Amplitude of the diurnal temperature cycle as observed by thermal infrared and microwave radiometers

Authors: Holmes, Thomas; Crow, Wade; HAIN, C. - University Of Maryland; Anderson, Martha; Kustas, William - Bill

Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/3/2014
Publication Date: 3/1/2015
Citation: Holmes, T.R., Crow, W.T., Hain, C., Anderson, M.C., Kustas, W.P. 2015. Amplitude of the diurnal temperature cycle as observed by thermal infrared and microwave radiometers. Remote Sensing of Environment. 158:110-125.

Interpretive Summary: Over the past decade, satellite observations have been successfully leveraged to provide new information about the hydrologic cycle over land and improve our ability to characterize the availability of water for agricultural and rangeland management. New long term and global datasets of soil moisture and land evaporation have been developed to supplement more established datasets of precipitation. Physical retrieval algorithms to translate the satellite observations into these hydrological parameters (used in agricultural drought monitoring) require knowledge of land surface temperature (LST). Even though global measurements of LST are provided by many satellite platforms, separate remote sensing technologies are not well reconciled. This paper aims to reconcile two such technologies for estimating LST from space: passive microwave (MW) and thermal infrared (TIR) LST approaches. Although MW has a lower spatial resolution, its temporal record is more complete as it is more tolerant to clouds. Moreover, because TIR and MW potentially provide two independent estimates of LST, their merger should reduce the level of random error in LST. To that end, we compile a 5-year MW-LST dataset over the African continent and Europe by combining observations from seven inter-calibrated low orbiting satellites. Results suggest that the diurnal LST amplitude over much of the African domain can be estimated from TIR and MW with similar levels of random error for clear sky days, except over dry sandy desert areas. Together with the temporal sampling advantage of MW these results suggest great benefit to merging MW and TIR LST for examining the hydrological cycle over land. This research may eventually be used to improve the ability of USDA to operationally monitor the event and severity of agricultural drought around the world.

Technical Abstract: Land surface temperature (LST) is a key input to physically-based retrieval algorithms of hydrological states and fluxes, and global measurements of LST are provided by many satellite platforms. Passive microwave (MW) observations offer an alternative to conventional thermal infrared (TIR) LST retrieval approaches. Although MW has a lower spatial resolution, its temporal record is more complete as it is more tolerant to clouds. Moreover, merging TIR and MW LST with independent random errors should result in enhanced LST products. Despite these benefits, MW-based LST retrievals are not widely adopted for land applications except as an input to soil moisture retrieval algorithms. This research aims to facilitate expanded use of MW-based LST by formulating a model to explain the structural differences in comparison to TIR-LST, and quantifying random errors in each datastream. To that end, we compile a 5-year MW-LST dataset over the African continent and Europe, combining observations from seven inter-calibrated low orbiting satellites equipped with suitable microwave radiometers. We compare the diurnal timing and amplitude of this dataset to TIR-LST time series produced by LSA-SAF from the Meteosat Second Generation geostationary satellite. A third independent data source, the skin temperature as modeled by the Modern Era Reanalysis for Research and Applications (MERRA) as produced at NASA’s Global Modeling and Assimilation Office, is included in a triple collocation analysis to calculate the random error in the amplitude anomaly. Results suggest that the diurnal LST amplitude over much of the African domain can be estimated from TIR and MW with similar levels of random error for clear sky days, except over dry sandy desert areas where the MW sensing depth extends too far below the surface. The temporal sampling advantage of MW, due to higher cloud tolerance in comparison with TIR retrievals, appears to be substantial but will need to be verified by follow-on studies. Overall the results of this study present a significant step forward in reconciling the structural differences in LST, preparing the way for a global merger of MW and TIR LST time series.