Forecast of PM10 time-series data: A study case in Caribbean cities

Özet

PM10 is one of the most important environmental problems in urban areas that may have a significant impact on human health. Therefore, forecasting is an important tool for warning and mitigation procedures for environmental control. PM10 data of 24 h mean concentration collected over 16 years from the air quality monitoron stations of several carribean cities of the northern coast of Colombia were analyzed. The aim of this article were the (i) estimation of missing observations time-series with intervention analysis, (ii) adjustment and selection of the best model, and (iii) forecast model for the PM10 concentrations in the study area. Data was used for testing the SARIMA model (mathematical model for Seasonal Auto-Regressive Integrated Moving Average) in order to forecast PM10 levels considering the time-series intervention analysis. This model was used for missing data intervention, as well. Moreover, a methodology was developed based on the assignment of extreme data and missing observations. Short-term forecasts, for the year 2018, were made from the modified SARIMA model and validated with monthly data of PM10 of the same year. The overall errors, measured by means of Root of the Mean Square Error (RMSE), was found to be almost zero. The primary contribution of this paper is that the developed approach significantly improves the forecasting accuracy of PM10