Predictability of Week-3–4 Average Temperature and Precipitation over the Contiguous United States

Abstract

Abstract This paper demonstrates that an operational forecast model can skillfully predict week-3–4 averages of temperature and precipitation over the contiguous United States. This skill is demonstrated at the gridpoint level (about 1° × 1°) by decomposing temperature and precipitation anomalies in terms of an orthogonal set of patterns that can be ordered by a measure of length scale and then showing that many of the resulting components are predictable and can be predicted in observations with statistically significant skill. The statistical significance of predictability and skill are assessed using a permutation test that accounts for serial correlation. Skill is detected based on correlation measures but not based on mean square error measures, indicating that an amplitude correction is necessary for skill. The statistical characteristics of predictability are further clarified by finding linear combinations of components that maximize predictability. The forecast model analyzed here is version 2 of the Climate Forecast System (CFSv2), and the variables considered are temperature and precipitation over the contiguous United States during January and July. A 4-day lagged ensemble, comprising 16 ensemble members, is used. The most predictable components of winter temperature and precipitation are related to ENSO, and other predictable components of winter precipitation are shown to be related to the Madden–Julian oscillation. These results establish a scientific basis for making week-3–4 weather and climate predictions.