The data on pollutants, SO
2 and suspended particulates, collected at three locations in Osaka Prefecture were analyzed by the autoregressive model and the spectra obtained (synonymous with forecasting equation) were applied to studies of the following items with satisfactory results.
1) Differences by year or season of measurment
2) Differences by location of observation
3) Differences by pollutant
4) Method for extracting a dominant frequency component in each frequency zone by changing the sampling interval
5) Method for eliminating low frequency components represented by the daily cycle and the like
6) Handling of a generating mechanism of pollutants as a statistical dynamic system
This technique was selected from statistical models for the reason that it has the following practical value in comparison with the autoregressive, moving average model which is regarded as theoretically better. The autoregressive moving average model is better than the autoregressive model in the point of providing the minimum expression of model. In the process of model building, however, the autoregressive model can determine uniquely the optimal forecasting equation by means of statistical quantity of forecasting errors called Akaike's information criterion whereas the autoregressive moving average model lacks such standards and has been regarded as a costly and hard-to-use technique. Furthermore, the autoregressive model can treat pollutants as a statistical dynamic system and can be expanded easily in the future to a multi-dimensional autoregressive model for analysis of a causal relation among pollutants.
A further study is needed to see whether or not a statistical model, in comparison with a physical model, can be put to practical use on the field as a forecasting technique, but it was concluded that a statistical model can be sufficiently practical in retrospective studies such as objective assessment of a vast quantity of data accumulated in the past by local and national govermental agencies.
抄録全体を表示