Refining the Shuttleworth-Wallace Model with Particle Swarm Optimization and Genetic Algorithm for Evapotranspiration Simulation in the Ecotone of the Eastern Margin of the Tibetan Plateau

Abstract

As a global climate amplifier, the peripheral ecotone of Qinghai-Tibet Plateau plays a crucial role in shaping Northern Hemisphere’s atmospheric circulation. To address the fundamental conflict between static parameterization of the traditional Shuttleworth-Wallace (SW) model and the dynamic ecosystem with heterogeneous surfaces, this study proposed a global optimization hybrid modelling method by integrating Particle Swarm Optimization (PSO) and Genetic Algorithm (GA) for the evapotranspiration (ET) simulation. By implementing dynamic calibration of two critical biophysical parameters including minimum stomatal resistance and surface roughness, we developed the enhanced SW-PG model adapted to alpine gradient environments. The results demonstrated that the SW_PG model significantly outperformed the traditional SW model in terms of both accuracy and stability across diverse ecosystems, exhibiting superior spatiotemporal adaptability over complex surfaces. Regarding the monthly optimization, it demonstrated superior performance in decreasing RMSE and improving the coefficient of determination, thereby effectively capturing fine-grained temporal dynamics, particularly for cropland ecosystems: 30.7% reduction in Root Mean Square Error (RMSE), 67.1% decrease in BIAS, and 2.5% increase in R value. Grassland ecosystems also showed considerable optimization, with 24.7% RMSE reduction, 44.3% BIAS improvement, and 14.4% R-value enhancement. Seasonal optimization better balanced systematic under-/overestimation tendencies, revealing seasonal response to environmental stressors. Parameter sensitivity analysis demonstrates that monthly-optimized minimum stomatal resistance exerts the strongest influence on evapotranspiration in wetland (0.0616) and grassland (0.056) ecosystems, and surface roughness significantly regulates ET in grassland ecosystems (0.0033 m). Overall, these improvements underscore the SW_PG model’s superior performance to the unique environmental gradients of the eastern Qinghai-Tibet Plateau.

Keywords: Refined Shuttleworth-Wallace model, Particle swarm optimization, Genetic algorithm, ET, Ecotone, Eastern Margin of the Tibetan Plateau