Skip to main content
ARS Home » Northeast Area » Beltsville, Maryland (BARC) » Beltsville Agricultural Research Center » Hydrology and Remote Sensing Laboratory » Research » Publications at this Location » Publication #376060

Research Project: Integrating Remote Sensing, Measurements and Modeling for Multi-Scale Assessment of Water Availability, Use, and Quality in Agroecosystems

Location: Hydrology and Remote Sensing Laboratory

Title: Global scale error assessments of soil moisture estimates from microwave-based active and passive satellites and land surface models over forest and mixed irrigated/dryland agriculture regions

item KIM, H. - University Of Virginia
item WIGNERON, J. - National Institute For Agricultural Research (INIAP)
item KUMAR, S. - Nasa Marshall Space Flight Center
item DONG, J. - US Department Of Agriculture (USDA)
item WAGNER, W. - Vienna University
item Cosh, Michael
item Bosch, David - Dave
item Holifield Collins, Chandra
item Starks, Patrick
item Seyfried, Mark
item LAKSHMI, V. - University Of Virginia

Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 8/15/2020
Publication Date: 12/15/2020
Citation: Kim, H., Wigneron, J., Kumar, S., Dong, J., Wagner, W., Cosh, M.H., Bosch, D.D., Holifield Collins, C.D., Starks, P.J., Seyfried, M.S., Lakshmi, V. 2020. Global scale error assessments of soil moisture estimates from microwave-based active and passive satellites and land surface models over forest and mixed irrigated/dryland agriculture regions. Remote Sensing of Environment. 251:112052.

Interpretive Summary: Triple Collocation (TC) analysis is used to understand the random errors associated with various methods of soil moisture estimation. The current methods for estimation are analyzed for different landscapes, specifically, forests and agricultural regions with a mixture of irrigated and dryland. A variety of independent models and remote sensing platforms were assessed along with in situ data. A new dataset was generated combined from the various inputs to minimize the amount of random error. This is a useful study for product development and algorithm calibration.

Technical Abstract: Over the past four decades, satellite systems and land surface models have been developed to estimate global-scale surface soil moisture (SSM). Before researchers make use of satellite and model-based SSM estimates over challenging areas such as forest and irrigated regions, it is important to understand the accuracy and error characteristics of SSM products from different data sources. Since high-density ground-based SSM networks are limited in space, methods of cross-comparing two or three independent collocated SSM data sets have recently been developed for evaluating SSM model/retrieval errors on a global scale. In this study, we utilized triple collocation (TC) analysis and in-situ SSM measurements to investigate SSM error characteristics collected at a spatial resolution of 0.25 degrees from the most widely used active and passive satellites and reanalysis data: The Advanced Scatterometer (ASCAT), the Soil Moisture and Ocean Salinity (SMOS), the Advanced Microwave Scanning Radiometer 2 (AMSR2), the Soil Moisture Active Passive (SMAP), the European Centre for Medium-Range Weather Forecasts Reanalysis 5 (ERA5), and the Global Land Data Assimilation System (GLDAS). We considered all possible triplets from 6 different products and proposed to consider the standard deviation of TC-based numbers to produce robust results from TC analysis. The study period is set to Apr. 2015 - Dec. 2019, which is an overlapping period of currently available SSM data from four major satellites. Over forested areas, it was expected that model-based SSM data could potentially provide less erroneous SSM estimates than satellites due to the intrinsic limitations of microwave-based SSM retrieval systems. In contrast to this expectation, satellite-based SSM estimates from ASCAT, SMAP, and SMOS showed fewer errors than the two land surface model (LSM) based SSM products over vegetated conditions. In addition, over irrigated regions, observation-based SSM data (i.e., satellites) were expected to be more accurate than model-based products because LSMs cannot capture artificial signals caused by human activities. It was shown that ASCAT, SMOS, and SMAP outperform other SSM products; however, reanalysis data from ERA5 and GLDAS showed better performance than AMSR2 over irrigated areas. These results emphasize that a wrong conclusion may be reached if satellite-based SSM data are considered as alternatives to model-based SSM data over irrigated areas. Similarly, referring to model data as a means to overcome the limitations of satellite-based SSM data in forested areas can also lead to erroneous conclusions. Selectively using multi-source SSM data according to the TC results, we were able to obtain a merged SSM data set with the highest signal-to-noise ratio; however, this "optimized" data set still included some limitations over densely forested and actively irrigated areas such as China and central Africa.