critical issues of exposure assessment for human health studies of air pollution
DESCRIPTION
Critical Issues of Exposure Assessment for Human Health Studies of Air Pollution. SAMSI September 15, 2009. Michelle L. Bell Yale University. Outline. Basic health effects model Methods of measuring exposure Key challenges in assessing exposure Spatial misalignment - PowerPoint PPT PresentationTRANSCRIPT
-
Critical Issues of Exposure Assessment for Human Health Studies of Air PollutionMichelle L. BellYale UniversitySAMSISeptember 15, 2009
-
OutlineBasic health effects modelMethods of measuring exposureKey challenges in assessing exposureSpatial misalignmentMultiple pollutant exposuresSpecial case of particulate matterCurrent and upcoming approaches to estimating exposureOther challenges
-
Exposure Assessment for Studies of Air Pollution and HealthBasic health effects modelMethods of measuring exposureKey challenges in assessing exposureSpatial misalignmentMultiple pollutant exposuresSpecial case of particulate matterCurrent and upcoming approaches to estimating exposureOther challenges
-
Example Air Pollution and Health Effects ModelTime-series model / Acute exposureCommunity-aggregated health dataCommunity-aggregated exposure data
-
Example Air Pollution and Health Effects ModelEstimated Air Pollution Exposure
-
Exposure Assessment for Studies of Air Pollution and HealthBasic health effects modelMethods of measuring exposureKey challenges in assessing exposureSpatial misalignmentMultiple pollutant exposuresSpecial case of particulate matterCurrent and upcoming approaches to estimating exposureOther challenges
-
Use of ambient monitors+ cost-effective+ can provide large sample sizenot present in all times or locations of interestlocations based on regulatory, not scientific, purposesobscures between-person differences
Personal monitors+ individualized datashort timeframesmall populationlogistical concernsTraditional Approaches to Exposure AssessmentSource: Louisville, KY governmentJiang and Bell EHP 2008
-
Traditional Approaches to Exposure AssessmentUse of ambient monitors+ cost-effective+ can provide large sample sizenot present in all times or locations of interestlocations based on regulatory, not scientific, purposesobscures between-person differences
Personal monitors+ individualized data+ measure pollutant characteristics of interestshort timeframe and small populationlogistical concerns and expensive
-
Exposure Assessment for Studies of Air Pollution and HealthBasic health effects modelMethods of measuring exposureKey challenges in assessing exposureSpatial misalignmentMultiple pollutant exposuresSpecial case of particulate matterCurrent and upcoming approaches to estimating exposureOther challenges
-
Spatial MisalignmentSpatially heterogeneity in the concentration surface fieldMismatch between data used to estimate exposure and actual subjects locations
-
Correlation of PM2.5 components by distancePeng and Bell Biostatistics Accepted
-
Spatial Disconnect in Data
-
Error ModelPollution data measured at a single or multiple fixed locationsPeng and Bell Biostatistics Accepted
-
Spatial Misalignment AdjustmentMonitor average ( ) good proxy for true value ( xt ) with good monitor coverage and/or low spatial heterogeneitySpatial misalignment adjustments useful when:Pollutant very spatially heterogeneous (e.g., EC)Poor monitor coverage within area of interest, but monitors elsewhere
-
Multi-Pollutant ConceptsPhysiologically we respond to a complex mixture of air pollutantsMany studies focus on the effects of a single pollutantAdditional pollutants typically considered with respect to confounding, not complex effectsAir pollution policy set for single pollutantsBased on single pollutant science
-
Major Air Pollution Emissions Sources Industrial SourcesDomestic SourcesPowerGenerationDieselGasolineSTATIONARYSOURCESVEHICLESOURCES
-
Particulate MatterWhat is represented by an exposure estimate for PM?Only pollutant regulated without regard to chemical formMay vary in:SizeShapeChemical structureWater contentAcidityAgeEtc.
-
Different chemical components by PM sizeDifferent sources by PM size
-
PM as a Pollutant MixturePM2.5 sulfate (2000-2002)Bell et al. EHP 2007
-
Bell et al. EHP 2007
-
% Change in Hospital Admissions per 10 mg/m3 PM2.5Respiratory InfectionCOPDHeart FailureHeart RhythmIschemic Heart Disease Peripheral Vascular DiseaseCerebrovascular DiseaseDominici et al. JAMA 2006PM2.5 and Medicare Hospital Admissions
-
Allow Temporal Variation in Effect Estimates Based on Variation in Exposure
-
Season Interaction ModelAllows different effect estimates by seasonHarmonic ModelAllows effect estimates to differ throughout the yearBell et al. Am J Epidemiol 2008
-
Seasonal Variation in PM Total MassHealth Effect EstimatesBell et al. Am J Epidemiol 2008Day of the year
-
Exposure Assessment for Studies of Air Pollution and HealthBasic health effects modelMethods of measuring exposureKey challenges in assessing exposureSpatial misalignmentMultiple pollutant exposuresSpecial case of particulate matterCurrent and upcoming approaches to estimating exposureOther challenges
-
Example: Air Quality Modeling to Estimate ExposureCMAQ / MM5Aug 15 18, 19952,168 cells with 4 km horizontal resolution8 monitors for ozoneBell Environ Int 2006
-
Bell and Ellis J Air Waste Manage Assoc 2003
Chart1
6064.141
5252.889
5149.654
5153.184
5149.157
4747.141
4148.317
4146.502
3243.277
3140.952
2738.577
2834.934
3231.522
4635.968
6146.305
7055.664
8068.408
8980.43
9987.333
10290.288
10391.152
9891.801
8891.71
7589.205
6784.783
6575.076
6272.638
6169.875
5663.918
4460.251
3621.716
296.573
1414.804
151.694
2023.83
2937.367
3841.058
5042.468
33052.374999999950.087
33052.416666666659.919
6762.557
7063.638
7266.194
7069.214
6571.173
6172.435
6571.504
6069.978
3568.375
5666.49
6260.338
6049.165
5064.446
5963.712
5360.677
4256.361
4452.923
3547.329
2840.469
3932.08
4443.023
6263.815
7282.429
7294.179
81103.699
96110.599
118113.655
131101.288
12596.651
11397.685
9398.788
79101.818
71101.579
Location: Millington, Maryland
Monitor Measurements
Model Estimates
Local Time
O3 (ppb)
Figure4
Figure 4. Comparison of Model-Estimated Ozone Concentrations and Monitor Measurements
Figure4
6064.141
5252.889
5149.654
5153.184
5149.157
4747.141
4148.317
4146.502
3243.277
3140.952
2738.577
2834.934
3231.522
4635.968
6146.305
7055.664
8068.408
8980.43
9987.333
10290.288
10391.152
9891.801
8891.71
7589.205
6784.783
6575.076
6272.638
6169.875
5663.918
4460.251
3621.716
296.573
1414.804
151.694
2023.83
2937.367
3841.058
5042.468
33052.374999999950.087
33052.416666666659.919
6762.557
7063.638
7266.194
7069.214
6571.173
6172.435
6571.504
6069.978
3568.375
5666.49
6260.338
6049.165
5064.446
5963.712
5360.677
4256.361
4452.923
3547.329
2840.469
3932.08
4443.023
6263.815
7282.429
7294.179
81103.699
96110.599
118113.655
131101.288
12596.651
11397.685
9398.788
79101.818
71101.579
Monitor Measurements
Model Estimates
Local Time
O3 [ppb]
Millington, MD
Figure5a
3815.19
5216.981
4628.224
4628.456
4417.655
432.097
400.004
380
330
280
210.381
33051.251.53
2110.788
1034.956
4851.042
8559.422
9865.219
8868.839
8370.157
7969.372
7064.506
4844.508
2539.712
3950.407
3310.093
210.181
200
220
300
273.947
1722.298
1718.679
1910.125
122.08
120.856
142.905
1115.792
2832.009
3743.765
5654.447
7964.591
7669.907
9473.034
9874.987
10477.482
11979.266
10374.606
9166.267
7137.139
420.178
400.055
434.473
310.383
360
380
450
390
240
110.287
33053.24999999990.883
1012.539
3545.704
6973.26
8678.622
9676.689
9177.194
8379.87
8081.913
7784.409
7783.831
7653.609
705.595
577.132
Monitor Measurements
Model Estimates
Local Time
O3 [ppb]
S. 18th and Hayes St., VA
Figure5b
8011.32
710.212
610
620
480
410
490
500
330
280
200.191
201.571
158.387
2119.99
5541.625
7257.881
9369.173
12077.435
13776.216
12972.916
11979.043
11676.916
10567.073
9349.122
8026.21
596.196
550.52
410.108
260.023
160
90
90
170
100
120.451
153.74
3118.112
4936.207
7650.642
9166.044
9578.281
10383.053
10082.427
10585.274
10890.823
11188.624
11590.619
10196.177
8881.63
6852.769
5227.43
4513.693
4314.087
3517.298
270.856
402.866
373.053
330.937
411.982
519.48
6226.353
8143.012
10258.873
11777.252
12280.982
12782.111
11581.325
11878.134
12268.707
12072.73
11979.546
10979.956
10277.214
Monitor Measurements
Model Estimates
Local Time
O3 [ppb]
Lake Clifton, MD
Figure5c
9257.534
8146.163
7841.695
6735.202
5230.264
5429.942
5433.163
5232.455
4431.709
3830.2
3826.195
3822.34
3919.561
4523.061
5842.666
7660.525
9572.344
11085.23
10995.055
10898.801
10795.769
10089.53
9782.586
8972.188
8162.8
7553.78
7046.078
5441.961
4741.113
4140.254
3536.031
2524.497
284.707
250.014
241.057
224.991
3117.251
3839.027
6859.773
8379.168
9495.355
97103.322
106108.933
109111.774
106116.06
132122.841
143124.735
127116.777
117108.021
9099.506
7873.015
6938.762
6326.701
5647.842
5557.154
5557.182
6148.737
6041.342
6134.835
6431.46
7240.424
9055.23
10271.375
11484.471
11483.272
10980.5
10780.649
12384.633
17987.377
12790.737
10994.107
112112.693
81123.287
Monitor Measurements
Model Estimates
Local Time
O3 [ppb]
Aldino, MD
Figures6a_6b
Figure 5a. Maximum Absolute Error for Comparison of Model-Estimated Ozone
Concentrations and Monitor Measurements
Figures6a_6b
52.021Garrison
71.697Ft. Holabird
Lake Clifton70.788
63.49157.37
62.72460.541
39.98964.216
40.62572.538
40.76964.676
33.37554.235
62.7861.476
56.34562.657
50.90360.18
52.99264.422
48.95954.182
54.84291.623
Cub Run Treatment Plant84.943
James S. Long Park97.371
Widewater Elementary School80.576
67.456Phelps Wildlife Management Area
64.40561.986
56.08260.357
5963.927
60.83571.057
65.82148.894
52.06936.678
59.20273.266
79.97435.135
Maryland
Virginia
Delaware
1990
1995
Max |Monitor Measurement - Model Estimate|[O3 ppb]
Figure7a
Figure 5b. Minimum Absolute Error for Comparison of Model-Estimated Ozone
Concentrations and Monitor Measurements
Figure7a
0.183Garrison
0.79Ft. Holabird
Lake Clifton1.01
0.5260.809
0.020.036
0.1420.137
0.291.133
1.0451.302
0.1410.308
0.0450.684
0.0110.981
0.2890.003
0.0570.331
0.870.76
0.4420.503
Cub Run Treatment Plant0.303
James S. Long Park0.509
Widewater Elementary School0.08
0.282Phelps Wildlife Management Area
1.5531.514
0.0510.445
0.2140.577
0.4680.432
0.5890.034
0.1890.286
1.7520.519
0.3640.053
Maryland
Virginia
Delaware
1990
1995
Min |Monitor Measurement - Model Estimate|[O3 ppb]
Figure7b
Figure 5c. Average Absolute Error for Comparison of Model-Estimated Ozone
Concentrations and Monitor Measurements
Figure7b
15.5787123288Garrison
27.1985753425Ft. Holabird
Lake Clifton31.5496164384
16.235569444420.2590555556
16.789808219216.7752739726
13.895589041115.071
15.650071428626.2787
16.013178082219.612369863
10.233605633816.6181232877
19.047136986318.4495205479
13.769847222219.062
16.827902777818.322369863
18.212666666721.0755342466
16.699333333321.0051780822
19.173042857118.5597123288
Cub Run Treatment Plant24.1999857143
James S. Long Park23.986109589
Widewater Elementary School25.6748472222
18.0699305556Phelps Wildlife Management Area
20.739676056324.6700416667
17.839986111117.7650416667
18.501126760618.8837183099
19.595971428620.446109589
17.287859154920.0214027778
13.705382352913.2268356164
17.625382352915.9515555556
20.333512.5694647887
Maryland
Virginia
Delaware
1990
1995
Avg |Monitor Measurement - Model Estimate|[O3 ppb]
Figures8a_8b
Figure 6a. Comparison of Model-Estimated Ozone Concentrations From
12-km and 4-km Resolution Domains for Greenbelt, Maryland Monitor Location
Figure 6b. Comparison of Model-Estimated Ozone Concentrations From
12-km and 4-km Resolution Domains for Suitland, Maryland Monitor Location
Figures8a_8b
49.73143.92951
45.74338.45752
44.33525.39754
42.15633.05351
38.31436.82243
34.335.61536
28.62529.42933
22.07819.44832
16.3280.86432
11.046025
6.8450.97418
9.2372.44119
19.9579.12525
26.16533.21437
44.4659.13162
68.42274.37584
81.30976.729103
81.26176.49107
81.10777.791103
83.63380.18594
86.39781.70496
86.83383.04583
77.4972.12261
58.17358.7237
40.22959.75623
24.14160.69619
18.49845.2069
10.50417.5544
0.50407
0.0021.5836
0.00237.98311
0.02644.1365
0.14541.9694
0.02322.8774
1.3255.0915
4.69813.31613
17.81926.23827
32.70936.90941
44.73551.58757
58.06462.96878
67.87871.75287
75.19373.139107
78.4273.428120
78.7574.131129
78.89177.214120
81.0182.22145
83.03482.368139
81.42280.897111
58.30866.76391
43.96463.09969
31.70166.12148
22.10156.75935
9.44653.08227
0.2352.04519
0.01840.54922
0.00828.03915
0.0021.02634
0027
1.2030.63819
3.5811.67318
31.2644.61928
51.78218.80440
72.54464.0860
97.7794.07896
101.20899.77198
98.55293.066102
94.25292.44695
95.35996.79190
98.22100.89789
100.575100.67391
101.65895.0885
97.67599.3280
83.85393.58266
Model Estimates - 12 km grid cells
Model Estimates - 4 km grid cells
Monitor Measurements
Local Time
O3 [ppb]
Figures9a_9b
62.9739.61265
45.73330.86667
32.60719.74665
25.3826.78662
22.46836.39563
22.85233.93161
20.9227.93558
18.17319.50351
14.22112.01945
7.6449.73937
2.2329.37625
4.93416.8225
16.67531.7238
34.01145.41153
53.6956.98267
68.01970.62279
80.95181.21480
89.90285.04677
96.43387.57583
99.60791.05590
100.16696.41995
99.55497.06593
96.0489.58886
88.08975.38672
72.90971.94463
67.53975.18152
66.09569.77545
56.91553.31644
40.36535.08947
30.69322.52849
26.37121.17443
23.18628.03240
21.13932.55642
15.12932.05338
6.97227.41834
13.6225.90736
27.04932.9137
34.61240.29628
46.38352.31458
63.14568.81182
77.92184.214105
92.91895.041120
104.381102.534124
109.844105.709119
113.003109.184108
113.676112.3105
113.09113.49498
111.656109.54185
99.824100.3266
80.26397.70450
63.01897.3740
59.04988.89438
54.363.135
46.06729.31235
38.36910.26632
32.9160.48329
35.1750.0131
32.597025
15.9050.39124
16.0873.09124
33.01533.96544
50.04448.61361
65.87143.08476
80.95259.578101
94.96390.346123
105.83106.537134
111.433113.097117
111.862111.191112
108.599116.02634895.6249999998
104.698127.72973
99.85125.08775
95.46396.65253
81.24694.13456
Model Estimates - 12 km grid cells
Model Estimates - 4 km grid cells
Monitor Measurements
Local Time
O3 [ppb]
49.73143.92951
45.74338.45752
44.33525.39754
42.15633.05351
38.31436.82243
34.335.61536
28.62529.42933
22.07819.44832
16.3280.86432
11.046025
6.8450.97418
9.2372.44119
19.9579.12525
26.16533.21437
44.4659.13162
68.42274.37584
81.30976.729103
81.26176.49107
81.10777.791103
83.63380.18594
86.39781.70496
86.83383.04583
77.4972.12261
58.17358.7237
40.22959.75623
24.14160.69619
18.49845.2069
10.50417.5544
0.50407
0.0021.5836
0.00237.98311
0.02644.1365
0.14541.9694
0.02322.8774
1.3255.0915
4.69813.31613
17.81926.23827
32.70936.90941
44.73551.58757
58.06462.96878
67.87871.75287
75.19373.139107
78.4273.428120
78.7574.131129
78.89177.214120
81.0182.22145
83.03482.368139
81.42280.897111
58.30866.76391
43.96463.09969
31.70166.12148
22.10156.75935
9.44653.08227
0.2352.04519
0.01840.54922
0.00828.03915
0.0021.02634
0027
1.2030.63819
3.5811.67318
31.2644.61928
51.78218.80440
72.54464.0860
97.7794.07896
101.20899.77198
98.55293.066102
94.25292.44695
95.35996.79190
98.22100.89789
100.575100.67391
101.65895.0885
97.67599.3280
83.85393.58266
Model Estimates - 12 km grid cells
Model Estimates - 4 km grid cells
Monitor Measurements
Local Time
O3 [ppb]
Figure 7a. Difference in Model-Estimated Ozone Concentrations Using Baseline Emissions Scenario 1
and Modified Emissions Scenario 2
6.1116.5636.5218.6613.4230.8560.6642.99210.3427.9417.6316.2929.6193.54
2.0315.9942.3424.8432.2572.1331.4071.2543.1964.2255.9058.1732.7930.256
1.5483.171.4650.0022.3476.8841.3881.1830.8560.1433.2912.6252.020.001
1.5631.8864.49901.9724.0821.111.3480.9050.0912.422.0672.1920
1.4322.0042.27401.9041.750.3411.6521.1620.0822.8211.9632.1890
1.0341.7451.79100.9282.451-0.6911.2281.3550.1651.871.8462.080
0.491.412.03200.0012.429-1.8610.5351.1050.2850.2461.6110.9640
-0.290.9722.247002.266-2.445-0.0360.3520.210.030.7150.20
-1.4520.450.198001.737-2.588-0.047-0.018-0.4140.0030.055-0.0040
-1.9010.0020.009000.779-2.71300.012-1.22300-0.0310
-1.3180.0460.4750.0670.0060.599-2.4750.0080.017-1.8530.2040-0.2070.007
-0.3960.2841.7840.820.0362.167-1.4470.0310.059-1.8650.580.104-0.2030.05
0.9793.2915.7885.5521.8573.7390.1470.4010.628-0.5365.7271.8651.8181.915
6.44711.2179.2579.42210.5586.1832.9716.7386.7563.55510.1949.427.6210.323
14.416.9298.3018.89215.995.0768.90715.41114.3629.5019.91715.95713.10316.007
17.7889.9240.671.30916.8211.39712.16620.8586.61914.5134.51510.53112.614.703
11.7833.198-0.863-1.11912.5430.7188.69423.4571.46315.8991.0633.17410.2819.825
7.6710.477-0.748-1.2079.877-0.0327.5523.7321.0559.2820.6310.9478.4677.032
5.515-0.208-0.703-0.83210.279-0.3336.83718.2531.144.9640.780.4817.8585.883
4.7450.082-0.873-0.88310.508-0.1437.55612.082.0143.2740.9880.6847.615.421
2.3350.475-0.911-0.95510.7781.088.2449.7142.4152.281.3921.1846.5185.657
2.6310.902-0.231-0.51511.173.8598.4867.33.3862.5912.0773.7785.1066.318
4.3244.8361.4620.6568.9958.1875.315.1685.5353.323.6914.0367.4898.005
2.1672.27514.9548.8690.26212.41-0.5672.1523.4682.2282.7032.6484.6730.004
1.9962.27412.68311.6670.00415.328-0.4873.132.2330.662.1982.0854.1920.008
2.9282.43713.38712.6160.00212.85-1.2880.0353.8721.6372.4073.5453.0430.02
0.814.345.0744.76606.657-2.1927.910.2581.51612.9360.0010.0010.002
0.5879.8152.1412.29202.4111.0494.8174.4181.1458.125.34400
9.0245.1661.521.5984.4740.6520.2242.1581.6594.3454.6266.29810.2030.54
4.8621.8931.5451.5832.9260.4921.0522.0741.0376.0212.5982.1915.323.501
2.9042.2771.7861.0812.1720.7160.0731.8110.964.8561.7011.6542.7432.019
0.0180.3131.2080.8251.3861.2221.0691.2120.8012.5841.0331.9241.7721.394
0.0570.3630.9350.8060.9141.0274.7660.8550.7541.4541.4511.2580.9920.966
0.5550.6331.161.0791.0770.8671.6561.0410.9930.921.3841.310.9451.05
1.4121.241.4951.3921.5451.1041.6531.4441.161.3041.7011.4151.7271.552
2.3642.7012.2962.2552.4681.8563.7822.0981.7442.2013.1542.772.8632.634
4.7875.7634.0063.6483.7173.3866.0653.3182.5093.6515.9286.5975.9154.281
8.7857.7045.6985.1644.2263.4539.8223.4792.5443.4875.4068.8259.3795.52
10.2976.5145.0485.8825.3542.48113.0883.3292.5362.2763.9786.4910.0417.428
7.7513.442.4424.1455.3761.66414.9843.983.2651.6473.23.947.2297.669
5.7692.6171.3172.2473.9111.48712.6014.1063.9071.9233.0883.2196.6026.248
4.8412.3541.0081.2342.8571.98811.4943.5433.9032.5824.4122.9826.8595.408
5.072.7371.2351.0092.1334.44412.1722.0443.5273.128.653.5797.9753.942
6.195.3451.3471.123.5268.19313.8791.2222.9623.27315.2076.6410.4264.212
7.9179.3392.8141.3285.63715.2814.791.3722.7323.09322.53711.31714.9516.5
6.76213.86110.0514.1566.19918.96812.9281.8692.012.37629.05616.28218.0518.205
5.54521.80613.4458.718.59217.1038.9064.1911.6542.28333.0116.44211.3959.727
5.0261.6516.28711.99310.17817.1457.274.5341.252.25431.1627.2526.9720.063
4.6489.20718.90915.1340.69120.1317.627.4810.8031.60526.6535.0737.5970
4.977.3619.8816.875024.5717.449.1895.3831.30219.4814.9167.8210
5.1555.48617.35813.884025.8166.9070.4316.755.7440.1315.3127.2050
5.1175.4944.8390018.76.11808.7225.0530.0085.7026.9040
4.9435.5774.9360011.6055.81109.2874.0204.7233.3410
5.8214.8562.139009.4365.30201.5033.6304.8830.3130
6.3024.6320.0490.00308.0025.66100.2544.16402.8336.1740
5.9220.3320.0010.00300.0224.37400.0284.81004.9040
4.8920.2090.5750.9840.0221.09-4.3570.0280.8415.430.1770.1654.6750.028
5.4910.3061.4633.1320.1191.989-0.0370.111.7816.3510.6710.361.610.139
5.6711.744.4344.0141.4496.8092.1570.432.0277.4751.9331.1224.5721.579
15.197.90112.57811.4439.96919.6426.5532.7296.0748.55417.5788.61912.718.594
20.39121.02520.06817.51616.48919.11510.06111.17913.36710.00416.23222.76912.8416.783
8.42729.08225.60525.09225.7833.4938.88822.58420.2624.719-0.16222.3513.42725.943
-2.9183.3516.36116.89828.154-1.1517.27437.70829.2143.11-3.0784.7680.80621.348
-4.151-2.85-2.201-0.3825.7-2.6173.88141.9930.993.531-4.253-0.2790.8542.982
-4.074-4.096-3.313-2.9041.426-4.0274.57115.65910.6083.751-4.983-1.5860.1420.269
-4.23-4.317-4.484-3.9540.407-4.9996.7943.2352.732-2.365-5.0630.354-0.443-0.423
-3.857-3.817-5.064-5.051-0.36-4.0019.964.362-1.016-4.159-4.5435.1250.131-1.197
-3.708-2.663-4.149-4.752-1.8120.00310.754.151-1.936-4.518-2.8099.1911.431-2.431
-3.643-2.1280.541-2.3023.0433.3076.131.8122.55-4.0664.7766.8181.7161.632
-1.6641.9434.0053.438.2686.458-1.246-0.406-0.3280.3096.2855.9371.8282.991
-3.2262.947.0446.6086.5377.123-1.6134.454-0.7543.8794.2624.8132.3555.387
Davidsonville
Ft. Meade
Garrison
Padonia
Essex
S. Carroll
S. MD
Edgewood
Aldino
Millington
Rockville
Greenbelt
Suitland
Ft. Holabird
Local Time
Figure 7b. Difference in Model-Estimated Ozone Concentrations Using Baseline Emissions Scenario 1
and Modified Emissions Scenario 3
Model Estimated O3 (w/ adjsuted emissions scenario) - O3 (w/ unadjusted emissions) [ppb]
6.1116.5636.5218.6613.4230.8560.6642.99210.3427.9417.6316.2929.6193.54
2.0315.9942.3424.8432.2572.1331.4071.2543.1964.2255.9058.1732.7930.256
1.5483.171.4650.0022.3476.8841.3881.1830.8560.1433.2912.6252.020.001
1.5631.8864.49901.9724.0821.111.3480.9050.0912.422.0672.1920
1.4322.0042.27401.9041.750.3411.6521.1620.0822.8211.9632.1890
1.0341.7451.79100.9282.451-0.6911.2281.3550.1651.871.8462.080
0.491.412.03200.0012.429-1.8610.5351.1050.2850.2461.6110.9640
-0.290.9722.247002.266-2.445-0.0360.3520.210.030.7150.20
-1.4520.450.198001.737-2.588-0.047-0.018-0.4140.0030.055-0.0040
-1.9010.0020.009000.779-2.71300.012-1.22300-0.0310
-1.3180.0460.4750.0670.0060.599-2.4750.0080.017-1.8530.2040-0.2070.007
-0.3960.2841.7840.820.0362.167-1.4470.0310.059-1.8650.580.104-0.2030.05
0.9793.2915.7885.5521.8573.7390.1470.4010.628-0.5365.7271.8651.8181.915
6.44711.2179.2579.42210.5586.1832.9716.7386.7563.55510.1949.427.6210.323
14.416.9298.3018.89215.995.0768.90715.41114.3629.5019.91715.95713.10316.007
17.7889.9240.671.30916.8211.39712.16620.8586.61914.5134.51510.53112.614.703
11.7833.198-0.863-1.11912.5430.7188.69423.4571.46315.8991.0633.17410.2819.825
7.6710.477-0.748-1.2079.877-0.0327.5523.7321.0559.2820.6310.9478.4677.032
5.515-0.208-0.703-0.83210.279-0.3336.83718.2531.144.9640.780.4817.8585.883
4.7450.082-0.873-0.88310.508-0.1437.55612.082.0143.2740.9880.6847.615.421
2.3350.475-0.911-0.95510.7781.088.2449.7142.4152.281.3921.1846.5185.657
2.6310.902-0.231-0.51511.173.8598.4867.33.3862.5912.0773.7785.1066.318
4.3244.8361.4620.6568.9958.1875.315.1685.5353.323.6914.0367.4898.005
2.1672.27514.9548.8690.26212.41-0.5672.1523.4682.2282.7032.6484.6730.004
1.9962.27412.68311.6670.00415.328-0.4873.132.2330.662.1982.0854.1920.008
2.9282.43713.38712.6160.00212.85-1.2880.0353.8721.6372.4073.5453.0430.02
0.814.345.0744.76606.657-2.1927.910.2581.51612.9360.0010.0010.002
0.5879.8152.1412.29202.4111.0494.8174.4181.1458.125.34400
9.0245.1661.521.5984.4740.6520.2242.1581.6594.3454.6266.29810.2030.54
4.8621.8931.5451.5832.9260.4921.0522.0741.0376.0212.5982.1915.323.501
2.9042.2771.7861.0812.1720.7160.0731.8110.964.8561.7011.6542.7432.019
0.0180.3131.2080.8251.3861.2221.0691.2120.8012.5841.0331.9241.7721.394
0.0570.3630.9350.8060.9141.0274.7660.8550.7541.4541.4511.2580.9920.966
0.5550.6331.161.0791.0770.8671.6561.0410.9930.921.3841.310.9451.05
1.4121.241.4951.3921.5451.1041.6531.4441.161.3041.7011.4151.7271.552
2.3642.7012.2962.2552.4681.8563.7822.0981.7442.2013.1542.772.8632.634
4.7875.7634.0063.6483.7173.3866.0653.3182.5093.6515.9286.5975.9154.281
8.7857.7045.6985.1644.2263.4539.8223.4792.5443.4875.4068.8259.3795.52
10.2976.5145.0485.8825.3542.48113.0883.3292.5362.2763.9786.4910.0417.428
7.7513.442.4424.1455.3761.66414.9843.983.2651.6473.23.947.2297.669
5.7692.6171.3172.2473.9111.48712.6014.1063.9071.9233.0883.2196.6026.248
4.8412.3541.0081.2342.8571.98811.4943.5433.9032.5824.4122.9826.8595.408
5.072.7371.2351.0092.1334.44412.1722.0443.5273.128.653.5797.9753.942
6.195.3451.3471.123.5268.19313.8791.2222.9623.27315.2076.6410.4264.212
7.9179.3392.8141.3285.63715.2814.791.3722.7323.09322.53711.31714.9516.5
6.76213.86110.0514.1566.19918.96812.9281.8692.012.37629.05616.28218.0518.205
5.54521.80613.4458.718.59217.1038.9064.1911.6542.28333.0116.44211.3959.727
5.0261.6516.28711.99310.17817.1457.274.5341.252.25431.1627.2526.9720.063
4.6489.20718.90915.1340.69120.1317.627.4810.8031.60526.6535.0737.5970
4.977.3619.8816.875024.5717.449.1895.3831.30219.4814.9167.8210
5.1555.48617.35813.884025.8166.9070.4316.755.7440.1315.3127.2050
5.1175.4944.8390018.76.11808.7225.0530.0085.7026.9040
4.9435.5774.9360011.6055.81109.2874.0204.7233.3410
5.8214.8562.139009.4365.30201.5033.6304.8830.3130
6.3024.6320.0490.00308.0025.66100.2544.16402.8336.1740
5.9220.3320.0010.00300.0224.37400.0284.81004.9040
4.8920.2090.5750.9840.0221.09-4.3570.0280.8415.430.1770.1654.6750.028
5.4910.3061.4633.1320.1191.989-0.0370.111.7816.3510.6710.361.610.139
5.6711.744.4344.0141.4496.8092.1570.432.0277.4751.9331.1224.5721.579
15.197.90112.57811.4439.96919.6426.5532.7296.0748.55417.5788.61912.718.594
20.39121.02520.06817.51616.48919.11510.06111.17913.36710.00416.23222.76912.8416.783
8.42729.08225.60525.09225.7833.4938.88822.58420.2624.719-0.16222.3513.42725.943
-2.9183.3516.36116.89828.154-1.1517.27437.70829.2143.11-3.0784.7680.80621.348
-4.151-2.85-2.201-0.3825.7-2.6173.88141.9930.993.531-4.253-0.2790.8542.982
-4.074-4.096-3.313-2.9041.426-4.0274.57115.65910.6083.751-4.983-1.5860.1420.269
-4.23-4.317-4.484-3.9540.407-4.9996.7943.2352.732-2.365-5.0630.354-0.443-0.423
-3.857-3.817-5.064-5.051-0.36-4.0019.964.362-1.016-4.159-4.5435.1250.131-1.197
-3.708-2.663-4.149-4.752-1.8120.00310.754.151-1.936-4.518-2.8099.1911.431-2.431
-3.643-2.1280.541-2.3023.0433.3076.131.8122.55-4.0664.7766.8181.7161.632
-1.6641.9434.0053.438.2686.458-1.246-0.406-0.3280.3096.2855.9371.8282.991
-3.2262.947.0446.6086.5377.123-1.6134.454-0.7543.8794.2624.8132.3555.387
Davidsonville
Ft. Meade
Garrison
Padonia
Essex
S. Carroll
S. MD
Edgewood
Aldino
Millington
Rockville
Greenbelt
Suitland
Ft. Holabird
Local Time
Figure 8a. Maximum Absolute Error for Comparison of Model-Estimated Ozone
Concentrations and Monitor Measurements, Using Various Emissions Scenarios
Figure 8b. Average Absolute Error for Comparison of Model-Estimated Ozone
Concentrations and Monitor Measurements, Using Various Emissions Scenarios
Model Estimated O3 (w/ adjsuted emissions scenario) - O3 (w/ unadjusted emissions) [ppb]
59.64660.1563.491
72.41367.09862.724
46.22743.17539.989
48.02248.18840.625
57.36852.76240.769
43.02135.65833.375
49.96156.1462.78
61.52658.31152.021
56.21554.69556.345
64.73468.70671.697
49.4750.21250.903
52.89452.95752.992
46.92446.30348.959
45.89452.23854.842
Max |Monitor Measurement - Model Estimate|[O3 ppb]
Model Estimates:
Using Adjusted Emissions Scenario 2
Using Adjusted Emissions Scenario 1
Using Unadjusted Emissions Scenario
Monitor Location
16.235569444416.481763888916.7168472222
16.789808219217.941931506821.2543835616
13.895589041115.423452054815.6892191781
15.650071428617.814971428619.4380285714
16.013178082217.365465753416.8973835616
10.233605633810.818352112713.5702535211
19.047136986319.265479452118.9338493151
15.578712328818.258410958919.6601643836
13.769847222213.379902777814.107375
27.198575342526.698068493225.6354794521
16.827902777818.080972222219.6541527778
18.212666666717.432055555618.2276111111
16.699333333316.960063492117.9296984127
19.173042857118.249142857117.8094
Max |Monitor Measurement - Model Estimate|[O3 ppb]
Model Estimates:
Using Adjusted Emissions Scenario 2
Using Adjusted Emissions Scenario 1
Using Unadjusted Emissions Scenario
Monitor Location
Figure 9a. Comparison of 8-Hour Average Model-Estimated Ozone Concentrations
and Monitor Measurements for Millington, Maryland
Figure 9b. Comparison of 8-Hour Average Model-Estimated Ozone Concentrations
and Monitor Measurements for Davidsonville, Maryland
49.2551.373125
45.7548.765125
43.12547.273
40.12545.888375
37.2543.607125
34.87541.40275
34.7540.006125
37.2539.754625
40.87540.899875
46.87544.04125
54.12548.976
63.12555.0705
72.37561.98975
81.2569.4435
87.7576.422625
91.12582.09825
91.7586.290875
90.12588.33775
87.12587.6685
82.585.831625
77.37583.28
71.579.87575
64.7575.932
58.2567.18275
52.556.85375
45.87548.106375
39.62538.933625
34.37532.832625
30.37528.769125
28.12525.911625
28.87523.68875
27.857142857127.235125
27.666666666733.903375
36.539.8725
45.666666666747.6155
54.333333333352.911
61.166666666756.891875
65.666666666760.65625
67.564.402125
67.142857142967.07925
66.2568.336625
62.2569.063875
60.569.420375
59.2568.688375
5866.18225
56.12565.341375
55.87564.251
54.37562.897625
52.12561.1955
53.2559.264
50.62556.868875
46.37554.38525
43.7552.249625
4349.57175
43.37549.584625
45.7552.303625
49.557.030875
54.12563.377875
61.7571.286625
7380.434875
84.589.085875
94.62595.789375
101100.023125
103.625102.068
104.5103.022875
103.25102.757875
Monitor Measurements
Model Estimates
Local Time
8-hr avg O3 [ppb]
Millington, Maryland
72.62535.079875
6830.6825
63.2526.503375
58.2523.353125
53.2521.419375
49.37521.20125
48.2523.40625
48.7528.18475
51.2535.287625
5543.134125
62.551.090875
72.12559.50575
83.12568.257125
92.62576.307875
99.582.05325
104.37584.821
106.87584.53325
107.583.051125
103.7580.404625
97.2575.63175
8968.715875
80.87560.343375
72.7552.406375
65.546.084625
5940.83875
5336.00075
48.37531.502625
45.37528.0835
43.526.277125
43.2527.427875
4330.865375
43.2535.185625
47.540.840625
55.7547.988875
6656.6305
77.87566.661625
9177.19975
102.586.481125
113.2594.365875
121101.014875
123.375105.496
121.25107.793125
116.125108.10625
107.625107.56725
96106.83875
84.25104.932625
70.5100.063625
57.87590.884375
47.2580.803
37.12571.759625
28.37563.262625
21.37553.19775
17.37542.578875
16.62535.599375
20.87531.880625
27.533.425
35.539.297625
45.87547.699375
5857.421125
7268.448
85.579.092625
96.12588.286
103.87598.299125
108.25108.35925
109115.8485
107.875117.47425
Monitor Measurements
Model Estimates
Local Time
8-hr avg O3 [ppb]
Davidsonville, Maryland
-
Modeling estimatesMonitor in countySpatial interpolation of monitors1-Hour max O3 (ppb)County-level Exposure Estimates Bell Environ Int 2006
-
1-Hour max O3 (ppb)Nearest monitorIndividual-Level Exposure EstimatesBell Environ Int 2006
-
Ozone monitors in Georgia 2000 Persons / Sq. Mile
-
Holford et al. Statistics in Medicine AcceptedExample: Traffic Modeling to Estimate Exposure
-
Estimated NO2 (traffic) Levels for New Haven County Area (2002)Holford et al. Statistics in Medicine Accepted
-
Exposure Assessment for Studies of Air Pollution and HealthBasic health effects modelMethods of measuring exposureKey challenges in assessing exposureSpatial misalignmentMultiple pollutant exposuresSpecial case of particulate matterCurrent and upcoming approaches to estimating exposureOther challenges
-
Other ChallengesOther factors affecting certainty of monitor valuesDetection limits of monitorsMeasurement error (see co-located monitors)Other factors that affect exposure and variation of exposureMovement through the communityIndoor/outdoor activity patternsBehaviors and activities (e.g., AC, jogging)Differences between exposure and dose
-
Thank youKey CollaboratorsFrancesca Dominici, Harvard UniversityRoger D. Peng, Johns Hopkins UniversityKeita Ebisu, Yale UniversitySponsorsNational Institute for Environmental Health Sciences (NIEHS)Health Effects InstituteU.S. Environmental Protection Agency-sponsored Johns Hopkins Particulate Matter Research Center
****Time-series model, assessing acute exposuresKey confounders: day of the week, temporal trends, temperatureCould talk more about why weather is important*Time-series model, assessing acute exposuresKey confounders: day of the week, temporal trends, temperatureCould talk more about why weather is important*Ambient monitor measures PM10 and PM2.5 (photo 2)Photo 1 is at a school.
Hybrid approaches (time activity diary, monitors at homes).*Ambient monitors more likely to be in urban locations.
Neither of these are biomarkers or actual dose.**Another issue is time: people move around. Time activity patterns.*Correlations plotted on Fishers z-transform scale.
Based on national U.S. data over a several year period.*Beta and z are vectors representing other confounders (day of the week, etc.)*We are calculating the estimated beta w based on wt (estimated exposure), but we want the beta we would get if we had xt (the true exposure).
What is the problem with E[ut] = 0.To estimate tau^2 (spatial misalignment error variance) var we need multiple mons. This var represents the spatial misalignment error variance for a specific area.This value goes down as the number of monitors increases.*Effect of spatial misalignment is function of spatial variability of pollutant and monitor coverage
******What is the relative toxicity of various PM components?Can the spatial and temporal variation in PM component concentrations explain spatial and temporal differences in PM-health relationships?What sources of PM are most harmful?
****These differences potentially explain spatial and temporal differences in PM health effects estimates. As an example, Roger Peng found higher effect estimates for PM10 and mortality in the Northeast and in Summer, whereas this work led by Francesca Dominici finds strong East/West patterns in the effect estimates for PM2.5 and cause-specific hospital admissions.
204 US Counties, 1999-2002 Cerebrovascular disease: any abnormality of the brain resulting from a pathologic process of the blood vessels Ischemic heart disease: narrowed heart arteriesPeripheral vascular disease: diseases of blood vessels outside the heart and brain. It's often a narrowing of vessels that carry blood to the legs, arms, stomach or kidneysCOPD: slow gradual disease characterized by loss of lung function. Includes chronic bronchitis
Beta and z are vectors representing other confounders (day of the week, etc.)*Iw = 0/1 indicator variable for winter, etc.Seasonal interaction model: also replaced ns(Time) with interaction terms (time x season)*Figure 1. Percent increase in CVD hospital admissions rate per 10 microggm/m3 increase in lag 0 PM2.5Note: Seasonal interaction model results are shown in red and harmonic model results in black. Dashed lines reflect 95% posterior interval.
***Domain 1: 108-km gridcell resolutionDomain 2: 36-km gridcell resolutionDomain 3: 12-km gridcell resolutionDomain 4: 4-km gridcell resolution
Areas of trouble:Nighttime averages too lowDifficulty capturing highest peaksSo what exposure window you consider is critical.
This is not a statistical model. Physical/chemical processes.**Why county? Often health data is available on an aggregated scale often county.
Spatial interpolation is inverse distance weighting. Also applied kriging, but same idea.(correlation between model county estimates and Fig 2 0.61, and Fig 3 0.66)Only 12% of the counties have monitors. Those counties have statistically higher population density and modeled ozone concentrations.
Also note that for Fig. 2 (very common approach), is based on monitors within the county, regardless of where they are.*Case study episode August 15, hour 00 (GMT) to August 18, hour 00 1995 in N. Georgia. Highest hourly value recorded. Eight monitors.
Fig1. Displays strong spatial heterogeneity.Fig2. Nearest monitor. Does include monitors outside the domain, so if an area is closest to a monitor other than these 8, those values are used.
Example: Henry County: modeling 53 ppb, monitor: 12 ppb.Issue with some locations being far away from monitors. Some populations in Bibb county (SE portion) are >90 km from the nearest monitor, and some parts of southern Georgia are even further.
*Linked with urban settings with high population:+ useful for hh studies, but if we had estimates at other locations:Larger sample size (# of people)Exploration of urban vs. rural differences (different pollution mixture)
By 2000, Georgia had 21 ozone monitors in counties averaging the top 12th percentile of population.0,0 point is study subject in center of buffer.This buffer shape assumes an isotropic surface field for pollution levels (unaffected by direction). Could imagine a different shape.
D = specified buffer distance around study subjectd = distance between a node (location on road) and study subjectOpen circles represent nodes dividing roads into segments (C), each of which have a value for traffic volumeS divides each Segment (C) into sub-segments, which each have a distance to study subject and a certain lengthAnnual estimate of traffic volume available for each segment
If we new the dispersion parameters, could use a Gaussian function, but they used a step function, among other approaches, for different distances. Compared to NO2 levels.*Circles are NO2 monitoring locations**We have better estimates of exposure for some pollutants than others.*