Accurate evapotranspiration (ET) measurement over large scales with remote sensing can contribute to effective water resource management. However, ET measurement is challenging, particularly in complex agricultural landscapes. ET calculated with the water balance method using data from the Gravity Recovery and Climate Experiment (GRACE)/GRACE Follow-On (GRACE-FO) missions has been found to exceed estimates from land surface models, although there has been no definitive explanation. Here, we show using high resolution ET from ECOSTRESS, that this bias is attributed to fine-scale irrigation not captured by models but which mass conservation detects. The percentage of irrigated area scales linearly with the bias between observed and modeled ET (R2 = 0.80, p < 0.01). We estimate that irrigation accounts for 35% of ET during the irrigated summer months in arid/semi-arid basins in North America. This study presents the synergies of using GRACE-FO and ECOSTRESS to improve remote sensing of agricultural water use.