Automated Selection of Anchor Pixels for Landsat Based Evapotranspiration Estimation

Abstract

When managing local and regional water resources, the estimation of evapotranspiration is important and has generally been one of the components of the hydrological cycle that has the greatest uncertainty. With the development of suitable models and algorithms applied to high resolution (30 m) satellite imagery, evapotranspiration may be estimated with greater accuracy, and in a cost effective and time efficient manner. The METRIC image processing procedure calculates net radiation, soil heat flux and sensible heat flux through a number of steps before estimating evapotranspiration as the residual from the energy balance. Sensible heat flux is calibrated using the so-called "cold" and "hot" anchor pixels. These pixels are selected by the user, which may introduce some operator dependency or human errors on the estimation of sensible heat flux and subsequently error in the final map of evapotranspiration. A procedure for automated selection of the anchor pixels is presented. The automated pixel selection procedure will reduce the user dependency of the estimations of sensible heat flux. Additionally, it may allow more novice users to obtain good results when applying METRIC.