Analysis of meteorology and emission in haze episode prevalence over mountain-bounded region for early warning.

This study investigated the main causes of haze episodes in the northwestern Thailand to provide early warning and prediction. In an absence of emission input data required for chemical transport modeling to predict the haze, the climatological approach in combination with statistical analysis was used. An automatic meteorological classification scheme was developed using regional meteorological station data of 8years (2001-2008) which classified the prevailing synoptic patterns over Northern Thailand into 4 patterns. Pattern 2, occurring with high frequency in March, was found to associate with the highest levels of 24h PM(10) in Chiangmai, the largest city in Northern Thailand. Typical features of this pattern were the dominance of thermal lows over India, Western China and Northern Thailand with hot, dry and stagnant air in Northern Thailand. March 2007, the month with the most severe haze episode in Chiangmai, was found to have a high frequency of occurrence of pattern 2 coupled with the highest emission intensities from biomass open burning. Backward trajectories showed that, on haze episode days, air masses passed over the region of dense biomass fire hotspots before arriving at Chiangmai. A stepwise regression model was developed to predict 24h PM(10) for days of meteorology pattern 2 using February-April data of 2007-2009 and tested with 2004-2010 data. The model performed satisfactorily for the model development dataset (R(2)=87%) and test dataset (R(2)=81%), which appeared to be superior over a simple persistence regression of 24h PM(10) (R(2)=76%). Our developed model had an accuracy over 90% for the categorical forecast of PM(10)>120μg/m(3). The episode warning procedure would identify synoptic pattern 2 and predict 24h PM(10) in Chiangmai 24h in advance. This approach would be applicable for air pollution episode management in other areas with complex terrain where similar conditions exist.