S WVRFf - Var Implementation for Data Assimilation

The goal of this Masters project is to implement the WRF model with 3D variational assimilation (3DVAR) at MIT. A working version of WRF extends the scope of experimentation to mesoscale problems in both real and idealized scenarios. A state-of-the-art model and assimilation package can now be used to conduct science or as a benchmark to compare new methods with. The second goal of this project is to demonstrate MIT's WRF implementation in an ongoing study of the impact of position errors on contemporary data assimilation (DA) methods [21]. In weather forecasting, accurately predicting the position and shape of small scale features can be as important as predicting their strength. Position errors are unfortunately common in operational forecasts [2, 14, 21, 27] and arise for a number of reasons. It is difficult to factor error into its constituent sources [21]. Traditional data assimilaton methods are amplitude adjustment methods, which do not deal with position errors well [4, 21]. In this project, we configured the WRF-Var system for use at MIT to extend experimentation on data assimilation to mesoscale problems. We experiment on position errors with the WRF-Var system by using a standard WRF test; a tropical cyclone. The results for this identical twin experiment show the common distorted analysis from 3DVAR in dealing with position errors. A field alignment solution proposed by Ravela et al. [21] explicitly represents and minimizes position errors. We achieve promising results in testing this algorithm with WRF-Var by aligning WRF fields from the identical twin. Thesis Supervisor: Kerry A. Emanuel Title: Professor Research Supervisor: Sai Ravela Title: Research Scientist Committee Member: Christopher Hill Title: Principal Research Scientist