環保專案成果報告查詢系統

專案計畫編號	MOENV054112001
經費年度	111
計畫經費	3,150 千元
預算科目	空氣品質監測
專案開始日期	2022/09/01
專案結束日期	2023/09/30
專案主持人	張士昱
主辦單位	環境部監測資訊司
承辦人	陳香宇
執行單位	中山醫學大學
專案分類	大氣空氣
中文關鍵字	消光係數、散光係數、PM2.5化學成分、相對濕度
英文關鍵字	Extinction coefficient, Scattering coefficient, PM2.5 chemical component, Relative humidity
協同主持人	李崇德；曾明性；陳韡鼐
共同主持人
計畫聯絡人	江庭瑜
計畫聯絡電話	04-36097086#
計畫聯絡信箱	tiffany19971229@gmail.com

中文摘要	本計畫執行環境部能見度監測網板橋、西屯、小港三測站能見度設備操作與維護，定期完成資料品質保證與品質管制作業，整合能見度監測網與環境部空氣品質測站等監測資料，並提供光學監測儀器監測結果應用及解析。自2019至今年計畫執行期間，監測結果顯示三測站能見度呈現改善趨勢。資料分析結果顯示：(一) 依據PM2.5化學組成推估消光係數，板橋、忠明與小港測站使用大氣消光係數經驗估算式估算消光係數值比使用revised IMPROVE推估公式更符合實際消光係數值。(二) 在吸光係數的應用上，藉由環境部公告污染事件驗證CO/NOx (Carbon monoxide / Nitrogen Oxides, 一氧化碳與氮氧化物比值)、AAE (Absorption Ångström Exponent, 吸收波長指數) 與BC/PM2.5 (Black Carbon / PM2.5, 黑碳與細懸浮微粒的比值) 三維指標，三維指標污染事件辨別結果經比對環境部公告污染事件，可應用於迅速辨別境外傳輸、境內污染與交通排放事件。(三) 透過PMF (Positive Matrix Factorization, 正矩陣因子法)解析過去四年三測站污染源對能見度貢獻影響，顯示硝酸鹽污染源管制對於改善能見度極具重要性。 (四) 歷年高解析影像經人工智慧深度學習辨識能見度，採用ResNet50預訓練模型（特徵萃取器）白天影像可達八成分類準確率。(五) 使用中研院LiDAR (Light Detection And Ranging)系統進行近地層大氣觀測，可輔助觀測本計畫既有儀器LPV-4缺乏的垂直方向數據，更能補足環境部光達觀測網近地250公尺內資料的不足，並提供更高空間解析度且仰角掃描0 ~ 180度的大氣垂直剖面能見度變化監測數據。
中文關鍵字	消光係數、散光係數、PM2.5化學成分、相對濕度

中文摘要

本計畫執行環境部能見度監測網板橋、西屯、小港三測站能見度設備操作與維護，定期完成資料品質保證與品質管制作業，整合能見度監測網與環境部空氣品質測站等監測資料，並提供光學監測儀器監測結果應用及解析。自2019至今年計畫執行期間，監測結果顯示三測站能見度呈現改善趨勢。資料分析結果顯示：(一) 依據PM2.5化學組成推估消光係數，板橋、忠明與小港測站使用大氣消光係數經驗估算式估算消光係數值比使用revised IMPROVE推估公式更符合實際消光係數值。(二) 在吸光係數的應用上，藉由環境部公告污染事件驗證CO/NOx (Carbon monoxide / Nitrogen Oxides, 一氧化碳與氮氧化物比值)、AAE (Absorption Ångström Exponent, 吸收波長指數) 與BC/PM2.5 (Black Carbon / PM2.5, 黑碳與細懸浮微粒的比值) 三維指標，三維指標污染事件辨別結果經比對環境部公告污染事件，可應用於迅速辨別境外傳輸、境內污染與交通排放事件。(三) 透過PMF (Positive Matrix Factorization, 正矩陣因子法)解析過去四年三測站污染源對能見度貢獻影響，顯示硝酸鹽污染源管制對於改善能見度極具重要性。 (四) 歷年高解析影像經人工智慧深度學習辨識能見度，採用ResNet50預訓練模型（特徵萃取器）白天影像可達八成分類準確率。(五) 使用中研院LiDAR (Light Detection And Ranging)系統進行近地層大氣觀測，可輔助觀測本計畫既有儀器LPV-4缺乏的垂直方向數據，更能補足環境部光達觀測網近地250公尺內資料的不足，並提供更高空間解析度且仰角掃描0 ~ 180度的大氣垂直剖面能見度變化監測數據。

中文關鍵字

消光係數、散光係數、PM2.5化學成分、相對濕度

類型	檔名	檔案大小	說明	最新版公開日期
期末報告	定稿公開版_111-112期末成果報告.pdf	34MB		2023/11/30 上午 12:00:00

摘要	The work of this project included maintaining and operating the Ministry of Environment visibility monitoring network (VMN) in Banqiao, Xitun, and Xiaogang stations, implementing the data quality assurance and quality control, investigating the measured information between Ministry of Environment air quality monitoring stations and VMN, and providing the analysis and application of the VMN optical monitoring data. From 2019 to the implementation of the project this year, monitoring results showed that the visibility of the three measuring stations showed an improving trend. The air visibility was estimated by the visual range constant divided the extinction coefficient. The multivariate regression relationship between the extinction coefficient and regular air quality data was established to assess the feasibility for simply and rapidly simulating visibility variations. The extinction coefficient is estimated based on the chemical composition of PM2.5. Banqiao, Zhongming, and Xiaogang stations use the empirical estimation formula for the atmospheric extinction coefficient to estimate the extinction coefficient value ratio using the revised IMPROVE estimation formula It is more in line with the actual local extinction coefficient value. In the application of the absorbance coefficient, the three-dimensional indicators of CO/NOx, AAE and BC/PM2.5 are verified by the pollutant events of Ministry of Environment announcements to identify the characteristics of long-range transport pollutants, local pollutants, and traffic emissions. The positive matrix factorization (PMF) was used at three stations for source apportionment of pollution. The extinction coefficients were then reconstructed from PMF factors modified with RH to understand the source contributions of the visibility degradation in the field atmospheric environment. During the heavy pollution period in winter and spring, the highest contributor to the reconstructed extinction coefficient at three stations was the RH-modified nitrate. The results highlight the importance of nitrate pollution source control in improving visibility policies. Using high-resolution images to identify visibility, currently using the ResNet50 pre-trained model (feature extractor), most of the daytime images can achieve 80% classification accuracy. Using the LiDAR (Light Detection And Ranging) system of Academia Sinica for near-surface atmospheric observation, in addition to observing the vertical data lacking in the existing instrument LPV-4 of this project, it can also complement the 250-meter near-earth observation network of the Environmental Protection Agency. Insufficient data, and provide higher resolution monitoring data of visibility changes in the vertical profile of the atmosphere.
英文關鍵字	Extinction coefficient, Scattering coefficient, PM2.5 chemical component, Relative humidity

摘要

The work of this project included maintaining and operating the Ministry of Environment visibility monitoring network (VMN) in Banqiao, Xitun, and Xiaogang stations, implementing the data quality assurance and quality control, investigating the measured information between Ministry of Environment air quality monitoring stations and VMN, and providing the analysis and application of the VMN optical monitoring data. From 2019 to the implementation of the project this year, monitoring results showed that the visibility of the three measuring stations showed an improving trend. The air visibility was estimated by the visual range constant divided the extinction coefficient. The multivariate regression relationship between the extinction coefficient and regular air quality data was established to assess the feasibility for simply and rapidly simulating visibility variations. The extinction coefficient is estimated based on the chemical composition of PM2.5. Banqiao, Zhongming, and Xiaogang stations use the empirical estimation formula for the atmospheric extinction coefficient to estimate the extinction coefficient value ratio using the revised IMPROVE estimation formula It is more in line with the actual local extinction coefficient value. In the application of the absorbance coefficient, the three-dimensional indicators of CO/NOx, AAE and BC/PM2.5 are verified by the pollutant events of Ministry of Environment announcements to identify the characteristics of long-range transport pollutants, local pollutants, and traffic emissions. The positive matrix factorization (PMF) was used at three stations for source apportionment of pollution. The extinction coefficients were then reconstructed from PMF factors modified with RH to understand the source contributions of the visibility degradation in the field atmospheric environment. During the heavy pollution period in winter and spring, the highest contributor to the reconstructed extinction coefficient at three stations was the RH-modified nitrate. The results highlight the importance of nitrate pollution source control in improving visibility policies. Using high-resolution images to identify visibility, currently using the ResNet50 pre-trained model (feature extractor), most of the daytime images can achieve 80% classification accuracy. Using the LiDAR (Light Detection And Ranging) system of Academia Sinica for near-surface atmospheric observation, in addition to observing the vertical data lacking in the existing instrument LPV-4 of this project, it can also complement the 250-meter near-earth observation network of the Environmental Protection Agency. Insufficient data, and provide higher resolution monitoring data of visibility changes in the vertical profile of the atmosphere.

英文關鍵字

Extinction coefficient, Scattering coefficient, PM2.5 chemical component, Relative humidity