Next Article in Journal
An Effectiveness Assessment of China’s WEEE Treatment Fund
Previous Article in Journal
Smoking, Blood Pressure, and Cardiovascular Disease Mortality in a Large Cohort of Chinese Men with 15 Years Follow-up
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Sensitivity Analysis of Weather Variables on Offsite Consequence Analysis Tools in South Korea and the United States

Department of Health Science, Korea University, Anam-ro 145, Seongbuk-gu, Seoul 02841, Korea
*
Author to whom correspondence should be addressed.
Int. J. Environ. Res. Public Health 2018, 15(5), 1027; https://doi.org/10.3390/ijerph15051027
Submission received: 30 March 2018 / Revised: 8 May 2018 / Accepted: 15 May 2018 / Published: 18 May 2018

Abstract

:
We studied sensitive weather variables for consequence analysis, in the case of chemical leaks on the user side of offsite consequence analysis (OCA) tools. We used OCA tools Korea Offsite Risk Assessment (KORA) and Areal Location of Hazardous Atmospheres (ALOHA) in South Korea and the United States, respectively. The chemicals used for this analysis were 28% ammonia (NH3), 35% hydrogen chloride (HCl), 50% hydrofluoric acid (HF), and 69% nitric acid (HNO3). The accident scenarios were based on leakage accidents in storage tanks. The weather variables were air temperature, wind speed, humidity, and atmospheric stability. Sensitivity analysis was performed using the Statistical Package for the Social Sciences (SPSS) program for dummy regression analysis. Sensitivity analysis showed that impact distance was not sensitive to humidity. Impact distance was most sensitive to atmospheric stability, and was also more sensitive to air temperature than wind speed, according to both the KORA and ALOHA tools. Moreover, the weather variables were more sensitive in rural conditions than in urban conditions, with the ALOHA tool being more influenced by weather variables than the KORA tool. Therefore, if using the ALOHA tool instead of the KORA tool in rural conditions, users should be careful not to cause any differences in impact distance due to input errors of weather variables, with the most sensitive one being atmospheric stability.

1. Introduction

Chemical accidents are caused by worker errors, facility defects, or aging. Chemical leakage accidents can cause damage to humans or the environment [1]. Each country operates a chemical management system to prevent accidents caused by chemical substances. In 1982, the Seveso directive [2] was enacted in the European Union (EU) after the Seveso dioxin leak accident of 1976 in Italy. After the Texas explosion of 1989 in the United States (US), the US revised the Clean Air Act [3] in 1990. After the hydrogen fluoride (HF) leak of 2012 in South Korea (KOR), KOR enacted the Chemicals Control Act [4] in 2013 [5].
The US Clean Air Act requires that risk management plan (RMP) reports be submitted. The KOR Chemicals Control Act requires that offsite consequence analysis (OCA) reports be submitted. RMP and OCA reports include analysis of the impact on residents and the environment near the workplace, which consists of worst-case scenarios and alternative scenarios that can be expected from leakage accidents [6,7].
KOR’s National Institute of Chemical Safety (NICS) developed the Korea Offsite Risk Assessment (KORA) tool for OCA. In KOR, OCA tools being used include not only the KORA tool, but also the Areal Location of Hazardous Atmospheres (ALOHA) tool, developed by the United States Environmental Protection Agency (USEPA). The KORA and ALOHA tools were developed by applying various dispersion models, namely Gaussian models, SLAB (an atmospheric dispersion model for denser than air releases) models, and dense gas dispersion (DEGADIS) models [8,9]. Dispersion models require not only leakage data, but also weather data. Atmospheric dispersion is influenced by variables such as wind speed, atmospheric stability, and surface roughness. Dispersion models compute phenomena that disperse with physical space and time, such as weather conditions [10]. Analysis of these algorithms is generally not easy for those who use OCA tools. Furthermore, analyzing the source of OCA tools is not possible for the average user, unless they are a programmer.
RMP and OCA reports require accident scenarios. An accident scenario assumes an accident that could lead to an offsite impact through chemical leakage or serious injury to a worker. When a scenario is selected, weather data are used for determining the area’s weather information. Surface roughness takes the degree of obstacles such as buildings into account [11]. Input data of weather and surface roughness are determined by the user handling the OCA tool. Input data may vary depending on general weather patterns in the area, as well as the significance level of obstacles. Variation in input data of weather variables affects the impact distance [7,11].
Research related to OCA tools includes qualitative analysis studies of the models’ reliability improvement, as well as dispersion evaluation [12]. The aims have been to study the sensitivity of the SLAB model [13], to study land cover data to improve real time forecasting [14], and to study atmospheric dispersion evaluation characteristics [15]. The research on the KORA and ALOHA tools is devoted to predicting impact distances and deriving management plans based on actual accident cases [16,17,18,19,20]. Most people who use offsite analysis tools are general managers who manage chemical processes. Impact distance varies depending on input information of weather variables during those same processes. The general user of the KORA and ALOHA tools cannot easily analyze the sources and algorithms used in the tools. If we examine statistical results based on the user’s input information, we can see which factors play a significant role in deriving damage distances from the KORA and ALOHA tools. Therefore, in order to reduce the error of distance results, it is necessary to examine the correlation between weather variables on the user side. Therefore, in this study, we performed sensitivity analysis of OCA results according to weather variables, using user-side OCA tools.

2. Materials and Methods

2.1. Offsite Consequence Analysis

The OCA tools used were KORA version 2.0.0.0 [21] from KOR’s NICS, and ALOHA version 5.4.4 [22] from the USEPA. In terms of hazardous chemical substances, the frequency of accidents and the amounts used were ranked in the following order, based on accidents from 2009 to 2015: hydrogen chloride, nitric acid, ammonia, and hydrofluoric acid [23,24]. The concentrations of aqueous solutions were taken into account for the accident cases [25], along with concentration ranges supported by the KORA and ALOHA tools [21,22]. Therefore, the chemicals used were 28% ammonia (NH3), 35% hydrogen chloride (HCl), 50% hydrofluoric acid (HF), and 69% nitric acid (HNO3).
The scenarios were based on contents of leakage accident cases provided by the National Institute of Environmental Research [26]. Accident scenarios are defined as those in which 10-mm-diameter holes occur in storage tanks (operating temperature of 25 °C, operating pressure of 1.0 kg/cm2) containing 30 tons of hazardous chemicals in an aqueous solution, with leakage occurring for 10 min. The endpoint concentration for estimating the extent of the impact was applied to emergency response planning guidelines (ERPG)-2 values of the American Industrial Hygiene Association (AIHA) [27]. Table 1 shows input information of the KORA and ALOHA tools for modeling, according to scenarios tested.
The common weather variables in the KORA and ALOHA tools are wind speed, air temperature, humidity, and atmospheric stability. Variable conditions for wind speed, air temperature, and humidity were based on Korean weather information provided by the Korean Statistical Information Service (KOSIS) [28]. Air temperature was altered between −5 °C and 35 °C, at a rate of 5 °C. Wind speed was varied from 1 m/s to 16 m/s (maximum daily wind speed), in 1 m/s increments. Humidity was varied from 10% to 90%, at a rate of 10%. Atmospheric stability was altered from unstable (A) to stable (F). Surface roughness was divided into urban and rural [11]. We obtained a total of 640 results using the OCA tools. Table 2 shows a sample of some OCA results.

2.2. Sensitivity Analysis

Sensitivity analysis data utilized impact distances, which were the results of OCA by the KORA and ALOHA tools. Data reduction of impact distances was carried out using a Microsoft Excel spreadsheet [29]. The sample size used for sensitivity analysis was 640. There were no missing values. Sensitivity analysis was applied to regression analysis [30], which is a technique for determining the relationship between one or more independent variables and one dependent variable. Since impact distance is category-type data, the data for analysis were transformed into dummy variables (1 for coincidence and 0 for non-coincidence) [31], which were indicator variables representing each category for regression analysis. Table 3 shows a sample of some of the data sources converted to dummy variables. Regression analysis for the sensitivity analysis was performed using the Statistical Package for the Social Sciences (SPSS) program, version 24 [32], from the International Business Machines Corporation (IBM).
In this study, the independent variables were the chemicals used, air temperature, wind speed, humidity, atmospheric stability, surface roughness, and the OCA tools used, while the dependent variable was the impact distance. Therefore, the regression equation [33] derived for multiple regression analysis was as follows:
Impact   distance   =   Constant   +   ( β 1 × Air   temperature ) + ( β 2 × Wind   speed ) + ( β 3 × Humidity ) + ( β 4   ×   OCA   tool   used ) + ( β 5 × Surface   roughness ) + ( β 6 × Chemical   used ) .

3. Results

The results of the normality test of the impact distances, according to weather variables in urban and rural conditions, are shown in Figure 1 and Figure 2. The impact distance data, according to weather variables in urban and rural conditions, followed a normal distribution. The results of the detection of outliers in the impact distances, according to weather variables in urban and rural conditions, are shown in Figure 3 and Figure 4. Outliers existed in all cases, except in impact distance data analyzed by the KORA tool in urban conditions.
Regression analysis for the sensitivity analysis between the impact distance and weather variables yielded results as follows. Table 4 shows the results of the regression analysis. The regression model showed significant correlation between variables (p < 0.001). Air temperature, wind speed, the OCA tool used, surface roughness, and the chemical used significantly affected the impact distance (p < 0.05), while the effect of humidity was not deemed significant (p = 0.112).
The regression equation for the sensitivity analysis according to weather variables, using the KORA and ALOHA tools in urban conditions, was as follows:
Impact   distance = 122.093 + ( 24.628 ) × Air   temperature + ( 27.788 ) × Wind   speed + ( 10.885 ) × Humidity + ( 53.122 ) × KORA + ( 67.766 ) × Urban + ( 32.556 ) × 50 %   HF + ( 134.613 ) × 28 %   NH 3 + ( 126.775 ) × 35 %   HCl .
The sensitivity analysis results are shown in Table 5. Using the KORA tool, the impact distances for 28% NH3 were 255.91 m with atmospheric stability, 231.28 m with air temperature, and 228.12 m with wind speed. The impact distances for 35% HCl were 248.07 m with atmospheric stability, 223.45 m with air temperature, and 220.29 m with wind speed. The impact distances for 50% HF were 153.86 m with atmospheric stability, 129.23 m with air temperature, and 126.07 m with wind speed. The impact distances for 69% HNO3 were 121.30 m with atmospheric stability, 96.67 m with air temperature, and 93.51 m with wind speed. The impact distances for 28% NH3, according to weather variables, were the largest. When using the KORA tool, 28% NH3 was the most sensitive chemical in urban conditions. The impact distances according to atmospheric stability were the largest. Atmospheric stability was the most sensitive weather variable when using the KORA tool in urban conditions. Using the ALOHA tool, the impact distances for 28% NH3 were 309.03 m with atmospheric stability, 284.40 m with air temperature, and 281.25 m with wind speed. The impact distances for 35% HCl were 301.20 m with atmospheric stability, 276.57 m with air temperature, and 273.41 m with wind speed. The impact distances for 50% HF were 206.98 m with atmospheric stability, 182.35 m with air temperature, and 179.19 m with wind speed. The impact distances for 69% HNO3 were 174.42 m with atmospheric stability, 149.79 m with air temperature, and 146.63 m with wind speed. Sensitivity analysis results for the ALOHA tool were the same as those for the KORA tool. When using the ALOHA tool, 28% NH3, which had the greatest impact distances, was the most sensitive substance in urban conditions. Atmospheric stability, which had the greatest impact distances, was the most influential variable when using the ALOHA tool in urban conditions. Additionally, sensitivity analysis showed that impact distances using the ALOHA tool were larger than those using the KORA tool. Therefore, the ALOHA tool was more sensitive to weather variables than the KORA tool.
The regression equation for the sensitivity analysis according to weather variables, using the KORA and ALOHA tools in rural conditions, was as follows:
Impact   distance = 122.093 + ( 24.628 ) × Air   temperature + ( 27.788 ) × Wind   speed + ( 10.885 ) × Humidity + ( 53.122 ) × KORA + ( 32.556 ) × 50 %   HF + ( 134.613 ) × 28 %   NH 3 + ( 126.775 ) × 35 %   HCl .
The sensitivity analysis results are shown in Table 6. Using the KORA tool, the impact distances for 28% NH3 were 325.68 m with atmospheric stability, 301.05 m with air temperature, and 297.89 m with wind speed. The impact distances for 35% HCl were 317.84 m with atmospheric stability, 293.21 m with air temperature, and 290.05 m with wind speed. The impact distances for 50% HF were 223.62 m with atmospheric stability, 198.99 m with air temperature, and 195.83 m with wind speed. The impact distances for 69% HNO3 were 191.06 m with atmospheric stability, 166.44 m with air temperature, and 163.28 m with wind speed. When using the KORA tool, 28% NH3 was the chemical most affected by the weather variables. Among the meteorological variables, atmospheric stability had the greatest damage distances, making it the most influential variable when using the KORA tool in rural conditions. Sensitivity analysis using the KORA tool in rural conditions showed that the impact distances were larger in rural conditions than in urban conditions. Therefore, the KORA tool was more sensitive to weather variables in rural conditions than in urban conditions. Using the ALOHA tool, the impact distances for 28% NH3 were 378.80 m with atmospheric stability, 354.17 m with air temperature, and 351.01 m with wind speed. The impact distances for 35% HCl were 370.96 m with atmospheric stability, 346.33 m with air temperature, and 343.17 m with wind speed. The impact distances for 50% HF were 276.74 m with atmospheric stability, 252.11 m with air temperature, and 248.95 m with wind speed. The impact distances for 69% HNO3 were 244.19 m with atmospheric stability, 219.56 m with air temperature, and 216.40 m with wind speed. The sensitivity analysis results for the ALOHA tool in rural conditions were the same as for the KORA tool. When using the ALOHA tool, 28% NH3, which had the greatest impact distances, was the most sensitive substance in rural conditions. Atmospheric stability, which had the greatest impact distances among weather variables, was the most influential variable when using the ALOHA tool. Sensitivity analysis using the ALOHA tool in rural conditions showed that the impact distances were larger in rural conditions than in urban conditions. Therefore, the ALOHA tool was more sensitive to weather variables in rural conditions than in urban conditions. Additionally, the impact distances when using the ALOHA tool were larger than those when using the KORA tool. Therefore, the ALOHA tool was more sensitive to weather variables than the KORA tool.

4. Discussion

This study was relevant in terms of statistical analysis of the OCA tools used in KOR and the US, in accordance with weather variables on the user side. The parameters of the scenarios applied in this study may be limited to each facility. This is because the scenarios selected situations with the greatest impact distances following chemical leakages due to equipment damage, breakage, fracture. [34,35]. Moreover, temperature, wind speed, humidity, and atmospheric stability interact with each other in actual atmospheric dispersion [36,37]. However, in this study, these interactions were not considered when analyzing the independent sensitivity of OCA tools to common weather variables.
The sample size used for multiple regression analysis was 640. The normality test used a quantile–quantile (Q–Q) plot. If the number of samples is large, the normality of its distribution can be assumed by the representative approximation theory and the central limit theorem [38]. Central limit theorems have been used in various fields of research [39,40,41,42]. One study in the field of info-communications assumed normality by using the central limit theorem to estimate sensor location in a non-uniform network environment [42], as the central limit theorem proved that many sensor network environments meet a normal distribution model. Therefore, in this study, the central limit theorem could be used to prove that a sufficiently large sample (n ≥ 30) conformed to a normal distribution. Box plots were used to find outliers, of which there were six. Outliers occurred when measurement values were correctly observed, but they were rare cases. In other words, outliers are impact distances that rarely occur in OCA. Outliers are not objects that need to be removed, but rather objects that contain important information about the entire dataset [43]. Not all outliers are a problem [43]. The analysis of outliers in categorical data should be subject to separate subjective considerations [43]. In this study, we did not eliminate outliers, as they were realistic outcomes when using the KORA and ALOHA tools. Therefore, the regression equation used in the sensitivity analysis was a reflection of reality.
Sensitivity analysis of the KORA and ALOHA tools for each substance, according to weather conditions, indicated that 28% NH3 showed the greatest damage distances. Moreover, sensitivity was found to decrease in the following order: 35% HCl, 50% HF, and 69% HNO3. Assuming that emissions per hour and evaporation remain constant, along with temperature, wind speed, and humidity, specific gravity is a physicochemical factor that affects the diffusion of chemicals [37,44,45]. The specific gravity of 28% NH3 is lower than one, and is lower than that of 35% HCl, 50% HF, and 69% HNO3. The specific gravity of 35% HCl is lower than that of 50% HF and 69% HNO3. The specific gravity of 50% HF is lower than that of 69% HNO3 [46]. Therefore, we can objectively observe, using statistical analysis, that substances with low specific gravity are more sensitive to weather variables than those with higher specific gravity.
The effect of humidity on impact distances was not significant (p = 0.112). In other words, there was no change in impact distances according to humidity. Unlike humidity, temperature, wind speed, and atmospheric stability affected damage distances significantly (p < 0.05). The impact distances in urban and rural conditions were largest according to atmospheric stability, while those according to air temperature were greater than those according to wind speed. Moreover, the ALOHA tool had larger impact distances than the KORA tool. Therefore, atmospheric stability in urban and rural areas was the most sensitive factor for the KORA and ALOHA tools, while air temperature was a more sensitive factor than wind speed. The KORA and ALOHA tools were more sensitive to atmospheric stability, air temperature, and wind speed in rural conditions than in urban conditions. In addition, the ALOHA tool was more affected by atmospheric stability, air temperature, and wind speed than the KORA tool. Users of the KORA and ALOHA tools need to be careful to not commit input errors in atmospheric stability, which was the most sensitive factor, so as to reduce variation in damage distances due to those errors. In particular, the ALOHA tool was more sensitive to atmospheric stability, air temperature, and wind speed than the KORA tool. Therefore, users should be more careful to not commit input errors regarding weather variables when using the KORA tool than when using the ALOHA tool.

5. Conclusions

Regression analysis showed that the KORA and ALOHA tools had large impact distances when using the following chemicals, in descending order: 28% NH3, 35% HCl, 50% HF, and 69% HNO3. Therefore, 28% NH3 was the most sensitive chemical to weather variables. Meanwhile, 35% HCl was a more sensitive chemical than 50% HF and 69% HNO3, while 50% HF was a more sensitive chemical than 69% HNO3. From the results of the sensitivity analysis of weather variables, humidity was found to neither affect the KORA nor ALOHA tools. Unlike humidity, atmospheric stability was the most sensitive factor for the KORA and ALOHA tools, with air temperature being a more sensitive factor than wind speed. The weather variables (atmospheric stability, air temperature, and wind speed) were more sensitive factors in rural conditions than in urban conditions, while the ALOHA tool was more sensitive than the KORA tool. Therefore, users of the ALOHA tool need to be careful to not affect impact distances due to input errors in atmospheric stability in rural conditions, as this was the most sensitive weather variable in OCA.

Author Contributions

M.-U.K. and S.-H.B. conceived and designed the experiment; M.-U.K. performed the task of data processing and coding, analyzed the results, and wrote the manuscript; K.W.M. and J.-R.S. were involved in the analysis process and contributed to the data processing work; S.-H.B. edited the manuscript. All authors contributed to and have approved the final manuscript.

Acknowledgments

This study was supported by the development of chemical accident behavior analysis/damage prediction models, and the environmental risk map technology of the Ministry of Environment.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Cho, M.S.; Yoon, Y.S.; Kim, K.J.; Park, Y.S.; Kwon, Y.H.; Chun, K.S.; Yoon, J.H. Study on improved environmental impact assessment measures for chemical accidents. Korean J. Hazard. Mater. 2013, 1, 11–16. Available online: http://www.dbpia.co.kr/Article/NODE02270603 (accessed on 17 September 2017).
  2. Cort, R.D. The development of UK and European major hazards legislation and the review of the Seveso directive-the implications for industry. Disaster Prev. Manag. 1994, 3, 8–14. [Google Scholar] [CrossRef]
  3. Innes, R.; Mitra, A. Parties, politics, and regulation: Evidence from clean air act enforcement. Econ. Inq. 2015, 53, 522–539. [Google Scholar] [CrossRef]
  4. Jung, H.G.; Ma, J.G. A study on legal systems and politics to control chemicals: Focus on regulation of hazardous chemicals. Adm. Law J. 2016, 44, 191–222. Available online: http://uci.or.kr/G704-001312.2016..44.007 (accessed on 11 September 2017).
  5. Park, J.G.; Suh, Y.W.; Gan, S.Y.; Lee, S.W. Improvement Measures for Chemical Accident Policies in the Chemicals Control Act and Measures to Support the Industry(I); Korea Environment Institute: Seoul, Korea, 2013; pp. 1–24. ISBN 979-11-5980-070-2. [Google Scholar]
  6. Kim, D.J.; Lee, I.B.; Moon, J.Y.; Chun, Y.W. Offsite consequence analysis and safety management system process integration plan of safety management system. J. Korea Saf. Manag. Sci. 2016, 18, 63–70. Available online: http://www.ndsl.kr/ndsl/search/detail/article/articleSearchResultDetail.do?cn=JAKO201631642279961 (accessed on 8 April 2017). [CrossRef]
  7. McCoy Associates, Inc. RMP Offsite Consequence Analysis Data Distribution Method Finalized; McCoy Associates, Inc.: Arlington, VA, USA, 2000; pp. 2.23–2.26. [Google Scholar]
  8. Lim, D.Y.; Seo, J.M.; Yoon, G.J.; Shim, M.S.; Lee, G.W.; Baek, U.S. Development of Off-Site Risk Assessment Program Improvement; National Institute Chemical Safety: Daejeon, Korea, 2015; pp. 1–59. [Google Scholar]
  9. Park, Y.K.; Kim, T.O. Evaluation of ammonia gas release in the solar cell manufacturing process using the ALOHA model. J. Odor Indoor Environ. 2015, 14, 136–149. [Google Scholar] [CrossRef]
  10. Jung, S.H.; Yoon, D.Y.; Ghim, Y.S. Usage characteristics of publicly-available accidental release models. J. KOSAE 1999, 15, 687–696. Available online: http://www.ndsl.kr/ndsl/search/detail/article/articleSearchResultDetail.do?cn=JAKO199911919857189 (accessed on 28 March 2017).
  11. National Institute Chemical Safety. Technical Guidelines for Selection of Accident Scenario; National Institute Chemical Safety: Daejeon, Korea, 2014; pp. 1–16. [Google Scholar]
  12. Yu, J.A.; Moon, J.Y.; Chun, K.S.; Hwang, M.S.; Noh, H.R.; Yang, H.S.; Lee, M.S. Characteristic analysis of the atmospheric dispersion assessment for the toxic gases belong to the accident precaution chemicals. In Proceedings of the 45th Meeting of KOSAE, Cheongju, Korea, 25 October 2007; Korean Society for Atmospheric Environment: Seoul, Korea, 2007; pp. 453–454. [Google Scholar]
  13. Won, G.M.; Lee, H.W.; Yu, J.A.; Hong, H.S.; Hwang, M.S.; Chun, K.S.; Choi, K.S.; Lee, M.S. Applicable evaluation of the latest land-use data for developing a real-time atmospheric field prediction of RAMS. J. KOSAE 2008, 24, 1–15. Available online: http://www.dbpia.co.kr/Article/NODE00959499 (accessed on 19 October 2017). [CrossRef]
  14. Yu, J.A.; Hwang, M.S.; Chun, K.S.; Kwon, Y.H.; Moon, J.Y.; Lee, J.S.; Yoon, I.; Park, C.H.; Park, Y.S.; Shin, S.I.; et al. Model sensitivity study on various ambient meteorological conditions using SLAB for hazardous chemical accidents. In Proceedings of the 43rd meeting of KOSAE, Mokpo, Korea, 26 October 2006; Korean Society for Atmospheric Environment: Seoul, Korea, 2006; pp. 101–102. [Google Scholar]
  15. Ministry of Environment. The Effect of Meteorological and Surface Characteristic Factors on the Diffusion of Hazardous Chemical Mass Leaks; Ministry of Environment: Sejong, Korea, 2010; pp. 1–111.
  16. Kim, J.H.; Jung, S.H. Offsite consequence modeling for HF accidental release scenarios. Theor. Appl. Chem. Eng. 2015, 21, 1176–1179. [Google Scholar] [CrossRef]
  17. Shin, C.H.; Park, J.H. An evaluation of the off-site risk of spill from a storage tank of nitric acid. Crisisonomy 2016, 12, 187–200. [Google Scholar] [CrossRef]
  18. Jung, S.H.; Lee, Y.S.; Lim, O.J.; Yoo, J.M. A Study on the Improvement of Environmental Impact Assessment of Industrial Complexes Based on Risk Assessment of Chemical Leakage Accidents; Korea Environment Institute: Seoul, Korea, 2013; pp. 3–225. ISBN 978-89-8464-786-2. [Google Scholar]
  19. Park, J.K.; Seo, Y.W. A Study on the Improvement of the Chemical Accident Response System; Korea Environment Institute: Seoul, Korea, 2013; pp. 1–122. [Google Scholar]
  20. Shin, C.H.; Lee, C.S.; Kang, J.E.; Ma, B.C.; Yoon, Y.; Yoon, J.H.; Park, J.H. Review on the safety management system of facilities handling hazardous chemicals under the chemical control act. Crisisonomy 2015, 11, 19–33. Available online: http://uci.or.kr/G704-SER000001473.2015.11.7.003 (accessed on 16 May 2017).
  21. National Institute Chemical Safety. Korea Off-Site Risk Assessment Supporting Tool User’s Manual; National Institute Chemical Safety: Daejeon, Korea, 2017; pp. 4–64. [Google Scholar]
  22. U.S. Environmental Protection Agency. ALOHA User’s Manual; U.S. Environmental Protection Agency: Washington, DC, USA; National Oceanic and Atmospheric Administration: Seattle, WA, USA, 2007; pp. 11–187.
  23. Ahn, S.R.; Kim, S.B.; Lee, J.H.; Chun, K.S. Study on chemical incident response plan identified as chemical accident statistics. Korean J. Hazard. Mater. 2014, 2, 50–54. Available online: http://www.dbpia.co.kr/Article/NODE02423733 (accessed on 6 July 2017).
  24. Lee, D.J.; Lyu, S.W.; Song, C.G. Improvement strategy for management of accident preparedness substances causing chemical accidents. J. Korean Soc. Saf. 2017, 32, 47–52. Available online: http://www.ndsl.kr/ndsl/search/detail/article/articleSearchResultDetail.do?cn=JAKO201726868680895 (accessed on 22 November 2017). [CrossRef]
  25. You, J.S.; Chung, Y.J. Case analysis of the harmful chemical substances spill. Fire Sci. Eng. 2014, 28, 90–98. Available online: http://www.dbpia.co.kr/Article/NODE06070955 (accessed on 6 July 2017). [CrossRef]
  26. National Institute of Environmental Research. Hazardous Chemical Substance Accident Case Book; National Institute of Environmental Research: Incheon, Korea, 2007; pp. 3–197.
  27. American Industrial Hygiene Association. ERPG/WEEL Handbook; American Industrial Hygiene Association: Akron, OH, USA, 2017; pp. 1–31. ISBN 978-1-935082-81-1. [Google Scholar]
  28. Korean Statistics Information Service. Available online: http://kosis.kr (accessed on 2 January 2017).
  29. Lee, Y.M.; Chung, H.J.; Lee, S.J. A study on edit order of text cells on the MS Excel files. J. Korea Inst. Inf. Secur. Cryptol. 2014, 14, 319–325. Available online: http://www.ndsl.kr/ndsl/search/detail/article/articleSearchResultDetail.do?cn=JAKO201418342936736 (accessed on 15 April 2017). [CrossRef]
  30. Hong, S.H.; Jung, S. Testing the interaction effects in regression and structural equation models: Theories and procedures. Korean J. Hum. Dev. 2014, 21, 1–24. [Google Scholar] [CrossRef]
  31. Lee, J.Y.; Lee, H.G. Multifactor dimensionality reduction (MDR) analysis by dummy variables. Korean J. Appl. Stat. 2009, 22, 435–442. Available online: http://www.ndsl.kr/ndsl/search/detail/article/articleSearchResultDetail.do?cn=JAKO200917639072399 (accessed on 20 February 2017). [CrossRef]
  32. International Business Machines Corporation. IBM SPSS Statistics Base 24; IBM Corp.: New York, NY, USA, 2016; pp. 33–160. [Google Scholar]
  33. Park, H.S.; Jeong, S.M.; Chung, G.H. Frequency analysis of future maximum fresh snow depth using multiple regression model with interaction. J. Korean Soc. Hazard Mitig. 2016, 16, 369–376. Available online: http://www.ndsl.kr/ndsl/search/detail/article/articleSearchResultDetail.do?cn=JAKO201621650493832 (accessed on 8 January 2017). [CrossRef]
  34. Korea Occupational Safety and Health Agency. Technical Guidelines for Selecting the Worst Leak Scenario; Korea Occupational Safety and Health Agency: Ulsan, Korea, 2012; pp. 1–14. [Google Scholar]
  35. Park, KS. Offsite risk assessment on toxic release. J. Korean Inst. Gas 2017, 21, 9–16. Available online: http://www.ndsl.kr/ndsl/search/detail/article/articleSearchResultDetail.do?cn=JAKO201732663193045 (accessed on 17 August 2017). [CrossRef]
  36. Klein, T.; Kukkonen, J.; Dahl, A.; Bossioli, E.; Baklanov, A.; Vik, A.F.; Agnew, P.; Karatzas, K.D.; Sofiev, M. Interactions of physical, chemical, and biological weather calling for an integrated approach to assessment, forecasting, and communication of air quality. AMBIO 2012, 41, 851–864. [Google Scholar] [CrossRef] [PubMed]
  37. Hwang, W.T.; Kim, E.H.; Jeong, H.S.; Jeong, H.J.; Han, M.H. Influence of modelling approaches of diffusion coefficients on atmospheric dispersion factors. J. Radiat. Prot. 2013, 38, 60–67. Available online: http://www.ndsl.kr/ndsl/search/detail/article/articleSearchResultDetail.do?cn=JAKO201322045996209 (accessed on 4 June 2017). [CrossRef]
  38. Datasolution Consulting Team. SPSS Statistics Descriptive Statistics and Correlation Analysis; Datasolution: Seoul, Korea, 2016; pp. 1–118. ISBN 978-89-8839-397-0. [Google Scholar]
  39. Kwon, J.S.; Kim, Y.K.; Joo, S.Y.; Choi, G.S. Central limit theorems for fuzzy random sets. J. Fuzzy Log. Intell. Syst. 2005, 15, 337–342. [Google Scholar] [CrossRef]
  40. Choi, H.S.; Kim, T.Y. Computer simulation program for central limit theorem—Dynamic MS Excel program. J. Korean Data Inf. Sci. Soc. 2005, 16, 359–369. [Google Scholar]
  41. Park, B.J.; Roh, C.G.; Kim, J.S. A case study of panoramic section image collection method for measuring density—With matched image in the Seoul beltway Sapaesan Tunnel. J. Korea Inst. Intell. Transp. Syst. 2014, 13, 20–29. [Google Scholar] [CrossRef]
  42. Park, H.; Hwang, D.K.; Park, J.H.; Seong, D.O.; Yoo, J.S. Sensor positioning scheme using density probability models in non–uniform wireless sensor networks. J. Korea Contents Assoc. 2012, 12, 55–66. [Google Scholar] [CrossRef]
  43. Jun, S.H. An outlier data analysis using support vector regression. J. Korean Inst. Intell. Syst. 2008, 18, 876–880. [Google Scholar] [CrossRef]
  44. Kim, M.S.; Park, J.K. Modelling methodology for estimation of the harmful hazardous distance and release time on release of hazardous toxic substances. J. Korean Soc. Environ. Adm. 2001, 7, 1–11. Available online: http://www.dbpia.co.kr/Article/NODE01409802 (accessed on 28 August 2016).
  45. Risk Management Program Guidance for Offsite Consequence Analysis. Available online: https://www.epa.gov/rmp/rmp-guidance-offsite-consequence-analysis (accessed on 5 January 2017).
  46. Globally Harmonized System-Material Safety Data Sheet (GHS-MSDS). Available online: http://msds.kosha.or.kr (accessed on 20 April 2017).
Figure 1. Normal quantile–quantile (Q–Q) plots of impact distances according to weather variables in urban conditions: (a) using the Areal Location of Hazardous Atmospheres (ALOHA) tool; (b) using the Korea Offsite Risk Assessment (KORA) tool.
Figure 1. Normal quantile–quantile (Q–Q) plots of impact distances according to weather variables in urban conditions: (a) using the Areal Location of Hazardous Atmospheres (ALOHA) tool; (b) using the Korea Offsite Risk Assessment (KORA) tool.
Ijerph 15 01027 g001
Figure 2. Normal Q–Q plots of impact distances according to weather variables in rural conditions: (a) using the ALOHA tool; (b) using the KORA tool.
Figure 2. Normal Q–Q plots of impact distances according to weather variables in rural conditions: (a) using the ALOHA tool; (b) using the KORA tool.
Ijerph 15 01027 g002
Figure 3. Box plot of impact distances according to weather variables in urban conditions.
Figure 3. Box plot of impact distances according to weather variables in urban conditions.
Ijerph 15 01027 g003
Figure 4. Box plot of impact distances according to weather variables in rural conditions.
Figure 4. Box plot of impact distances according to weather variables in rural conditions.
Ijerph 15 01027 g004
Table 1. Input data provided to the tools.
Table 1. Input data provided to the tools.
ToolsClassificationInformation
Areal Location of Hazardous Atmospheres (ALOHA)Measurement height above ground3 m
Cloud coverPartly cloudy
SourcePuddle
Puddle diameter10 m
Mass of puddle0.54 ton (50% hydrofluoric acid, HF), 0.42 ton (28% ammonia, NH3)
0.54 ton (35% hydrogen chloride, HCl), 0.6 ton (69% nitric acid, HNO3)
Ground typeConcrete
Korea Offsite Risk Assessment (KORA)Equipment appearanceVertical cylinder (drum)
Equipment diameter3 m
Equipment height6 m
Storage amount30 ton
Operating pressure1.0 kg/cm2
Bonded pipe diameter50 mm
Leakage typeStorage tank leakage
Height of leakage hole0.5 m
Diameter of leakage hole10 mm
Table 2. Offsite consequence analysis (OCA) results for 50% HF and 28% NH3 by weather variables (N = 640).
Table 2. Offsite consequence analysis (OCA) results for 50% HF and 28% NH3 by weather variables (N = 640).
Weather Variables50% HF28% NH3
KORAALOHAKORAALOHA
Urban (m)Rural (m)Urban (m)Rural (m)Urban (m)Rural (m)Urban (m)Rural (m)
Air Temperature−5 °C367738718418195156
0 °C3677427784182106175
5 °C3678478485184118194
10 °C3779539186186132217
15 °C3780589987188147244
20 °C37806410888189164273
25 °C38817111988191183306
30 °C38828113689193200336
35 °C38829215390194208351
Wind Speed1 m/s439288145101217215339
2 m/s39858013393200199318
3 m/s38817111988191183306
4 m/s36786811486185169283
5 m/s35766611183180160266
6 m/s35756410882176152252
7 m/s34736310680173145241
8 m/s34726110479170139231
9 m/s33716010278168135222
10 m/s33705910077166130215
11 m/s3270589976164126208
12 m/s3269579875162122203
13 m/s3268569775161119197
14 m/s3168559674160116193
15 m/s3167549573158114189
16 m/s3167539473157111184
Humidity10%38817111988191183306
20%38817111988191183306
30%38817111988191183306
40%38817111988191183306
50%38817111988191183306
60%38817111988191183306
70%38817111988191183306
80%38817111988191183306
90%38817111988191183306
Atmospheric StabilityA2026333745598191
B20393856459095138
C2658498359133128207
D38817111988191183306
E60130118198140307305461
F60221145351140532378734
Table 3. Some of the data sources from the OCA results converted to dummy variables (N = 604).
Table 3. Some of the data sources from the OCA results converted to dummy variables (N = 604).
50% HF28% NH330% HCl69% HNO3KORAALOHAUrbanRuralATWSHUASID (m)
10001010100036
10001010100036
10001010100036
10001010100037
10001010100037
10001010100037
10001010100038
10001010100038
10001010100038
10001010010043
10001010010039
10001010010038
10001010010036
10001010010035
10001010010035
10001010010034
10001010010034
10001010010033
10001010010033
10001010010032
10001010010032
10001010010032
10001010010031
10001010010031
10001010010031
10001010001038
10001010001038
10001010001038
10001010001038
10001010001038
10001010001038
10001010001038
10001010001038
10001010001038
10001010000120
10001010000120
10001010000126
10001010000138
10001010000160
10001010000160
10001001100077
10001001100077
AT: air temperature. WS: wind speed. HU: humidity. AS: atmospheric stability. ID: impact distance.
Table 4. Regression analysis for the sensitivity analysis of weather variables.
Table 4. Regression analysis for the sensitivity analysis of weather variables.
Variableβt
Air temperature−24.628−3.602 *
Wind speed−27.788−4.475 *
Humidity−10.885−1.592
KORA−53.122−12.950 *
Urban−69.766−17.008 *
50% HF32.5565.612 *
28% NH3134.61323.205 *
35% HCl126.77521.854 *
Constant122.09317.428 *
F Value161.716 *
Adjusted R20.668
* p < 0.05.
Table 5. Sensitivity analysis of weather variables according to chemicals in urban conditions.
Table 5. Sensitivity analysis of weather variables according to chemicals in urban conditions.
ChemicalsWeather VariablesKORA (m)ALOHA (m)
28% NH3Atmospheric stability255.91309.03
Air temperature231.28284.40
Wind speed228.12281.25
35% HClAtmospheric stability248.07301.20
Air temperature223.45276.57
Wind speed220.29273.41
50% HFAtmospheric stability153.86206.98
Air temperature129.23182.35
Wind speed126.07179.19
69% HNO3Atmospheric stability121.30174.42
Air temperature96.67149.79
Wind speed93.51146.63
Table 6. Sensitivity analysis of weather variables according to chemicals in rural conditions.
Table 6. Sensitivity analysis of weather variables according to chemicals in rural conditions.
ChemicalsWeather VariablesKORA (m)ALOHA (m)
28% NH3Atmospheric stability325.68278.80
Air temperature301.05354.17
Wind speed297.89351.01
35% HClAtmospheric stability317.84370.96
Air temperature293.21346.33
Wind speed290.05343.17
50% HFAtmospheric stability223.62276.74
Air temperature198.99252.11
Wind speed195.83248.95
69% HNO3Atmospheric stability191.06244.19
Air temperature166.44219.56
Wind speed163.28216.40

Share and Cite

MDPI and ACS Style

Kim, M.-U.; Moon, K.W.; Sohn, J.-R.; Byeon, S.-H. Sensitivity Analysis of Weather Variables on Offsite Consequence Analysis Tools in South Korea and the United States. Int. J. Environ. Res. Public Health 2018, 15, 1027. https://doi.org/10.3390/ijerph15051027

AMA Style

Kim M-U, Moon KW, Sohn J-R, Byeon S-H. Sensitivity Analysis of Weather Variables on Offsite Consequence Analysis Tools in South Korea and the United States. International Journal of Environmental Research and Public Health. 2018; 15(5):1027. https://doi.org/10.3390/ijerph15051027

Chicago/Turabian Style

Kim, Min-Uk, Kyong Whan Moon, Jong-Ryeul Sohn, and Sang-Hoon Byeon. 2018. "Sensitivity Analysis of Weather Variables on Offsite Consequence Analysis Tools in South Korea and the United States" International Journal of Environmental Research and Public Health 15, no. 5: 1027. https://doi.org/10.3390/ijerph15051027

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop