- Email: [email protected]
Comparison of AERMOD, ADMS and ISC3 for incomplete upper air meteorological data (case study: Steel plant)
Atmospheric Pollution Research xxx (2017) 1e6
H O S T E D BY
Contents lists available at ScienceDirect
Atmospheric Pollution Research journal homepage: http://www.journals.elsevier.com/locate/apr
Comparison of AERMOD, ADMS and ISC3 for incomplete upper air meteorological data (case study: Steel plant) Mostafa Kalhor a, *, Mehrshad Bajoghli b a b
HPA Company, 2d Floor, No 4, Building 314, Valiasr Square, 1439914153, Tehran, Iran Department of Environmental Science, Isfahan (Khorasgan) Branch, Islamic Azad University, Isfahan, Iran
a r t i c l e i n f o
a b s t r a c t
Article history: Received 3 January 2017 Received in revised form 31 May 2017 Accepted 4 June 2017 Available online xxx
In this paper three well known Gaussian dispersion models have been evaluated for a case study of a steel plant using complete and incomplete upper air meteorological data. In developing countries, the availability of surface and upper air meteorological data is limited. AMS/ EPA Regulatory Model (AERMOD), Advanced Dispersion Modeling System (ADMS) and Industrial Source Complex Model (ISC3) have been evaluated for both real and estimated upper meteorological data and the results have been compared with field measurements both in the horizontal and vertical directions. The results show significant differences in predicted concentrations when modeling with real (actual) and estimated upper meteorological data. The differences ranged from 100% to 450%. Comparison of model performance suggests that AERMOD and ADMS with real meteorological data produce consistent results in the horizontal direction while ISC3 output over-predicts in general. In AERMOD and ISC3 the predicted concentrations have a similar trend of variation in the vertical direction but in ADMS the concentration variation in the vertical direction exhibited different trends. In general, the ADMS predicted concentrations under-estimated field observations. The paper suggests that upper data must be used for modeling and the default values must be used with care. In absence of upper meteorological data, users could estimate upper meteorological data by different available algorithm rather than only default option of models. © 2017 Turkish National Committee for Air Pollution Research and Control. Production and hosting by Elsevier B.V. All rights reserved.
Keywords: AERMOD ADMS ISC3 Upper air meteorological data Steel plant
1. Introduction Gaussian dispersion models have been used widely for concentration prediction, sampling network design, EIA (Environmental Impact Assessment) and environmental management scenarios (EPA, 2015). The relative simplicity of use, quick setup, acceptable accuracy and wide applicability in different atmospheric condition are the advantages of these models (EPA, 2015). In developing countries, the using of these models has several constraints in both source emission determination and availability of meteorological data. In these countries accurate and sequential meteorological data are not available (especially upper meteorological data). Another problem is the location of industries. Most
* Corresponding author. E-mail address: [email protected] (M. Kalhor). Peer review under responsibility of Turkish National Committee for Air Pollution Research and Control.
plants and industrial facilities have been constructed outside of large cities where meteorological sites are not available for modeling (Carbonell et al., 2010). In this paper three different models (i.e., ADMS, AERMOD, and ISC3) have been evaluated. 1.1. AERMOD (AMS/EPA Regulatory Model) AERMOD is the Gaussian air dispersion model which incorporates building downwash algorithms, advanced depositional parameters, local terrain effects, and advanced meteorological turbulence calculations. AERMOD could be implemented with both real and estimated upper meteorological data (EPA, 2004). The minimum surface meteorological data for running AERMOD are (EPA, 2004): 1. 2. 3. 4.
Year, Month, Day, Hour Wind Speed Wind Direction Dry Bulb Temperature
http://dx.doi.org/10.1016/j.apr.2017.06.001 1309-1042/© 2017 Turkish National Committee for Air Pollution Research and Control. Production and hosting by Elsevier B.V. All rights reserved.
Please cite this article in press as: Kalhor, M., Bajoghli, M., Comparison of AERMOD, ADMS and ISC3 for incomplete upper air meteorological data (case study: Steel plant), Atmospheric Pollution Research (2017), http://dx.doi.org/10.1016/j.apr.2017.06.001
2
M. Kalhor, M. Bajoghli / Atmospheric Pollution Research xxx (2017) 1e6
5. Cloud Cover (tenths)
1.2. Advanced Dispersion Modeling System (ADMS) ADMS is a Gaussian air dispersion model used to evaluate industrial impacts on air quality. ADMS incorporates building effects, complex terrain, coastlines and variations in surface roughness; dry and wet deposition; chemistry schemes; short-term releases (puffs); calculation of fluctuations of concentration on short timescales, odors and condensed plume visibility; and allowance for radioactive decay including g-ray dose. One can use default values for upper air data or use a file includes vertical meteorological in format of _prf (CERC, 1998). The minimum ADMS meteorological data inputs (CERC, 1998): 1. 2. 3. 4.
Year, Month, Day, Hour Wind speed, Wind direction One of the following: the reciprocal of the Monin-Obukhov length, the surface heat flux, or the cloud cover
1.3. Industrial Source Complex Model version 3 (ISC3) ISC3 is another Gaussian plume model which can be used to predict pollutant concentrations from industrial facilities. ISC3 had been the US-EPA preferred air dispersion model for regulatory purposes until December 2006 when it was replaced by AERMOD. The major advantage of AERMOD and ADMS over ISC3 is related to the state of art algorithm of turbulent dispersion (EPA, 1995). The minimum ISC3 meteorological data are (EPA, 1995): 1. 2. 3. 4.
Wind speed Wind direction A stability class determination Mixing depth
Table 1 PM10 emission rates for MOBARAKEH steel facility. Process and unit
Method of emission calculation
Contribution
g/sa
Pellet making Direct reduction Steel making Wind erosion Miscellaneous Total
measurement measurement measurement CFD-Fluentc measurement
29% 12% 46% 8.4% 4.6%
118.32 48.96 187.68b 34.272 18.768 408
and emission factor and emission factor and emission factor and emission factor
a The value of total emissions from point, area and line sources for each process units has been converted to g/s. b High value of emission in steel making unit is due to uncontrolled dust emission from roof of units and above EAF. c The Fluent CFD Model was used for PBL and the default coefficient of the model for turbulence and surface roughness and wall function has been modified before use (Torano et al., 2006; Ashrafi et al., 2015).
1. Direct measurement 2. Indirect measurement (using measurement results in other same companies) 3. The use of modeling tools like CFD 4. The use of ap42 emission rates There are about 400 miscellaneous sources of PM10 in this facility. The sources and emissions are mentioned in Table 1. The modeling area has a radius of 7.5 km around the facility and the topographical conditions include flat and elevated terrain. The modeling area of was determined to be urban based on Gimson et al. (2007). The models were ran for the 2014e2015 meteorological period. Maximum wind speeds at a 10 m anemometer tower was 35 m/s for 2014e2015 period. The wind rose of the area is shown in Fig. 1.
3. Methodology The performance of the models was evaluated using different meteorological inputs. AERMOD was evaluated for the following three meteorological cases:
In this paper, these models have been evaluated for complete and incomplete upper meteorological data and all of them have been verified with measurements.
1. AERMOD_REAL: AERMOD modeling with real upper meteorological data.
2. Area of study
In this case, hourly surface and upper data has been used for modeling. These data measurements from the nearby airport meteorological station.
The case study is the MOBARAKEH steel complex located at 539733.79 m E and 3567898.85 m N zone 39 S. It is the largest steel maker of MENA1 region. This facility made 52 percent of all the steel produced in Iran. Its products consist of hot and cold rolled sheets and coils, pickled coils, narrow strip coil, tinplate sheet and coil galvanized coil, pre-painted coil and slab (MSC, 2015). There are plenty of buoyant and non-buoyant sources in the facility include point, line and area sources. The site included point sources (like.g., stacks), line sources (e.g., conveyor and roads), area sources (e.g., piles and wind erosion from surrounding areas). The main pollutant from the facility is PM10 which was evaluated in this paper. The emissions from piles were evaluated by using Fluent CFD modeling (in conjunction with AP-42) and wind erosion fluxes have been estimated. Gambit model has been used to generate iron ore piles and the wind profile is calculated by Fluent (EPA, 2006; Torano et al., 2006; Ashrafi et al., 2015). The method of estimating emission rates of sources follows these steps sequentially:
1
Middle East & Northern Africa.
2. AERMOD_ESTIMATOR: AERMOD modeling with default upper meteorological estimator of AERMOD. In this case, hourly surface data was used but upper data has been estimated with AERMOD default option estimator (The et al., 2001). 3. AERMOD_ALGORITHM: AERMOD modeling with upper meteorological data based on the algorithms proposed by Batchvarova and Gryning (1991). In this case, hourly surface data has been used but upper data has been estimated with Batchvarova and Gryning algorithm. For details of this algorithm and implementation please refer to (Carbonell et al., 2010; Gill, 1982; Thomson, 2000). The ADMS model was evaluated for the following cases: 1. ADMS_PRF: ADMS modeling for real vertical profile meteorological data (_prf)
Please cite this article in press as: Kalhor, M., Bajoghli, M., Comparison of AERMOD, ADMS and ISC3 for incomplete upper air meteorological data (case study: Steel plant), Atmospheric Pollution Research (2017), http://dx.doi.org/10.1016/j.apr.2017.06.001
M. Kalhor, M. Bajoghli / Atmospheric Pollution Research xxx (2017) 1e6
340°
0°
350°
10°
4000
330°
700
20°
ADMS
30°
320°
600
40° 3000
300°
50°
500 60°
2000
290°
70° 1000
280°
80°
270°
90°
260°
100°
250°
microgram/m3
310°
3
ISC3 400 300 200 AER_EST
110°
240°
AER_ALG
100
AER_REAL ADMS_PRF
0
Specified points
120°
230°
Fig. 3. Modeling results for maximum 1hr PM10 at z ¼ 0 m.
130° 220°
140° 210°
150° 200°
190°
0
3
0
1.5
180° 170° 6 10 16
160° (knots) Wind speed
3.1
5.1
8.2
predicted concentrations, verification also has been performed by using the statistical procedure proposed by Hanna (1989). Seven specified points (A to G) have been defined along a single wind direction for this statistical procedure (Fig. 2).
(m/s)
Fig. 1. Wind rose of MOBARAKEH steel facility in 2014e2015.
2. ADMS: ADMS modeling for default upper meteorological data estimator of model The ISC3 model was evaluated for the real upper meteorological data case only. The results from the modeling performed were compared directly to field measurements for verification. There are four measurement points (S1, S2, S3, and S4) with continuous samplers in the facility (Fig. 2). Measurement has been made around facility according to EPA CFR40 Appendix J to Part 50 methods with an OMNITM FT Ambient Air Sampler. The following points depicted in Fig. 2 were selected for direct verification (the background concentration of PM10 substituted from measurement data). For understanding how well the models
4. Results and discussion There is great interest in investigating a model's ability to accurately predict the highest concentrations because it is used in assessing compliance with air quality standards. The maximum 1hr concentration in points A to G is shown in Figs. 3 and 4. The results were predicted for height levels of z ¼ 0 m, z ¼ 5 m, z ¼ 20 m and z ¼ 100 m, however, only z ¼ 0 m, z ¼ 100 m are depicted here the results for other heights are available in the Supporting materials. Figs. 3 and 4 shows a descending trend as distances from the facility increased. As shown in the figures above, the model differences from using results of real upper meteorological data and estimated data are significant. The major discrepancies occur for ADMS. For AERMOD the differences between real and estimated upper meteorological data are about 130%, for ADMS it can be about 450%. The concentration pattern over A to G is the same for all models and there is a peak near the facility and concentration decreases with distance. The results of AERMOD with Batchvarova and Gryning algorithm (1991) located somewhere between real and default estimator option. In other words, the algorithm predicted
700
ADMS 600
microgram/m3
500 400 300
ISC3 200
AER_REA AER_ALG
100 0 Fig. 2. Specified points for modeling comparison.
ADMS_PRF AER_EST Specified points
Fig. 4. Modeling results for maximum 1hr PM10 at z ¼ 100 m.
Please cite this article in press as: Kalhor, M., Bajoghli, M., Comparison of AERMOD, ADMS and ISC3 for incomplete upper air meteorological data (case study: Steel plant), Atmospheric Pollution Research (2017), http://dx.doi.org/10.1016/j.apr.2017.06.001
4
M. Kalhor, M. Bajoghli / Atmospheric Pollution Research xxx (2017) 1e6
Fig. 5. Vertical variation of maximum 1hr PM10 at point A.
differently than ISC3 and AERMOD in the vertical direction. The AERMOD results for algorithm Batchvarova and Gryning (1991) are located somewhere between default estimator and real meteorological data. The maximum 1hr concentration of PM10 for all modeling scenarios at all heights for points A to G have been shown in Table 2. Table 2 suggests that maxima occurs at z ¼ 0 m or z ¼ 20 m at points D or E. As shown in Table 2, for all models, maxima occurs at 1.8 km from the center of the facility while ISC3 predicted a maximum at 0.5 km which is in agreement with results of Hanna et al. (1999). The 2D contours of the maximum 1hr concentration of PM10 in 7.5 km around facility have been studied (not shown here, available in Supporting material). The maximum points of concentration in the center of the domain are repeated in all models results and are the same for all models. The AEMOD plumes dispersion show a good trend with the direction of prevailing wind. For verification, the results from the modeling were compared with S1, S2, S3 and S4 points. The maximum of 1hr and annual average measurements were compared to modeling results as shown in Table 3. As shown in Table 3, results for real and estimated upper air have been closer to each other in annual average concentration. The differences between maximum 1hr concentrations are significant but for the annual average (due to wide range of data) it is expected that the variance of data decreases. The results show good correlation near the facility and show an under-prediction tendency as distance decreases. ADMS_PRF tends to under-predict by a larger amount than other models. In comparison to annual results, ADMS had a scatter of about a factor of 2, AERMOD had a scatter of about factor of 2.4, and ISC3 has a scatter about 3 in average. The statistical model suggested by Hanna (1989) has been used for quantify each model performance. His model suggests several performance factors for quantify decision about models accuracy. Statistical factors of FAC2,2 MG3 and VG4 have been calculated for all measurement points for each modeling software as following:
FAC2 ¼ fraction of data that satisfy 0:5
Cp 2:0 C0
MG ¼ exp ln C0 ln Cp
Fig. 6. Vertical variation of maximum 1hr PM10 at point E.
concentration more realistic than AERMOD default estimator. The results of modeling in the vertical direction for A and E points have been depicted in Figs. 5 and 6. The results for other points have been available in the Supporting materials. As shown in the figures above, ISC3 and AERMOD results have a similar pattern (increasing first then decreasing with height) but the ADMS trend of concentration is completely different. For AERMOD and ADMS, the differences between results for real and estimated upper air data are obvious. For ADMS_PRF concentration value changes between A to G point but at each point, there is no change in vertical concentration. This behavior has been expected because the variation of meteorological data has been defined and fixed in _prf file. These figures suggest that, as height increases the predicted concentrations of ISC3 and AERMOD decrease. In contrast, ADMS, concentrations decrease with height but increase rapidly at 100 m. The only clear trend that is seen is that ADMS performs
VG ¼ exp ln C0 ln Cp
2
where Cp is model predictions, C0 is observations and C is average over the dataset. The median of statistical factors for all models in all measurement points is shown in Table 4: Table 4 suggests that, modeling with real upper meteorological data show better performance. AERMOD_REAL and ADMS_PRF are comparable and ISC3 show poor results. 5. Conclusion In this paper three Gaussian air pollution models (i.e., AERMOD, ADMS and ISC3) were evaluated with real and estimated upper air meteorological data. The results of modeling performances both in vertical and horizontal directions show a significant difference
2 3 4
Fraction of predictions within a factor of two of the observations points. Geometric Mean. Geometric Variance.
Please cite this article in press as: Kalhor, M., Bajoghli, M., Comparison of AERMOD, ADMS and ISC3 for incomplete upper air meteorological data (case study: Steel plant), Atmospheric Pollution Research (2017), http://dx.doi.org/10.1016/j.apr.2017.06.001
M. Kalhor, M. Bajoghli / Atmospheric Pollution Research xxx (2017) 1e6
5
Table 2 Rank 1 of 1hr PM10 concentration. Models
Rank 1 (mg/m3)
Points
Height (m)
Distance from center of facility (km)
AERMOD_ESTIMATOR AERMOD_REAL AERMOD_ALGORITHM ADMS_PRF ADMS ISC3
644 1059 778 860 601 1135
D E E E E Between C and D
20 20 20 0 0 0
1.8 1.8 1.8 1.8 1.8 0.5
Table 3 Comparison of annual and maximum 1hr concentration of PM10 for modeling and measurements.
AERMOD_ESTIMATOR AERMOD_REAL AERMOD_ALGORITHM ISC3 ADMS ADMS_PRF
S1
S2
S3
S4
1HR MAX (mg/m3) ANNUAL (mg/ m3)
1HR MAX (mg/m3) ANNUAL (mg/ m3)
1HR MAX (mg/m3) ANNUAL (mg/ m3)
1HR MAX (mg/m3) ANNUAL (mg/ m3)
M/O ratio 3.077 0.590 2.167 4.705 1.295 0.654
3.024 1.422 2.651 5.506 0.301 0.916
3.447 2.466 3.573 4.718 0.573 1.583
1.713 0.538 1.329 2.804 0.469 0.420
1.870 0.348 1.304 3.000 1.043 0.435
1.548 0.677 1.355 2.903 1.000 0.387
1.605 1.163 1.651 2.163 1.605 0.744
0.959 0.224 0.714 1.531 0.531 0.227
Table 4 Median performance factor over all measurements points.
Max Cp/C0 MG VG FAC2
AERMOD_REAL
AERMOD_ESTIMATOR
AERMOD_ALGORITHM
ADMS
ADMS_PRF
ISC3
1.72 0.537 1.18 0.19
2.23 0.288 1.95 0.25
2.31 0.316 1.77 0.22
0.61 2.23 1.325 0.1
1.05 0.79 1.02 0.02
2.35 0.4 1.44 0.35
when using real (actual) and estimated upper meteorological data results. The models' results in the horizontal direction have a similar trend of variation, however, in the vertical direction, ADMS has a different trend. In comparing field measurement for annual average concentrations, AERMOD and ADMS under-estimate predictions in most points but ISC3 results show an over-estimation tendency. The performance of models does show better results near the facility but accuracy decreased with distance. The AERMOD_ALGORITHM's performance is slightly better than AERMOD_ESTIMATOR. The results from the ADMS modeling general were under-estimated especially ADMS_PRF which under-estimated in all points. The differences between real and estimated upper meteorological data for ADMS are significantly larger than for the AERMOD one. Considering only the highest 1 h predicted and observed concentrations, ISC3 over-predicts by a factor of 380% on average, ADMS over-predicted by a factor of 10%, ADMS_PRF underpredicted by a factor of 11%, AERMOD_REAL over-predicts by a factor of 25%, AERMOD_ESTIMATOR by a factor of 180%, AERMOD_ALGORTHM by a factor of 140%. The paper suggests that estimated upper air meteorological data should be used with care. In cases when upper air meteorological data are not available, the users should provide upper meteorological data by measuring itself in local site or using the best fitting algorithm and should not trust the models defaults values. The algorithms proposed by Batchvarova and Gryning (1991) shows better result than AERMOD default estimator option and the results were closer to measurements. Acknowledgements Thanks to Mohsen Sattari and management of MOBARAKEH steel facility. They encouraged us to complete the paper and gave us
necessary information about the facility. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.apr.2017.06.001. References Ashrafi, Kh, Kalhor, M., Esfahaniyan, V., 2015. Numerical simulation of aerodynamic suspension of particles during wind erosion. Environ. Earth Sci. 74 (2), 1569e1578. Batchvarova, E., Gryning, S.E., 1991. Applied model for the growth of the daytime mixed layer. Bound. Layer. Meteorol. 56, 261e274. Cambridge Environmental Research Consultants, 1998. ADMS Technical Specification, CERC, 3. Kings Parade, Cambridge, U.K.. CB2 1SJ. Carbonell, L.M.T., Gacita, M.S., Oliva, J.D.J.R., Garea, L.C., Rivero, N.D., Ruiz, E.M., 2010. Methodological guide for implementation of the AERMOD system with incomplete local data. Atmos. Pollut. Res. 1, 102e111. EPA, 2015, 40 CFR Part 51, Federal Register. Revision to the Guideline on Air Quality Models: Enhancements to the AERMOD Dispersion Modeling System and Incorporation of Approaches to Address Ozone and Fine Particulate Matter; Proposed Rule, vol. 80 No. 145, pp. 45340e45387. EPA, 2006. Industrial Wind Erosion. AP-42, CH 13.2.5. EPA, 2004. User's Guide for the AMS/EPA Regulatory Model e AERMOD. EPA 454/B 03 001. EPA, 1995. User's Guide for the Industrial Source Complex (ISC3) Dispersion Model (Revised). Volume II - Description of Model Algorithms. EPA-454/b-95e0036. Gill, A.E., 1982. Atmosphere Ocean Dynamics. Academic Press. Gimson, N., Olivares, G., Khan, B., Zawar-Reza, P., 2007. Dispersion Modeling in New Zealand, Part1eassessment of Meteorological Models. FRST program Protecting New Zealand’s Clean Air. Hanna, S.R., Egan, B.A., Purdum, J., Wagler, J., 1999. Evaluation of ISC3, AERMOD, and ADMS Dispersion Models with Observations from Five Field Sites. HC Report P020, API, 1220 LSt. NW, Washington, DC 20005e4070. Hanna, S.R., 1989. Confidence limits for air quality model evaluations, as estimated by bootstrap and jackknife resampling methods. Atmos. Environ. 23, 1385e1398. MSC, 2015. Mobarkeh Steel Facility Annual Report. The, J.L., Lee, R., Brode, 2001. Worldwide data quality effects on PBL short-range
Please cite this article in press as: Kalhor, M., Bajoghli, M., Comparison of AERMOD, ADMS and ISC3 for incomplete upper air meteorological data (case study: Steel plant), Atmospheric Pollution Research (2017), http://dx.doi.org/10.1016/j.apr.2017.06.001
6
M. Kalhor, M. Bajoghli / Atmospheric Pollution Research xxx (2017) 1e6
regulatory air dispersion models. In: Proceedings 7th International Conference on Harmonization within Atmospheric Dispersion Modelling for Regulatory Purposes, Belgirate, Italy, pp. 202e206. Thomson, D.J., 2000. The Meteorological Input Module, ADMS 3 Technical
Specification. CERC. Torano, J.A., Rodriguez, R., Diego, I., Rivas, J.M., Pelegry, A., 2006. Influence of pile shape on wind erosion CFD emission simulation. Appl. Math. Model. 31, 2487e2502.
Please cite this article in press as: Kalhor, M., Bajoghli, M., Comparison of AERMOD, ADMS and ISC3 for incomplete upper air meteorological data (case study: Steel plant), Atmospheric Pollution Research (2017), http://dx.doi.org/10.1016/j.apr.2017.06.001
H O S T E D BY
Contents lists available at ScienceDirect
Atmospheric Pollution Research journal homepage: http://www.journals.elsevier.com/locate/apr
Comparison of AERMOD, ADMS and ISC3 for incomplete upper air meteorological data (case study: Steel plant) Mostafa Kalhor a, *, Mehrshad Bajoghli b a b
HPA Company, 2d Floor, No 4, Building 314, Valiasr Square, 1439914153, Tehran, Iran Department of Environmental Science, Isfahan (Khorasgan) Branch, Islamic Azad University, Isfahan, Iran
a r t i c l e i n f o
a b s t r a c t
Article history: Received 3 January 2017 Received in revised form 31 May 2017 Accepted 4 June 2017 Available online xxx
In this paper three well known Gaussian dispersion models have been evaluated for a case study of a steel plant using complete and incomplete upper air meteorological data. In developing countries, the availability of surface and upper air meteorological data is limited. AMS/ EPA Regulatory Model (AERMOD), Advanced Dispersion Modeling System (ADMS) and Industrial Source Complex Model (ISC3) have been evaluated for both real and estimated upper meteorological data and the results have been compared with field measurements both in the horizontal and vertical directions. The results show significant differences in predicted concentrations when modeling with real (actual) and estimated upper meteorological data. The differences ranged from 100% to 450%. Comparison of model performance suggests that AERMOD and ADMS with real meteorological data produce consistent results in the horizontal direction while ISC3 output over-predicts in general. In AERMOD and ISC3 the predicted concentrations have a similar trend of variation in the vertical direction but in ADMS the concentration variation in the vertical direction exhibited different trends. In general, the ADMS predicted concentrations under-estimated field observations. The paper suggests that upper data must be used for modeling and the default values must be used with care. In absence of upper meteorological data, users could estimate upper meteorological data by different available algorithm rather than only default option of models. © 2017 Turkish National Committee for Air Pollution Research and Control. Production and hosting by Elsevier B.V. All rights reserved.
Keywords: AERMOD ADMS ISC3 Upper air meteorological data Steel plant
1. Introduction Gaussian dispersion models have been used widely for concentration prediction, sampling network design, EIA (Environmental Impact Assessment) and environmental management scenarios (EPA, 2015). The relative simplicity of use, quick setup, acceptable accuracy and wide applicability in different atmospheric condition are the advantages of these models (EPA, 2015). In developing countries, the using of these models has several constraints in both source emission determination and availability of meteorological data. In these countries accurate and sequential meteorological data are not available (especially upper meteorological data). Another problem is the location of industries. Most
* Corresponding author. E-mail address: [email protected] (M. Kalhor). Peer review under responsibility of Turkish National Committee for Air Pollution Research and Control.
plants and industrial facilities have been constructed outside of large cities where meteorological sites are not available for modeling (Carbonell et al., 2010). In this paper three different models (i.e., ADMS, AERMOD, and ISC3) have been evaluated. 1.1. AERMOD (AMS/EPA Regulatory Model) AERMOD is the Gaussian air dispersion model which incorporates building downwash algorithms, advanced depositional parameters, local terrain effects, and advanced meteorological turbulence calculations. AERMOD could be implemented with both real and estimated upper meteorological data (EPA, 2004). The minimum surface meteorological data for running AERMOD are (EPA, 2004): 1. 2. 3. 4.
Year, Month, Day, Hour Wind Speed Wind Direction Dry Bulb Temperature
http://dx.doi.org/10.1016/j.apr.2017.06.001 1309-1042/© 2017 Turkish National Committee for Air Pollution Research and Control. Production and hosting by Elsevier B.V. All rights reserved.
Please cite this article in press as: Kalhor, M., Bajoghli, M., Comparison of AERMOD, ADMS and ISC3 for incomplete upper air meteorological data (case study: Steel plant), Atmospheric Pollution Research (2017), http://dx.doi.org/10.1016/j.apr.2017.06.001
2
M. Kalhor, M. Bajoghli / Atmospheric Pollution Research xxx (2017) 1e6
5. Cloud Cover (tenths)
1.2. Advanced Dispersion Modeling System (ADMS) ADMS is a Gaussian air dispersion model used to evaluate industrial impacts on air quality. ADMS incorporates building effects, complex terrain, coastlines and variations in surface roughness; dry and wet deposition; chemistry schemes; short-term releases (puffs); calculation of fluctuations of concentration on short timescales, odors and condensed plume visibility; and allowance for radioactive decay including g-ray dose. One can use default values for upper air data or use a file includes vertical meteorological in format of _prf (CERC, 1998). The minimum ADMS meteorological data inputs (CERC, 1998): 1. 2. 3. 4.
Year, Month, Day, Hour Wind speed, Wind direction One of the following: the reciprocal of the Monin-Obukhov length, the surface heat flux, or the cloud cover
1.3. Industrial Source Complex Model version 3 (ISC3) ISC3 is another Gaussian plume model which can be used to predict pollutant concentrations from industrial facilities. ISC3 had been the US-EPA preferred air dispersion model for regulatory purposes until December 2006 when it was replaced by AERMOD. The major advantage of AERMOD and ADMS over ISC3 is related to the state of art algorithm of turbulent dispersion (EPA, 1995). The minimum ISC3 meteorological data are (EPA, 1995): 1. 2. 3. 4.
Wind speed Wind direction A stability class determination Mixing depth
Table 1 PM10 emission rates for MOBARAKEH steel facility. Process and unit
Method of emission calculation
Contribution
g/sa
Pellet making Direct reduction Steel making Wind erosion Miscellaneous Total
measurement measurement measurement CFD-Fluentc measurement
29% 12% 46% 8.4% 4.6%
118.32 48.96 187.68b 34.272 18.768 408
and emission factor and emission factor and emission factor and emission factor
a The value of total emissions from point, area and line sources for each process units has been converted to g/s. b High value of emission in steel making unit is due to uncontrolled dust emission from roof of units and above EAF. c The Fluent CFD Model was used for PBL and the default coefficient of the model for turbulence and surface roughness and wall function has been modified before use (Torano et al., 2006; Ashrafi et al., 2015).
1. Direct measurement 2. Indirect measurement (using measurement results in other same companies) 3. The use of modeling tools like CFD 4. The use of ap42 emission rates There are about 400 miscellaneous sources of PM10 in this facility. The sources and emissions are mentioned in Table 1. The modeling area has a radius of 7.5 km around the facility and the topographical conditions include flat and elevated terrain. The modeling area of was determined to be urban based on Gimson et al. (2007). The models were ran for the 2014e2015 meteorological period. Maximum wind speeds at a 10 m anemometer tower was 35 m/s for 2014e2015 period. The wind rose of the area is shown in Fig. 1.
3. Methodology The performance of the models was evaluated using different meteorological inputs. AERMOD was evaluated for the following three meteorological cases:
In this paper, these models have been evaluated for complete and incomplete upper meteorological data and all of them have been verified with measurements.
1. AERMOD_REAL: AERMOD modeling with real upper meteorological data.
2. Area of study
In this case, hourly surface and upper data has been used for modeling. These data measurements from the nearby airport meteorological station.
The case study is the MOBARAKEH steel complex located at 539733.79 m E and 3567898.85 m N zone 39 S. It is the largest steel maker of MENA1 region. This facility made 52 percent of all the steel produced in Iran. Its products consist of hot and cold rolled sheets and coils, pickled coils, narrow strip coil, tinplate sheet and coil galvanized coil, pre-painted coil and slab (MSC, 2015). There are plenty of buoyant and non-buoyant sources in the facility include point, line and area sources. The site included point sources (like.g., stacks), line sources (e.g., conveyor and roads), area sources (e.g., piles and wind erosion from surrounding areas). The main pollutant from the facility is PM10 which was evaluated in this paper. The emissions from piles were evaluated by using Fluent CFD modeling (in conjunction with AP-42) and wind erosion fluxes have been estimated. Gambit model has been used to generate iron ore piles and the wind profile is calculated by Fluent (EPA, 2006; Torano et al., 2006; Ashrafi et al., 2015). The method of estimating emission rates of sources follows these steps sequentially:
1
Middle East & Northern Africa.
2. AERMOD_ESTIMATOR: AERMOD modeling with default upper meteorological estimator of AERMOD. In this case, hourly surface data was used but upper data has been estimated with AERMOD default option estimator (The et al., 2001). 3. AERMOD_ALGORITHM: AERMOD modeling with upper meteorological data based on the algorithms proposed by Batchvarova and Gryning (1991). In this case, hourly surface data has been used but upper data has been estimated with Batchvarova and Gryning algorithm. For details of this algorithm and implementation please refer to (Carbonell et al., 2010; Gill, 1982; Thomson, 2000). The ADMS model was evaluated for the following cases: 1. ADMS_PRF: ADMS modeling for real vertical profile meteorological data (_prf)
Please cite this article in press as: Kalhor, M., Bajoghli, M., Comparison of AERMOD, ADMS and ISC3 for incomplete upper air meteorological data (case study: Steel plant), Atmospheric Pollution Research (2017), http://dx.doi.org/10.1016/j.apr.2017.06.001
M. Kalhor, M. Bajoghli / Atmospheric Pollution Research xxx (2017) 1e6
340°
0°
350°
10°
4000
330°
700
20°
ADMS
30°
320°
600
40° 3000
300°
50°
500 60°
2000
290°
70° 1000
280°
80°
270°
90°
260°
100°
250°
microgram/m3
310°
3
ISC3 400 300 200 AER_EST
110°
240°
AER_ALG
100
AER_REAL ADMS_PRF
0
Specified points
120°
230°
Fig. 3. Modeling results for maximum 1hr PM10 at z ¼ 0 m.
130° 220°
140° 210°
150° 200°
190°
0
3
0
1.5
180° 170° 6 10 16
160° (knots) Wind speed
3.1
5.1
8.2
predicted concentrations, verification also has been performed by using the statistical procedure proposed by Hanna (1989). Seven specified points (A to G) have been defined along a single wind direction for this statistical procedure (Fig. 2).
(m/s)
Fig. 1. Wind rose of MOBARAKEH steel facility in 2014e2015.
2. ADMS: ADMS modeling for default upper meteorological data estimator of model The ISC3 model was evaluated for the real upper meteorological data case only. The results from the modeling performed were compared directly to field measurements for verification. There are four measurement points (S1, S2, S3, and S4) with continuous samplers in the facility (Fig. 2). Measurement has been made around facility according to EPA CFR40 Appendix J to Part 50 methods with an OMNITM FT Ambient Air Sampler. The following points depicted in Fig. 2 were selected for direct verification (the background concentration of PM10 substituted from measurement data). For understanding how well the models
4. Results and discussion There is great interest in investigating a model's ability to accurately predict the highest concentrations because it is used in assessing compliance with air quality standards. The maximum 1hr concentration in points A to G is shown in Figs. 3 and 4. The results were predicted for height levels of z ¼ 0 m, z ¼ 5 m, z ¼ 20 m and z ¼ 100 m, however, only z ¼ 0 m, z ¼ 100 m are depicted here the results for other heights are available in the Supporting materials. Figs. 3 and 4 shows a descending trend as distances from the facility increased. As shown in the figures above, the model differences from using results of real upper meteorological data and estimated data are significant. The major discrepancies occur for ADMS. For AERMOD the differences between real and estimated upper meteorological data are about 130%, for ADMS it can be about 450%. The concentration pattern over A to G is the same for all models and there is a peak near the facility and concentration decreases with distance. The results of AERMOD with Batchvarova and Gryning algorithm (1991) located somewhere between real and default estimator option. In other words, the algorithm predicted
700
ADMS 600
microgram/m3
500 400 300
ISC3 200
AER_REA AER_ALG
100 0 Fig. 2. Specified points for modeling comparison.
ADMS_PRF AER_EST Specified points
Fig. 4. Modeling results for maximum 1hr PM10 at z ¼ 100 m.
Please cite this article in press as: Kalhor, M., Bajoghli, M., Comparison of AERMOD, ADMS and ISC3 for incomplete upper air meteorological data (case study: Steel plant), Atmospheric Pollution Research (2017), http://dx.doi.org/10.1016/j.apr.2017.06.001
4
M. Kalhor, M. Bajoghli / Atmospheric Pollution Research xxx (2017) 1e6
Fig. 5. Vertical variation of maximum 1hr PM10 at point A.
differently than ISC3 and AERMOD in the vertical direction. The AERMOD results for algorithm Batchvarova and Gryning (1991) are located somewhere between default estimator and real meteorological data. The maximum 1hr concentration of PM10 for all modeling scenarios at all heights for points A to G have been shown in Table 2. Table 2 suggests that maxima occurs at z ¼ 0 m or z ¼ 20 m at points D or E. As shown in Table 2, for all models, maxima occurs at 1.8 km from the center of the facility while ISC3 predicted a maximum at 0.5 km which is in agreement with results of Hanna et al. (1999). The 2D contours of the maximum 1hr concentration of PM10 in 7.5 km around facility have been studied (not shown here, available in Supporting material). The maximum points of concentration in the center of the domain are repeated in all models results and are the same for all models. The AEMOD plumes dispersion show a good trend with the direction of prevailing wind. For verification, the results from the modeling were compared with S1, S2, S3 and S4 points. The maximum of 1hr and annual average measurements were compared to modeling results as shown in Table 3. As shown in Table 3, results for real and estimated upper air have been closer to each other in annual average concentration. The differences between maximum 1hr concentrations are significant but for the annual average (due to wide range of data) it is expected that the variance of data decreases. The results show good correlation near the facility and show an under-prediction tendency as distance decreases. ADMS_PRF tends to under-predict by a larger amount than other models. In comparison to annual results, ADMS had a scatter of about a factor of 2, AERMOD had a scatter of about factor of 2.4, and ISC3 has a scatter about 3 in average. The statistical model suggested by Hanna (1989) has been used for quantify each model performance. His model suggests several performance factors for quantify decision about models accuracy. Statistical factors of FAC2,2 MG3 and VG4 have been calculated for all measurement points for each modeling software as following:
FAC2 ¼ fraction of data that satisfy 0:5
Cp 2:0 C0
MG ¼ exp ln C0 ln Cp
Fig. 6. Vertical variation of maximum 1hr PM10 at point E.
concentration more realistic than AERMOD default estimator. The results of modeling in the vertical direction for A and E points have been depicted in Figs. 5 and 6. The results for other points have been available in the Supporting materials. As shown in the figures above, ISC3 and AERMOD results have a similar pattern (increasing first then decreasing with height) but the ADMS trend of concentration is completely different. For AERMOD and ADMS, the differences between results for real and estimated upper air data are obvious. For ADMS_PRF concentration value changes between A to G point but at each point, there is no change in vertical concentration. This behavior has been expected because the variation of meteorological data has been defined and fixed in _prf file. These figures suggest that, as height increases the predicted concentrations of ISC3 and AERMOD decrease. In contrast, ADMS, concentrations decrease with height but increase rapidly at 100 m. The only clear trend that is seen is that ADMS performs
VG ¼ exp ln C0 ln Cp
2
where Cp is model predictions, C0 is observations and C is average over the dataset. The median of statistical factors for all models in all measurement points is shown in Table 4: Table 4 suggests that, modeling with real upper meteorological data show better performance. AERMOD_REAL and ADMS_PRF are comparable and ISC3 show poor results. 5. Conclusion In this paper three Gaussian air pollution models (i.e., AERMOD, ADMS and ISC3) were evaluated with real and estimated upper air meteorological data. The results of modeling performances both in vertical and horizontal directions show a significant difference
2 3 4
Fraction of predictions within a factor of two of the observations points. Geometric Mean. Geometric Variance.
Please cite this article in press as: Kalhor, M., Bajoghli, M., Comparison of AERMOD, ADMS and ISC3 for incomplete upper air meteorological data (case study: Steel plant), Atmospheric Pollution Research (2017), http://dx.doi.org/10.1016/j.apr.2017.06.001
M. Kalhor, M. Bajoghli / Atmospheric Pollution Research xxx (2017) 1e6
5
Table 2 Rank 1 of 1hr PM10 concentration. Models
Rank 1 (mg/m3)
Points
Height (m)
Distance from center of facility (km)
AERMOD_ESTIMATOR AERMOD_REAL AERMOD_ALGORITHM ADMS_PRF ADMS ISC3
644 1059 778 860 601 1135
D E E E E Between C and D
20 20 20 0 0 0
1.8 1.8 1.8 1.8 1.8 0.5
Table 3 Comparison of annual and maximum 1hr concentration of PM10 for modeling and measurements.
AERMOD_ESTIMATOR AERMOD_REAL AERMOD_ALGORITHM ISC3 ADMS ADMS_PRF
S1
S2
S3
S4
1HR MAX (mg/m3) ANNUAL (mg/ m3)
1HR MAX (mg/m3) ANNUAL (mg/ m3)
1HR MAX (mg/m3) ANNUAL (mg/ m3)
1HR MAX (mg/m3) ANNUAL (mg/ m3)
M/O ratio 3.077 0.590 2.167 4.705 1.295 0.654
3.024 1.422 2.651 5.506 0.301 0.916
3.447 2.466 3.573 4.718 0.573 1.583
1.713 0.538 1.329 2.804 0.469 0.420
1.870 0.348 1.304 3.000 1.043 0.435
1.548 0.677 1.355 2.903 1.000 0.387
1.605 1.163 1.651 2.163 1.605 0.744
0.959 0.224 0.714 1.531 0.531 0.227
Table 4 Median performance factor over all measurements points.
Max Cp/C0 MG VG FAC2
AERMOD_REAL
AERMOD_ESTIMATOR
AERMOD_ALGORITHM
ADMS
ADMS_PRF
ISC3
1.72 0.537 1.18 0.19
2.23 0.288 1.95 0.25
2.31 0.316 1.77 0.22
0.61 2.23 1.325 0.1
1.05 0.79 1.02 0.02
2.35 0.4 1.44 0.35
when using real (actual) and estimated upper meteorological data results. The models' results in the horizontal direction have a similar trend of variation, however, in the vertical direction, ADMS has a different trend. In comparing field measurement for annual average concentrations, AERMOD and ADMS under-estimate predictions in most points but ISC3 results show an over-estimation tendency. The performance of models does show better results near the facility but accuracy decreased with distance. The AERMOD_ALGORITHM's performance is slightly better than AERMOD_ESTIMATOR. The results from the ADMS modeling general were under-estimated especially ADMS_PRF which under-estimated in all points. The differences between real and estimated upper meteorological data for ADMS are significantly larger than for the AERMOD one. Considering only the highest 1 h predicted and observed concentrations, ISC3 over-predicts by a factor of 380% on average, ADMS over-predicted by a factor of 10%, ADMS_PRF underpredicted by a factor of 11%, AERMOD_REAL over-predicts by a factor of 25%, AERMOD_ESTIMATOR by a factor of 180%, AERMOD_ALGORTHM by a factor of 140%. The paper suggests that estimated upper air meteorological data should be used with care. In cases when upper air meteorological data are not available, the users should provide upper meteorological data by measuring itself in local site or using the best fitting algorithm and should not trust the models defaults values. The algorithms proposed by Batchvarova and Gryning (1991) shows better result than AERMOD default estimator option and the results were closer to measurements. Acknowledgements Thanks to Mohsen Sattari and management of MOBARAKEH steel facility. They encouraged us to complete the paper and gave us
necessary information about the facility. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.apr.2017.06.001. References Ashrafi, Kh, Kalhor, M., Esfahaniyan, V., 2015. Numerical simulation of aerodynamic suspension of particles during wind erosion. Environ. Earth Sci. 74 (2), 1569e1578. Batchvarova, E., Gryning, S.E., 1991. Applied model for the growth of the daytime mixed layer. Bound. Layer. Meteorol. 56, 261e274. Cambridge Environmental Research Consultants, 1998. ADMS Technical Specification, CERC, 3. Kings Parade, Cambridge, U.K.. CB2 1SJ. Carbonell, L.M.T., Gacita, M.S., Oliva, J.D.J.R., Garea, L.C., Rivero, N.D., Ruiz, E.M., 2010. Methodological guide for implementation of the AERMOD system with incomplete local data. Atmos. Pollut. Res. 1, 102e111. EPA, 2015, 40 CFR Part 51, Federal Register. Revision to the Guideline on Air Quality Models: Enhancements to the AERMOD Dispersion Modeling System and Incorporation of Approaches to Address Ozone and Fine Particulate Matter; Proposed Rule, vol. 80 No. 145, pp. 45340e45387. EPA, 2006. Industrial Wind Erosion. AP-42, CH 13.2.5. EPA, 2004. User's Guide for the AMS/EPA Regulatory Model e AERMOD. EPA 454/B 03 001. EPA, 1995. User's Guide for the Industrial Source Complex (ISC3) Dispersion Model (Revised). Volume II - Description of Model Algorithms. EPA-454/b-95e0036. Gill, A.E., 1982. Atmosphere Ocean Dynamics. Academic Press. Gimson, N., Olivares, G., Khan, B., Zawar-Reza, P., 2007. Dispersion Modeling in New Zealand, Part1eassessment of Meteorological Models. FRST program Protecting New Zealand’s Clean Air. Hanna, S.R., Egan, B.A., Purdum, J., Wagler, J., 1999. Evaluation of ISC3, AERMOD, and ADMS Dispersion Models with Observations from Five Field Sites. HC Report P020, API, 1220 LSt. NW, Washington, DC 20005e4070. Hanna, S.R., 1989. Confidence limits for air quality model evaluations, as estimated by bootstrap and jackknife resampling methods. Atmos. Environ. 23, 1385e1398. MSC, 2015. Mobarkeh Steel Facility Annual Report. The, J.L., Lee, R., Brode, 2001. Worldwide data quality effects on PBL short-range
Please cite this article in press as: Kalhor, M., Bajoghli, M., Comparison of AERMOD, ADMS and ISC3 for incomplete upper air meteorological data (case study: Steel plant), Atmospheric Pollution Research (2017), http://dx.doi.org/10.1016/j.apr.2017.06.001
6
M. Kalhor, M. Bajoghli / Atmospheric Pollution Research xxx (2017) 1e6
regulatory air dispersion models. In: Proceedings 7th International Conference on Harmonization within Atmospheric Dispersion Modelling for Regulatory Purposes, Belgirate, Italy, pp. 202e206. Thomson, D.J., 2000. The Meteorological Input Module, ADMS 3 Technical
Specification. CERC. Torano, J.A., Rodriguez, R., Diego, I., Rivas, J.M., Pelegry, A., 2006. Influence of pile shape on wind erosion CFD emission simulation. Appl. Math. Model. 31, 2487e2502.
Please cite this article in press as: Kalhor, M., Bajoghli, M., Comparison of AERMOD, ADMS and ISC3 for incomplete upper air meteorological data (case study: Steel plant), Atmospheric Pollution Research (2017), http://dx.doi.org/10.1016/j.apr.2017.06.001