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 #293984
Title: Application of the Auto-Tuned Land Assimilation System (ATLAS) to ASCAT and SMOS soil moisture retrieval products

Author
item Crow, Wade
item YILMAZ, TUGRUL - Science Systems, Inc

Submitted to: International Geoscience and Remote Sensing Symposium Proceedings
Publication Type: Abstract Only
Publication Acceptance Date: 6/1/2013
Publication Date: 7/21/2013
Citation: Crow, W.T., Yilmaz, T. 2013. Application of the Auto-Tuned Land Assimilation System (ATLAS) to ASCAT and SMOS soil moisture retrieval products. Proceeding of the 2013 International Geoscience and Remote Sensing Symposium [abstract] . July 21-26, 2013, Melbourne, Australia. 2013 CDROM.

Interpretive Summary:

Technical Abstract: Land data assimilations are typically based on highly uncertain assumptions regarding the statistical structure of observation and modeling errors. Left uncorrected, poor assumptions can degrade the quality of analysis products generated by land data assimilation systems. Recently, Crow and van den Berg (2010) described the combined application of a triple collocation calculation and filter innovation analysis to optimally derive an observational error variance for remotely-sensed surface soil moisture observations (for their assimilation into a land surface model). In particular, they note the benefits of combining innovation analysis with a triple collocation theory since the key statistical assumption required for a triple collocation analysis (zero cross-correlation in error for various soil moisture products) is distinct from the key assumption required for a classical innovation analysis (zero auto-correlation in observation and model forecast errors). As a result, there are opportunities arising for designing more robust data assimilation procedures based on the application of both approaches. In parallel, Yilmaz and Crow (2013) describe a triple-collocation based approach for optimally rescaling remote-sensed surface soil moisture retrievals prior to their ingestion in to a data assimilation system. This presentation will describe an attempt to combine this previous work into a single system referred to as the Auto-Tuned Land Assimilation System (ATLAS). Based on the input of a single satellite-based precipitation product and two independent surface soil moisture products, ATLAS will attempt to solve simultaneously for: an optimal observation (or rescaling) operator required to map observations into model space, the correct forecast error and observation error variance (for both assimilated soil moisture products) and the correct model forecast parameters. In addition, ATLAS will attempt to detect, and mitigate, the impact of auto-correlated observation errors and/or cross-correlated errors in the two remotely-sensed surface soil moisture products. By providing an objective basis for estimating key assimilation parameters, and detecting the presence of cross- and/or auto-correlated errors which complicate data assimilation cases, ATLAS will provide a more robust basis for the assimilation of surface soil moisture retrievals into land surface models.