IMPROVEMENT OF A BOTTOM-UP MODEL FOR ESTIMATING CO₂ EMISSIONS IN RESIDENTIAL SECTORS BY MUNICIPALITY: COMPLEMENTATION OF MISSING DATA AND APPLICATION OF NON LINEAR REGRESSION

Abstract

 This paper aims to improve models estimating CO₂ emissions in the residential sectors of municipalities. A model with higher fitness was constructed by applying missing data processing using the multiple imputation method and nonlinear multiple regression. Further, a case study was conducted for Utsunomiya City, considering the usability of the model in local governments. As a result of the model improvements, the Akaike's Information Criterion (AIC) was improved over the existing models in all regional divisions and the rationality of the positive and negative signs of the regression coefficient was also improved. The confidence intervals of the regression coefficients were also narrowed. Due to this change, the difference between the national average value of CO₂ emissions per household calculated from the bottom up using the model parameters and statistics of each municipality and the national average value published on the portal site for Japanese government statistics shrunk from approximately 7% in the existing model to approximately 4% in the improved model. Preliminary results of sensitivity analysis for Utsunomiya City indicated the following. If the average household size decreases from the current value of 2.62 to 2.00, CO₂ emissions per household will decrease by approximately 8% but CO₂ emissions per capita will increase by approximately 9%. If the percentage of households living in detached dwellings increases from 60% to 80% of the current value, CO₂ emissions per household will increase by approximately 3%.