Sensitivity of evapotranspiration retrievals from the METRIC processing algorithm to improved radiometric resolution of Landsat 8 thermal data and to calibration bias in Landsat 7 and 8 surface temperature

Authors

Ayse Kilic, University of Nebraska-LincolnFollow
Richard G. Allen, University of IdahoFollow
Ricardo Trezza, University of Idaho Research and Extension Center, KimberlyFollow
Ian Ratcliffe, University of Nebraska-LincolnFollow
Baburao Kamble, University of Nebraska-LincolnFollow
Clarence W. Robison, University of IdahoFollow
Doruk Ozturk, University 0f Nebraska-LincolnFollow

Document Type

Article

Date of this Version

2016

Citation

Published in Remote Sensing of Environment 185 (2016), pp 198–209.

doi:10.1016/j.rse.2016.07.011

Comments

Copyright © 2016; published by Elsevier Inc. Used by permission.

Abstract

We made an assessment on the use of 12-bit resolution of Landsat 8 (L8) on evapotranspiration (ET) retrievals via the METRIC process as compared to using 8-bit resolution imagery of previous Landsat missions. METRIC (Mapping Evapotranspiration at high Resolution using Internalized Calibration) is an ET retrieval system commonly used in water and water rights management where the surface energy balance process is coupled with an extreme- end point calibration process to remove most impacts of systematic bias in remotely sensed inputs. We degraded L8 thermal images by grouping sequential digital numbers to reduce the apparent numerical resolution and then recomputed ET using METRIC and compared to nondegraded ET products. The use of 8-bit thermal data did not substantially impair the accuracy of ET retrievals derived from METRIC, as compared to the use of 12-bit thermal data. The largest error introduced into ET was <1%. We also compared ET retrieved from images processed during the L8 and Landsat 7 (L7) March 2013 underfly to assess differences in ET caused by differences in signal to noise ratio (SNR) and scaling of the two systems.

We evaluated the impact of bias in land surface temperature (LST) retrievals on ET determination using the CIMEC calibration approach (Calibration using Inverse Modeling using Extreme Member Calibration) employed in METRIC by introducing globally systematic biases into LST retrievals from L7 and L8 and comparing to ET from non-biased retrievals. The impacts of the introduction of both additive and multiplicative biases into surface temperature on ET were small for the three regions of the US studied, and for both L7 and L8 satellite systems. An independent study showed that METRIC-produced ET compared to within 3% of measured ET for the California site. The study assessed the impact of the February 2014 recalibration of L8 thermal data that caused a 3 K downward shift in LST estimation and changed reflectance values by about 0.7%. We found that the use of the recalibrated LST and shortwave data sets in METRIC did not change the accuracy of ET retrievals due to the automatic compensation for systematic biases employed by METRIC.