|
論文名稱
|
使用空間混合模型分析PM2.5的長期變化趨勢(1994年至2020年)－以臺中市為例
|
|
Title
|
The Long-term Trend Analysis of PM2.5 Variability From 1994 to 2020 Using a Hybrid Spatial Model: A Case Study of Taichung City, Taiwan
|
|
作者
|
林祐如、許家瑋、李佳禾、曾于庭、翁佩詒、陳保中、陳裕政、吳治達
|
|
Author
|
Yu-Ju Lin, Chia-Wei Hsu, Chia-Ho Lee, Yu-Ting Zeng, Pei-Yi Wong, Pau-Chung Chen, Yu-Cheng Chen, Chih-Da Wu
|
|
中文摘要
|
研究以臺中市為例，利用空間模型推估1994至2020年PM2.5濃度趨勢，並評估城市開發對空氣品質的影響。研究使用PM2.5相關汙染物、氣象資料、土地利用、地標、路網、地形、植生指數等作為預測變數。結合土地利用迴歸和機器學習方法，使用隨機森林、梯度提升機、極限梯度提升、輕量梯度提升機和基於梯度提升的決策樹模型擬合預測模型。通過數據拆分、十折交叉和外部驗證確認模型穩健性，結果顯示模型穩定且可信，Adj-R2為0.93。結果表明多數地點的「年份」變數係數為負，顯示過去25年空氣污染顯著改善。研究強調在城市開發規劃中管理和控制空氣污染的重要性。
|
|
Abstract
|
This study takes Taichung City as an example and aims to investigate the long-term impact of urban development on air pollution. By establishing a spatial model, we estimate the concentration trends of fine particulate matter (Particulate Matter 2.5, PM2.5) over the past 25 years (from 1994 to 2020) and further assess the influence of urban development on air quality. Various databases were utilized as sources of spatial predictor variables, including the Environmental Resources Database, meteorological database, land-use inventory, landmark database, digital road network map, digital terrain model, MODIS Normalized Difference Vegetation Index (NDVI) database, and power plant distribution database. The spatial hybrid model in this study combines Hybrid Kriging/Land-Use Regression and machine learning methods. Initially, important predictor variables were determined using traditional Land-Use Regression (LUR) and Hybrid Kriging-LUR. Subsequently, prediction models based on the selected variables from LUR models were fitted using Random Forest (RF), Gradient Boosting Machine (GBM), eXtreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (Light GBM), and CatBoost algorithms. Validation methods such as data splitting, 10-fold cross-validation, and external data verification were employed to confirm the robustness of the developed models. The results indicate that the model is stable and reliable, with an Adj-R2 of 0.93. Through linear regression, it was observed that the estimated values of the predictor variable ‘year’ for most locations in the city are negative, indicating a significant improvement in air pollution over the past 25 years. This study emphasizes the importance of managing and controlling air pollution in urban development planning.
|
|
關鍵字
|
PM2.5、都市開發、空間混合模型、趨勢分析
|
|
Keywords
|
PM2.5, Urban Development, Hybrid Spatial Model, Trend Analysis
|
|
|
|