The NCEP Climate Forecast System

Abstract

The Climate Forecast System (CFS), the fully coupled ocean–land–atmosphere dynamical seasonal prediction system, which became operational at NCEP in August 2004, is described and evaluated in this paper. The CFS provides important advances in operational seasonal prediction on a number of fronts. For the first time in the history of U.S. operational seasonal prediction, a dynamical modeling system has demonstrated a level of skill in forecasting U.S. surface temperature and precipitation that is comparable to the skill of the statistical methods used by the NCEP Climate Prediction Center (CPC). This represents a significant improvement over the previous dynamical modeling system used at NCEP. Furthermore, the skill provided by the CFS spatially and temporally complements the skill provided by the statistical tools. The availability of a dynamical modeling tool with demonstrated skill should result in overall improvement in the operational seasonal forecasts produced by CPC. The atmospheric component of the CFS is a lower-resolution version of the Global Forecast System (GFS) that was the operational global weather prediction model at NCEP during 2003. The ocean component is the GFDL Modular Ocean Model version 3 (MOM3). There are several important improvements inherent in the new CFS relative to the previous dynamical forecast system. These include (i) the atmosphere–ocean coupling spans almost all of the globe (as opposed to the tropical Pacific only); (ii) the CFS is a fully coupled modeling system with no flux correction (as opposed to the previous uncoupled “tier-2” system, which employed multiple bias and flux corrections); and (iii) a set of fully coupled retrospective forecasts covering a 24-yr period (1981–2004), with 15 forecasts per calendar month out to nine months into the future, have been produced with the CFS. These 24 years of fully coupled retrospective forecasts are of paramount importance to the proper calibration (bias correction) of subsequent operational seasonal forecasts. They provide a meaningful a priori estimate of model skill that is critical in determining the utility of the real-time dynamical forecast in the operational framework. The retrospective dataset also provides a wealth of information for researchers to study interactive atmosphere–land–ocean processes.