A Bias in Skill in Forecasts Based on Analogues and Antilogues

H. M. van den Dool Department of Meteorology, University of Maryland, College Park, MD 20742

Search for other papers by H. M. van den Dool in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

A bias in skill may exist in statistical forecast methods in which the verification datum is withheld from the developmental data (cross-validation methods). Under certain circumstances this bias in skill can become troublesome. By way of example, it is shown that the judgment of the quality of forecasts based on analogues and anti-analogues may severely suffer from a bias in skill. A cure to the problem is discussed. Some implications for published results of long-range weather forecasting models based on analogues are discussed.

Abstract

A bias in skill may exist in statistical forecast methods in which the verification datum is withheld from the developmental data (cross-validation methods). Under certain circumstances this bias in skill can become troublesome. By way of example, it is shown that the judgment of the quality of forecasts based on analogues and anti-analogues may severely suffer from a bias in skill. A cure to the problem is discussed. Some implications for published results of long-range weather forecasting models based on analogues are discussed.

Save