Application of Triple Collocation to Parameterize Two-Dimensional Land Data Assimilation Systems (Invited Presentation)

Recent growth in the operational availability of soil moisture observations from ground-based networks has fueled interest in the design of data assimilation systems to simultaneously assimilate both (spatially-sparse) ground-based observations and remote-sensing-based soil moisture retrievals into a land surface model. However, in order to function efficiently, these systems must be two-dimensional and capable of laterally-translating sparse ground-based observations into neighboring grids not containing ground observation sites. In addition, observation error statistics must properly account for uncertainty incurred with utilizing a point-scale soil moisture observation to describe a much-coarser grid scale mean. Meeting these challenges requires the development of new approaches for accurately estimating the variance and spatial auto-correlation structure of errors afflicting both land surface models and assimilated observations. In this presentation, we will describe the development of a triple-collocation-based technique capable of estimating the error variance and spatial auto-correlation parameters required to simultaneously assimilate both remotely-sensed surface soil moisture retrievals (ASCAT) and sparse surface soil moisture observations acquired from existing ground-based soil moisture networks (NOAA CRN and USDA SCAN) into a simple land surface model. Preliminary results will be used to assess the relative value of existing ASCAT, CRN, and SCAN soil moisture data products for spatially-continuous operational agricultural drought monitoring within the contiguous United States. In addition, the approach will be applied to quantify the spatial sampling density required in ground-based networks to match the information content of (more spatially-continuous) remote sensing products.