Significant Advances in Statistical Long-Range Forecasting

摘要

It is purpose of this report to demonstrate that there exists meaningful predictability of monthly and seasonal U.S. temperature anomalies at rather long range. Successful experiments along two separate approaches, one employing pressure predictors and the other using temperature predictors, lead to an accomplishment of this purpose. Monthly precipitation anomalies are shown to be statistically related to monthly temperature anomalies on a contemporary basis. The first chapter provides a preliminary discussion on mathematical tools. Chapter II displays Empirical Orthogonal Function Decompositions of pressure, temperature, and precipitation, and contains the statistical results on dependent sample lag relations of these fields. The last chapter presents the methods and results of independent sample temperature to temperature prediction. (Author)