Improvements in statistical forecasts of monthly and two‐monthly surface air temperatures using a stratospheric predictor

Abstract

Anomalies in the extratropical stratospheric circulation are known to affect climate at the surface for up to 2 months during winter. Previously, several studies have proposed simple statistical models based on a stratospheric predictor to forecast surface climate; however, none of these studies considered a possibility, nor evaluated the skill, of categorical weather forecasts at individual sites. Here we evaluate the skill of the three‐category (below‐normal, near‐normal and above‐normal) monthly and two‐monthly mean surface air temperature statistical forecasts, both deterministic and probabilistic, which use, as predictors, normalized mean polar cap geopotential height at 150 hPa and mean surface air temperature over the preceding 30‐day period (persistence). We show that, during the period 1957–2013, at several sites influenced by the Northern Annular Mode variability, the monthly and two‐monthly forecasts based on the stratospheric predictor possess a moderate skill in December–March at lead times of several days, but in many cases they only marginally improve forecasts based on persistence only. Nevertheless at some sites in northeastern Europe and northwestern Russia the improvements indicate that statistical climate forecasting at these sites can benefit from addition of stratospheric information. The results of this study further demonstrate usefulness of simple statistical models based on stratospheric predictors, both as a complementary source of forecasts, and as providing benchmarks for evaluating the skill of more complex dynamical systems.