Forecasting the NAO: December through February Period

[donsutherland1]

Many of the approaches employed for seasonal forecasting follow some combination of lagged SSTAs with some input for Northern Hemisphere snowcover.

http://www.metoffice.com/research/seaso ... ional/nao/

My hypothesis suggests that such an approach is too simplistic. Rather, a global look should produce better results. For one reason, just as teleconnections play a role in ridge-trough positions, I believe that there is some teleconnective impact with regard to the ridge-trough positions attributable to the North Atlantic Oscillation (NAO).

After examining a large number of global indices, I have taken a small number of them and tested whether they would forecast an NAO that averaged below or above 0 for the December-February period.

The results are below:

Code: Select all

Winter   Forecast   Actual
1950-51   Positive   0.467
1951-52   Positive   0.767
1952-53   Positive   0.067
1953-54   Positive   0.700
1954-55   Negative   -0.033
1955-56   Positive   -0.467
1956-57   Positive   0.733
1957-58   Negative   -0.700
1958-59   Negative   -0.400
1959-60   Negative   -0.400
1960-61   Positive   0.600
1961-62   Positive   0.033
1962-63   Negative   -0.700
1963-64   Negative   -1.333
1964-65   Positive   -0.333
1965-66   Negative   -0.467
1966-67   Positive   0.600
1967-68   Negative   -0.300
1968-69   Negative   -0.967
1969-70   Negative   -0.200
1970-71   Positive   -0.767
1971-72   Positive   0.067
1972-73   Positive   0.800
1973-74   Positive   0.533
1974-75   Negative   0.000
1975-76   Positive   0.033
1976-77   Negative   -1.233
1977-78   Negative   -0.567
1978-79   Negative   -1.333
1979-80   Negative   0.067
1980-81   Positive   1.000
1981-82   Negative   -0.400
1982-83   Positive   0.700
1983-84   Positive   1.133
1984-85   Negative   -0.300
1985-86   Positive   0.100
1986-87   Positive   -0.300
1987-88   Positive   0.300
1988-89   Positive   0.733
1989-90   Negative   0.000
1990-91   Positive   0.867
1991-92   Positive   0.133
1992-93   Positive   0.567
1993-94   Positive   0.767
1994-95   Positive   0.900
1995-96   Negative   -0.467
1996-97   Positive   -0.067
1997-98   Negative   -0.567
1998-99   Positive   1.033
1999-00   Positive   1.200
2000-01   Negative   0.400
2001-02   Positive   0.333
2002-03   Negative   -0.300
2003-04   Positive   -0.100

Errors occurred in the following winters:

1955-56, 1964-65, 1970-71, 1974-75, 1979-80, 1986-87, 1989-90, 1996-97, 2000-01, and 2003-04

Consequently, in 44 of 54 cases (81.5%), the global approach correctly forecast whether the NAO would average above or below zero during the December-February period.

The link for one site that carries a seasonal NAO forecast pointed to a success rate of 66%. If this preliminary look at a global approach continues to bear out, it would mark a qualitative improvement over the simpler approaches, as it had an accuracy rate 23.5% above the 66% rate noted.

Winter 2004-05 could be a first test. The conventional approach points to a positive NAO. Right now, the global approach is pointing to an NAO that would average below 0 for the winter and that idea should be finalized by mid-November.