All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 32 7 0
PDF Downloads 9 3 0

Predictability of Monthly North Pacific Sea Level Pressure from Monthly Sea Surface Temperature for the Period 1933–1976

View More View Less
  • 1 Department of Meteorology and Physical Oceanography, Cook College and New Jersey Agricultural Experiment station, Rutgers, The State University of New Jersey, New Brunswick 08903
Full access

Abstract

Statistical models were developed to specify and predict the mean monthly sea level pressure (SLP) distribution over the central and eastern North Pacific Ocean from the mean monthly sea surface temperature (SST) distribution for the same area. These models were derived from data for the period 1947–71, with data from two additional periods (1933–41 and 1972–76) retained for independent testing.

The earlier period SST data is taken from a new data set compiled at the National Climatic Center and processed for use in examining large-scale air-sea interactions. This procedure is described.

Empirical orthogonal function (EOF) analysis was used to represent each field (SST and SLP) by a small number of composite variables. Regression analysis. was then used in which SST EOF amplitudes were the predictors and SLP EOF amplitudes were the predictands. The analyses were stratified by month, with lags from 0–3 months considered. Of the 84 models developed, 18 were statistically significant at the 10% level. The number of significant relationships was found to decrease with increasing tag, being greatest for SST contemporaneous with SLP. All statistically significant models involved SST's from the period June–January.

Each of the significant models was tested on the independent samples, using the reduction of error (RE) statistic as the measure of skill. An adjustment was made to the 1933–41 SST data to remove a systematic bias, and the RE scores recomputed. Using the adjusted data, RE scores for the 1933–41 period improved, with 10 of 18 models demonstrating skill overall. Most of the skill, especially at longer lags, was associated with models using late autumn and early winter SST as predictors.

Possible reasons for the seasonal distribution of the skillful relationships are discussed.

Abstract

Statistical models were developed to specify and predict the mean monthly sea level pressure (SLP) distribution over the central and eastern North Pacific Ocean from the mean monthly sea surface temperature (SST) distribution for the same area. These models were derived from data for the period 1947–71, with data from two additional periods (1933–41 and 1972–76) retained for independent testing.

The earlier period SST data is taken from a new data set compiled at the National Climatic Center and processed for use in examining large-scale air-sea interactions. This procedure is described.

Empirical orthogonal function (EOF) analysis was used to represent each field (SST and SLP) by a small number of composite variables. Regression analysis. was then used in which SST EOF amplitudes were the predictors and SLP EOF amplitudes were the predictands. The analyses were stratified by month, with lags from 0–3 months considered. Of the 84 models developed, 18 were statistically significant at the 10% level. The number of significant relationships was found to decrease with increasing tag, being greatest for SST contemporaneous with SLP. All statistically significant models involved SST's from the period June–January.

Each of the significant models was tested on the independent samples, using the reduction of error (RE) statistic as the measure of skill. An adjustment was made to the 1933–41 SST data to remove a systematic bias, and the RE scores recomputed. Using the adjusted data, RE scores for the 1933–41 period improved, with 10 of 18 models demonstrating skill overall. Most of the skill, especially at longer lags, was associated with models using late autumn and early winter SST as predictors.

Possible reasons for the seasonal distribution of the skillful relationships are discussed.

Save