Application of a machine learning algorithm in generating evapotranspiration data product from thermal infrared and microwave coupled satellite observations

Authors: FANG, L. - University Of Maryland; ZHAN, X. - National Oceanic & Atmospheric Administration (NOAA); KALLURI, S. - National Oceanic & Atmospheric Administration (NOAA); YU, P. - University Of Maryland; HAIN, C. - Nasa Marshall Space Flight Center; Anderson, Martha; LASZLO, I. - National Oceanic & Atmospheric Administration (NOAA)

Submitted to: Frontiers in Big Data
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/28/2022
Publication Date: 5/20/2022
Citation: Fang, L., Zhan, X., Kalluri, S., Yu, P., Hain, C., Anderson, M.C., Laszlo, I. 2022. Application of a machine learning algorithm in generating evapotranspiration data product from thermal infrared and microwave coupled satellite observations. Frontiers in Big Data. https://doi.org/10.3389/fdata.2022.768676.
DOI: https://doi.org/10.3389/fdata.2022.768676

Interpretive Summary: Thermal infrared satellite observations, describing the temperature of the land-surface, have proven to be an effective input for mapping evapotranspiration (ET) or crop water use. Thermal-based energy balance models connect the rate of evaporative cooling to the observed land-surface temperature, enabling estimates of ET without needing ancillary information about rainfall or irrigation management. Thermal sensors, however, have the disadvantage of not being able to see through clouds; hence, ET estimates can be generated under clear-sky conditions only, which can be very limiting in more humid climates. This paper investigates a method for augmenting the thermal observations with estimates of land-surface temperature acquired from microwave imagery under both clear and cloudy conditions. A landsurface model and machine learning are used to harmonize these two temperature datasets. In comparison with ground-based observations, the combined thermal-microwave all-sky ET model performs as well as the traditional thermal-based model. However, the allsky model provides significantly better coverage, both spatially and temporally. These methods can be applied at continental to global scales to provide satellite-based estimates of this important component of the hydrologic budget.

Technical Abstract: Land surface evapotranspiration (ET) is one of the main energy sources for atmospheric dynamics and a critical component of the local, regional, or global water cycle. Consequently, accurately measuring or estimating ET is one of the most active topics in hydro-climatology research since the last century. With massive and spatially distributed observational data sets of land surface properties and environmental conditions being collected from the ground, airborne or space-borne platforms daily since the past few decades, many research teams have started to use big data science to advance the ET estimation methods. The Geostationary satellite Evapotranspiration and Drought (GET-D) product system has been developed at the National Oceanic and Atmospheric Administration (NOAA) in 2016 to generate daily ET and drought maps operationally. The primary inputs of the current GET-D system are the thermal infrared (TIR) observations from NOAA GOES satellite series. Because of the cloud contamination to the TIR observations, the spatial coverage of the daily GET-D ET product has been severely impacted. Based on the most recent advances, we have tested a machine learning algorithm to estimate all-weather land surface temperature (LST) from TIR and microwave (MW) combined satellite observations. With the regression tree machine learning approach, we can combine the high accuracy and high spatial resolution of GOES TIR data with the better spatial coverage of passive microwave observations and LST simulations of land surface models (LSMs). The regression tree model combines the three LST data sources for both clear and cloudy days, which enables the GET-D system to derive an all-weather ET product. This paper reports how the all-weather LST and ET are generated in the upgraded GET-D system and how these LST and ET estimates are validated against ground measurements. The results illustrate that the regression tree machine learning method is proven to be feasible and effective for generating daily ET under all weather conditions with satisfactory accuracy from the big volume of satellite observations.