Development and Application of an Automated Calibration Algorithm for Estimating Evapotranspiration from Agriculture Using a Remotely Sensed Surface Energy Balance Model

The purpose of this research was to develop an automated calibration algorithm for the Mapping Evapotranspiration with Internalized Calibration at High Resolution (METRIC) remotely sensed surface energy balance model. The automated calibration algorithm was used to characterize the uncertainty of calibration of the METRIC model and to explore the potential of using METRIC for operational evapotranspiration (ET) estimates. METRIC is being used by the Nevada Division of Water Resources and the Desert Research Institute to estimate historical consumptive water use (CWU) from agriculture in western Nevada using Landsat imagery. Each METRIC ET estimate must be calibrated by a trained user in a time consuming process. The automated calibration algorithm was designed to generate ET estimates comparable to those from trained users by mimicking the user calibration process. The automated calibration would allow for METRIC ET estimates with minimal user intervention and allow for uncertainty and sensitivity analysis of the model and rapid generation of operational ET estimates.;For the historical CWU project, ET estimates have been made for different image footprints and dates by multiple users. In order to assess the uncertainty of calibration by multiple users, the automated calibration algorithm was used to generate 100 ET estimates for Landsat TM images from 2006. ET estimates were also generated using an automated pixel selection algorithm developed by the University of Idaho. The automated ET estimates were compared to user METRIC ET estimates of the same image dates made by five trained users. The variation in ET estimates generated by the automated calibration algorithm was found to be similar to the natural variation between user ET estimates. The calibration uncertainty was the highest for fields with low ET levels and lowest for fields with high ET levels. The seasonal calibration 95% confidence interval was found to be approximately 5% for the automated calibration algorithm. In order to assess the accuracy of the automated algorithms, 100 ET estimates were generated for Landsat TM images from 2003--2004 and 2005--2006. Automated daily and seasonal ET estimates compared well with measured ET data at multiple sites, which indicates that automated methods could be used for generating operational ET estimates that are similar to time-intensive manual efforts.