Water use is usually metered in municipal and industrial sectors but not in agriculture. The survey data of irrigation water use are highly variable in quality, depending on reporting requirements at the level of the state, the watershed, or the irrigation project. In hydrologic-agronomic models, irrigation is either treated as an estimated fixed input or determined by the model through certain prescribed empirical criteria. This paper presents an approach for estimating irrigation water use based on a hydro-agronomic model (Soil Water Atmosphere Plant model-SWAP) and crop evapotranspiration (ET) assessed from remote sensing utilizing the data surface energy balance algorithm for land. A coupled forward-inverse procedure is implemented for the analysis. The forward procedure is to assimilate the ET estimation into the hydro-agronomic model through the ensemble Kalman filter, which is an efficient data assimilation method for complex nonlinear dynamic models. The inverse procedure is to search irrigation scheduling using a genetic algorithm (GA). The two procedures are tightly coupled through the objective function, which is based on the likelihood function assessed from the data assimilation framework by maximizing the joint distribution of all "observed" crop ET. The forward-inverse framework incorporates model errors and observation errors, which allows the assessment of the inaccuracy of the irrigation estimate. The soil hydraulic properties of the case study area, which are the most sensitive parameters of the SWAP model, are estimated by genetic algorithm based on the available soil survey data and the in-site soil moisture measurement in a grass land.