城市內(nèi)部尺度PM2.5傳輸關(guān)聯(lián)方法研究—以北京市為例

尹 浩,胡冬梅,閆雨龍,彭 林,王 凱,張可可,鄧萌杰

城市內(nèi)部尺度PM2.5傳輸關(guān)聯(lián)方法研究—以北京市為例

尹 浩,胡冬梅*,閆雨龍,彭 林,王 凱,張可可,鄧萌杰

(華北電力大學(xué)環(huán)境科學(xué)與工程學(xué)院,資源環(huán)境系統(tǒng)優(yōu)化教育部重點(diǎn)實(shí)驗(yàn)室,北京 102206)

基于數(shù)據(jù)驅(qū)動思想,以城市內(nèi)部環(huán)境空氣質(zhì)量監(jiān)測站點(diǎn)為研究對象,建立了目標(biāo)站點(diǎn)與周邊站點(diǎn)間PM2.5濃度、風(fēng)向、風(fēng)速、歐幾里得距離等參數(shù)的多元線性關(guān)聯(lián)回歸模型,使用梯度下降算法學(xué)習(xí)得到各參數(shù)權(quán)重系數(shù),計(jì)算得出周邊站點(diǎn)對目標(biāo)站點(diǎn)PM2.5傳輸貢獻(xiàn),并評估了模型的可行性.以北京市豐臺花園(FT)為目標(biāo)站點(diǎn)的應(yīng)用研究結(jié)果顯示,2016年FT站點(diǎn)PM2.5濃度為82μg/m3,周邊站點(diǎn)大興(DX)、房山(FS)、亦莊(YZ)、東四環(huán)(DS)、古城(GC)和萬柳(WL)濃度分別為93,82,80,79,77,71μg/m3; FT站點(diǎn)PM2.5濃度與上一時刻周邊站點(diǎn)WL、GC、DX、YZ的相關(guān)性分別為0.634、0.631、0.608和0.601,顯示其對FT站點(diǎn)PM2.5污染傳輸顯著;建立的4個季節(jié)關(guān)聯(lián)回歸模型RMSE值分別為13.22、11.74、12.51和13.22, PM2.5模擬濃度與監(jiān)測濃度變化趨勢一致,驗(yàn)證了模型的可行性;WL、DX、YZ、GC分別是對應(yīng)春、夏、秋、冬4個季節(jié)對FT 站點(diǎn)PM2.5污染傳輸貢獻(xiàn)較大的站點(diǎn),其貢獻(xiàn)值分別為1.61%、1.71%、2.20%和8.57%.該模型解析的結(jié)果可為北京市未來城市規(guī)劃、建設(shè)提供依據(jù),提出的PM2.5傳輸多元線性關(guān)聯(lián)回歸方法同樣可用來解析其他城市內(nèi)部尺度PM2.5傳輸關(guān)聯(lián),為挖掘城市內(nèi)部PM2.5傳輸路徑、精準(zhǔn)溯源提供基礎(chǔ).

城市內(nèi)部尺度；PM2.5；多元線性關(guān)聯(lián)回歸；傳輸貢獻(xiàn)

研究表明,城市某地區(qū)PM2.5濃度受自身上一時刻污染物濃度積累、周邊區(qū)域濃度、氣象條件和歐幾里得距離等多種因素影響[1-6], PM2.5濃度預(yù)測需綜合考慮眾多要素. 目前,對于城市PM2.5污染傳輸研究,主要分為數(shù)值分析、氣象場分析和統(tǒng)計(jì)分析3大類,其中Models-3/CMAQ[7]、WRF-Chem[8]和軌跡分析法[9]運(yùn)用相對較多.薛文博等[10]利用CAMx空氣質(zhì)量模型的顆粒物來源追蹤技術(shù)(PSAT)定量模擬得到全國PM2.5跨區(qū)域輸送規(guī)律; 齊孟姚[11]利用WRF-Chem模擬并得到了河北南部城市邊界的PM2.5通量,并由此推導(dǎo)出模擬期間PM2.5的主要傳輸路線及通道區(qū)域;任傳斌等[12]利用HYSPLIT后向軌跡模型分析了北京市不同輸送途徑的空間特征及其對北京城區(qū)PM2.5聚集的貢獻(xiàn);余創(chuàng)等[13]利用氣流后向軌跡聚類分析法研究了銀川市PM2.5的輸送路徑及潛在源分布.但此類模型主要適用于大尺度城市間的污染傳輸研究,運(yùn)用在小尺度傳輸研究中可能會出現(xiàn)流場解析不準(zhǔn)確、化學(xué)機(jī)制不確定性大等因素導(dǎo)致結(jié)果存在較大不確定性問題[14].

本文綜合考慮環(huán)境空氣質(zhì)量監(jiān)測站點(diǎn)污染物濃度、風(fēng)向、風(fēng)速、歐幾里得距離等影響污染傳輸?shù)臅r空因素特征,建立基于時空傳輸特征的城市內(nèi)部尺度PM2.5傳輸關(guān)聯(lián)研究方法.以北京市豐臺花園站點(diǎn)為目標(biāo)站點(diǎn),以其周邊相關(guān)聯(lián)的6個站點(diǎn)為傳輸變量,研究不同季節(jié)6個關(guān)聯(lián)站點(diǎn)對豐臺花園站點(diǎn)PM2.5的污染傳輸貢獻(xiàn),旨在為挖掘城市內(nèi)部PM2.5污染傳輸路徑、精準(zhǔn)溯源提供基礎(chǔ).

1 數(shù)據(jù)來源及方法建立

1.1 數(shù)據(jù)來源

圖1 7個空氣質(zhì)量站點(diǎn)空間分布

選取北京市2016年7個空氣質(zhì)量站點(diǎn)PM2.5小時濃度數(shù)據(jù),7個氣象監(jiān)測站點(diǎn)的風(fēng)向、風(fēng)速小時數(shù)據(jù)以及站點(diǎn)經(jīng)緯度數(shù)據(jù).其中環(huán)境空氣PM2.5濃度數(shù)據(jù)來源于北京市環(huán)境監(jiān)測中心站實(shí)時數(shù)據(jù)(http://zx.bjmemc.com.cn),共計(jì)69656條.氣象數(shù)據(jù)來源于國家氣象科學(xué)數(shù)據(jù)中心(http://data.cma.cn/),共計(jì)122990條,站點(diǎn)經(jīng)緯度數(shù)據(jù)來源于高德地圖.對于PM2.5濃度、氣象數(shù)據(jù)中檢測到的異常、缺失值采用基于時間序列的拉格朗日插值法填充處理[15].

考慮北京市PM2.5濃度南高北低特征、站點(diǎn)空間分布特征以及目標(biāo)站點(diǎn)可能受周邊站點(diǎn)影響程度大小,選取北京市南部的、周邊關(guān)聯(lián)站點(diǎn)數(shù)量較多且PM2.5濃度相對較高的豐臺花園站點(diǎn)(FT)為目標(biāo)站點(diǎn),探討其周邊6個相關(guān)站點(diǎn)對豐臺花園站點(diǎn)PM2.5時空傳輸貢獻(xiàn).6個站點(diǎn)包括萬柳(WL)、古城(GC)、房山(FS)、大興(DX)、亦莊(YZ)和東四環(huán)(DS),涵蓋了北京市朝陽區(qū)、海淀區(qū)、大興區(qū)、豐臺區(qū)、石景山區(qū)和房山區(qū)6個區(qū)域,空間分布如圖1所示.

1.2 方法研究

城市內(nèi)部PM2.5污染傳輸受濃度梯度、氣象條件、站點(diǎn)間歐幾里得距離等多種因素共同制約[16-19].因此,綜合考慮監(jiān)測站點(diǎn)污染物濃度、風(fēng)向、風(fēng)速、歐幾里得距離等時空因素特征,定義目標(biāo)站點(diǎn)1時刻PM2.5濃度,如公式(1):

式中:w為擴(kuò)散系數(shù);F()為時刻站點(diǎn)間的風(fēng)力系數(shù)值;為2個空氣質(zhì)量監(jiān)測站連線方向與風(fēng)向之間的夾角;為污染物可以進(jìn)行傳輸?shù)淖畲缶嚯x(km).

圖2 站點(diǎn)間風(fēng)力系數(shù)示意

圖3 夾角q示意圖及風(fēng)向十六方位

1.2.5 模型結(jié)果誤差計(jì)算方法 為定量評估模型的計(jì)算準(zhǔn)確度,采用均方根誤差(RMSE)對模型進(jìn)行定量評估[27].

1.2.6 時空傳輸來源計(jì)算方法 根據(jù)建立的時空傳輸關(guān)聯(lián)模型,定義時空傳輸貢獻(xiàn)率,如公式(10).

式中:C分別為關(guān)聯(lián)站點(diǎn)時刻PM2.5濃度(μg/m3);為對應(yīng)影響權(quán)重;DF為空氣質(zhì)量監(jiān)測站點(diǎn)之間時刻的風(fēng)力系數(shù)差值;d為空氣質(zhì)量監(jiān)測站點(diǎn)之間的歐幾里得距離(km);為污染傳輸?shù)淖畲缶嚯x(km);(t+1)為模型計(jì)算目標(biāo)站點(diǎn)濃度值(μg/m3).

2 結(jié)果與討論

2.1 環(huán)境空氣質(zhì)量數(shù)據(jù)時空特征分析

圖4 北京市35空氣質(zhì)量站點(diǎn)PM2.5濃度空間分布

北京市共35個環(huán)境空氣質(zhì)量監(jiān)測站點(diǎn),各站點(diǎn)PM2.5濃度整體呈南高北低的空間分布特征[28],如圖4所示,南部站點(diǎn)更易受其周邊高PM2.5濃度--站點(diǎn)影響,站點(diǎn)間PM2.5輸送貢獻(xiàn)也更為明顯[29-30].2016年各站點(diǎn)PM2.5年均濃度分別為DX 93μg/m3、FS 82μg/m3、FT 81μg/m3、YZ 80μg/m3、DS 79μg/m3、GC 77μg/m3和WL 71μg/m3.

從圖5中可以看出,7個站點(diǎn)PM2.5峰值濃度隨時間推移而逐步升高,相鄰站點(diǎn)間PM2.5濃度變化趨勢具有明顯協(xié)同性.

圖5 2016年一次空氣重污染過程時刻各站點(diǎn)PM2.5濃度時間序列

圖6 關(guān)聯(lián)站點(diǎn)與目標(biāo)站點(diǎn)PM2.5濃度相關(guān)性

為定量解析任意關(guān)聯(lián)站點(diǎn)與目標(biāo)站點(diǎn)PM2.5濃度的相關(guān)性,基于最大信息系數(shù)理論[31],計(jì)算各關(guān)聯(lián)站點(diǎn)時刻PM2.5濃度與目標(biāo)站點(diǎn)1時刻PM2.5濃度的MIC值.從圖6可以看出,WL(t)、GC(t)、DX(t)、YZ(t)與FT(t+1)的MIC值相對較大,分別為0.6340、0.6305、0.6079和0.6007,表明站點(diǎn)之間的關(guān)聯(lián)性較大,PM2.5傳輸貢獻(xiàn)顯著.探究其可能的原因,WL、GC與目標(biāo)站點(diǎn)FT直線距離相對較小,分別為13.81和9.89km,傳輸相對容易且傳輸中損耗較小,故WL(t)、GC(t)與FT(t+1)關(guān)聯(lián)性較大;DX、YZ與FT的傳輸距離優(yōu)勢相對較小,但DX、YZ位于FT的東南方向,且2016年北京市的主導(dǎo)風(fēng)向?yàn)闁|南風(fēng),風(fēng)頻為28.39%,氣象條件有利于傳輸,故DX(t)、YZ(t)與FT(t+1)關(guān)聯(lián)性較大.

2.2 時空傳輸關(guān)聯(lián)模型結(jié)果解析

城市內(nèi)部大氣污染物的傳輸具有復(fù)雜多變的動態(tài)特征[32-33],季節(jié)變化顯著[34].本文基于2016年7個站點(diǎn)PM2.5小時濃度數(shù)據(jù)、小時氣象數(shù)據(jù)、站點(diǎn)空間位置數(shù)據(jù),應(yīng)用建立的時空傳輸關(guān)聯(lián)模型分別得到2016年4個季節(jié)的目標(biāo)站點(diǎn)PM2.5濃度,如式(12)～式(15).

2016年春季結(jié)果:

2016年夏季結(jié)果:

2016年秋季結(jié)果:

2016年冬季結(jié)果:

以FT為目標(biāo)站點(diǎn)的各個季節(jié)關(guān)聯(lián)回歸模型結(jié)果顯示, 4個季節(jié)PM2.5模擬濃度與監(jiān)測濃度的RMSE值分別為13.22、11.74、12.51和13.22,模型準(zhǔn)確性較高.分別隨機(jī)選取4個季節(jié)部分歷史時刻數(shù)據(jù)對模型進(jìn)行驗(yàn)證,模型模擬值與實(shí)際監(jiān)測濃度變化趨勢一致,進(jìn)一步驗(yàn)證了模型的準(zhǔn)確性,如圖7所示.該模型解析的結(jié)果可為北京市未來城市規(guī)劃、建設(shè)提供依據(jù).提出的PM2.5傳輸多元線性關(guān)聯(lián)回歸模型方法同樣可用來解析其他城市內(nèi)部尺度PM2.5傳輸關(guān)聯(lián).

圖7 2016年4個季節(jié)模型結(jié)果與實(shí)際數(shù)據(jù)對比

2.3 應(yīng)用時空傳輸關(guān)聯(lián)模型解析PM2.5傳輸特征

應(yīng)用時空傳輸關(guān)聯(lián)模型解析2016年4個季節(jié)PM2.5傳輸特征,如圖8所示.春季周邊傳輸對FT的貢獻(xiàn)率為2.56%,其中WL、YZ貢獻(xiàn)相對較大,分別為1.61%和0.81%;夏季周邊傳輸對FT的貢獻(xiàn)率為3.47%,其中DX、WL貢獻(xiàn)相對較大,分別為1.71%和1.69%;秋季周邊傳輸對FT的貢獻(xiàn)率為4.08%,其中YZ、WL貢獻(xiàn)相對較大,分別為2.20%和1.77%;冬季周邊傳輸對FT的貢獻(xiàn)率為8.98%,其中GC貢獻(xiàn)最大,為8.57%;DS全年對FT貢獻(xiàn)最小.綜合比較發(fā)現(xiàn),4個季節(jié)中周邊站點(diǎn)對FT的貢獻(xiàn)由大到小分別為冬季>秋季>春季>夏季;WL在春、夏、秋3個季節(jié)對FT貢獻(xiàn)率均排在前3位,冬季對FT基本無貢獻(xiàn);GC在冬季對FT貢獻(xiàn)明顯,春、夏、秋3個季節(jié)基本無貢獻(xiàn);DX在夏季對FT傳輸相對較多,春、秋、冬3個季節(jié)相對較少;YZ在春、秋季對FT貢獻(xiàn)相對較多,夏、冬季相對較少.分析PM2.5傳輸結(jié)果呈現(xiàn)的站點(diǎn)差異、季節(jié)差異原因知,不同季節(jié)的氣象條件、污染物濃度各有區(qū)別,不同站點(diǎn)與目標(biāo)站點(diǎn)的歐幾里得距離、風(fēng)力差值大小不同,會對PM2.5傳輸產(chǎn)生不同程度影響.

圖8 2016年4個季節(jié)周邊站點(diǎn)對FT傳輸貢獻(xiàn)

3 結(jié)論

3.1 基于數(shù)據(jù)驅(qū)動思想,以城市內(nèi)部環(huán)境空氣質(zhì)量監(jiān)測站點(diǎn)為研究對象,提出并建立了目標(biāo)站點(diǎn)與周邊站點(diǎn)間PM2.5濃度、風(fēng)向、風(fēng)速、歐幾里得距離等參數(shù)的多元線性關(guān)聯(lián)回歸模型,使用梯度下降算法學(xué)習(xí)得到各參數(shù)權(quán)重系數(shù),計(jì)算得到周邊站點(diǎn)對目標(biāo)站點(diǎn)PM2.5傳輸貢獻(xiàn).

3.2 2016年FT站點(diǎn)PM2.5濃度為82μg/m3,周邊站點(diǎn)DX、FS、YZ、DS、GC和WL濃度分別為93, 82,80,79,77,71μg/m3. FT站點(diǎn)PM2.5濃度與上一時刻周邊站點(diǎn)WL、GC、DX、YZ的MIC值分別為0.634、0.631、0.608和0.601.

3.3 建立了以FT為目標(biāo)站點(diǎn)的2016不同季節(jié)關(guān)聯(lián)回歸模型, PM2.5模擬濃度與實(shí)際監(jiān)測濃度變化趨勢一致,4個季節(jié)模型的RMSE值分別為13.22、11.74、12.51和13.22.

3.4 WL,DX,YZ,GC分別是對應(yīng)春、夏、秋、冬4個季節(jié)對FT 站點(diǎn)PM2.5污染傳輸貢獻(xiàn)較大的站點(diǎn),其貢獻(xiàn)值分別為1.61%,1.71%,2.20%和8.57%.

[1] Bove M C, Brotto P, Cassola F, et al. An integrated PM2.5source apportionment study: positive matrix factorisation vs. the chemical transport model CAMx [J]. Atmospheric Environment, 2014,94:274-286.

[2] Sulaymon I D, Zhang Y, Hu J, et al. Evaluation of regional transport of PM2.5during severe atmospheric pollution episodes in the Western Yangtze River Delta, China [J]. Journal of Environmental Management, 2021,293(18):112827.

[3] 鄧萌杰,閆雨龍,胡冬梅,等.長治市沙塵天氣污染特征及傳輸路徑解析[J]. 中國環(huán)境科學(xué), 2020,40(10):4275-4283.

Deng M J, Yan Y L, Hu D M, et al. Analysis of dust weather pollution characteristics and transmission paths in Changzhi, China [J]. China Environmental Science, 2020,40(10):4275-4283.

[4] Li R, Wang J, Xue K, et al. Spatial and temporal distribution characteristics and influencing factors analysis of particulate matter pollution in Jinan City [J]. Air Quality Atmosphere & Health, 2021,14:1267-1278.

[5] 吳健生,廖 星,彭 建,等.重慶市PM2.5濃度空間分異模擬及影響因子[J]. 環(huán)境科學(xué), 2015,36(3):759-767.

Wu J S, Liao X, Peng J, et al. Simulation and influencing factors of spatial distribution of PM2.5concentrations in Chongqing [J]. Environmental Science, 2015,36(3):759-767.

[6] Hou X, Zhu B, Kumar K R, et al. Inter-annual variability in fine particulate matter pollution over China during 2013～2018: Role of meteorology [J]. Atmospheric Environment, 2019,214:116842.

[7] Solazzo E, Hogrefe C, Colette A, et al. Advanced error diagnostics of the CMAQ and Chimere modelling systems within the AQMEII3 model evaluation framework [J]. Atmospheric Chemistry and Physics, 2017,17(17):10435-10465.

[8] 周廣強(qiáng),謝 英,吳劍斌,等.基于WRF-Chem模式的華東區(qū)域PM2.5預(yù)報(bào)及偏差原因[J]. 中國環(huán)境科學(xué), 2016,36(8):2251-2259.

Zhou G Q, Xie Y, Wu J B, et al. Based on WRF-Chem model, the PM2.5forecast and deviation reason in East China region [J]. Chinese Environmental Science, 2016,36(8):2251-2259.

[9] Qin Y G, Yi C, Dong G L, et al. Investigating the influence of meteorological factors on particulate matters: A case study based on path analysis [J]. Energy and Environment, 2019,31(3):479-491.

[10] 薛文博,付 飛,王金南,等.中國PM2.5跨區(qū)域傳輸特征數(shù)值模擬研究[J]. 中國環(huán)境科學(xué), 2014,34(6):1361-1368.

Xue W B, Fu F, Wang J N, et al. Numerical study on the characteristics of regional transport of PM2.5in China [J]. China Environmental Science, 2014,34(6):1361-1368.

[11] 齊孟姚.應(yīng)用WRF-Chem模型解析河北南部城市PM2.5來源及傳輸路徑[D]. 邯鄲:河北工程大學(xué), 2020.

Qing M Y. Source apportionment and inter-city transport of PM2.5in southern Hebei cities using WRF-Chem [D]. Handan: Hebei University of Engineering, 2020.

[12] 任傳斌,吳立新,張媛媛,等.北京城區(qū)PM2.5輸送途徑與潛在源區(qū)貢獻(xiàn)的四季差異分析[J]. 中國環(huán)境科學(xué), 2016,36(9):2591-2598.

Ren C B, Wu L X, Zhang Y Y, et al. Analyze to the seasonal differences of transport pathways and potential source-zones of Beijing Urban PM2.5[J]. China Environmental Science, 2016,36(9): 2591-2598.

[13] 余 創(chuàng),張玉秀,陳 偉.銀川市PM2.5的輸送路徑及潛在源解析[J]. 中國環(huán)境科學(xué), 2021,41(7):3055-3065.

Yu C, Zhang Y X, Chen W. The transport pathways and potential source regions of PM2.5in Yinchuan, China [J]. China Environmental Science, 2021,41(7):3055-3065.

[14] 呂宗璞.基于CMAQ模型的京津冀地區(qū)污染物傳輸路徑和輸送通道研究[D]. 石家莊:河北科技大學(xué), 2019.

Lv Z P. Study on pollutant transmission path and conveying channel in Beijing-Tianjin-Hebei area based on CMAQ model [D]. Shijiazhuang: Hebei University of Science and Technology, 2019.

[15] Resende D C O D, Santana D L D, Fábio Manoel Frana Lobato. Time Series Imputation Using Genetic Programming and Lagrange Interpolation [Z]. 2016 5th Brazilian Conference on Intelligent Systems (BRACIS). IEEE, 2017.

[16] Wang, Yu F, Hai Y, et al. A weighted higher-order network analysis of fine particulate matter PM2.5transport in Yangtze River Delta [J]. Physica, A. Statistical Mechanics and its Applications, 2018,496:654-662.

[17] Zhao G, Huang G, He H, et al. Innovative Spatial-Temporal Network Modeling and Analysis Method of Air Quality [J]. IEEE Access, 2019,7:26241-26254.

[18] Qbadou M, Hajji M, Samadi A, et al. Improving the selection process of students in higher education based on data warehouse and data mining techniques [Z]. International Technology, Education and Development Conference, 2016.

[19] 柳萌萌,趙書良,韓玉輝,等.多尺度數(shù)據(jù)挖掘方法[J]. 軟件學(xué)報(bào), 2016,27(12):3030-3050.

Liu M M, Zhao S L, Han Y H, et al. Research on multi-scale data mining method [J]. Journal of Software, 2016,27(12):3030-3050.

[20] Wen H F, Dang Y M, Li L W. Short-term PM2.5concentration prediction by combining GNSS and meteorological factors [J]. IEEE Access, 2020,8:115202-115216.

[21] Zhao G, Huang G, HE H, et al. Mining key stations by constructing the air quality spatial-temporal propagation network [J]. IEEE Access, 2020,PP(99):97485-97502.

[22] 吳 瑤.區(qū)域大氣污染物關(guān)鍵傳播路徑和重要節(jié)點(diǎn)挖掘方法研究[D]. 秦皇島:燕山大學(xué), 2018.

Wu Y. Research on key propagation path and important node mining method of regional air pollutants [D]. Qinhuangdao: Yanshan university, 2018.

[23] Liberti L, Lavor C, Maculan N, et al. Euclidean distance geometry and applications [J]. SIAM Review, 2012,56(1):3-69.

[24] Quentin M, Fabrice P, Jeng-antoine D. Stochastic multiple gradient descent algorithm [J]. European Journal of Operational Research, 2018,271(3):808-817.

[25] Lee G, Yi G, Youn B D. Special issue: a comprehensive study on enhanced optimization-based model calibration using gradient information [J]. Struct Multidiscip Optim., 2018,57(5):2005-2025.

[26] 王 丹.隨機(jī)梯度下降算法研究[D]. 西安:西安建筑科技大學(xué), 2020.

Wang D. A research of stochastic gradient descent algorithm [D]. Xi￠an: Xi'an University of Architecture and Technology, 2020.

[27] Chai T, Draxler R R. Root mean square error (RMSE) or mean absolute error (MAE)-Arguments against avoiding RMSE in the literature [J]. Geosci. Model Dev., 2014,7(3):1247-1250.

[28] 馮春莉,李潤奎.基于土地利用回歸模型的北京市2013～2019年大氣污染時空變化分析[J]. 環(huán)境科學(xué)學(xué)報(bào), 2021,41(4):1231-1238.

Feng C L, Li R K. Spatiotemporal variation analysis of air pollution from 2013 to 2019 in Beijing based on land use regression model [J]. Acta Scientiae Circumstantiae, 2021,41(4):1231-1238.

[29] 聶 滕,李 璇,王占山,等.APEC期間北京市PM2.5時空分布與過程分析[J]. 中國環(huán)境科學(xué), 2016,36(2):349-355.

Nie T, Li X, Wang Z S, et al. Spatial and temporal distribution and process analysis of PM2.5pollution over Beijing during APEC [J]. China Environmental Science, 2016,36(2):349-355.

[30] 李珊珊,程念亮,徐 峻,等.2014年京津冀地區(qū)PM2.5濃度時空分布及來源模擬[J]. 中國環(huán)境科學(xué), 2015,35(10):2908-2916.

Li S S, Chen N L, Xu J, et al. Spatial and temporal distrubions and source simulation of PM2.5in Beijing-Tianjin-Hebei region in 2014 [J]. China Environmental Science, 2015,35(10):2908-2916.

[31] Reshef D N, Reshef Y A, Finucane H K, et al. Detecting novel associations in large data sets [J]. Science, 2011,334(6062):1518-1524.

[32] Zhang H H, Hou Y G, Wu K J, et al. Study on Mechanism of Series-Flow of Pollutants between Consecutive Tunnels by Numerical Simulation [J]. Appl. Sci.-Basel., 2019,9(19):4125.

[33] Narasimhachar V, Poostindouz A, Gour G. Uncertainty, joint uncertainty, and the quantum uncertainty principle [J]. New Journal of Physics, 2016,18(3):033019.

[34] 盧 文,王紅磊,朱 彬,等.南京江北2014～2016年P(guān)M2.5質(zhì)量濃度分布特征及氣象和傳輸影響因素分析[J]. 環(huán)境科學(xué)學(xué)報(bào), 2019,39(4): 1039-1048.

Lv W, Wang H L, Zhu B, et al. Distribution characteristics of PM2.5mass concentration and their impacting factors including meteorology and transmission in North Suburb of Nanjing during 2014 to 2016 [J]. Acta Scientiae Circumstantiae, 2019,39(4):1039-1048.

[35] Zheng Z, Ren G, Hong W, et al. Relationship Between Fine-Particle Pollution and the Urban Heat Island in Beijing, China: Observational Evidence [J]. Boundary-Layer Meteorology, 2018,169(1):93-113.

Study on the transport correlation method of PM2.5at urban scale—taking Beijing as an example.

YIN Hao, HU Dong-mei*, YAN Yu-long, PENG Lin, WANG Kai, ZHANG Ke-ke, DENG Meng-jie

(Key Laboratory of Resources and Environmental System Optimization, Ministry of Education, College of Environmental Science and Engineering, North China Electric Power University, Beijing 102206, China)., 2022,42(2)：550～556

Based on data driven, internal environment in city air quality monitoring sites as the research object, multiple linear correlation regression models were established for PM2.5concentration, wind direction, wind speed, Euclidean distance and other parameters between target stations and surrounding stations. The weight coefficients of each parameter were obtained by gradient descent algorithm, the PM2.5transmission contribution of surrounding stations to target stations was calculated and the feasibility of the model was evaluated. Taken Feng Tai Garden (FT) in Beijing as the target site, the results showed that the PM2.5concentration of FT site in 2016 was 82μg/m3, Da Xing (DX), Fang Shan (FS), Yi Zhuang (YZ), Dong Sihuan (DS), Gu Cheng (GC) and Wan Liu (WL) sites were93, 82, 80, 79, 77和71μg/m3. The correlation between PM2.5concentration of FT station and WL, GC, DX and YZ of surrounding stations at the last moment was 0.634, 0.631, 0.608 and 0.601, respectively, which indicates the significantly transmitted PM2.5pollution to FT station. RMSE values of the four seasonal correlation regression models were 13.22, 11.74, 12.51 and 13.22, respectively. The variation trend of PM2.5simulated concentration was consistent with that of the monitored concentration, which verified the feasibility of the model. WL, DX, YZ and GC were the stations that contribute more to PM2.5pollution transmission of FT station in spring, summer, autumn and winter respectively, and their contribution values were 1.61%, 1.71%, 2.20% and 8.57%, respectively. The model results can provide a basis for the future urban planning and construction of Beijing. The proposed multiple linear correlation regression method of PM2.5transmission can also be used to analyze the PM2.5transmission correlation of other urban scales, providing a basis for the mining of PM2.5transmission path and accurate traceability within the city.

intra-city scale；PM2.5；multiple linear correlation regression；transmission contribution

X513

A

1000-6923(2022)02-0550-07

尹 浩(1996－),男,河北張家口人,華北電力大學(xué)碩士研究生,主要從事大氣污染控制研究.發(fā)表論文1篇.

2021-07-14

國家重點(diǎn)研發(fā)計(jì)劃項(xiàng)目(2019YFC0214202,2019YFC0214203);國家自然科學(xué)基金資助項(xiàng)目(21976053);中央高?；究蒲袠I(yè)務(wù)費(fèi)專項(xiàng)資金資助項(xiàng)目(2019MS043)

* 責(zé)任作者, 講師, huhu3057@ 163.com