Scale-Selective Ridge Regression for Multimodel Forecasting

Timothy DelSole George Mason University, Fairfax, Virginia, and Center for Ocean-Land-Atmosphere Studies, Calverton, Maryland

Search for other papers by Timothy DelSole in
Current site
Google Scholar
PubMed
Close
,
Liwei Jia Center for Ocean-Land-Atmosphere Studies, Calverton, Maryland

Search for other papers by Liwei Jia in
Current site
Google Scholar
PubMed
Close
, and
Michael K. Tippett International Research Institute for Climate and Society, Palisades, New York, and Center of Excellence for Climate Change Research, Department of Meteorology, King Abdulaziz University, Jeddah, Saudi Arabia

Search for other papers by Michael K. Tippett in
Current site
Google Scholar
PubMed
Close
Restricted access

Abstract

This paper proposes a new approach to linearly combining multimodel forecasts, called scale-selective ridge regression, which ensures that the weighting coefficients satisfy certain smoothness constraints. The smoothness constraint reflects the “prior assumption” that seasonally predictable patterns tend to be large scale. In the absence of a smoothness constraint, regression methods typically produce noisy weights and hence noisy predictions. Constraining the weights to be smooth ensures that the multimodel combination is no less smooth than the individual model forecasts. The proposed method is equivalent to minimizing a cost function comprising the familiar mean square error plus a “penalty function” that penalizes weights with large spatial gradients. The method reduces to pointwise ridge regression for a suitable choice of constraint. The method is tested using the Ensemble-Based Predictions of Climate Changes and Their Impacts (ENSEMBLES) hindcast dataset during 1960–2005. The cross-validated skill of the proposed forecast method is shown to be larger than the skill of either ordinary least squares or pointwise ridge regression, although the significance of this difference is difficult to test owing to the small sample size. The model weights derived from the method are much smoother than those obtained from ordinary least squares or pointwise ridge regression. Interestingly, regressions in which the weights are completely independent of space give comparable overall skill. The scale-selective ridge is numerically more intensive than pointwise methods since the solution requires solving equations that couple all grid points together.

Current affiliation: NOAA/Geophysical Fluid Dynamics Laboratory, Princeton, New Jersey, and University Corporation for Atmospheric Research, Boulder, Colorado.

Corresponding author address: Timothy DelSole, Center for Ocean-Land-Atmosphere Studies, 4041 Powder Mill Rd., Suite 302, Calverton, MD 20705. E-mail: delsole@cola.iges.org

Abstract

This paper proposes a new approach to linearly combining multimodel forecasts, called scale-selective ridge regression, which ensures that the weighting coefficients satisfy certain smoothness constraints. The smoothness constraint reflects the “prior assumption” that seasonally predictable patterns tend to be large scale. In the absence of a smoothness constraint, regression methods typically produce noisy weights and hence noisy predictions. Constraining the weights to be smooth ensures that the multimodel combination is no less smooth than the individual model forecasts. The proposed method is equivalent to minimizing a cost function comprising the familiar mean square error plus a “penalty function” that penalizes weights with large spatial gradients. The method reduces to pointwise ridge regression for a suitable choice of constraint. The method is tested using the Ensemble-Based Predictions of Climate Changes and Their Impacts (ENSEMBLES) hindcast dataset during 1960–2005. The cross-validated skill of the proposed forecast method is shown to be larger than the skill of either ordinary least squares or pointwise ridge regression, although the significance of this difference is difficult to test owing to the small sample size. The model weights derived from the method are much smoother than those obtained from ordinary least squares or pointwise ridge regression. Interestingly, regressions in which the weights are completely independent of space give comparable overall skill. The scale-selective ridge is numerically more intensive than pointwise methods since the solution requires solving equations that couple all grid points together.

Current affiliation: NOAA/Geophysical Fluid Dynamics Laboratory, Princeton, New Jersey, and University Corporation for Atmospheric Research, Boulder, Colorado.

Corresponding author address: Timothy DelSole, Center for Ocean-Land-Atmosphere Studies, 4041 Powder Mill Rd., Suite 302, Calverton, MD 20705. E-mail: delsole@cola.iges.org
Save
  • Barnston, A. G., and H. M. van den Dool, 1993: A degeneracy in cross-validated skill in regression-based forecasts. J. Climate, 6, 963977.

    • Search Google Scholar
    • Export Citation
  • DelSole, T., 2007: A Bayesian framework for multimodel regression. J. Climate, 20, 28102826.

  • DelSole, T., and J. Shukla, 2006: Specification of wintertime North America surface temperature. J. Climate, 19, 26912716.

  • Hastie, T., R. Tibshirani, and J. H. Friedman, 2003: Elements of Statistical Learning. Corrected ed. Springer, 552 pp.

  • Kistler, R., and Coauthors, 2001: The NCEP–NCAR 50-Year Reanalysis: Monthly means CD-ROM and documentation. Bull. Amer. Meteor. Soc., 82, 247267.

    • Search Google Scholar
    • Export Citation
  • Peña, M., and H. van den Dool, 2008: Consolidation of multimodel forecasts by ridge regression: Application to Pacific surface temperature. J. Climate, 21, 65216538.

    • Search Google Scholar
    • Export Citation
  • Penland, C., and P. D. Sardeshmukh, 1995: The optimal-growth of tropical sea-surface temperature anomalies. J. Climate, 8, 19992024.

  • Phelps, M. W., A. Kumar, and J. J. O'Brian, 2004: Potential predictability in the NCEP CPC dynamical seasonal forecast system. J. Climate, 17, 37753785.

    • Search Google Scholar
    • Export Citation
  • Press, W. H., S. A. Teukolsky, W. T. Vetterling, and B. P. Fannerty, 1992: Numerical Recipes. Cambridge University Press, 693 pp.

  • Robertson, A. W., U. Lall, S. E. Zebiak, and L. Goddard, 2004: Improved combination of multiple atmospheric GCM ensembles for seasonal prediction. Mon. Wea. Rev., 132, 27322744.

    • Search Google Scholar
    • Export Citation
  • van den Dool, H., and L. Rukhovets, 1994: On the weights for an ensemble-averaged 6–10-day forecast. Wea. Forecasting, 9, 457465.

  • Weisheimer, A., and Coauthors, 2009: ENSEMBLES: A new multi-model ensemble for seasonal-to-annual prediction—Skill and progress beyond DEMETER forecasting tropical Pacific SSTs. Geophys. Res. Lett.,36, L21711, doi:10.1029/2009GL040896.

  • Yun, W. T., L. Stefanova, and T. N. Krishnamurti, 2003: Improvement of the multimodel superensemble technique for seasonal forecasts. J. Climate, 16, 38343840.

    • Search Google Scholar
    • Export Citation
  • Zwiers, F. W., 1987: A potential predictability study conducted with an atmospheric general circulation model. Mon. Wea. Rev., 115, 29572974.

    • Search Google Scholar
    • Export Citation
All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 255 88 6
PDF Downloads 148 61 3