A new statistical modeling and optimization framework for establishing high-resolution PM10 emission inventory – I. Stepwise regression model development and application

Abstract

A new stepwise regression method was proposed in this study to develop a high-resolution emission inventory. Utilizing PM10 emission inventory as an example, a group of regression models for various industrial and non-industrial sectors were developed based on an emission case study of Handan region in northern China. The main data requirements of the regression models for industrial sectors were coal consumption, electricity consumption, other solid fuel consumption, and annual operating cost of exhaust gas control devices. The data requirements for non-industrial sector emission estimations were the area of construction sites, the length of transportation routes, the vehicle population, and the cultivated land area. The models were then applied to Tangshan region in northern China, and the results revealed that the developed regression models had relatively satisfactory performance. Modeling error at the regional level and county level was 17.0% and 30.4%, respectively. The regression models were also applied to other regions in northern China. The results indicated that the new method could generate emission estimations with significantly lower error than found in previous emission inventory studies. The modeling uncertainty due to the allocation of modeling input parameter value, from regional level to county level, was also discussed in this study. It was concluded that the new statistical method presented is a promising technique for the development and updating of high-resolution emission inventories based on easily obtained statistical data. It can be performed with data available from the current statistical reporting system in China. It does not require a detailed data investigation and survey, as is necessary by conventional “bottom-up” emission inventory investigation approach.