Data assimilation and land surface modelling: The role of open source platforms

Abstract

The use of data assimilation (DA) techniques is well established in environmental sciences, but most applications are non-transferrable and, as such, model-dependent. We trialled an open source platform to design a DA experiment, tailored to a specific land surface model and geographical context. Once set up, we tested the system by selecting one of a number of generic DA algorithms available. The specifics of the experiment relate to the Australian Water Resources Assessment Landscape (AWRA-L) model, in which we assimilated lumped, catchment-average satellite-derived soil moisture retrievals for the Campaspe River Catchment, Victoria, in the Murray-Darling Basin, Australia. We found the use of a generic DA algorithm feasible, although technical challenges may constitute a threshold in the initial design stages of an experimental set-up, and further refinement and analysis of preliminary results is necessary. Further complications with regard to efficiency and speed are anticipated in refining the tested lumped experimental set-up to one with gridded data and multi-variable constraints. The experiment shows transferrable DA methods offered by an open source platform can be used to set-up a specific model-observation configurations for the Australian Water Resource Assessment system. While this in principle allows for testing of various DA algorithms and configurations to make the best choice, it potentially cuts time and development costs.