Application of the Kalman Filter to Cyclone Forecasting 3. Hurricane Forecasting 4. Additional Typhoon Forecasting

Takeuchi, K. (1976). Application of the Kalman Filter to Cyclone Forecasting 3. Hurricane Forecasting 4. Additional Typhoon Forecasting. IIASA Research Memorandum. IIASA, Laxenburg, Austria: RM-76-062

Preview

Text RM-76-062.pdf Download (932kB) | Preview

Abstract

This is the second part of a report on application of the Kalman filter to cyclone forecasting. Following the preliminary experiments of typhoon forecasting, this paper presents the results of hurricane experiments and further typhoon experiments.

The 12 and 24 hour forecasting NHC72 model and the 24 hour forecasting SNT model developed by the National Hurricane Center, NOAA, USA and the Japan Meteorological Agency, respectively, were examined. The improvements obtained by using the Kalman filter over the original models were found to be roughly 10% for hurricane forecasting and 20% for 24 hour typhoon forecasting, on the average, in terms of vector errors.

The conclusion drawn by the previous experiments was reconfirmed. That is, the application of the Kalman filter to utilize better simple linear regression models is effective when the original regression model gives consecutively biased forecasts for a considerably long time; it is not effective when the performance of the original model is poor, yet its residual errors are not highly correlated.

In addition to this conclusion, a statistical test of the validity of forecasting regression models showed that the structure of the model should be further improved before considering application of the Kalman filter.

Actions (login required)

View Item