Water withdrawals for agriculture represent the single largest consumptive use for many parts of the U.S., bearing a large anthropogenic footprint on the water and energy cycles. Our study analyzes the potential of remotely sensed soil moisture to inform models - through the process of data assimilation – in order to estimate irrigation. The study employs an observation simulation system experiment (OSSE) where a model simulation forced with prescribed irrigation– termed as the truth simulation - produces soil moisture outputs used in the place of observations in the data assimilation system. These synthetic observations are merged with the Variable Infiltration Capacity (VIC) model in a synthetic data assimilation experiment that employs a particle batch smoother (PBS) algorithm. The posterior weights from the PBS are used to calculate the expected incoming precipitation. Understanding that this expected value for precipitation also incorporates irrigation, the known precipitation from NLDAS-2 is subtracted to derive estimated irrigation. A single grid cell from the Nebraskan Silver Creek watershed was chosen to evaluate the ability of a PBS to estimate daily and seasonal irrigation. 

Experiments were conducted to determine the technique’s sensitivity to: return interval of observations, observational uncertainty, timing of irrigation, and model parameter uncertainty - sensitivities that will manifest in non-synthetic applications of the method.