Quantifying net irrigation across the North China Plain through dual modelling of evapotranspiration

Knowledge of irrigation water use is crucial for ensuring food and water security in water scarce regions. Even though irrigation is one of the most important direct human interferences with the terrestrial water cycle, there exists limited knowledge on the extent of irrigated areas and in particular the amount of water applied for irrigation. In this study, we develop a novel approach that estimates net water loss due to irrigation and apply it over the North China Plain domain, which is a global hotspot for severe groundwater depletion caused by extensive irrigation practices. Our goal is to retrieve spatio-temporal patterns of net irrigation amounts, constituted as evaporative loss of irrigated water, at monthly timescale at 1km² spatial resolution. The analysis is based on a direct comparison of two alternative evapotranspiration (ET) models: (1) A remote sensing based model (PT-JPL-thermal) using various MODIS products as input and (2) a one-dimensional, free drainage hydrological model (mHM). The hydrological model is purely driven by rainfall and will therefore naturally show a strong disagreement with the remote sensing based ET during periods of extensive irrigation. We use this systematic residual term that reflects a non-precipitation-based water source, as quantification of net irrigation. The hydrological model is calibrated against the remote sensing based ET at grids that are not affected by irrigation and discharge records representing natural flow. Total water storage anomalies retrieved by GRACE are utilized to evaluate the derived net irrigation amounts over the North China Plain. We find, that irrigation peaks in May, which corresponds to the peak of the growing season of winter wheat. Moreover total irrigation amounts to 116 mm per year (14km³), which is in good agreement with previous studies. The net irrigation estimates are at an unprecedented spatial and temporal resolution and are extremely valuable input for water resources management as well as for subsequent groundwater modelling where net irrigation can be utilized as pumping boundary condition.