1 results
Chapter 20 - Weather and seasonal climate forecasts using the superensemble approach
-
- By T. N. Krishnamurti, Department of Meteorology, Florida State University, Tallahassee, T. S. V. Vijaya Kumar, Department of Meteorology, Florida State University, Tallahassee, Won-Tae Yun, Department of Meteorology, Florida State University, Tallahassee, Arun Chakraborty, Department of Meteorology, Florida State University, Tallahassee, Lydia Stefanova, Department of Meteorology, Florida State University, Tallahassee
- Edited by Tim Palmer, Renate Hagedorn
-
- Book:
- Predictability of Weather and Climate
- Published online:
- 03 December 2009
- Print publication:
- 27 July 2006, pp 532-560
-
- Chapter
- Export citation
-
Summary
In this chapter we present a short overview of the Florida State University (FSU) superensemble methodology for weather and seasonal climate forecasts and cite some examples on application for hurricanes, numerical weather prediction (NWP) and seasonal climate forecasts. This is a very powerful method for producing a consensus forecast from a suite of multimodels and the use of statistical algorithms. The message conveyed here is that the superensemble reduces the errors considerably compared with those of the member models and of the ensemble mean. This is based on results from several recent publications, where varieties of skill scores such as anomaly correlation, root-mean-square (rms) errors and threat scores have been examined. The improvements in several categories such as seasonal climate prediction from coupled atmosphere–ocean multimodels and NWP forecasts for precipitation exceed those of the best models in a consistent manner and are more accurate compared with the ensemble mean. It is difficult to state, soon after a forecast is made, as to which among the member models would have the highest skill. The superensemble is very consistent in this regard and is thus more reliable. In this study, we show walk-through tables that illustrate the workings of the superensemble for a hurricane track and heavy rain forecast for a flooding event. A number of features of the superensemble – number of training days, behaviour as the number of models increased, reduction of systematic errors and use of a synthetic superensemble – illustrate the strength of this new forecast experience.