Identification of California Climate Division Rain Year Precipitation Anomaly Patterns (1895-96 to 2013-14 Seasons) with Bayesian Analyses of Occurrence Probabilities Relative to El Nino, Neutral, or La Nina Episodes

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Tuesday, 6 January 2015: 4:30 PM
124B (Phoenix Convention Center - West and North Buildings)
Charles J. Fisk, Naval Base Ventura County, Point Mugu, CA
Manuscript (545.3 kB)

The State of California, nearly 800 miles long and 250 miles wide, is divided into seven NCDC Climate Divisions. Based on area-averaging techniques, single-valued month-to-month precipitation statistics have been compiled, division-by-division, since 1895, and just recently, the National Climatic Data Center, using new, improved areal averaging techniques, has recalculated the entire division-by-division precipitation statistics, by year. With such huge distances between the northern and southern borders, and the great topographical variation, it would seem likely that the character of rain year (July-June) anomalies would be non-uniform across divisions, one season to the next. The nature of these contrasts, and relationships to El Nino, Neutral, or La Nina episodes should make for interesting study. Given the above, the existence and relative frequencies of California Climate Division rain year anomaly variation patterns ("or modes") is investigated using K-Means Clustering Analysis integrated with the V-Fold Cross Validation Algorithm. Period of record is 1895-96 thru 2013-14, some 119 seasons. As applied to K-Means, the V-Fold Cross Validation Algorithm is an iterative training sample type procedure that tends to optimize the number of clusters created, depending on the choice of statistical distance metric (Euclidean, Squared Euclidean, etc.) and a percent improvement cutoff threshold (e.g., 5 percent). The present study performs the cluster analysis on the 119 seasons' (normalized) data utilizing the Squared Euclidean distance metric combined with the 5 percent distance improvement cutoff threshold. Then, with the "optimal" K number of clusters determined (in this case six), and through referencing of two lists from the NOAA Climate Prediction Center online site which identify past ENSO episode types (El Nino, Neutral, or La Nina) back through 1895-96, a Bayesian statistical analysis is performed that addresses the following questions: given an impending El Nino, Neutral, or La Nina episode, what are conditional probabilities that each of the six inter-divisional anomaly patterns will be expressed for a given July-June rain season. Results are described and interpreted, with the Bayesian probabilities compared among episode types.