Process-based characterization of evapotranspiration sources over the North American monsoon region

Evapotranspiration (ET) is a poorly constrained flux in the North American monsoon (NAM) region, leading to potential errors in land-atmosphere feedbacks. We quantified the spatiotemporal variations of ET using the Variable Infiltration Capacity (VIC) model, modified to account for soil evaporation (E_soil), irrigated agriculture, and the variability of land surface properties derived from the Moderate Resolution Imaging Spectroradiometer during 2000-2012. Simulated ET patterns were compared to field observations at 59 eddy covariance towers, water balance estimates in nine basins, and six available gridded ET products. The modified VIC model performed well at eddy covariance towers representing the natural and agricultural land covers in the region. Simulations revealed that major sources of ET were forested mountain areas during the summer season and irrigated croplands at peak times of growth in the winter and summer, accounting for 22% and 9% of the annual ET, respectively. Over the NAM region, E_soil was the largest component (60%) of annual ET, followed by plant transpiration (T, 32%) and evaporation of canopy interception (8%). E_soil and T displayed different relationships with P in natural land covers, with E_soil tending to peak earlier than T by up to 1 month, while only a weak correlation between ET and P was found in irrigated croplands. Based on the model performance, the VIC-based estimates are the most realistic to date for this region. Furthermore, spatiotemporal patterns reveal new information on the magnitudes, locations, and timing of ET in the North American monsoon region with implications on land-atmosphere feedbacks.