Towards an Operational Data Assimilation System for the National Water Model: Progress and Plans

Abstract:

Accurate water prediction across much of the nation requires an accurate analysis of snow states. We are working towards an operational, next-generation national snow analysis with the broad goals of introducing a snow analysis capability to the National Water Model (NWM) and improving the existing National Snow Analysis (NSA). Such a system contains three main components: (i) models; (ii) observations; and (iii) data assimilation (DA) methods. Our multi-agency team includes NCAR’s Research Applications Lab with WRF-Hydro/NWM model development expertise, NOAA’s Office of Water Prediction (OWP) who produce the NSA and have a long history of processing and understanding observations (in-situ and airborne gamma) and the Joint Center for Satellite Data Assimilation (JCSDA) who can leverage cutting-edge developments in data assimilation using the Joint Effort for Data assimilation Integration (JEDI).

The JEDI software suite provides the backbone for our DA system, which is being developed incrementally starting from the snow variables defined in this project. To date we have developed: (i) interfaces to couple JEDI and the NWM; (ii) converters to ingest example NSA in-situ observations and from model restart files for experiments; (iii) forward operators for a variety of 2D and 3D model states; and (iv) an approach to implementing DA solvers for use in testing and development. During this initial stage we are using existing workflow capabilities to control, start and advance the model, and a file-based or “pseudo-model” approach to the interface with JEDI. Here we present some preliminary results from observing system simulation experiments in the small Taylor Park catchment.

Ultimately, this project will result in a quasi-operational NWM snow data assimilation capability that integrates stakeholder feedback, the most promising operational methodologies, and the currently available observations (from NSA) to produce improved snow analyses and run off prediction. From the start, our system has been designed to be a capable research to operations to research (R2O2R) testbed able to support scientific and methodological research and development, drawing upon the many generic, interchangeable capabilities of JEDI, and leveraging existing efforts at OWP developing evaluation and diagnostic suites.