A STATISTICAL WARNING MODEL OF TRANS-BOARDER HAZE DISASTER: WITH REFERENCE TO REGIME SWITCHING AND LONG MEMORY

Abstract

The forest fires in Sumatra Island in Indonesia cause the haze disasters, damaging the human daily lives in the neighboring countries. Thus, the Malaysian government is now building the haze warning systems, which are designed to monitor the quantity of the air pollutants on real time, and to provide the Malaysian citizens with the precautionary warning messages in advance. In providing the warning messages, the statistical forecasting model plays an essential role in forecasting the air pollutant levels based upon the satellite data and the on-site monitoring data. In this paper, a Regime Switching Periodically Spatio-Temporal AutoRegressive Fractionally Integrated Moving Average (RS-P-STARFIMA) model is formulated to simulate the generation mechanism of the haze disaster. The applicability of the methodology presented in this paper is analyzed by an empirical case study conducted in Peninsula Malaysia.