Sequential storm-scale data assimilation of polarimetric radar observations using an ensemble Kalman filter

The United States Weather Research Program recognized quantitative precipitation forecasting as one of the most important yet challenging problems currently facing the operational meteorological community. Among their recommended research focuses was “designing new data-gathering strategies for numerical model initialization”, which included greater utilization of available observations through advanced data assimilation systems along with improvements to cloud microphysical process representation within numerical models. Toward this goal we will demonstrate a prototype assimilation system which ingests synthetic polarimetric radar products.

Polarimetric radar products have been utilized by microphysical retrieval algorithms, yet few efforts have been made to directly assimilate polarimetric data. A previous polarimetric data assimilation initiative used a 4DVAR framework, and encountered difficulties with convergence under the system’s strong constraint. Developing an adjoint (required for 4DVAR) is a difficult and time consuming process for complex microphysics, making the problem less tractable. This study will take advantage of the ensemble Kalman filter (EnKF) system, which benefits from a both a weak constraint and does not require the development of an adjoint. As such, it provides a simpler framework for developing assimilation systems with advanced microphysical parameterizations. Several prior storm-scale data assimilation studies have demonstrated the EnKF system is able to capture the qualitative storm structure following a few sequential cycles of radial velocity and reflectivity assimilation. This method is currently limited, however, by simplified microphysics used to improve the convergence characteristics. Convergence should improve with the inclusion of additional data sets that provide information about the precipitation particle characteristics, allowing for the use of more complex microphysical models.

We will offer results from a simplified demonstration system, which incorporates dual moment warm rain microphysics and assimilates differential reflectivity and specific differential phase polarimetric moments in addition to radial velocity and reflectivity. This will be conducted via observing system simulation experiments where the “observations” are made from a model simulation. The same model configuration is used for the assimilation system, such that the exact solution is known and available for comparison. Moreover, this proof-of-concept demonstration will provide a framework for future efforts that will include more advanced microphysical models (i.e. ice processes), additional polarimetric variables as well as real data experiments.