A comprehensive analysis of seasonal predictability for weather extremes and daily forecast for air pollution in Hong Kong, China

Abstract

Nowadays the society is very concerned about reliable forecast for extreme weathers and air pollution since it could significantly facilitate the resources management and the planning of precautionary measures. This study conducts a comprehensive analysis of feasibility to provide seasonal forecast for weather extremes, as well as daily forecast for air quality in Hong Kong. The existing forecast methods worldwide are reviewed and the forecast models for Hong Kong are proposed. At last the understanding on the mechanism of the local weather extremes is sharpened. First of all, the high-humidity weather is noticed when the warm air from South China Sea and the cold air from north reach a balance in south China. The weather event is identified as high-humidity days with relative humidity above 90 and no rainfall in the period from February to April. Then the characteristics of meteorological conditions during the high-humidity days are studied. In the last, the multiple regression and classification and regression tree (CART) models are built to forecast the number of high-humidity days in Hong Kong. The results show that both the local weather conditions and large-scale climate patterns have a great impact on the occurrences of high-humidity days. Secondly, the characteristics of tropical cyclones (TCs) affecting Hong Kong are explored. The Spectral analysis is conducted to disclose the relationship between the TC activities in western North Pacific (WNP) and El Niño Southern Oscillation (ENSO). Then through the statistical analysis, the frequency, intensity, track and TC-induced rainfall under different ENSO phases (namely, El Niño, La Niña and normal phases) are explored. Based on the understanding, the forecasting for the number of TCs affecting Hong Kong is carried out by climatological probability matrix and CART modelling. The performance of two methods is good in the validation step, while the CART models can catch more of the complex interaction between the meteorological variables and climate patterns. In addition, the forecasting for the onset of the TC season in Hong Kong is also provided. Then, the variability of the annual rainfall and summer rainfall in Hong Kong is studied. The effects of ENSO on the annual rainfall and NAO on the summer rainfall are explored by statistical analysis and the physical explanation. Then the forecasting of the rainfall in Hong Kong is carried out by climatological probability matrix and CART modelling. The performances of the forecasting models are evaluated accordingly. The overall accuracy reaches 70% with the best agreement in extreme rainfall cases. Last but not least, the weather conditions favorable to air pollution in Hong Kong are studied and the long-term trends of different air pollutants are investigated. Based on the statistical analysis, the key air pollutants are identified and the current air quality reporting system in Hong Kong, namely the air quality health index (AQHI), is reviewed. In addition, the variation of AQHI is found to follow different rules in terms of different seasons. The CART forecasting models for daily AQHI has been built at the seasonal scale.