In the previous paper, the authors showed that the AR and VAR models were applicable to analyse the data sets of seawater quality that have been measured in Osaka bay for long period of time. They pointed out, at the same time, that there were several defeats in those models, such that coefficients derived in the model are not necessarily related with the fluctuations of measured data. In the present paper, the authors proposed an improved model taking the long term trend, seasonal changes and short period fluctuation into account separately. They showed that the proposed model is applicable for analysing data sets of seawater quality measured at various points in Osaka bay.