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 #399657

Research Project: From Field to Watershed: Enhancing Water Quality and Management in Agroecosystems through Remote Sensing, Ground Measurements, and Integrative Modeling

Location: Hydrology and Remote Sensing Laboratory

Title: An introduction to Bayesian Machine Learning with an application in global-scale active and passive satellite-based soil moisture error pattern analysis

item KIM, H. - Oak Ridge Institute For Science And Education (ORISE)
item WAGNER, W. - Technische Universität Wien
item Crow, Wade
item LI, X. - Collaborator
item LAKSHMI, V. - University Of Virginia

Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 7/8/2023
Publication Date: 8/8/2023
Citation: Kim, H., Wagner, W., Crow, W.T., Li, X., Lakshmi, V. 2023. An introduction to Bayesian Machine Learning with an application in global-scale active and passive satellite-based soil moisture error pattern analysis. Remote Sensing of Environment. 296.

Interpretive Summary: Remotely sensed soil moisture retrievals are of potentially great value for efforts to improve the monitoring of agricultural drought; however, we do not completely understand how the quality of these estimates varies as a function of environmental variables (e.g., land cover, soil type and climate characteristics). This work utilizes new mathematical tools to better understand how soil moisture uncertainty varies across space and, therefore, how valuable these retrievals are for agricultural and hydrological applications. Our proposed methodology is unique in that it can simultaneously measure the relative importance of multiple environmental factors on the quality of remotely sensed data simultaneously. Results from this study will eventually be used by drought analysts to properly weigh drought estimates provided by soil moisture remote sensing in the context of broader information provided by other sources.

Technical Abstract: Estimating accurate surface soil moisture (SM) dynamics from space and knowing the error characteristics of the data are of great importance for the application of satellite-based SM data throughout many Earth Science/Environmental Engineering disciplines. Here, we introduce the Bayesian inference approach to analyze the error characteristics of most widely used passive and active microwave satellite-derived SM data sets for different overpass times: Soil Moisture Active Passive (SMAP), Soil Moisture and Ocean Salinity (SMOS), and Advanced Scatterometer (ASCAT). Specifically, we propose the application of Bayesian hierarchical modeling to investigate the relative importance of different environmental factors and human activities on the accuracy of satellite-based data. We conducted an uncertainty analysis on model parameters with a full range of uncertainties to assess various environmental factors’ impact on SM data accuracy. We focused on investigating human-induced error sources such as radio frequency interference (RFI) and the degree of irrigation on microwave satellite systems, and naturally introduced error sources such as vegetation and soil organic matter. We found that, in contrast to our expectations, RFI can either be an informative or uninformative predictor in inferring the uncertainties of different microwave satellite systems – with the same intensity of RFI, the SMOS system can be more vulnerable than the SMAP and ASCAT systems. In addition, the irrigation fraction of lands is relatively unimportant in predicting the uncertainties of satellite-based SM data. However, other factors such as soil temperature, vegetation, and soil organic matter have a more significant effect on the accuracy of satellite-based SM data. The current study provides a useful framework for applying Bayesian theory to investigate the error characteristics of not only satellite-based SM data, but also any other time-varying geophysical variable.