Determining the feasibility of using remote sensing to estimate irrigation water use in South Dakota at the field scale

Remote sensing of evapotranspiration (ET) offers the possibility to monitor irrigation water use at the field scale in a consistent manner over a large area through the use of satellite imagery. Several existing ET remote sensing models rely on a thermal band and/or red and near infrared bands of the electromagnetic spectrum. In this study, two models that use thermal data were incorporated into a larger model to produce seasonal irrigation maps over two areas of South Dakota. The remote sensing models used were the Mapping Evapotranspiration at High Resolution with Internalized Calibration (METRIC) model and the Simplified Surface Energy Balance (SSEB) model. The METRIC irrigation results were compared to reported irrigation volumes. Definitive conclusions were difficult to reach due to the lack of a clear relationship between the remotely sensed and reported values. The amount of error in both the remotely sensed and the reported values remains unknown. It was found that a seasonal irrigation map produced by applying the SSEB model to Landsat-sharpened MODIS images compared well to one produced by applying the METRIC model to Landsat images for a season and location that was well covered by Landsat images. The SSEB and METRIC based seasonal irrigation maps did not compare well when there was not good Landsat image coverage throughout the season. A single day ET map produced using the METRIC model was also compared to a purchased ET map of the same day and location produced using the Surface Energy Balance Algorithm for Land (SEBAL) model. The limited comparison showed a clear relationship between the METRIC and SEBAL models. Normalized Difference Vegetation Index (NDVI) images from the MODIS satellite were also briefly explored as a means to map ET at the field scale. The results indicated that it may be possible to incorporate MODIS NDVI images into a seasonal irrigation mapping model. The number and timing of Landsat images necessary to produce an accurate seasonal irrigation map using the METRIC model was also investigated by running several scenarios for two seasons. Each scenario involved removing one or two Landsat images from the seasonal analysis. Comparisons of the different scenarios indicated that the irrigation mapping model was less sensitive to Landsat images from later in the growing season.