Reducing ET Modeling Uncertainty Using Thermal Infrared Radiometers.

Estimation of crop water use with thermal infrared sensing is potentially the best way to model the surface energy balance, resolve evapotranspiration (ET), and distinguish stressed and un-stressed vegetation. However, in many instances ET modeling with the Penman-Monteith (PM) equation is more robust, accurate and simpler than the energy balance approach, such as for extensive, well-watered, full-canopy conditions. Because of trade-offs between model simplicity and potential accuracy, a cotton field experiment was conducted in Maricopa, Arizona to quantitatively evaluate ET modeling uncertainties encountered with each. In 2014 and 2015, cotton was grown under linear move irrigation (sprinkler early, then drop hoses for remainder of season) in a 5 ha field with four irrigation treatment effects imposed: nominal FAO-56, NDVI FAO-56, DSSAT-CSM CROPGRO-Cotton Model, and Surface Energy Balance (SEB). Three wireless-based thermal infrared radiometers were located within the SEB treatment. Bayesian data analyses were performed to measure confidence intervals for modeled ET with prior probabilities estimated from local weather data and PM. Posterior probabilities provided confidence intervals after incorporation of plant canopy temperature data and a surface energy balance model. Uncertainty results from early, mid, and late season growth stages will be discussed.