Type of Document	Thesis
Author	Wolf, Linden
URN	etd-08242009-163026
Title	Statistical Forecasting of Florida Monthly Rainfall
Degree	Master of Science
Department	Meteorology, Department of
Advisory Committee	Advisor Name Title Jon E. Ahlquist Committee Chair Philip Sura Committee Member Robert Hart Committee Member
Keywords	Florida Rainfall Statistical Forecasting Long-Range Forecasting Seasonal Forecasting
Date of Defense	2009-08-06
Availability	unrestricted
Abstract This study computes statistical forecasts of monthly and three-monthly rainfall for seven regions of Florida defined by the National Climatic Data Center. First, time-lagged auto- and cross-correlations are computed involving monthly regional rainfall time series and various potential predictors. Various statistical monthly forecasting models are then built for each of the seven regions based on teleconnection indices and principal components of monthly heights of the global 500 hPa pressure surface. To compare these forecasts to those of the Climate Prediction Center (CPC), the forecasts are categorized into terciles, corresponding to the upper, middle, and lower thirds of the climatological distribution of rainfall for each of the twelve months for each region. Following CPC, these are scored with the Heidke Skill Score. The variability of model coefficients and forecast skill is measured using cross-validation. The monthly Heidke Skill Score is low but generally better than a climatological forecast, which is CPC's standard of comparison. For most months and forecast regions, the Heidke Skill Score increases if a forecast for the middle tercile is replaced by a forecast that all three terciles are equally likely. Averaged over the year, the Florida Panhandle has the lowest monthly forecast skill, and Southwest Florida has the highest. April and May as well as September and October have low skill statewide. These times of year are associated with shifts in the prevailing winds as well as the El Nino-Southern Oscillation (ENSO) phase. Higher skills are obtained when forecasting the next three month's total precipitation than the next month's total precipitation. This increase in skill is largely due to the important of ENSO as a predictor and that ENSO is less noisy across three months than one month. A summer low in the forecast skill for three month’s rain is due to the minimum in time-lagged correlation between late spring and summer. A middle tercile forecast for three-month rainfall is more likely to verify than a middle tercile forecast for one-month rainfall.
Files	Filename       Size       Approximate Download Time (Hours:Minutes:Seconds)     28.8 Modem   56K Modem   ISDN (64 Kb)   ISDN (128 Kb)   Higher-speed Access    Wolf_L_Thesis_2009.pdf 1.29 Mb 00:05:57 00:03:03 00:02:40 00:01:20 00:00:06