Thermodynamic Retrieval experiments with a 2-D model

For these experiments, we view data assimilation as a method for analyzing storm-scale observations. Assimilation of Doppler wind observations into a numerical model is an approach for building time continuity into the dataset and retrieving unobserved parameters.

As they are typically formulated, model and adjoint data assimilation methods produce a numerical simulation that matches the observations in a least squares sense. Unfortunately, this approach for producing an optimal simulation requires considerable expense in terms of design, coding, and computation.

We are currently conducting experiments with new, near-optimal data assimilation strategies that involve integrating a numerical model forwards and backwards. For the experiments, we use simulated radar observations from a 2-D model of a cold downburst. We will compare the results of the retrievals to those from traditional methods.