AMSR-E-Based soil moisture retrieval algorithms and transferability to AMSR2

Authors: Mladenova, Iliana; Jackson, Thomas; NJOKU, ENI - Jet Propulsion Laboratory; BINDLISH, R. - Science Systems, Inc; Cosh, Michael; CHAN, S. - Jet Propulsion Laboratory

Submitted to: International Geoscience and Remote Sensing Symposium Proceedings
Publication Type: Abstract Only
Publication Acceptance Date: 4/1/2014
Publication Date: 7/13/2014
Citation: Mladenova, I., Jackson, T.J., Njoku, E., Bindlish, R., Cosh, M.H., Chan, S. 2014. AMSR-E-Based soil moisture retrieval algorithms and transferability to AMSR2. International Geoscience and Remote Sensing Symposium Proceedings. July 13-18, 2014. Quebec, Canada. 2014 CDROM.

Interpretive Summary:

Technical Abstract: The launch of the Advanced Microwave Scanning Radiometer on NASA’s Earth Observing System Aqua satellite (AMSR-E) in June of 2002 has led to major advancements in the routine global mapping of soil moisture. The wide availability of AMSR-E data has promoted development of a number of global soil moisture products available from different sources, including JAXA (Japan Aerospace Exploration Agency), NASA (National Aeronautics and Space Administration), and other groups. These AMSR-E based soil moisture products have been evaluated and inter-compared in a number of studies, under a range of ground and climate conditions and using a variety of metrics. These evaluations have shown differences between the AMSR-E products in terms of biases, sensitivities and temporal responses. In our research we have investigated these differences and approaches for improving the algorithms and products. Two algorithms that have received primary attention in our studies are the Normalized Polarization Difference (NPD) algorithm, used in the AMSR-E product available through the National Snow and Ice Data Center (NSIDC) and the Single Channel Algorithm (SCA) used in the AMSR-E data product available through the U.S. Department of Agriculture (USDA). Other algorithms evaluated included the Land Parameter Retrieval Model (LPRM), the University of Montana soil moisture retrieval algorithm (UMT) and the HydroAlgo-Artificial Neural Network-based approach (HA-ANN). Since this work was initiated the first Water-related Global Change Observation Mission (GCOM-W1) was launched by JAXA in May of 2012. GCOM-W1 carries onboard the AMSR2 instrument, the successor of AMSR-E. AMSR2 provides continuity following AMSR-E and the opportunity to generate a global long-term satellite soil moisture data record from the same instrument type. The algorithms investigated in our research study have therefore been evaluated in addition for their potential in processing AMSR-E and AMSR2 data jointly to generate such a long-term data record. In order to evaluate options, and to improve or modify the theoretical basis of the algorithms, the algorithms were first inter-compared. All the algorithms studied are based on the same microwave principles. However, as indicated earlier they produce different results. Most passive-based soil moisture retrieval algorithms have two main components: a Radiative Transfer Model (RTM) that relates brightness temperature to soil dielectric properties and a dielectric mixing model that relates the soil dielectric properties to volumetric soil moisture. The two are connected through the Fresnel reflectivity equations. The microwave signal observed by the sensor at the top of the atmosphere is a composite measure that includes information on all of the constituents between the soil surface and the satellite. This requires that the modifying effects of the land surface temperature, vegetation, surface roughness and soil properties are properly accounted for in the inversion of the RTM and dielectric model to retrieve soil moisture. Our research focused on five well-established global approaches that were capable of retrieving soil moisture using the same set of AMSR-E observations as the NPD algorithm. Examination of the operational codes of these five approaches confirmed that they all followed the above outlined scheme. However, the specific algorithm solutions and final soil moisture retrievals differed depending on how the individual developers approached the dimensionality problem, i.e. number of system variables and ancillary datasets used, and assumptions and simplifications undertaken to make the problem solvable. A series of sensitivity analyzes were performed to evaluate the impact of the assumptions and identify the theoretical components that define the algorithm’s sensitivity and response. It was determined that the latter strongly