Assessing Climate-Induced Variability in Evapotranspiration Dynamics in Arid Zone of Rajasthan, India

Abstract

Climate change is increasingly reshaping hydrological processes in arid environments, where evapotranspiration (ET) plays a critical role in regulating water availability and agricultural sustainability. In the arid zone of western Rajasthan, recent warming trends and rainfall variability are intensifying atmospheric water demand, necessitating a refined understanding of ET dynamics under changing climatic conditions. This study updates earlier assessments by integrating long-term observations (1971–2025) with both physically based and data-driven approaches, combining the Penman–Monteith method with machine learning models, including Random Forest and Long Short-Term Memory networks, to capture non-linear interactions and temporal variability. Sensitivity analyses are further strengthened using emerging CMIP7 climate projections under contrasting emission pathways. The results indicate a clear intensification of ET, with baseline values increasing by about 15% over the historical period and projections suggesting a further rise of ~22–30% by mid-century under high-emission scenarios. Temperature emerges as the dominant driver (explaining over half of the variability), while solar radiation and humidity exert secondary but significant controls, particularly influencing seasonal responses. Machine learning outputs reveal amplified summer ET and strong non-linear behavior that is not captured in conventional approaches. Although limited reductions in ET may occur under specific humidity–radiation feedbacks, the overall trajectory points toward increasing evapotranspiration and heightened drought risk across the region. These findings underscore growing pressure on already fragile water resources in arid Rajasthan and highlight the need for adaptive strategies such as precision irrigation and climate-resilient agriculture. Notably, this study provides a novel contribution by integrating machine learning–based ET modeling with next-generation CMIP7 projections, enabling a more realistic assessment of non-linear climate–hydrology interactions under future scenarios.