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 #310149
Title: The impact of temporal auto-correlation mismatch on the assimilation of satellite-derived surface soil moisture retrievals

Author
item QUI, J. - Collaborator
item Crow, Wade
item MO, XINGGUO - Chinese Academy Of Sciences
item LIU, SUXIA - Chinese Academy Of Sciences

Submitted to: IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 12/1/2014
Publication Date: 12/15/2014
Publication URL: http://handle.nal.usda.gov/10113/60049
Citation: Qui, J., Crow, W.T., Mo, X., Liu, S. 2014. The impact of temporal auto-correlation mismatch on the assimilation of satellite-derived surface soil moisture retrievals. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. 7(8):3534-3542. DOI: 10.1109/JSTARS.2014.2349354.

Interpretive Summary: Remotely-sensed surface soil moisture estimates (acquired, for example, from earth-orbiting satellites) can be combined with dynamic soil water balance model predictions to obtain high-quality estimates of root-zone soil moisture availability for agricultural applications including: drought monitoring, fertilizer application, and irrigation scheduling. This integration of static observations with a dynamic model is commonly referred to as data assimilation. Unfortunately, a number of factors can degrade the performance of a data assimilation system designed to predict root-zone soil moisture availability. In particular, this paper examines the sensitivity of root-zone soil moisture estimates obtained from a data assimilation system to systematic inconsistencies in the dynamics of remotely-sensed versus modeled soil moisture time series and demonstrates that these inconsistencies can degrade the ability of a data assimilation system to accurately predict root-zone soil moisture levels. In response, the paper derives a simple diagnostic that can be applied to the detect the presence of these inconsistencies. Ultimately, the concepts developed in this paper will be applied to improve our ability to accurately monitor global agricultural drought.

Technical Abstract: Satellite-based surface soil moisture retrievals are commonly assimilated into eco-hydrological models in order to obtain improved profile soil moisture estimates. However, differences in temporal auto-correlation structure between these retrievals and comparable model-based predictions can potentially undermine the efficiency of such assimilation. Here, we conduct a series of synthetic experiments to examine the magnitude of this problem and the potential for detecting the presence of retrieval/model auto-correlation differences using a simple diagnostic procedure. Our synthetic experiments are based on modifying the observation operator within a data assimilation system to artificially induce auto-correlation differences between assimilated surface soil moisture retrievals and the surface soil moisture background created by an eco-hydrological model. Results demonstrate that neglecting a mismatch in retrieval/model auto-correlation can reduce the benefit of surface soil moisture data assimilation. The impact is especially large for soil profiles with limited vertical coupling. However, the presence of this source of retrieval/model auto-correlation misfit is detectable using a simple diagnostic index derived from a time series of soil moisture retrievals and open loop model predictions. Using relatively short data sets (~2 years in length), the diagnostic is capable of identifying worst-case scenarios leading to the most significant degradation of assimilation results.