A SIF-Constrained Noah-MP land surface model for simulating water use efficiency and drought Resilience: Application in a Semi-Arid Basin

Land surface models (LSMs) systematically overestimate evapotranspiration (ET) in water-limited ecosystems due to inadequate representation of plant physiological water stress, undermining reliable water resource projections. We introduce a novel, mechanistic integration of solar-induced fluorescence (SIF) directly into the Ball-Berry stomatal resistance scheme of WRF-Hydro/Noah-MP. In the absence of eddy covariance flux tower measurements within the basin, model evaluation is conducted through inter-comparison with multiple independent gridded reference products (MODIS, GLEAM, GLDAS), and results are interpreted as physically consistent shifts in model behaviour rather than validated improvements in absolute accuracy. The methodology is designed to be transferable, relying only on publicly available SIF products and the widely used Noah–MP model. We developed a self-calibrating SIF factor (0–1) from GOSIF v2 to modulate the net photosynthesis rate, thereby coupling observed photosynthetic activity to stomatal resistance and transpiration. This SIF-enhanced model was applied to the Awash Basin, Ethiopia (2000–2024), and evaluated through inter-comparison with MODIS, GLEAM, and GLDAS. SIF integration increases consistency with the multi–product ensemble: mean annual ET decreased by 36.1 % (from 8.45 to 5.40 mm day⁻ 1 , p < 0.001), thereby reducing the ET discrepancy relative to the ensemble of satellite–based reference products (with α = 0.10 calibrated for this basin). The transpiration fraction of ET increased most during the critical dry season (+7.0 %, p < 0.001), reflecting more realistic water-conservation behaviour. Consequently, Water Use Efficiency (WUE) increased under both normal and drought conditions, and the SIF-constrained Model showed substantially smaller WUE losses during drought than the Control (a 5 % decline versus a 13 % decline, respectively), consistent with a more water-conservative stomatal strategy that reduces transpiration by a greater proportion than photosynthesis under stress. This simulated response aligns with observed plant behaviour in water-limited ecosystems, yielding a drought resilience index of 0.95, compared with 0.87 for the Control. Monthly flux correlations confirmed strong carbon–water coupling (GPP-ET r = 0.93), while annual-scale analysis revealed a developing trade-off in which high-productivity years were least efficient (WUE-ET r = -0.88). Furthermore, unsupervised clustering of SIF-constrained WUE and resilience outputs yielded the first physiologically based agroecological zonation of the basin. We identified four distinct functional zones, including a productive but drought-vulnerable core (21,425 km 2 , WUE = 2.49 g C kg⁻ 1 H 2 O) and resilient but inefficient lowlands (8,175 km 2 , WUE = 0.05 g C kg⁻ 1 H 2 O). This scalable SIF-informed physiological constraint framework reduces a structural LSM bias, yielding more credible ecohydrological projections and providing dynamic, plant-informed targets for climate-resilient land and water planning.