The auto-tuned land data assimilation system (ATLAS)

Authors: Crow, Wade; YILMAZ, M - Collaborator

Submitted to: Water Resources Research
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 12/12/2013
Publication Date: 1/17/2014
Publication URL: http://handle.nal.usda.gov/10113/58520
Citation: Crow, W.T., Yilmaz, M.T. 2014. The auto-tuned land data assimilation system (ATLAS). Water Resources Research. 50(1):371-384. DOI:10.1002/2013WR0145502014

Interpretive Summary: Agricultural drought has enormous consequences for domestic economic interests and international food security concerns. These implications can be minimized by the effective detection and mitigation of drought impacts. Land data assimilation systems are increasingly being tasked with optimally combining hydrologic information acquired from a variety of sources (e.g., hydrologic modeling, ground-based observations and remotely-sensed observations) into integrated predictions of root-zone soil moisture availability. These predictions, in turn, form the basis of agricultural drought monitoring systems. The performance of these systems requires detailed information about errors in various sources of soil moisture information (since accurate products with low errors should be weighted more than inaccurate products with high errors). This paper describes a new mathematical technique for deriving soil moisture error information required for data assimilation systems to accurately predict the onset and evolution of drought. Results demonstrate the superiority of the approach versus existing techniques. As a result, it provides an important contribution to current USDA ARS efforts to improve our ability to track agricultural drought.

Technical Abstract: Land data assimilation systems are tasked with the merging remotely-sensed soil moisture retrievals with information derived from a soil water balance model driven (principally) by observed rainfall. The performance of such systems is frequently degraded by the imprecise specification of parameters describing modeling and observation errors. Here, a new land data assimilation procedure is proposed - coined the Auto-tuned Land Data Assimilation System (ATLAS) - which simultaneously solves for all parameters required for the application of simple land data assimilation system to merge satellite-based rainfall and soil moisture retrievals into a soil moisture analysis for drought monitoring applications. The approach is based on the merger of a triple collocation (TC) strategy with a filtering innovation analysis and designed specifically to leverage the simultaneous availability of satellite-based soil moisture retrievals products acquired are both active and passive microwave remote sensing techniques. A number of variants of the ATLAS approach - each based on a different strategy for leveraging TC and innovation analysis within an adaptive filtering framework - are derived analytically and evaluated via the application of a synthetic twin experiment. In addition, a preliminary real data analysis is conducted using real satellite-base products and evaluated against independent ground-based observations. Results illustrate the potential of the ATLAS approach to improve the analysis of soil moisture anomalies using data products derived from the Global Precipitation Measurement (GPM) and the NASA Soil Moisture Active/Passive missions.