Ordinary Kalman filter (OKF) and adaptive Kalman filter (AKF) are used to analyze the long-term pattern fluctuations in a precipitation sequence. They are applied to the smoothed logarithmic-transformed monthly average precipitation sequence at Fukuoka City.
An abnormality detection index φ calculated by OKF is used to detect the abnorma precipitation period and to quantitatively estimate the magnitude of the abnormality. The structure of these abnormal periods is studied.
The system parameters of the precipitation sequence are identified by AKF. The time of occurrence of parameter change and magnitude of change are estimated. The sequence can then be divided into several periods, where each period consists of the same parameters. Characteristics of the precipitation pattern in each period divided by AKF and of the change from one period to the other are investigated.