Analogs in the Wintertime 500 mb Height Field

View More View Less
  • 1 Goddard Laboratory for Atmospheric Sciences, NASA/Goddard Space Flight Center, Greenbelt, MD 20771
© Get Permissions
Full access

Abstract

A 15-winter sample of daily gridded values of Northern Hemisphere 500 mb heights is examined for the existence of recurrent flow patterns (“analogs”). The analog search is repeated several times after degrees of freedom are successively removed from the data by spatial filtering, temporal averaging, and consideration of smaller sectors of the hemisphere. The root mean square difference (or “rms error”) between the most closely analogous maps, defined over the middle latitudes (30–70°N), is slightly greater than half the average error between randomly chosen maps, with an estimated rms error doubling time of nearly 8 days. If the analog search is conducted using only the longwave component (zonal wavenumbers 0–4) of each map, the rms error between the best analog pairs is reduced to less than half of the rms error between long-wave anomalies on randomly chosen maps, but the doubling time is also reduced to less than 7 days. If the analog search is further restricted to a limited region over North America or Europe, the rms error between the best analog pairs is less than 40% of the rms error (for the sme region) between randomly chosen maps, but the error doubling time is further reduced to 4–5 days. In all cases, the degradation of analog quality is so rapid that a forecasting scheme based on the analogs would fail to produce more skillful forecasts than simple persistence.

Abstract

A 15-winter sample of daily gridded values of Northern Hemisphere 500 mb heights is examined for the existence of recurrent flow patterns (“analogs”). The analog search is repeated several times after degrees of freedom are successively removed from the data by spatial filtering, temporal averaging, and consideration of smaller sectors of the hemisphere. The root mean square difference (or “rms error”) between the most closely analogous maps, defined over the middle latitudes (30–70°N), is slightly greater than half the average error between randomly chosen maps, with an estimated rms error doubling time of nearly 8 days. If the analog search is conducted using only the longwave component (zonal wavenumbers 0–4) of each map, the rms error between the best analog pairs is reduced to less than half of the rms error between long-wave anomalies on randomly chosen maps, but the doubling time is also reduced to less than 7 days. If the analog search is further restricted to a limited region over North America or Europe, the rms error between the best analog pairs is less than 40% of the rms error (for the sme region) between randomly chosen maps, but the error doubling time is further reduced to 4–5 days. In all cases, the degradation of analog quality is so rapid that a forecasting scheme based on the analogs would fail to produce more skillful forecasts than simple persistence.

Save