contrast, we develop a scale-aware constraint framework that selects the locally preferred predictor at each grid point and outperforms either single-predictor approach. These findings highlight the importance of scale-aware observational constraints and may inform constrained projections in other high-elevation and topographically complex regions.