Why offline aridity diagnostics overestimate future drying: the role of feedback-inflated evaporative demand

Offline aridity and drought indices have often implied widespread terrestrial drying under a warming environment, while Earth system models (ESMs) have projected modest changes in land-surface water fluxes. This persistent divergence has been typically attributed to missing vegetation physiological processes in offline frameworks. However, we here show that a more foundational cause is a structural inconsistency embedded in those diagnostics. Conventional potential evapotranspiration (PET) formulations can violate the assumption that precipitation (P) and atmospheric evaporative demand act as independent climatic constraints in the Budyko framework. Using open-water Penman and vegetation-responsive Penman–Monteith formulations forced by reanalysis data and ESM projections, we found that uncorrected PET strongly reflected land–atmosphere feedbacks, leading to pronounced negative P–PET correlations (-0.45 ± 0.29; mean ± s.d.). When PET was thermodynamically deflated, this dependence was largely removed (-0.02 ± 0.42), restoring consistency with the theoretical basis of Budyko-type diagnostics. This structural correction reduced inflation of the aridity index and substantially moderated projected evapotranspiration (ET) trends. Under a business-as-usual scenario, the trend of Budyko-based ET from uncorrected PET (+0.61 mm yr^-2) exceeded that of CMIP6 ensemble mean (+0.28 mm yr^-2) by more than a factor of two. CEP-deflated PET narrowed this discrepancy (+0.39 mm yr^-2), while additional physiological adjustments provided comparatively smaller improvements. We suggest that violations of structural assumptions, rather than missing physiological processes alone, can play a central role in the divergence between offline aridity diagnostics and ESM hydrological projections.

Acknowledgement: This work was jointly supported by the National Research Foundation of Korea (NRF) grants funded by the Korea government (MSIT) (RS-2025-16070291 & RS-2024-00416443).