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 #300759

Research Project: Leveraging Remote Sensing, Land Surface Modeling and Ground-based Observations ... Variables within Heterogeneous Agricultural Landscapes

Location: Hydrology and Remote Sensing Laboratory

Title: Beyond triple collocation: Applications to satellite soil moisture

Author
item Su, Chun-hsu - University Of Melbourne
item Ryu, D - University Of Melbourne
item Crow, Wade
item Western, A - University Of Melbourne

Submitted to: Journal of Geophysical Research Atmospheres
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 3/1/2014
Publication Date: 6/14/2014
Publication URL: http://handle.nal.usda.gov/10113/60043
Citation: Su, C., Ryu, D., Crow, W.T., Western, A. 2014. Beyond triple collocation: Applications to satellite soil moisture. Journal of Geophysical Research Atmospheres. 119(11):6419-6439. DOI: 10.1102/2013JD021043.

Interpretive Summary: Accurate monitoring of soil moisture conditions is important for a range of agricultural applications including: drought monitoring, fertilizer scheduling and irrigation management. Remotely-sensed surface soil moisture retrievals offer an extremely promising technique for measuring soil moisture conditions over large geographic areas. However, before this data can be used with confidence in agricultural applications, it must be fully verified and validated. Unfortunately, such verification is challenging due to a severe lack of ground-based observations to serve as an objective ground-truth for validation activities. Recently a statistical technique called "triple collocation analysis" has been developed which allows for the estimation of errors in remotely-sensed surface soil moisture retrievals without any reliance on ground-based observations. However, as the name applies, triple collocation analysis requires the availability of three independent soil moisture data products. This stringent requirement limits it's applicability. In response, this paper generalizes the triple collocation analysis concept and presents an alternative version that can be applied using only two independent soil moisture data products. As a result, it provides an important advancement in our ability to widely validate remotely-sensed surface soil moisture retrievals for agricultural applications.

Technical Abstract: Triple collocation is now routinely used to resolve the exact (linear) relationships between multiple measurements and/or representations of a geophysical variable that are subject to errors. It has been utilized in the context of calibration, rescaling and error characterisation to allow comparisons of diverse data records from different direct and indirect measurement techniques ranging from in situ, remote sensing and model-based approaches. However successful application of triple collocation requires sufficiently large numbers of coincident data points from three independent time-series and, within the analysis period, linearity between data sets and stationarity in their relationships. These conditions are difficult to realise in practice due to infrequent spatiotemporal sampling of satellite and ground-based sensors. This work explores a potential strategy to ease these constraints by proposing the use of lagged variables as instruments to resolve the linear relationship between two data sets. This represents a broader class of methods known as the method of instrumental variables. It is well-suited for circumstances, where spatiotemporally collocated and independent data records are sparse and the geophysical variable of interest is sampled at time intervals shorter than their temporal coherence length. As a demonstration of utility, we apply the method to satellite soil moisture to recover the spatial error structures of active and passive microwave soil moisture products over Australia, and to estimate temporal properties of errors and scaling against in situ and model data. These are compared against the triple collocation estimator.