| Summary: | 碩士 === 國立成功大學 === 環境醫學研究所 === 101 === As people spend most of their time in different buildings, indoor temperature is thought to be better associated with human health performance compared to ambient levels. Current study is aimed to establish the statistical model to predict indoor temperature based on outdoor information from atmospheric monitoring stations. Several factors including building characteristics, outdoor climate data and adaptive behavior of occupants, such as energy consumption through using air conditioner, are further evaluated.
Thirty homes, where people spend about 73% of their daily time, were selected from 11 cities in Taiwan. Real-time monitors for temperature were set inside the rooms where occupants spent most of their time when inside the house. Building characteristics, human activities and electricity consumption data were recorded weekly, and corresponding outdoor climate data from local station was collected accordingly. Mixed effect model has been conducted for model establishment after adjusting for above-mentioned determinants. The model is further cross-validated afterwards.
Study buildings are with average age and area of 22.3 years and 121 m2, respectively, and 67% of them are below five-floor-high. During the period of investigation from September 2012 to July 2013, average indoor temperature was 25.2°C (range from 14.0 to 35.0°C) while outdoors was 23.4°C (range from 8.0 to 36.3°C). Correlation coefficient between hourly indoor and outdoor temperatures is 0.84 (p〈 0.001) and varies by month (range from 0.4 to 0.68) and city (range from 0.55 to 0.87). Four-hour lag of outdoor temperature is found to be with the best estimation compared to the in-time and other lag ones. Information collected from ambient station including temperature, relative humidity, atmospheric pressure, wind speed, sunshine hours and wind direction are significant predictors associated with the variation of indoor temperature. Building characteristics including number of windows and floors were related to the outcome of interest as well. Factors such as energy consumption of electricity and behavior records of occupants were yet to result in the significant effects on the change of indoor temperature which might due to the short study period and limit sample size. Above-mentioned variables with significant effects were prioritized to consider in the establishment of model fixed with the variables of month of investigation, city and interaction term of outdoor temperature × month of investigation. As to the model for whole Taiwan, the 4-hour-lag outdoor temperature explains 36.2% (R2) of the variance for the in-time indoor temperature. When taking building characteristics into account, it’s found that every increase of number of window is associated with an increase of 0.22°C for indoor temperature (R2 = 36.8%). In order to enhance the predictive ability and availability of the model, we further separated the dataset into northern, central and southern region defined by latitudes which reflected geographical climatic differences. Results of the northern model (Hsinchu, Taipei, Keelung and Yilan) show that every increment of one floor number is associated with an increase of 0.30°C for indoor temperature. As to the model for central region (Chiayi, Taichung, and Hualien) and southern region (Tainan, Kaohsiung, Hengchuen/Pingtung and Taitung) were yet to result in the significant effects on the change of indoor temperature. Overall, the model established in northern Taiwan shown with the highest ability to explain the variation of indoor temperature (R2=44%), followed by models in central (R2=36.4%) and south area (R2=33.8%). Cross-validation is conducted afterward to confirm the predictive ability of models. Correlations between predicted and observed values of whole island, and northern, central and southern area are 0.73, 0.88, 0.75 and 0.70, respectively, while average difference are -1.85,-1.14,-1.53 and -2.15°C, respectively. Lower predictive ability during May to July and September to November might be related to frequently adaptive behavior of occupants, such as using air conditioner, which resulted into higher level of difference between indoor and outdoor temperature.
In Taiwan, the effect of outdoor temperature on the indoors seems to present apparent delay, and their associations are affected by the heat retention of building materials as well as occupants’ adaptive behaviors. In this study, at most, a total of 44% of the variance of indoor temperature can be explained by our model, depending on the region of interest, and the prediction will be more accurate during the period of December to April. The model can be used as a means of mapping city- or country-wide indoor temperatures under various scenarios of outdoor environments, to provide a better representation of individuals’ exposure levels in ecological studies examining the diversity of regional impaction between global warming and health effects. Furthermore, the model will be considered as the critical basis for the early warning system, through the prediction of indoor temperature in advance, to prevent potential health risk of susceptible group due to the sudden changes of the temperature.
|
|---|
