Date of Award

January 2016

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Atmospheric Sciences

First Advisor

Dr. Matthew S. Gilmore

Abstract

Weather radar (radio detection and ranging) is a specialized meteorological tool used to sample and track meteorological objects. This tool is critical for meteorologists and public decision-makers to inform and provide for their constituents in a timely manner, often with the protection of lives and property on the line. With the application of using meteorological and geospatial data in the realm of geographic information systems (G.I.S.), the task of blending the two sciences to inhibit further research and dissemination of information occurs.

This study focuses on the creation and implementation of a new geospatial tool, the Radar and Rainfall Analyzed in GIS (R2AIn-GIS) tool. The R2AIn-GIS tool was built upon the initial concepts from Zhang and Srinivasan’s (2010) NEXRAD validation and calibration (NEXRAD-VC) tool for G.I.S. R2AIn-GIS is updated to support the latest software features present in the geospatial world as well as analyze dual-polarization radar products.

To test the R2AIn-GIS tool, a warm seasonal precipitation study along with statistical analysis was performed over the Glacial Ridge National Wildlife Refuge in Minnesota, the largest prairie and wetland restoration site. Utilizing rain gauges operated by the United States Geological Survey, warm season precipitation events from 24 May 2012 to 31 August 2013 were analyzed using the R2AIn-GIS tool.

The R2AIn-GIS tool calculates the values from various dual-polarization radar products in conjunction with the recorded precipitation gauges to provide a detailed depiction of the weather event. Statistical tests including several iterations of multiple-linear regression of various combinations of dual-polarization radar variables allowed determination of rainfall rate prediction equations over the study area. This contributes to the body of radar literature regarding the best prediction equations for other locations. Unlike treatments in prior literature, most of the various assumptions in multiple linear regression are considered herein.

Based off the findings of the various statistical tests that adhere to the linear regression assumptions, regression models utilizing both reflectivity and correlation coefficient were the best models found during this study. These two variables had statistical significant p-values and their Durbin-Watson scores were among the highest even compared with the other radar variables of differential reflectivity and specific differential phase. Models including the radar variables reflectivity and correlation coefficient were found to be heteroscedastic along with the highest R Squared values. While the overall rainfall amounts were too small in terms of effective precipitation sampling, the results still positively contribute to the literature and provides the opportunity for future work.

Share

COinS