Attribution of Evapotranspiration Variations in 364 Large Irrigation Districts across China: Quantifying the effects of irrigation via an extended Budyko Framework

Evapotranspiration (ET ) in irrigation districts is a key proxy of actual water consumption, reflecting the shifting agricultural water demand under the combined pressures of climate change and human activities. Accurate attribution of ET variations is essential for sustainable agricultural water management but remains challenging due to the complex interplay of hydrological drivers. The Budyko framework provides a physically-based approach to partition these influences. However, the traditional Budyko framework is not applicable the irrigation area.

In this study, we used an extended Fu’s formula within the Budyko framework to explicitly incorporate the impacts of irrigation activities on the water-energy balance. We defined "equivalent precipitation" (P_e) —the sum of irrigation water ( I ) , precipitation (P ) , and groundwater evaporation (𝐸𝑇_gw ) , as the water availability. This framework was applied to 364 large irrigation districts across China to quantify ET from 2010 to 2017 and evaluate the extended Budyko framework’s applicability at the irrigation district scale.Then, we calculate the elasticity coefficients for I、P、Pe、ω (the optimal values of Budyko parameter）、potential evaporation (ET_o) to quantify the sensitivity of ET to each forcing factor of ET change via a dimensionless elasticity-based attribution method.

Our results indicate that:

• (1) Framework Applicability: We found that the extended Budyko framework demonstrates good applicability in irrigation districts. Compared with the ET results from water balance calculations and the results from the MOD16A3 dataset, the relative error of annual evapotranspiration was less than 10% in 87.91% (320 samples) of the irrigated areas, and the root mean square error was less than 60 mm in 80.77% of the irrigated areas (294 sample points). The calculation error was smallest in the humid region.
• (2) Parameter ω’s distribution: The optimal values of Budyko parameter w in the extended Fu’s formula exhibit significant regional distribution characteristics, showing a pattern different from that of natural watersheds. The more humid the irrigated area, the larger the value of w in the fitted equation, reflecting the impact of human irrigation activities on hydrological conditions.
• (3) Attribution Analysis: In arid and semi-arid regions, evapotranspiration is jointly limited by water and energy, exhibiting the highest sensitivity to parameter ω, with elasticity coefficients of 3.073 and 1.879, respectively. This is followed by sensitivity to energy, with elasticity coefficients of 0.754 and 0.413, and then sensitivity to irrigation, with elasticity coefficients of 0.709 and 0.239, respectively. As the climate becomes wetter, the system transitions to an energy-limited state; therefore, ET becomes more sensitive to ET_o, while its sensitivity to ω and irrigation significantly decreases. In humid regions, the elasticity coefficient to irrigation is only 0.091, to potential evapotranspiration is 0.752, and to ω is 0.722.