环境监测数据分析模式与空气质量预警预测

环境监测数据分析模式与空气质量预警预测

Ambient Air Quality Monitoring Data Anal ysis Mode and Air Quality Forecast and Warning Song Guojun [email protected] Institute of environmental policy and environmental planning, Renmin University of China Beijing Huitong Research Institute of environmental technology 24-25 June, 2014 Contents Status and Needs Air Quality Monitoring Data Analysis Stationary Pollution Source Monitoring Data Anal ysis Air Quality Forecast and Warning Based on Statis tical Analysis Urban Air Quality Daily Management 1. Status and Needs 12th Five Year Plan requirements 11th Five Year Plan status quo Monitoring data has not been managed effectively Monitoring information has not been fully utilized Low degree of automation for data processing insufficient information for air quality Air Quality Data Analysis, Forecast, and Warning Implement a comprehensive control plan on a variety of atmospheric pollutants, promote urban air pollution control, and increase efforts to reduce emissions of sulfur dioxide and nitrogen oxides. Construct the early warning and emergency response system, strength the foundation of environmental information, statistics and application

Implement the new "Ambient Air Quality Standards" Measure 9: establish a monitoring and early warning system to cope with heavy pollution properly. Ten National measures 2. Air Quality Monitoring Data Analysis Analytical Framework Air Quality Monitoring Data Analysis concentrat ion mean analysis AQI and air quality rating the degree of change Status Quo Assessment comprehe nsive pollution index exceed analysis Comparative analysis of different pollutants Feature Analysis Comparativ e analysis of different monitoring sites Changes on Quality

the treand of change comparativ e analysis of different time periods Assessment and Analysis Assessment basis: Ambient Air Quality Standard GB 3095-2012 Ambient Air Quality Index (AQI) Daily technic Assessment basis al requirements Ambient air quality evaluation of technical sp ecifications Assessment scope: Assessment Assessment Monitoring sites, urban area scope content Assessment and (the whole city) Analysis (the whole Assessment items: SO2, NO2, PM10, PM2.5, CO, O3 city and monitoring (1 hour and 8 hours) cites) Assessment index: Concentration of pollutants, Emissions compl Assessment iance, exceeding standard rate, integrated p

Assessment index ollution index items Change rate, rank correlation coefficient Assessment content: Pollution status, compliance assessment, ind ex assessment, and assessment on the degr ee and trend of the changes Multi-dimensional Features Analysis Multi-dimensional Features Analysis Analysis of spatial features Analysis scope monitoring sites Analysis indicators: exceeding standard rate Analysis Scale: hourly, daily Display: Spatial map with major pollutants highlighted, early warning Display content: Results and causes Analysis of time features Analysis scope The whole city, monitoring sites Analysis indicators Exceeding standard rate Analysis mode:

Same period comparison, year-on-year comparison, month on month comparison Display: Improvements showed on map Display content: Results and causes To identify key polluted location, time and pollutants Major pollutants identification Analysis scope The whole city, monitoring sites Analysis indicators Exceeding standard rate Analysis Scale: Hourly, Daily Display A map showing major pollutants for each monitoring sites Display content: Results and causes Technology Roadmap major pollutants identification Time Feature Analysis Pollutant Spatial Features Analysis

Variable Selection Results Output Data Comparison Table Index Data Comparison Chart Spatial Scale Statistical Analysis Table Time Scale Accurately Indentify Critical Management Period, Pollotants and Regions Results Case Study Comparative Analysis of Different Pollutants Major pollutants Identification: from the above charts, we can see that the daily average PM10 of this monitoring site severely exceeded the standard rate, followed by SO2. Comparative Analysis of Different Monitoring Sites Comparative Analysis of Different Time Periods Time period/year The whole year Heating period Non-heating period 2006

2007 2008 2009 2010 The 11 th five-years plan period average 69.9% 80.8% 86.1% 94.2% 95.6% 85.3% 50.6% 65.1% 72.5% 88.0% 91.0% 73.4% 85.9% 94.0% 97.5% 99.5% 99.5% 95.3% Although the overall level of average PM10 compliance rate remained low, daily average compliance rate has gradually increased

during 2006 to 2010, with 2010 being the best performed year. The average PM10 daily concentration compliance rate of the whole year was 85.3%. However, average PM10 daily concentration compliance rate during the heating period was 73.4%, while the date for the non-heating period was 95.3%, significantly higher than the heating period. 3. Pollution Source Monitoring D ata Analysis Frame Diagram Pollution Source Data Analysis Hourly Pollution Source Daily Average Concentratio n and Distribution Exceeding Standard Rate and Exceeding Standard Amount Discharg Amount Industrie s Monthly Citywide Annually Pollotion Source Industries Citywide

Case study: Concentration distribution 18000 16000 14000 12000 10000 8000 6000 4000 2000 0 Frequency distribution of SO2 concentration 120.00% As can be seen from the frequency distribution of the SO2 concentration, the SO2 concentration monitored was between 50mg/m3~150mg/m3 100.00% 80.00% 60.00% 40.00% 20.00% 0 50 100 150 200 250 300 350 400 450 500 0.00% mg/m3 Frequency distribution of NOx concentration 12000 120.00% 10000 100.00%

8000 80.00% 6000 60.00% 4000 40.00% 2000 20.00% 0 0 80 160 240 320 400 480 560 640 720 800 Concentration mg/m3 0.00% % As can be seen from the frequency distribution of NOx concentration, the NOx concentration monitored was between 80mg/m3~320mg/m3. Case study: Descriptive statistics Mean mode Standard N Stationary sources Minimum Maxima

90% quantile 98% quantile Emission limit mg/m3 corresponding percentile of emission limit % deviation valid Rainbow 91 thermoelectric Changrun 72 thermoelectric De Neng Power Plant 91 Gaotang 88 Thermoelectric Guodian biomass 91 power Huatong 91 thermoelectric (1,2) Huaxiang 60 thermoelectric (2) Lantian 91 thermoelectric Liaocheng Power 90 Plant (1) Liaocheng Power 87 Plant (2) Liaocheng Power 91 Plant (3,4) Liaocheng 86

thermoelectric (5,6) Liaocheng 91 thermoelectric (7,8) quanlin 90 Thermoelectric senquan 58 Thermoelectric Yuncheng 55 Thermoelectric yinhe Thermoelectric 56 huaxiang 60 Thermoelectric 1 dongchang 89 Coke 1 2 dongchang 89 Coke 3 a. munium mode is showed deficient 1 109.0619 89.73a 19.29271 80.63 170.15 129.8380 162.8504 200 20 174.0246 67.23a

55.04416 67.23 280.08 250.3690 274.1092 200 59 1 112.9148 117.33 4 23.44604 46.15 154.46 142.6900 153.2168 200 118.5203 a 89.54 17.19603 89.54 203.17 138.0470 162.0718

200 1 24.0015 12.71a 27.73760 1.50 179.29 33.6160 173.1664 200 1 55.9796 55.58a 27.92952 10.67 170.29 85.1680 163.0156 200 32 66.0432 10.67a 43.65629

10.67 243.35 134.5150 225.8666 200 99 1 136.4753 134.58a 13.27643 86.60 162.29 151.8140 159.7028 200 2 354.2126 74.44a 183.62869 74.44 995.02 618.7860 885.6402 200 18

5 122.5084 130.00 62.48748 7.00 282.00 204.0960 250.8248 200 89 1 332.2642 200.08 204.31884 134.15 1323.48 605.6080 954.0900 200 19 6 97.6080 20.71a 40.97408 20.71 345.75

144.3890 216.5904 200 99 1 91.0437 67.06a 21.12884 25.83 139.96 121.5640 132.1396 100 68 2 120.1841 129.38 17.41563 79.69 164.77 140.3130 163.2694 200 34 85.0226

62.06 17.30818 2.20 129.67 107.6840 127.1716 200 37 79.2582 68.54a 19.09984 42.69 158.42 99.9580 153.6800 200 36 104.1280 1.00a 77.30551 1.00 478.10 124.7180 476.9772

200 97 32 66.0432 a 10.67 43.65629 10.67 243.35 134.5150 225.8666 200 99 3 224.9425 38.75a 39.89132 38.75 290.90 269.8300 286.4280 200 26 3 202.0511 127.31a

31.27732 127.31 261.94 243.6700 260.3880 100 51 a The 90%, 98% fractile values of daily average SO2 emission concentrations from stationary sources such as Changrun thermoelectric, Liaocheng Power Plant, Huaxin Aluminum, Xinyuan Alumiunm, Huayu Aluminum and Xinfa Aluminum were significantly higher than the emission standards; Besides Huaxin Aluminum and Xinyuan Aluminum, other stationary sources kept the SO2 daily average concentration modes greatly lower than emission standards, indicating that with the current level of emission control, it is possible to cut emission and pollutant discharge. Excessive emission from stationary sources was mainly due to poor management. It is suggested to issue pollutant discharging permits for stationary sources. Case study: Exceeding standard rate Octorber November stationary sources Rainbow thermoelectric (16-18) Changrun thermoelectric (1-3) Deneng Power Plant (1-2) Dongchang Coke (1-2) Dongchang Coke (3) Dongchang cement (4-5) Gaotang thermoelectric (89) Guodian Biomass power Huatong thermoelectric (12) Huaxiang thermoelectric (2)

Huaxin Aluminum (12-13) Huaxin Aluminum (14-15) Watson Aluminum (9-11) Lantian thermoelectric (12) Liaocheng Power Plant (1) Liaocheng Power Plant (2) Liaocheng Power Plant (34) Liaocheng thermoelectric (5-6) December SO2exceedin g standard rate% NOx exceeding standard rate% SO2 exceeding standard rate % NOx exceeding standard rate% SO2 exceeding standard rate% NOx exceeding standard rate% 1.35 92.05 0.64 98.39 0 100 0.82 88.11 26.48 98.17

56.12 94.91 4.92 100 99.85 0.3 79.88 78.23 7.44 17.06 3.51 100 100 0.38 99.24 51.83 8.54 10.06 3.18 55.67 99.31 13.56 97.14 100 37.76 41.73 0.32 50.57 0 10.4 0.48 91.26 0 0.15 0

0.48 0 0.32 0.66 91.56 1.94 81.7 1.11 93.88 7.84 25.34 6.1 11.91 2.75 32.96 100 99.7 100 100 100 97.81 99.68 100 99.54 94.76 99.38 98.46 100 100 100 95.69

100 100 0.14 0 0 0.3 0 0.48 61.36 26.05 83.65 95.16 78.34 18.63 100 89.31 72.5 6.81 99.69 70.81 69.46 100 67.53 96.74 65.15 100 0.43 98.55 2.02 100

1.04 100 Case Study: Discharge Amount 10 10-12 SO2 (t)t) 6000.00 11 5000.00 4000.00 (t)t) 3000.00 2000.00 1000.00 0.00 (1 1 6- 8) (1 2)

(3 ) (8 9) (1 2) (1 1 2- 3)

( 1 9 1) (1 ) (3 4) (7 8) (2

5 ) (6 -8 ) (1 -3 ) (1 -3

) (1 -2 ) (1 -3 ) (4 -5 ) (1

-8 ) (4 -5 ) (1 -3 ) t Case Study: Excessive Discharge Amou nt 6000 10-12 SO2 10 5000 4000 3000 2000 1000 0 11 12 4. Quality Forecast and Warning Factors Affecting Air Quality

Pollution emissions: Stationary sources Mobile sources non-point sources Air Quality Weather conditions: Wind speed and direction Temperature and pressure... Existing Prediction Models and Predictive Effects Numberical Prediction Model of Institute of Atmospheric Physics, Chinese Academy of Sciences informations on sources of pollotion Air quality monitoring data Geographical information Meteorological data (including aerial and ground data) need high-performance computers to process data Model of Chinese Academy of Meteorological Sciences Air quality data Meteorological data included in TTAA message Model of the National Meteorological Center mode Air quality monitoring data (for the revise of model parameters); global modes of four-dimensional data assimilation data daily 8:00pm T106; Geomorphological conditions of the forecasting city;

Geographic Information Using emission intensities and wansportation and diffusion patterns of pollutant to calculate the concentration of pollutants Statistical Models Multiple regression methods Air quality data; Meteorological data (pressure, temperature, variable temperature 24 hours, 24 hours transformer, minimum temperature, wind direction, wind speed, precipitation, sunshine, cloudiness, etc.) Kalman filtering method Air quality data; Meteorological data (pressure, temperature, variable temperature 24 hours, 24 hours transformer, minimum temperature, wind direction, wind speed, precipitation, sunshine, cloudiness, etc.) T106 meteorological data Using historical data to establish statistical relationships between pollutant concentrations and meteorological conditions or non-meteorological conditions Sources: Huan Fa [2000] No. 231 Notices on Carrying out Air Quality Protection Measures in Major Cities Air quality forecast accuracy in47 major cities, from June 5, 2001 to December 31, 2004 Source: "China's major cities pollution forecasting and progress," Tong Yanchao Monitoring Station Advantages and disadvantages of the Existing prediction models Numberical Prediction Advantages Disadvantages High efficiency Higher spatial and temporal resolution High precision High requirements on hardware and software High Costs Need a variety of parameters, many cities failed to meet the

requirements with current conditions Statistical Models Advantages Low Cost Accumulate a large amount of historical data Easy access to relevant parameters Disadvantages low prediction accuracy 1. Predictive models should be customized based on local situation; 2. Samples can be classified based on uncontrollable factors (meteorological factors) to improve the accuracy of the model; 3. Predictive models need to consider factors such as emission and excessive mining to diagnose causes of pollution. Air quality forecast and diagostic model based on meteorological classification Step 1: data preprocessing and correlation analysis Classification of meteorological factors (transformed from continuous variables into categorical variable) Conduct correlation analysis: on meteorological factors an d concentrations of pollutants; select meteorological factors with high correlation ; Step 2: Sample classification based on the sample homogeneity of variance test, choos e different test methods for multiple comparisons;take mont h, atmospheric stability, wind direction and other meteorolog ical factors as categorical variables to conduct pollutant co ncentration classification. make corrections for sample classification base on experti se. Step 3: identify independent variables identify independent variables needed for stepwise regres sion with the consieration of literatures. . Calculate urban pollution emission and use it as an indep endent variable for stepwise regression. Step 4: build predictive models build predictive model for each catagories with considera tions on pollution sources. Step 5: test the predicting effects Data preprocessing and

correlation analysis Sample classification that based on meteorological factors Meteorol ogy Pollution sources discharg e amount build predictive models Case Study: Data Preprocessing Data explanation Air quality monitoring data and indicators, Caitun Monitoring site, Benxi, March 6, 2012 to December 17, 2012: Concentration of pollutants (hourly), meteorological factors (wind speed, wind direction, humidity, atmospheric pressure and temperature) Sample pre-processing Classify by temperature: heating period and non-heating period Classify by atmospheric stability: the atmospheric stability is categorize into six grades based on the Pascual atmospheric stability classification method Classify by wind directions: according to the standard classification, there are 16 wind directions; Case Study: Build Predictive Models Stepwise Regression using stepwise regression method to establish pollutant concentration predictive models for all types of samples. Dependent Variable Concentrations of pollutants selected independent variables (summarized from literatures) The average pollutants concentration of the previous 24h, daily average temperature, minimum temperature, maximum temperature, daily average temperature difference, average humidity, average air pressure, wind speed every ten minutes, 8:00 pressure, the highest temperature of the previous 24h, average temperature of the previous 24h, 14:00 wind speed, 14:00 temperature, pollution discharge amount.

Case study: SO2 sample classification and regression curves, Caitun monitoring site SO2 classification Heating/ Atmospheri nonc stability heating A Heating BCDE period F A Nonheating period BCDEF Model Stepwise regression, model building No samples y=-2.313+0.461x13+0.008x12-0.013x20.013x11+0.002x4+0.024x10 No samples No samples y=0.01+0.518x13-0.0001x1+5.77*10-5x140.001x9+0.001x3+0.003x7 Case study: NO2 sample classification and regression curves, Caitun monitoring site NO2 Classification Heating/ Atmospheric non-heating stability A ,B,C Heating D period E,F Non-heating period Regression curves Regression curves No samples y=0.023+0.001x12+0.00035x14 y=0.031+0.003x9 A,B,C,D

No samples E,F y=0.039-0.004x11 Case study: PM10 sample classification and regression curves, Caitun monitoring sit e Classification Stepwise regression Heating/non- Atmospheri Wind direction heating c stability Regression equation A Heating period B CDEF A,B PM10 Non-heating period C D,E,F Others ENE,E y=4.540+0.008x5-0.045x8-0.0008x9 No samples Others y=0.017+0.414x13+0.001x14-0.002x9 ENE,E No differences No differences N,NNE Others Others ENE,E

No samples y=0.115-0.025x11+0.406x13 No samples No samples y=0.086+0.582x13-0.0005x-0.004x11-0.0006x9 y=-5.409+0.055x10+0.577x13 y=-0.255+0.003x+0.001x9 Case study: NOx sample classification and regression curves, Caitun monitoring site Combination Heating/ Atmospheri nonc stability heating A,B,C Heating D period NOx E,F Non-heating A,B,C,D,E period F Stepwise regression Regression equation No sample y=0.039+0.001x14+0.002x4+0.002x12 y=0.04+0.004x9 y=1.495-0.01x11-0.001x1-0.014x6 No sample Case Study: Prediction Accuracy Sample size Accurately predicted samples Percentage SO2 287 253 88.15% NO2 287

227 79.09% PM10 (excluding wind) 287 197 68.64% PM10(including wind) 287 210 73.17% NOX 287 247 86.06% 5. Urban Air Quality Daily Man agement Defining Urban Air Quality Daily Man agement The scale of annual average air quality is too broad for manage ment, while hourly air quality data is too specific for the current m anagement capacity. environmental data of daily scale are relatively convenient to coll ect and process. They are consistent with the requirements of st ationary source emission permits, and also easy for the mobile s ources and non-point sources pollution management. According to the Standard provisions, by narrowing the time sc ale to daily data, systematic and statistical analysis on historical monitoring data such as air quality, pollution emissions and weat her conditions can be conducted to diagnose and summarize ca uses of the exceeded pollution. It can be seen as a early warning for air quality of the near future (for future days or week), and th us measures will be taken to control emissions from stationary, mobile, or non-point sources.

Emissions Control Plan: the Core of Urb an Air Quality Daily Management Since the daily data of pollution sources and emission condition can be seen as a warning alert for the air quality of the future one or two days, when the result shows standard exceeded air quality rate, authorities ca n implement an emission control plan and decide what control measures may be taken on which type of pollution source. Formulation of the emissions control plan is the core of urban air quality daily management. The plan should be negotiated and made by the gov ernment, enterprises and public together. As the stationary sources bei ng the culprit of urban air pollution, government can make emission cont rol plan for stationary sources days in advance based on the daily emiss ion data and emission standard exceeding warning; while enterprises sh ould implement the emission plan without causing significant economic i mpact. The public can provide corresponding feedback to the emission c ontrol plan to facilitate further amendments. Urban Air Quality Daily Management Mo del Daily air quality diagnosis and exceeding warning Historical daily data and information Data processing tools Air quality coordination meeting Stationary sources plan feedback s Emission control plan confirmation Feedback s Information disclosure and public participation Non-point sources plan Mobile sources plan Adjustments on the Management Mode of the City Environmental Protection Bureau

Urban air quality daily management needs to be supported by th e corresponding management mechanism to coordinate daily wo rks on emission management. It is suggested that the City Environmental Protection Bureau ca n establish an air quality management office to in charge with all relevant works, including making air quality plans, air quality mon itoring, stationary source emission permits management, mobile sources and non-point source management, air quality diagnosti c analysis and early warning, pollution emission control plan pre paration, as well as information disclosure and public participatio n. This office may coordinate urban air quality management, and is expected to reduce management costs while improving efficie ncy. Thank you! All comments and corrections are welcome.

Recently Viewed Presentations

  • Fats & Oils - Welcome to Mrs. Scherr's Class

    Fats & Oils - Welcome to Mrs. Scherr's Class

    Fats & Oils. Unit: Lipids. Warm-Up. ... There is an empty spot or Point of Unsaturation where hydrogens are missing. Generally from plant sources. ... Fats & Oils Crossword Puzzle and Fats & Oils in Food Preparation Word Scramble. worksheet.
  • Volcanoes and Igneous Activity Earth - Chapter 4

    Volcanoes and Igneous Activity Earth - Chapter 4

    Others may have been island arcs, submerged crustal fragments, extinct volcanic islands, or submerged oceanic plateaus. Terranes and Mountain Building Accretion and orogenesis As oceanic plates move, they carry embedded oceanic plateaus, island arcs, and microcontinents to Andean-type subduction zones.
  • Utopia and Dystopia - Mrs. Presnell's Class

    Utopia and Dystopia - Mrs. Presnell's Class

    A utopia refers to a perfect society that does not exist or can never exist Star Trek is an example of a utopian society because humankind has overcome sickness, racism, poverty, and warfare A dystopia is the opposite of a...
  • Business Essentials 5e. - Ebert and Griffin

    Business Essentials 5e. - Ebert and Griffin

    Describe data communication networks. Describe new options for organizational design stemming from communication technology. Describe application programs at various organizational levels. Discuss databases and the purpose of database software.
  • Deep Neck Spaces and Infections - ssu

    Deep Neck Spaces and Infections - ssu

    Deep Neck Spaces Dr.S.A.Mirvakili Shahid sadughi university Presentation/Origin Peritonsillar Space Fever, malaise Dysphagia, odynophagia "Hot-potato" voice, trismus, bulging of superior tonsil pole and soft palate, deviation of uvula Cause—extension from tonsillitis Presentation/Origin Masticator Temporal Space Pain, trismus Posterior FOM edema...
  • Simple Linear Regression - Stony Brook

    Simple Linear Regression - Stony Brook

    X. Y. is assumed to be a random variable while, even if ... Essentially, we are interested in knowing the behavior of . Y. given we know . X = x. Simple Linear Regression Model /69. Regression models attempt to...
  • This week in 206 Dr. Lloyd Sea urchin

    This week in 206 Dr. Lloyd Sea urchin

    UT-Austin Other titles: Times Helvetica Arial Blank PowerPoint Presentation PowerPoint Presentation PowerPoint Presentation PowerPoint Presentation PowerPoint Presentation PowerPoint Presentation PowerPoint Presentation PowerPoint Presentation PowerPoint Presentation PowerPoint Presentation ...
  • LINCOLN DOUGLAS DEBATE - Coaching Speech

    LINCOLN DOUGLAS DEBATE - Coaching Speech

    LINCOLN DOUGLAS DEBATE U.I.L./T.F.A./N.F.L. rules, requirements, expectations BRIEF DEFINITION Lincoln-Douglas debate is a one-on-one argumentation in which debaters attempt to convince the judge of the acceptability of their side of a proposition.