GFDL - Geophysical Fluid Dynamics Laboratory

Long-Range Forecasting


At the beginning of my career, while working in the Meteorology Department
at Rutgers University, the project which supported me was geared towards
the prediction of short-term (monthly and seasonal) climate anomalies over the
U.S. using advanced statistical techniques. I assisted in the issuance of
operational winter season long-range U.S. temperature forecasts using a
statistical model. One objective of this multi-year project was to expand the
existing scheme to other seasons and to precipitation as well. A summary of some
of the methods and results were given in Harnack et al. (1986a/b/c) for temperature
forecasts, in Lanzante and Harnack (1982) and Harnack and Lanzante (1985) for
precipitation, and in Harnack and Lanzante (1984) for atmospheric circulation.

The basic philosophy of the schemes was to use Empirical Orthogonal Functions
(EOFs) to extract the important signal contained in predictor fields and to
use these to build statistical models for the prediction of U.S monthly
and seasonal temperature, precipitation or atmospheric circulation.
Most frequently the models were constructed using screening multiple linear
regression, however other techniques such as the “analog approach” and
discriminant analysis were employed as well. In addition we used Monte Carlo
and resampling schemes to evaluate the practical utility of our models.

The principal predictor fields included tropical and extratropical SSTs,
700 hPa heights and sea-level pressure, along with antecedent conditions of
the predictands. We also performed specification studies in which regression
models were based on zero lag (simultaneous) relationships. The idea here
was to assess an upper limit on predictive capabilities. Although this
project terminated many years ago many of the methods and relationships found
to be useful, as well as the general level of skill of the forecasts have not
changed dramatically!

Although I am no longer active in this area some of my research is geared
towards diagnostics which may have relevance towards climate prediction.
Particularly in this regard is a long-standing project related to the
discontinuous nature of the seasonal cycle of temperature (see also SEASONAL CYCLE).
I hope to demonstrate its relation to both medium and long-range forecasting.

Return to

<a href="/john-lanzante-homepage/#Details">John Lanzante's Home Page</a>.