Degree Type


Date of Award


Degree Name

Master of Science


Geological and Atmospheric Sciences

First Advisor

William A. Gallus


Ensemble forecasts have been used to increase the accuracy of Quantitative Precipitation Forecasts (QPF). Because of the challenging nature of developing a QPF, forecasters express uncertainty through Probability of Precipitation (POP) forecasts. POP forecasts can be developed using the percentage of agreement among ensemble members for a given forecast point. POP forecasts can also be developed through a more intricate statistical interpretation of model output called post-processing. There are numerous post-processing approaches to create POP forecasts, which include ensemble member agreement, calibration, binning precipitation amounts, and neighborhood approaches.

The main purpose of this research is to expand upon previous works regarding the relationship between QPF and POP forecasts through the neighborhood approach. By redefining the traditional ensemble through the use of a neighborhood ensemble of points, previous works had found that a single deterministic model can achieve similar or better skill than traditional approaches. Ensemble forecasts provided by the Center for Analysis and Prediction of Storms from the 2007, 2008, and 2010 NOAA Hazardous Weather Testbed Spring Experiments were used with additional variations of the neighborhood approach in order to determine if even better skill could be obtained. Four neighborhood variation tests were conducted using Brier scores and Brier skill scores for both 20 km and 4 km horizontal grid spacing, and results were compared skill scores of traditional approaches. Results had shown that some neighborhood variations could provide better skill than previously obtained, as well as outperforming traditional methods. Using a single model for POP generation through the neighborhood variations shows operational forecasting potential by providing more accurate forecasts than traditional methods, as well as requiring fewer computational resources that could be focused on improving a single deterministic model.

Copyright Owner

Michael Charles Kochasic



File Format


File Size

76 pages

Included in

Meteorology Commons