Journal cover Journal topic
Nonlinear Processes in Geophysics An interactive open-access journal of the European Geosciences Union
Journal topic

Journal metrics

Journal metrics

  • IF value: 1.558 IF 1.558
  • IF 5-year value: 1.475 IF 5-year
    1.475
  • CiteScore value: 2.8 CiteScore
    2.8
  • SNIP value: 0.921 SNIP 0.921
  • IPP value: 1.56 IPP 1.56
  • SJR value: 0.571 SJR 0.571
  • Scimago H <br class='hide-on-tablet hide-on-mobile'>index value: 55 Scimago H
    index 55
  • h5-index value: 22 h5-index 22
Volume 4, issue 3
Nonlin. Processes Geophys., 4, 157–166, 1997
https://doi.org/10.5194/npg-4-157-1997
© Author(s) 1997. This work is licensed under
the Creative Commons Attribution-NonCommercial-ShareAlike 2.5 License.
Nonlin. Processes Geophys., 4, 157–166, 1997
https://doi.org/10.5194/npg-4-157-1997
© Author(s) 1997. This work is licensed under
the Creative Commons Attribution-NonCommercial-ShareAlike 2.5 License.

  30 Sep 1997

30 Sep 1997

A reexamination of methods for evaluating the predictability of the atmosphere

J. L. Anderson and V. Hubeny J. L. Anderson and V. Hubeny
  • Geophysical Fluid Dynamics Laboratory, P.O.Box 308 Princeton Universtity, Princeton, NJ 08542 USA

Abstract. Pioneering work by Lorenz (1965, 1968, 1969) developed a number of methods for exploring the limits of predictability of the atmosphere. One method uses an integration of a realistic numerical model as a surrogate for the atmosphere. The evolution of small perturbations to the integration are used to estimate how quickly errors resulting from a given observational error distribution would grow in this perfect model context.
In reality, an additional constraint must be applied to this predictability problem. In the real atmosphere, only states that belong to the atmosphere's climate occur and one is only interested in how such realizable states diverge in time. Similarly, in a perfect model study, only states on the model's attractor occur. However, a prescribed observational error distribution may project on states that are off the attractor, resulting in unrepresentative error growth. The 'correct' error growth problem examines growth for the projection of the observational error distribution onto the model's attractor.
Simple dynamical systems are used to demonstrate that this additional constraint is vital in order to correctly assess the rate of error growth. A naive approach in which this information about the model's 'climate' is not used can lead to significant errors. Depending on the dynamical system, error doubling times may be either underestimated or overestimated although the latter seems more likely for more realistic models. While the magnitude of these errors is not large in the simple dynamical systems examined, the impact could be much larger in more realistic forecast models.

Publications Copernicus
Download
Citation