Integrating Landsat7 ETM+ and MODIS Products for Improved Spatial and Temporal Evapotranspiration Estimates

Several remotely sensed evapotranspiration (ET) models with varying complexity have been developed to map surface conditions at various spatial scales. However, routine monitoring (daily~weekly) of ET at high spatial resolutions has not been feasible due to the long repeat cycle of high spatial resolution satellite data (i.e., Landsat, ASTER and Quickbird, etc). Consequently, high temporal resolution (daily or more frequent) but moderate spatial resolution satellite data (i.e., MODIS, AVHRR and GOES, etc) have been typically used to formulate algorithms for ET monitoring. In this study, we investigate a method to provide ET estimates that have high spatial (~30m) and temporal (~daily) resolution by combining the advantages from Landsat and MODIS satellites. The proposed ET model is based on the commonly applied surface temperature-vegetation index (Ts-VI) triangle method. Prior to applying the ET model, two main inputs (Ts and VI) are obtained by disaggregating relevant MODIS images to Landsat scale by using a subtraction method that applies the difference between two MODIS images to subsequent Landsat images, producing sub-pixel variability within the MODIS pixel. Another key input (i.e., net radiation) is obtained from a previously developed MODIS-based net radiation model. The derived ET model is formulated completely from remote sensing data and does not require ground-based observations for implementation. Evaluation of the resultant ET product is undertaken within the Southeastern Arizona region containing a range of flux tower sites. Initial validation yields insight on the ability of the proposed method to integrate multi-scale platforms and the usefulness of cross-sensor data streams for high resolution ET modeling.