Monday, 13 May 2002
The development of a long-range forecast tool for Shemya winds using CART analysis
The Ballistic Missile Defense Organization (BMDO) will construct, during a two-week period, an X-Band Radar (XBR) on Shemya, AK. Wind speeds must not, at any time during the construction, exceed a 25 knot limit set by the BMDO for lifting the massive dome panels into place. The goal of this research was to develop a forecast tool, or model, to determine the feasibility of predicting any upcoming two-week windows of opportunity, well in advance, that will ensure the successful completion of constructing the XBR on Shemya. In order to reach this goal, the following objectives were achieved; 1) a climatological wind study for Shemya to assess the optimal “climatological window” to build the XBR, 2) a detailed synoptic study over the North Pacific, to gain an understanding of how systems develop, move, and vary on an annual basis, 3) a traditional statistical analysis of the data first, then a Classification and Regression Tree (CART) analysis for pattern recognition of global teleconnection indices, and 4) a decision tree long-range forecasting tool was developed for use by the 11th Operational Weather Squadron (OWS) Alaskan forecast hub. The Aleutian Island chain is plagued by persistent strong winds, since the Aleutian Low and expanding polar vortex affect the region in the winter, as do tropical storms and frontal passages in the summer. This, combined with Shemya being located near the exit region of the climatological storm track off the East Asian continent, makes the island one of the most challenging forecast locations in the Northern Hemisphere. This study compares surface winds and teleconnection indices as computed by the Climate Prediction Center (CPC), to statistically analyze the data. A discussion of how the teleconnection indices are computed, plus a brief explanation of how rotated principle component analysis (RPCA) is used to compute the teleconnection indices are also included in the study. The data were analyzed using standard regression, including linear and multiple linear regression techniques, and then CART analysis was used for large-scale pattern recognition. The approach of the CART analysis theory used in this study was to determine which large-scale pressure patterns in the Northern Hemisphere are conducive to “low winds” over Shemya, and then to determine if these large scale patterns can be seasonally predicted from the known global teleconnection patterns which occurred a month before. CART was discovered to be the best method of analysis and a forecast decision matrix for the 11th OWS Alaskan forecast hub was then developed to assist forecasters in providing long-range wind forecasts to the BMDO.