Abstract:
Much of the progress in land data assimilation over the past decade has been made using so-called uni-variate systems. In such a system, only one particular type of observation is assimilated. For example, retrievals of surface soil moisture from space-borne passive microwave observations can be assimilated to obtain improved estimates of the soil moisture profile. Slightly more complex assimilation systems may include more than one type observation but are still focused on estimating one particular geophysical variable. For example, soil moisture retrievals from active and passive microwave observations can be used in a soil moisture assimilation system, or snow cover and snow water equivalent retrievals can be used in a snow assimilation system. While the benefit of assimilating more than one type of observation is already apparent in such systems, they can still be considered uni-variate systems because the analysis targets only one kind of geophysical variable. By contrast, multi-variate systems use several types of observations to analyze different kinds of geophysical variables. Progress towards such a multi-variate system is illustrated using examples from the NASA GEOS-5 land data assimilation system. The first example discusses the assimilation of satellite microwave brightness temperatures (Tbs) along with observations of large-scale terrestrial water storage (TWS) to estimate the TWS components at the scale of the model grid. The challenge is to configure the system such that the estimated uncertainties are realistic at the multiple spatial and temporal scales needed for a successful analysis. The second example addresses the assimilation of Tbs and freeze-thaw observations in the soil moisture and temperature analysis of the NASA SMAP L4_SM product, which involves the use of coarse-resolution radiometer Tbs and retrievals based on high-resolution radar backscatter.