Contribution To The Seasonalair Temperature Forecast In The northern Hemisphere; A Statistical Approach

摘要

A seasonal air temperature forecasting is attempted by using statistical methods. The data basis consists of seasonal surface air temperature values at 514 grid points of the northern hemisphere, for the period 1948-2006. At first, grid points with covariant seasonal air temperatures for the period 1948-1996 are objectively grouped, by using Factor Analysis. Then, Canonical Correlation Analysis is applied on the time series of factor scores for each of the 4 pairs of sequential seasons as well as for the 4 pairs of 'crosswise' seasons. The results show that the number of the statistically significant pairs of canonical variates (W_i, V_i) ranges between 4 and 9 and the correlation coefficients between the canonical variates are higher than 0.92. Then, for every analysis, the W_i time series is correlated to the air temperature ones of the predictor season and the V_i time series to the air temperature ones of the predictant season for all the grid points. By plotting the correlation coefficients on maps, the isopleths indicate the areas where seasonal air temperature can be forecasted. The best results (r> 0.70) are found for three low latitude areas, where persistence prevails: a) autumn-winter: western Indian Ocean -central and eastern Indian Ocean, b) autumn-winter: central Pacific - eastern Pacific and c) spring-autumn: eastern Pacific - eastern Pacific. The results in the middle and high latitudes are less significant and practically they cannot be used for a seasonal air temperature forecast. Finally, for the areas characterized by high correlation coefficients between the canonical variates and the temperature time series, a validation process is carried out by comparing the temperature anomalies time series of the corresponding seasons for the period 1997-2006. The results confirm that prediction may be considered practically satisfactory for some low latitude areas of the Indian, the Atlantic and the Pacific Oceans.