Bechini, Renzo, authorChandrasekar, V., advisorJayasumana, Anura, committee memberMielke, Paul, committee memberSun, Juanzhen, committee member2017-01-042018-12-302016http://hdl.handle.net/10217/178810Nowcasting is primarily a description of the near-future forecasted atmospheric state, relying heavily on observations. Besides routine meteorological observations (pressure, temperature, humidity, wind), dual-polarization weather radar provides a large amount of useful information due to the frequent-update (~5 min) and high-resolution (~500 m) three-dimensional sampling of the atmosphere. However, the atmospheric state variables are not readily invertible from radar remote observations, resulting in complexity in the numerical model data assimilation. This problem is normally dealt with by defining observation operators to simulate the radar variables from the model state vector. In this work the dual-polarization radar based retrievals are developed in order to demonstrate their potential for microphysics and dynamics retrievals. In particular the analysis of radar observations in convective storms and in stratiform ice clouds revealed that specific dual-polarization signatures can be successfully related to important dynamic properties such as vertical air motions, both in convective precipitation (strong updrafts, several m s-1) and in stratiform precipitation (large areas of weak updrafts, tenths of m s-1, associated with mid-tropospheric mesoscale forcing). Given the relevance of polarimetric signatures to dynamics retrievals, an improved hydrometeor classification method is developed based on a learn-from-data approach. In this technique, the traditional bin-based classification is replaced with a semi-supervised approach which combines cluster analysis, spatial contiguity, and statistical inference to assign the most likely class to a set of identified connected regions. The hydrometeor classification and relevant dual-polarization signatures establish a starting point to explore new means to improve the analysis of precipitation and near-surface winds, and their subsequent nowcasting. In particular the relevance of a well-known dual-polarization feature associated with deep convection (vertical columns of differential reflectivity) is illustrated by including the microphysics and dynamics-related information into a simple method for the analysis of surface winds. The goal of a physically consistent analysis is further pursued considering the Variational Doppler Radar Analysis System (VDRAS), an advanced four-dimensional data assimilation system based on a cloud-scale model, specifically designed for ingesting Doppler weather radar observations. The typical application using single-polarization observations from long-range S-band or C-band radars is here extended to high frequency (X-band), short range radars and dual-polarization observations. The combination of the hydrometeor classification and dual-polarization rainwater estimation allows to successfully assimilating the X-band observations, otherwise prone to relevant errors when using the reflectivity-based observation operator widely employed in numerical models. The feasibility of X-band data assimilation to contribute building a consistent analysis for nowcasting is demonstrated over the Dalls-Fort Worth test bed, where a dense network of dual-polarization X-band radars is being deployed. Eventually, a novel method for the nowcasting of precipitation and winds is built upon the VDRAS analysis, in an attempt to combine the robustness and consistency of data assimilation and the efficacy of extrapolation techniques for very short-term forecasting.born digitaldoctoral dissertationsengCopyright and other restrictions may apply. User is responsible for compliance with all applicable laws. For information about copyright law, please see https://libguides.colostate.edu/copyright.microphysicspolarimetryweathernowcastingdynamicsradarText