AERMOD performance evaluation for three coal-fired electrical generating units in Southwest Indiana

Figures & data

Figure 1. Site map of Gibson (a) and Petersburg/Ratts (b) study areas. The wind rose on the Gibson map represents wind speed and direction at the meteorological tower that is colocated with the East Road SO₂ monitor and the wind rose on the Petersburg/Ratts map indicates wind speed and direction at the meteorological tower that is colocated with the Pike SO₂ monitor.

Table 1. Summary of emissions data for Gibson, Petersburg, and Ratts generating stations

Source	Stack height (m)	Stack diameter (m)	Average annual hourly SO2 emission rate (g/sec)	Maximum annual hourly SO2 emissions (g/sec)	Exit temperature (K)	Average stack exit velocity (m/sec)
Gibson 1 (Boilers 1 and 2)	189.0	23.5	116.0	462.2	327.0	15.6
Gibson 2 (Boiler 3)	189.0	25.0	90.9	794.1	327.0	18.5
Gibson 3 (Boiler 4)	152.4	25.0	111.3	460.3	322.0	17.4
Gibson 4 (Boiler 5)	152.4	25.0	290.0	819.8	328.0	12.5
Petersburg 1	166.7	9.0	17.6	141.7	323.0	10.8
Petersburg 2	189.3	6.1	35.2	423.4	324.0	9.9
Petersburg 3	189.9	6.7	155.3	1340.3	342.0	19.5
Petersburg 4	186.2	6.7	517.9	2314.5	343.0	22.1
Ratts 1	91.4	3.4	320.5	631.5	412.0	17.7
Ratts 2	91.4	3.4	292.9	736.6	412.0	19.2

Table 2. Summary of surface characteristics for Gibson, P/R, and KEVV

	Average surface roughness			Average noontime albedo			Bowen ratio
Month	Gibson	P/R	KEVV	Gibson	P/R	KEVV	Gibson	P/R	KEVV
January	0.021	0.042	0.021	0.28	0.28	0.25	0.44	0.40	0.25
February	0.022	0.044	0.022	0.31	0.32	0.27	0.44	0.40	0.25
March	0.031	0.064	0.039	0.15	0.14	0.14	0.41	0.34	0.25
April	0.030	0.056	0.029	0.15	0.14	0.14	0.41	0.34	0.25
May	0.031	0.054	0.047	0.15	0.14	0.14	0.41	0.34	0.25
June	0.060	0.211	0.057	0.19	0.19	0.17	0.53	0.47	0.32
July	0.071	0.216	0.086	0.19	0.19	0.17	0.53	0.47	0.32
August	0.074	0.214	0.135	0.19	0.19	0.17	1.43	1.26	0.63
September	0.059	0.212	0.095	0.19	0.19	0.17	2.05	1.81	0.8
October	0.053	0.211	0.076	0.20	0.20	0.18	2.05	1.81	0.8
November	0.064	0.158	0.090	0.21	0.21	0.19	2.05	1.81	0.8
December	0.022	0.034	0.031	0.26	0.26	0.23	2.05	1.81	0.8

Table 3. DV ratios (predicted DV/observed DV) for 2010 at the six monitor–receptor sites under five modeling scenarios; “OS” and “NWS” refer to onsite and NWS meteorological data runs

		AER v12060		AER v12345 No ADJ_U*	AER v12345 ADJ_U*	AER v12345 ADJ_U* & LOWWIND 1	AER v12345 ADJ_U* & LOWWIND 2
Site	Observed DV (ppb)	DV ratio (OS)	DV ratio (NWS)	DV ratio (NWS)	DV ratio (NWS)	DV ratio (NWS)	DV ratio (NWS)
Mt Carmel	66	2.96	1.00	0.93	0.93	1.27	0.92
East Road	45	3.20	1.48	1.47	1.47	1.99	1.44
Coal Road	74	3.08	1.23	1.12	1.12	1.40	1.11
Schrodt	59	n/a	0.98	0.97	0.97	1.01	0.95
Gibson-wide average	61	n/a	1.15	1.10	1.10	1.38	1.08
Daviess	115	1.15	1.17	1.35	1.35	1.43	1.31
Pike	211	1.42	1.10	1.06	1.06	1.38	1.06
P/R-wide average	163	1.32	1.13	1.16	1.16	1.40	1.15

Table 4. Robust Highest Concentration ratios (predicted RHC/observed RHC) for 2010 at the six monitor–receptor sites under five modeling scenarios; “OS” and “NWS” refer to onsite and NWS meteorological data

Site		AER v12060		AER v12345: No ADJ_U*	AER v12345: ADJ_U*	AER v12345 ADJ_U* & LOWWIND 1	AER v12345 ADJ_U* & LOWWIND 2
	Observed RHC (ppb)	RHC ratio (OS)	RHC ratio (NWS)	RHC Ratio (NWS)	RHC Ratio (NWS)	RHC Ratio (NWS)	RHC Ratio (NWS)
Mt Carmel	144	2.08	0.54	0.56	0.56	0.80	0.54
East Road	63	3.37	1.36	1.36	1.36	1.71	1.34
Coal Road	113	2.47	1.03	0.93	0.93	1.17	0.93
Schrodt	88	n/a	0.97	0.97	0.97	1.04	0.96
Gibson-wide average	102	n/a	0.90	0.88	0.88	1.10	0.87
Daviess	156	1.15	1.35	1.38	1.38	1.69	1.30
Pike	282	1.32	1.28	1.33	1.33	1.50	1.32
P/R-wide average	219	1.25	1.30	1.34	1.34	1.57	1.31

Figure 2. Q–Q plots of monitor vs. model hourly SO₂ data for 2010 Gibson–Coal Road (a) and Petersburg/Ratts–Pike (b) data for five modeling scenarios: (1) AERMOD v12060, (2) AERMOD v12345 with No ADJ_U*, (3) AERMOD v12345 with ADJ_U*, (4) AERMOD v12345 with ADJ_U* and LOWWIND1, and 5() AERMOD v12345 with ADJ_U* and LOWWIND2. Coal Road and Pike data are presented because these two monitors are most often downwind of the source. Solid lines represent 1:1 agreement between the monitor and model, and the dotted lines represent a factor of 2 over- or underprediction.

Figure 3. Hourly scatterplots of Gibson–Coal Road (a) and Petersburg/Ratts–Pike (b) for AERMOD v12345 with No ADJ_U* scenario.

Table 5. Analysis of paired hourly monitor and model data using AERMOD v12345 No ADJ_U* model run with NWS meteorology (C_p = predicted concentration, C_o = observed concentration)

Site	Mean model bias (95% CI)	Fractional bias (95% CI)	Factor of 2	NMSE (95% CI)	Geometric mean model bias	Geometric variance
Mt Carmel (n = 3347)	−0.51 (–0.82, –0.20)	−0.14 (–0.16, –0.12)	0.80	3.73 (2.43, 5.04)	0.86	1.82
East Road (n = 4715)	−0.77 (–0.85, –0.69)	−0.36 (–0.38, –0.34)	0.78	2.18 (1.74, 2.62)	0.68	1.84
Coal Road (n = 7496)	−0.12 (–0.30, 0.05)	−0.40 (–0.41, –0.39)	0.74	3.67 (3.14, 4.21)	0.66	2.09
Schrodt (n = 5916)	−1.36 (–1.47, –1.25)	−0.53 (–0.55, –0.52)	0.68	2.02 (1.63, 2.41)	0.56	2.06
Daviess (n = 6816)	0.17 (–0.13, 0.46)	0.01 (–0.004, 0.02)	0.83	5.20 (4.29, 6.12)	1.02	1.73
Pike (n = 8077)	−0.71 (–1.10, –0.31)	−0.18 (–0.19, –0.16)	0.77	6.52 (5.07, 7.97)	0.82	2.14
Good model ^a ^, ^b		≤ ±0.3 ^a –0.5 ^b	≥ 0.5 ^a , ≥ 0.8 ^b	≤ 3 ^a , 0.5 ^b	0.75–1.25 ^b	1–1.25 ^b
Notes: ^aAs defined by Chang and Hanna (2004) for rural dispersion model evaluation.
^b Kumar et al. (2006).

Table 6. A summary of performance measures for the top 5% of hourly monitor and model data at all six sites using AERMOD v12345 No ADJ_U* model run with NWS meteorology (C_p = predicted concentration, C_o = observed concentration)

Site	Mean model bias (95% CI)	Fractional bias (95% CI)	Factor of 2	NMSE (95% CI)	Geometric mean model bias		Geometric variance
Mt Carmel (n = 626)	−1.19 (–2.83, 0.44)	−0.10 (–0.19, –0.01)	0.30	2.73 (1.77, 3.68)	0.87		10.20
East Road (n = 626)	0.24 (–0.94, 1.42)	−0.11 (–0.19, –0.02)	0.35	1.77 (1.41, 2.12)	0.88		7.47
Coal Road (n = 694)	7.04 (5.27, 8.82)	0.27 (0.19, 0.36)	0.27	1.91 (1.64, 2.17)	1.48		17.44
Schrodt(n = 680)	−1.56 (–0.63, –2.49)	−0.33 (–0.41, –0.26)	0.43	1.87 (1.47, 2.27)	0.66		5.47
Daviess (n = 719)	1.74 (–1.01,4.49)	−0.003 (–0.10, 0.10)	0.14	3.05 (2.54, 3.57)	1.09		48.30
Pike (n = 677)	−2.48 (–7.12, 2.15)	−0.02 (–0.12, 0.08)	0.22	2.29 (1.78, 2.80)	0.95		53.10
Good model ^a ^, ^b		≤ ±0.3 ^a –0.5 ^b	≥ 0.5 ^a , ≥ 0.8 ^b	≤ 3 ^a , <0.5 ^b	0.75–1.25 ^b		1–1.25 ^b
Notes: ^aAs defined by Chang and Hanna (2004) for rural dispersion model evaluation.
^b Kumar et al. (2006).

Table 7. A summary of performance measures for the top 5% of Gibson–Coal Road and P/R–Pike monitor and model concentrations using AERMOD v12345 No ADJ_U* model run with NWS meteorology (C_p = predicted concentration, C_o = observed concentration) under various meteorological conditions

Wind speed/stability regime	Mean model bias (95% CI)	Fractional bias (95% CI)	Factor of 2	NMSE (95% CI)	Geometric mean model bias	Geometric variance
Coal Road
<2 m/sec (n = 116)	6.52 (1.46, 11.58)	0.21 (–0.01, 0.44)	0.26	3.12 (1.75, 4.49)	1.44	20.43
>2 m/sec (n = 576)	7.23 (5.35, 9.12)	0.29 (0.19, 0.39)	0.27	1.72 (1.49, 1.95)	1.50	16.95
Unstable (n = 422)	10.46 (8.29, 12.64)	0.53 (0.43, 0.64)	0.26	2.09 (1.69, 2.48)	2.14	15.63
Stable (n = 73)	−14.50 (–18.15, –10.84)	−1.08 (–1.24, –0.93)	0.25	4.15 (1.81, 6.49)	0.23	27.75
Neutral (n = 196)	8.06 (4.59, 11.53)	0.24 (0.08, 0.41)	0.30	1.57 (1.25, 1.88)	1.40	18.74
<2 m/sec, Unstable (n = 99)	9.61 (3.90, 15.32)	0.46 (0.23, 0.69)	0.27	3.21 (1.77, 4.65)	2.03	19.16
<2 m/sec, Stable (n = 13)	−11.08 (–14.31, –7.85)	−1.18 (–1.47, –0.89)	0.23	2.53 (1.02, 4.05)	0.21	25.25
Pike
<2 m/sec (n = 140)	9.26 (–4.18, 22.70)	0.21 (–0.02, 0.45)	0.15	4.06 (1.77, 6.36)	0.86	48.90
>2 m/sec (n = 537)	−5.54 (–10.21, –0.87)	−0.08 (–0.19, 0.03)	0.23	1.86 (1.58, 2.13)	0.95	53.10
Unstable (n = 487)	10.61 (5.08, 16.14)	0.41 (0.30, 0.52)	0.24	2.29 (1.63, 2.95)	1.83	29.18
Stable (n = 38)	−38.55 (–44.61, –32.49)	−1.71 (–1.74, –1.68)	0.00	15.00 (7.79, 22.20)	0.07	1251.49
Neutral (n = 148)	−35.40 (–43.47, –27.32)	−0.97 (–1.14, –0.80)	0.20	2.63 (2.06, 3.20)	0.23	157.43
<2 m/sec, Unstable (n = 120)	16.47 (1.16, 31.77)	0.53 (0.30, 0.76 )	0.18	4.02 (1.69, 6.35)	2.27	51.58
<2 m/sec, Stable (n = 17)	−33.35 (–37.15, –29.55)	−1.69 (–1.73, –1.65)	0.00	10.90 (8.41, 13.40)	0.08	655.05
Good model ^a ^, ^b		≤ ±0.3 ^a –0.5 ^b	≥ 0.5 ^a , ≥ 0.8 ^b	≤ 3 ^a , <0.5 ^b	0.75–1.25 ^b	1–1.25 ^b
Notes: ^aAs defined by Chang and Hanna (2004) for rural dispersion model evaluation.
^b Kumar et al. (2006).

Figure 4. Adjusted box and whisker plots of C_p/C_o at Coal Road for the upper 5% of data for each of eight meteorological scenarios: (a) wind speed <2 m/sec, (b) wind speed >2 m/sec, (c) wind speed >4 m/sec, (d) unstable, (e) stable, (f) neutral, (g) low wind and unstable conditions, (h) low wind and stable conditions. The thick, dark line represents the median and the notches in each box represent the 95% confidence interval around the median. The unfilled circle corresponds to the population mean with error bars representing the 95% confidence interval around the mean. The shaded area represents the range of values acceptable for a “good model,” as defined by Chang and Hanna (2004) and Kumar et al. (2006).

Chang , J.C. and Hanna , S.R. 2004 . Air quality model performance evaluation . Meteorol. Atmos. Phys , 87 ( 1–3 ) : 167 – 196 . doi: 10.1007/s00703-003-0070-7

 Web of Science ®Google Scholar

Kumar , A. , Dixit , S. , Varadarajan , C. , Vijayan , A. and Masuraha , A. 2006 . Evaluation of the AERMOD dispersion model as a function of atmospheric stability for an urban area . Environ. Prog , 25 : 141 – 151 . doi: 10.1002/ep.10129

 Google Scholar