Forecasting tropical cyclone landfall using ECMWF's seasonal forecasts from System4

TitleForecasting tropical cyclone landfall using ECMWF's seasonal forecasts from System4
Publication TypeMiscellaneous
Year of Publication2017
AuthorsBergman, DL, Magnusson, L, Nilsson, J, Vitart, F
Secondary TitleTechnical Memorandum
Number811
Abstract

 

Forecasting tropical cyclone landfall using ECMWF’s seasonal forecasts from System 4 Abstract The European Centre for Medium-Range Weather Forecast’s seasonal forecast system 4 (System 4) has been used to issue basin-wide tropical cyclone forecasts since 2011.   This report describes a method developed to forecast seasonal landfall risk using the ensembles of cyclone tracks generated by System 4.  The method has been applied to analyze and retrospectively forecast the landfall risk along different segments of the North American coast, with a focus on the U.S. part.

The main result is that the method can be used to forecast landfall for some parts of the coast, but the method’s skill is, generally speaking, lower for landfall than for basin-wide forecasts of activity. The rank correlations between forecast and observation are 0.6 for basin-wide storm number, 0.5 for landfall anywhere along the coast, and 0.3 for landfall along the U.S. part of the coast. When we limit the forecast to the peak of the hurricane season (August, September, and October), the correlation increases to 0.6 for the entire coast whereas it remains close to 0.3 for the U.S. part.

The forecast error is substantial in all cases, in part due to model error, in part due to the chaotic dynamics embedded in the climate system.  A crude analysis suggests that the forecast error can be reduced by 10 to 25 percent (depending on the forecast) before reaching the ultimate limit set by the chaotic dynamics.  In conclusion, the quality of the forecasts is well in line with that obtained using other state-of-the-art methods, and it is sufficient to be of use for organizations, such as reinsurance companies, that plan and operate with a statistical mind-set on a multi-year horizon.