Empirical forecasts of 850 hPa air temperature anomalies over the North Atlantic

• Published in: Quarterly Journal of the Royal Meteorological Society
• Main Authors: Gómez, E. Sánchez, García, F. Álvarez, Beviá, M. J. Ortiz
• Format: Article in Journal/Newspaper
• Language: English
• Published: Wiley 2001
• Subjects: North Atlantic
• Online Access: http://dx.doi.org/10.1002/qj.49712757813; http://api.wiley.com/onlinelibrary/tdm/v1/articles/10.1002%2Fqj.49712757813; https://rmets.onlinelibrary.wiley.com/doi/pdf/10.1002/qj.49712757813

Abstract The empirical forecasts presented here were developed as a benchmark for dynamical forecasts provided by uncoupled and coupled general‐circulation models. The predictand field is the variability of 850 hPa air temperature anomalies over the North Atlantic basin at timescales greater than eight months. The predictor field is chosen among different datasets (sea surface temperature anomalies, the predictand field itself, etc.). Some of the experiments make use of a single predictor field, whereas others use several. The performance of the forecasts, measured by the correlation between observed and forecast fields, is assessed against the skill of forecasts performed on a simple persistence basis. The statistical scheme used to make the forecast is based on singular‐value decomposition (SVD) of the cross‐covariance matrix between the predictor and the predictand field at the lead requested for the forecast. Prior to the SVD calculations, we remove with a filter the variability of predictor and predictand fields at timescales below eight months. Forecasts of air temperature anomaly have been made over the period from 1970 to 1998, at different lead times. The highest skill is achieved when the predictors are computed from anomalies of sea surface temperature over roughly the same domain as the predictand. The skill of the forecast is important south of 45°N and is broadly consistent through the three decades. About 56% of the variance of the eight‐month filtered predictand field (36% of the variance of the original field) can be forecast nine months ahead. The best values of the forecast skill appear in two regions, one in the subtropics and the other around the Iberian peninsula. Only in this last region, and also in another region over the USA, does the skill of the forecasts beat the skill of persistence. In order to identify the origin of the skill we study its dependence on the target season. We also investigate its dependence on the phase of the El Niño‐Southern Oscillation (ENSO) and the North ...