Science of the Total Environment 472 (2014) 901–911
Contents lists available at ScienceDirect
Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv
Modelling the propagation of smoke from a tanker fire in a built-up area Lucyna Brzozowska University of Bielsko-Biala, Faculty of Management and Computer Science, Willowa 2, 43-309 Bielsko-Biała, Poland
H I G H L I G H T S • • • • •
The application of a custom-developed model of pollutants dispersion is presented. The application of model is connected with safety in road transport. Numerical simulations are performed for a case of smoke emission from a tanker fire. The results of simulation of smoke propagation in a built-up area are presented. The GIS is used to provide data for pre- and post-processing purposes.
a r t i c l e
i n f o
Article history: Received 2 September 2013 Received in revised form 13 November 2013 Accepted 25 November 2013 Available online 15 December 2013 Keywords: Numerical modelling Safety in road transport Environmental protection Tanker fire Lagrangian particle model Diagnostic wind field model
a b s t r a c t The paper presents the application of a Lagrangian particle model to problems connected with safety in road transport. Numerical simulations were performed for a hypothetical case of smoke emission from a tanker fire in a built-up area. Propagation of smoke was analysed for three wind directions. A diagnostic model was used to determine the air velocity field, whereas the dispersion of pollutants was analysed by means of a Lagrangian particle model (Brzozowska, 2013). The Idrisi Andes geographic information system was used to provide data on landforms and on their aerodynamic roughness. The presented results of computations and their analysis exemplify a possible application of the Lagrangian particle model: evaluation of mean (averaged over time) concentrations of pollutants and their distribution in the considered area (especially important due to the protection of people, animals and plants) and simulation of the propagation of harmful compounds in time as well as performing computations for cases of the potential effects of road incidents. © 2013 Elsevier B.V. All rights reserved.
1. Introduction Modelling the dispersion of pollutants belongs to a group of fundamental problems that are connected with evaluating the impact transport has on the environment. Although evaluation for longer periods of time and on larger scales (meso- and macro-) is relatively easy, in individual road accidents assessment of their effects proves to be difficult. This includes incidents resulting in the sudden release of pollutants. A tanker fire in a tunnel is a particularly important problem that has been discussed in a number of scientific papers (Beard and Carvel, 2005; Caliendo et al., 2012; Colella et al., 2010; Hu et al., 2008; Miles and Smithies, 2006; Xiaojun, 2008; PIARC, 1999). According to PIARC (1999), in this case models of the CFD class are predominantly used, as in: Caliendo et al. (2012), Beard and Carvel (2005), Hu et al. (2008), Miles and Smithies (2006), and Xiaojun, (2008); less often zone models (Hua et al., 2005; Jain et al., 2008; Xiaojun, 2008; Yao et al., 1999) and one-dimensional models play an auxiliary role (Colella et al., 2010).
E-mail address: [email protected]. 0048-9697/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.scitotenv.2013.11.130
It should be noted that road accidents and vehicle fires, including vehicles that transport fuel, do not only take place in tunnels, but also elsewhere. Fires occurring in close proximity to buildings are especially dangerous — this is the case that this article aims to analyse. Basically, there are two types of models that are used to model dispersion of smoke originating from a fire: microscale ones — such as the aforementioned CFD models (Novozhilov, 2001; Viskanta, 2008), which are predominantly used in problems related to fires in tunnels and closed spaces (e.g. a garage or an atrium (Qin et al., 2009)), and also in other cases when phenomena that spatially occur in the microscale (such as the fire of a fuel tank (Argyropoulos et al., 2010; Markatos et al., 2009)) must be taken into account; and mesoscale as well as macroscale ones, which are used most often to analyse forest fires (Lavrov et al., 2006; Mell et al., 2010). The following types may be distinguished in the group of mesoscale and macroscale models: empirical and theoretical (a comparison of models for the case of modelling dispersion of a smoke plume originating from a fire is presented in Fisher et al. (2001)), statistical (Yang et al., 2004) and numerical models in Euler (Vautard et al., 2007) or Lagrange (Stohl et al., 2005) coordinates. Another division of models of fires is possible: into probabilistic models (event trees, statistical models, Monte Carlo simulations)
902
L. Brzozowska / Science of the Total Environment 472 (2014) 901–911
Fig. 1. Example of approximation of land surface with buildings by using the continuous function s (Brzozowska, 2013).
and deterministic models (zone models, discrete ones (field models, CFD)). A taxonomy of models used in simulations of forest fires and their analysis can also be found in Pastor et al. (2003). Modelling propagation of smoke originating from fires in urban canyons and in close proximity to buildings is a problem that is rather seldom discussed. Nevertheless, there are papers (Hu et al., 2009, 2011; Liu et al., 2011) worth mentioning in which the problem of smoke propagation from a building on fire into the street is analysed
by using either (Hu et al., 2009, 2011) LES or experimental data (Liu et al., 2011) for computer simulation. Geographic Information Systems (GIS) are used in modelling the dispersion of pollutants in the mesoscale and macroscale (Pastor et al., 2003). They may prove useful in preprocessing or postprocessing. The current paper uses a Lagrangian numerical model combined with a diagnostic model of the air velocity field (Brzozowska, 2013). Problems related to transfer of heat were neglected in modelling the dispersion of smoke caused by an accident and the ignition of a tanker on a
Fig. 2. Map of the modelled area and the location of the accident (marked with a cross).
L. Brzozowska / Science of the Total Environment 472 (2014) 901–911
903
2. Numerical models
measurement data are interpolated at a given altitude above the ground; and, next, the initial vertical profile of air velocity is determined at node points of the discretisation net. Brzozowska (2013) contains a basic description of the model.
2.1. Diagnostic model of an air velocity field
2.2. Lagrangian model of dispersion of pollutants
Modelling the phenomenon of dispersion in the atmosphere is directly linked with the need to know the air velocity field. Prognostic models may be used for this purpose. These use complex physical phenomena while requiring huge computing power. Simpler diagnostic models (Ehrhard et al., 2000; Seaman, 2000) are also used. This paper employs a diagnostic model of the air velocity field for conditions involving complex topography. A variational algorithm (Homicz, 2002) was used to satisfy the criterion of mass balance. An equation of the elliptic type was formulated and then solved by using the method of conformal maps and finite differences for a nonequidistant discretisation net. The task was reduced to the problem of solving a system of algebraic linear equations. The iterative conjugate gradient method was subsequently used to solve it. The initial velocity field was determined by adjusting the air velocity field to the measurement data. The method that was used comprises two stages: first,
Particle models, besides their usefulness in analysing problems on larger spatial scales (macro and meso), may be used to predict concentrations of pollutants around areas of direct emission (roads, car parkings), as well as to assess the influence of proposed modifications of road infrastructure on the environment. These models also continually evaluate the impact of existing emission sources (Brzozowska, 2013; Clarke et al., 2004; Grašič et al., 2011; Stohl et al., 2005; Souto et al., 2001; Wilson and Sawford, 1996; Wang et al., 2008). Their role may also be significant in predicting and evaluating the propagation of pollutants emitted in sudden releases Kim et al. (2009) or in malfunctions occurring in road accidents. Location r of a particle representing a conventional mass of pollutants in space is determined by its velocity vector, which is the sum of the mean air velocity u and the fluctuation velocity u′ : u ¼ u þ u′ . The fluctuation velocity can be described with the Monte Carlo method
road. Idrisi Andes GIS (geographic information system) was used to provide data on buildings, landforms and their aerodynamic roughness.
Fig. 3. The modelled area: a) landforms, b) assumed directions of the wind vector and location of the source of smoke emission.
904
L. Brzozowska / Science of the Total Environment 472 (2014) 901–911
in which the behaviour of a particle fulfils one possibility out of an infinite set. Both the location r of a particle in space and its fluctuation velocity u′ are assumed to be driven by the Markov process, which is discrete in time (Mayer et al., 2008; Souto et al., 2001): h i ′ rnþ1 ¼ rn þ τ uðrn Þ þ unþ1 ;
ð1Þ
′
′
unþ1 ¼ ψðr n Þ un þ ω;
ð2Þ
where rn = r(tn), rn + 1 = r(tn + 1), un ¼ uðt n Þ, un′ = u′(tn), un′ + 1 = u′(tn + 1), tn + 1 = tn + τ, τ is the time step, ψ is the matrix defining the autoregressive summand ψ ⋅ un′ of the fluctuation velocity u′, and ω is the vector defining the partially random summand of the fluctuation velocity.
Fig. 4. Air velocity field in the modelled area at the 3rd layer of the discretisation net (at an altitude of about 2 m) for: Case 1 — a), Case 2 — b), Case 3 — c) of the initial air velocity vector and for the fragment of the area around the emission source.
L. Brzozowska / Science of the Total Environment 472 (2014) 901–911
905
the velocity field), the concentration of a pollutant in volume V ið1Þ ;ið2Þ ;ið3Þ is computed from the formula: −1
C ið1Þ ;ið2Þ ;ið3Þ ðt Þ ¼ V ið1Þ ;ið2Þ ;ið3Þ
N X
mp δp ;
ð3Þ
p¼1
where mp is the mass of particle number p, δp equals 1 for rp ∈V ið1Þ ;ið2Þ ;ið3Þ and 0 otherwise, and N denotes the total number of particles in the considered area. 3. Simulation of smoke propagation Fig. 5. Change of emission intensity in time.
Determining the matrix ψ and calculating its elements by using Cholesky decomposition (Janicke, 2000) was performed according to the procedure proposed in Mayer et al. (2008). A model of the motion of particles also requires a suitable algorithm that takes into account the influence of reflection and permeation at the upper boundary of the mixing layer and reflection on the land surface, as analysed in detail in Brzozowska (2013). Both for the diagnostic model and for the Lagrangian particle model, the terrain (including buildings) was assumed to be represented by an approximating function s = s(x(1)x(2)) (Brzozowska, 2013). This approach distinguishes the model that is used from the other models of this type, e.g. the QUIC-Plume model as described in Williams et al. (2004). Fig. 1 shows the way the surface is approximated along with the buildings. By applying function s, which can describe the domain with the buildings inside it, the presented model allows to model the dispersion of pollutants on an urban scale. The model can be especially useful in cases when it is necessary to obtain calculation results in a short period of time. Such urgent calculation results might be necessary when modelling the sudden release of dangerous gases in car accidents or in a fire. The dispersion model was described in detail in Brzozowska (2013). In Lagrangian particle model's concentration of a pollutant is computed by counting particles in isolated volumes of the atmosphere or by using so-called smoothing functions (Vitali et al., 2006). With the method of counting particles in isolated volumes (the volumes may be determined by the discretisation net used to model
Predicting the impacts of emissions for both hypothetical and real sources in a given area requires each time the selection of an appropriate model, whose accuracy has been determined under comparable conditions. For example, the location of an emissions source in a street canyon, or the application of the model used in this paper for development conditions in a city centre area with close high-rise buildings forming street canyons, is not allowable without a prior validation process. The analysed case is the ignition of a fuel-filled tanker due to a road accident on a bypass in Bielsko-Biała (Fig. 2), which is a medium-sized city (about 170,000 inhabitants) situated in southern Poland, in the Silesian foothills. The city features complex landforms and varied buildings. The considered area has low- and medium-sized building heights. The location of potential emission has been situated outside the immediate developed area in an open space comprising a sharp road curve (a route turn angle of over 120 degrees), constituting a ringroad. The road is a dual carriageway with two lanes for each direction. The closest buildings are located at a distance greater than 100 m from the adopted source location. At the same time, this location of the emission source relative to the area, for which calculations will be carried out, corresponds to the conditions for which the dispersion model has been validated. On these grounds it could be expected that the average concentration levels to be determined will be calculated with an accuracy comparable to that obtained in the validation process presented in another study (Brzozowska, 2013). In addition, the mapping of the surfaces limiting the air flow and smoke dissipation employed in the model using a single continuous function additionally restricts its application to areas of less dense urban development. These conditions are met by the area under examination, which comprises the outer areas of the city, through which the ring-road part under consideration passes.
Fig. 6. Change in time of the concentration of pollutants along a road with a distance from the source for Case 3.
906
L. Brzozowska / Science of the Total Environment 472 (2014) 901–911
Fig. 7. Field of a 5-minute mean concentration for Case 1: a) in the entire simulation area for t = 1500 s and in the marked fragment of the area for: b) t = 1500 s, c) t = 2400 s, d) t = 3600 s.
The total area of the analysed region was 9 km2, of which the builtup area constituted 17%. 3.1. Determining the air velocity field Three variants of the initial air velocity field were considered in the analysed case of smoke dispersion. In the first variant the air flow direction was assumed to be α = 140°, in the second variant it was α = 219° and in the third variant it was α = 180°, which was relative to the horizontal axis (x(1)); in all variants the wind speed was u = 2.06 m/s. A terrain map from GIS (Idrisi Andes) was imported into the model along with s(x,y) for buildings and aerodynamic roughness coefficients of the terrain. Fig. 3 shows the modelled area: landforms and both wind directions. Sensors are assumed to be placed at 9 points of the area (at an altitude of 8 m above the terrain), thus giving the initial profile of air velocity at the ends of the edges, midpoints of the edges and the centre of the area. The nodes of the discretisation net were assumed to be nonequidistant in the vertical direction (32 nodes up to an altitude of about 630 m) — according to the Chebyshev polynomial, whereas for the horizontal directions equidistant nodes were used with a 15 m step (200 nodes). In total, the analysed area was divided into 1,280,000 cells of different heights. The air velocity field in the x(1)x(2) plane at the altitude of about 2 m (the 3rd layer) used in computation above the ground is presented in Fig. 4. This altitude is variable depending on the landforms due to the
non-equidistant discretisation net that is used and varies as indicated with a grayscale from 280 m to 400 m. Wind speed was calculated after assuming atmospheric conditions corresponding to the neutral equilibrium of the atmosphere. 3.2. Simulation of the sudden release of smoke particles When modelling the phenomenon of the sudden release of particles, it was assumed that it was caused by the fire of a fuel-filled tanker. According to the Eureka HGV fire test, the peak power of the fire of a fuel-filled tanker reaches the range of 100, 120 MW — but only for a short period of time (PIARC, 1999). Smoke emission intensity is thereby estimated at 100–300 m3/s at a temperature of 300 °C (PIARC, 1999). This analysis was done with an assumed value of 200 m3/s, which for a smoke density of 300 mg/m3 gives a maximum value of smoke emission intensity that is equal to 60 g/s. Propagation of smoke is assumed to last 60 minutes, reaching its maximum after 1200 s and progressing in accordance with the curve as shown in Fig. 5. The subsequent figure presents the change of concentration in time at an altitude of 1 m above the road surface at different distances from the emission source for Case 3 (Fig. 6). Figs. 7–9 present momentary means over 5-minute periods of concentrations of pollutants in the entire simulation area for a simulation duration of 1500 s and in an area where x(1)∈b 2000;3000 N and x(2)∈b1000;2000N for selected moments. These concentrations occur at 1.5 m above the land surface.
L. Brzozowska / Science of the Total Environment 472 (2014) 901–911
907
Fig. 8. Field of a 5-minute mean concentration for Case 2: a) in the entire simulation area for t = 1500 s and in the marked fragment of the area for: b) t = 1500 s, c) t = 2400 s, d) t = 3600 s.
The width of the smoke cloud in which 5-minute mean concentrations exceeded 2 mg/m3 at the time of the most intense smoke emission (t = 1500 s) was 160 m, whereas an hour after the fire had started it was 50 m. Fig. 10. shows concentrations of pollutants in a selected crosssection at a distance of 200 m from the emission source for different moments of the simulation for Case 3. It is important, in terms of the point of view of safety, to examine the way in which a fire influences its surroundings. An example of plume propagation in selected cross-sections that are parallel to axes x(2) at time t = 1500 s for the analysed cases are shown in Figs. 11–13. The diagrams allow for the land topography and development. The built-up area covered by the smoke cloud in cases where the wind direction was α = 140° (Case 1), α = 219° (Case 2), and α = m2, 31,050 m2, and 13,730 m2, respectively, and 1 hour after the fire had started it was 0 for Case 1, 1350 m2 for Case 2 and 225 m2 for Case 3. In the first case no concentrations exceeding 5 mg/m3 were found in the built-up area at the time of peak emission, in the second case such concentrations occurred in areas of 11,900 m2 in total, whereas in the third case they occurred in areas of 3800 m2. Also, concentrations with values N 10 mg/m3 in the built-up area were registered in the second case, in areas of 4950 m2 in total and in areas of about 900 m2 in Case 3.
An hour into the simulation (fire) the determined maxima of concentration for the built-up area were 1 mg/m3 in Case 1, 9 mg/m3 in Case 2 and 3 mg/m3 in Case 3. As can be seen, the largest number of buildings is under the smoke cloud when the wind direction is α = 219°, i.e. northwest wind. The smoke cloud covers a smaller total built-up area in Case 3; however, the road itself was not taken into account as the built-up area in the computations. On the other hand, in the first case mostly forest areas, orchards and scrubland were exposed to smoke concentrations more then 2 mg/m3. In the third case the axis of the smoke plume coincides with the centre line of the road on which the accident happened. Fig. 14 shows a map of the smoke concentration for Case 3 at an altitude of 1 m above the land surface for t = 1500°s. In this area concentrations with values over 2 mg/m3 were found at distances of more than 1 km from the emission source, both at the fire's peak intensity and at the moment of 1 hour into the simulation. Higher concentrations, with values above 5 mg/m3 and 10 mg/m3, occur respectively in 800 m and 600 m strips at time t = 1500 s, and 700 m and 550 m at time t = 3600 s. Modelling of the propagation of the smoke plume formed as a result of the fire enables the identification of the areas, where the time of exposure to various chemicals contained in the smoke can jeopardize the human health. The currently applicable standards of air quality,
908
L. Brzozowska / Science of the Total Environment 472 (2014) 901–911
Fig. 9. Field of 5-minute mean concentration for Case 3: a) in the entire simulation area for t = 1500 s and in the marked fragment of the area for: b) t = 1500 s, c) t = 2400 s, d) t = 3600 s.
depending on the exposure time, define the relevant allowable levels for annual and 24 hours' mean values. In the case of intermittent emissions, limits defined for shorter averaging times are practically applied for the assessment of the level of risk to human health. In further calculations, a single indicator defined for total suspended particulate matter was decided to be adopted as the threshold value defining the allowable exposure level. Thus, it is not necessary to take the assumption concerning chemical composition. In such cases, a 30-minute averaging period has been used for calculation purposes in Poland, with taking the value of 350 mg/m3 for total suspended particles (TSP) as the allowable average 30-minute concentration limit (Official Gazette, 1998). The use of this indicator enables the identification of the zones of the area under analysis, in which the calculated average concentrations exceed the adopted limit. Fig. 15 shows the overall range of the smoke plume for each of the considered cases, and the identified areas where the concentration exceeds the threshold value. The presented data applies to the values calculated for the altitude of 1.5 m above ground level. Taking into account the type of development, the number of inhabitants in the area under examination (about 52,760 people) and in individual zones can be estimated based on the statistical data for the city. Depending on the wind direction, the maximum range of impact the smoke plume covers the examined area to a varying extent. In Case 1, the number of people who will potentially fall into the smoke plume is approximately 6.7% of the total population of the examined area; in Case 2 it is 23.5%, and in Case 3, 14.8%.
In turn, the number of people in zones, where the adopted allowable limit values are exceeded is, respectively, 2.5%, 5% and 3% of the area inhabitants for the analysed cases. Based on the performed calculations, the worst case in terms of the effects of the potential impact of the smoke plume is considered to be Case 2. For this wind direction, the number of people exposed to smoke concentrations above the threshold value is approximately 2640, while in the most favourable instance — Case 1, it is about 1320 people. And for Case 3, where the smoke plume axis largely coincides with the road axis, the number of people in the zone with the exceeded threshold value is about 1580. These differences are due to not only the density of development, but also its type. The smoke plume in Case 2 covers predominantly multi-family residential areas, while the development on the right side of the road (Case 1) is made up exclusively by single-family houses in a detached and row housing. 4. Summary The presented results of computations and their analysis exemplify a possible application of the Lagrangian particle model. In addition to standard simulations, whose subjects of evaluation are mean (averaged over time) concentrations of pollutants and their distribution in the considered area and especially taking into account areas that are important due to the protection of people, animals and plants, such models also lend themselves to other tasks. One of these is the simulation of the propagation of harmful compounds in time as well as performing
L. Brzozowska / Science of the Total Environment 472 (2014) 901–911
Fig. 10. Profile of concentration in cross-section for x(1) = 2640 m, for selected moments in Case 3.
Fig. 11. Cross-section parallel to axes x(2) at time t = 1500 s for Case 1.
Fig. 12. Cross-section parallel to axes x(2) at time t = 1500 s for Case 2.
909
910
L. Brzozowska / Science of the Total Environment 472 (2014) 901–911
Fig. 13. Cross-section parallel to the axes x(2) at time t = 1500 s for Case 3.
Fig. 14. Concentration of smoke above the road at an altitude of 1 m, t = 1500 s for Case 3.
Fig. 15. The overall range of the smoke plume and the zones for which the average 30-minute smoke concentrations exceed the allowable values (the squared area).
L. Brzozowska / Science of the Total Environment 472 (2014) 901–911
computations for cases of the potential effects of road incidents that take place. The numeric analyses that were performed allow us to conclude that obtaining prognostic results of computations is possible in a time period that is shorter than 1 hour. This is the case for dispersion of smoke or other harmful substances emitted into the air in selected vertical or horizontal profiles (the duration of a simulation for 10,000 cells of a discretisation net and a time period of 3600 s with step = 1 s is about 45 minutes on standard PC). To further accelerate the computations, a coarser discretisation net should be used or the analysed area should be further limited. In the analysed case (200 × 200 × 32 cells of the discretisation net), computations of the velocity field took about 20 minutes. Once computed, the air velocity field may be used for further simulations of propagation of pollutants without having to redo the computations. Employing parallel computing and an efficient computing unit, while using a non-equidistance computation grid computing in all directions, could significantly reduce the required computation time. An important feature of the model is its integration with a GIS (Idrisi Andes), thus enabling both preparation of initial data on landforms and aerodynamic roughness (preprocessing), and performing analyses related to the influence of the effects of a road incident on the area of simulation (postprocessing).
References Argyropoulos CD, Sideris GM, Christolis MN, Nivolianitou Z, Markatos NC. Modelling pollutants dispersion and plume rise from large hydrocarbon tank fires in neutrally stratified atmosphere. Atmos Environ 2010;44:803–13. http://dx.doi.org/10.1016/ j.atmosenv.2009. 11.034. Beard A, Carvel R. The handbook of tunnel fire safetyIn: Telford Thomas, editor. ; 2005. Brzozowska L. Validation of a Lagrangian particle model. Atmos Environ 2013;70:218–26. http://dx.doi.org/10.1016/j.atmosenv.2013.01.015. Caliendo C, Ciambelli P, De Guglielmo ML, Meo MG, Russo P. Numerical simulation of different HGV fire scenarios in curved bi-directional road tunnels and safety evaluation. Tunn Undergr Space Technol 2012;31:33–50. http://dx.doi.org/10.1016/j.tust.2012. 0.004. Clarke AG, Robertson LA, Hamilton RS, Gorbunov B. A Lagrangian model of the evolution of the particulate size distribution of vehicular emissions. Sci Total Environ 2004;334–335:197–206. http://dx.doi.org/10.1016/j.scitotenv.2004.04.038. Colella F, Rein G, Verda V, Borchiellini R, Carvel R, Steinhaus T, et al. Design of tunnel ventilations systems for fire emergencies using multiscale modelling. Fourth International Symposium on Tunnel Safety and Security. Germany: Frankfurt am Main; 2010. [March 17-19]. Ehrhard J, Khatib IA, Winkler C, Kunz R, Moussiopoulos N, Ernst G. The microscale model MIMO: development and assessment. J Wind Eng Ind Aerodyn 2000;85:163–76. http://dx.doi.org/10.1016/S0167-6105(99)00137-3. Fisher BEA, Metcalfe E, Vince I, Yates A. Modelling plume rise and dispersion from pool fires. Atmos Environ 2001;35:2101–10. http://dx.doi.org/10.1016/S1352-2310(00) 00495-7. Grašič B, Mlakar P, Zlata Božnar M. Method for validation of Lagrangian particle air pollution dispersion model based on experimental field data set from complex terrain. In: Nejadkoorki Farhad, editor. Advanced Air Pollution 2011; 2011. p. 535–56. http://dx.doi.org/10.5772/17286. [Chapter 27]. Homicz GF. Three-dimensional wind field modeling: a review. Sandia National Laboratories, Albuquerque, SAND Report; 20022002–597 [Source:http://prod.sandia.gov/techlib/ access-control.cgi/2002/022597.pdf]. Hu LH, Peng W, Huo R. Critical wind velocity for arresting upwind gas and smoke dispersion induced by bear-wall fire in road tunnel. J Hazard Mater 2008;150: 68–75. http://dx.doi.org/10.1016/j.jhazmat.2007.04.094. Hu LH, Huo R, Yang D. Large eddy simulation of fire-induced buoyancy driven plume dispersion in an urban street canyon under perpendicular wind flow. J Hazard Mater 2009;166:394–406. http://dx.doi.org/10.1016/j.jhazmat.2008.11.105. Hu LH, Xua Y, Zhub W, Wua L, Tanga F, LuaLarge KH. Large eddy simulation of pollutant gas dispersion with buoyancy ejected from building into an urban street canyon. J Hazard Mater 2011;192:940–8. http://dx.doi.org/10.1016/j.jhazmat. 2010.12.063.
911
Hua J, Wang J, Kumar K. Development of a hybrid field and zone model for fire smoke propagation simulation in buildings. Fire Safety J 2005;40:99–119. http://dx.doi.org/ 10.1016/j.firesaf.2004.09.005. Jain S, Kumar S, Kumar S, Sharma TP. Numerical simulation of fire in a tunnel: comparative study of CFAST and CFX predictions. Tunn Undergr Space Technol 2008;23:160–70. http://dx.doi.org/10.1016/j.tust.2007.04.004. Janicke L. A random walk model for turbulent diffusion. Reports on Environmental Physics, 1. ; 2000. p. 1–10. [Source:http://www.janicke.de/data/bzu/bzu-001-01.pdf]. Kim BG, Kim JS, Lee C. Lagrangian stochastic model for buoyant gas dispersion in a simple geometric chamber. J Loss Prev Process Ind 2009;22:995–1002. http://dx.doi.org/ 10.1016/j.jlp.2008.11.007. Lavrov A, Utkin AB, Vilar R, Fernandes A. Evaluation of smoke dispersion from forest fire plumes using lidar experiments and modelling. Int J Therm Sci 2006;45: 848–59. http://dx.doi.org/10.1016/j.ijthermalsci.2006.01.003. Liu XP, Niu JL, Kwok KCS. Analysis of concentration fluctuations in gas dispersion around high-rise building for different incident wind directions. J Hazard Mater 2011;192: 1623–32. http://dx.doi.org/10.1016/j.jhazmat.2011.06.090. Markatos NC, Christolis C, Argyropoulos C. Mathematical modeling of toxic pollutants dispersion from large tank fires and assessment of acute effects for fire fighters. Int J Heat Mass Transf 2009;52:4021–30. http://dx.doi.org/10.1016/j.ijheatmasstransfer. 2009.03.039. Mayer D, Reiczigel J, Rubel F. A Lagrangian particle model to predict the airborne spread of foot-and-mouth disease virus. Atmos Environ 2008;42:466–79. http://dx.doi.org/ 10.1016/j.atmosenv.2007.09.069. Mell WE, Manzello SL, Maranghides A, Butry D, Rehm RG. The wildland–urban interface fire problem — current approaches and research needs. Int J Wildland Fire 2010;19: 238–51. http://dx.doi.org/10.1071/WF07131. Miles S, Smithies N. CFD modelling of tunnel fires for UPTUN WP2. UPTUN and BRE Trust; 2006 [July]. Novozhilov V. Computational fluid dynamics modeling of compartment fires. Prog Energy Combust Sci 2001;27:611–66. http://dx.doi.org/10.1016/S0360-1285(01)00005-3. Pastor E, Zarate L, Planas E, Arnaldos J. Mathematical models and calculation systems for the study of wildland fire behaviour. Prog Energy Combust Sci 2003;29:139–53. http://dx.doi.org/10.1016/S0360-1285(03)00017-0. PIARC Committee on Road Tunnels (C5). Fire and smoke control in road tunnels, PIARC, 05.05.B. World Road Association; 1999. Qin TX, Guo YC, Chan CK, Lin WY. Numerical simulation of the spread of smoke in an atrium under fire scenario. Build Environ 2009;44:56–65. http://dx.doi.org/10.1016/ j.buildenv.2008.01.014. Seaman N. Meteorological modeling for air-quality assessments. Atmos Environ 2000;34: 2231–59. http://dx.doi.org/10.1016/S1352-2310(99)00466-5. Souto MJ, Souto JA, Pérez-Muñuzuri V, Casares JJ, Bermúdez JL. A comparison of operational Lagrangian particle and adaptive puff models for plume dispersion forecasting. Atmos Environ 2001;35:2349–60. http://dx.doi.org/10.1016/S1352-2310(00) 00537-9. Stohl A, Forster C, Frank A, Seibert P, Wotawa G. Technical note: the Lagrangian particle dispersion model FLEXPART version 6.2. Atmos Chem Phys 2005;5:2461–74. The Regulation of the Minister of the Environment. Natural Resources and Forestry of 28 April 1998 on the allowable concentrations limits of pollutants in the air. Official Gazette of 6 May 1998; 1998. Vautard R, Ciais P, Fisher R, Lowry D, Breon FM, Vogel F, et al. The dispersion of the Buncefield oil fire plume: an extreme accident without air quality consequences. Atmos Environ 2007;41:9506–17. http://dx.doi.org/10.1016/j.atmosenv. 2007.08.055. Viskanta R. Overview of some radiative transfer issues in simulation of unwanted fires. Int J Therm Sci 2008;47:1563–70. http://dx.doi.org/10.1016/j.ijthermalsci. 2008.01.008. Vitali L, Monforti F, Bellasio R, Bianconi R, Sachero V, Mosca S, et al. Validation of a Lagrangian dispersion model implementing different kernel methods for density reconstruction. Atmos Environ 2006;40:8020–33. http://dx.doi.org/10.1016/j.atmosenv. 2006.06.056. Wang J, Hiscox AL, Miller DR, Meyer TH, Sammis TW. A dynamic Lagrangian, field-scale model of dust dispersion from agriculture tilling operations. Trans ASABE 2008;51(5): 1763–74. Williams MD, Brown MJ, Singh B, Boswell D. QUIC-PLUME theory guide. Los Alamos National Laboratory; 2004 [LA-UR-04–0561(Source: http://lanl.gov/projects/quic/ open_files/QUICPLUME_theory.pdf)]. Wilson JD, Sawford BL. Review of Lagrangian stochastic models for trajectories in the turbulent atmosphere. Bound-Layer Meteorol 1996;78:191–210. http://dx.doi.org/ 10.1007/BF00122492. Xiaojun C. Simulation of temperature and smoke distribution of a tunnel fire based on modifications of multi-layer zone model. Tunn Undergr Space Technol 2008;23: 75–9. http://dx.doi.org/10.1016/j.tust.2006.10.005. Yang J, He HS, Gustafson EJ. A hierarchical fire frequency model to simulate temporal patterns of fire regimes in LANDIS. Ecol Model 2004;180:119–33. http://dx.doi.org/ 10.1016/j.ecolmodel.2004.03.017. Yao J, Fan WC, Kohyu S, Daisuke K. Verification and application of field–zone–network model in building fire. Fire Safety J 1999;33:35–44.
Contents lists available at ScienceDirect
Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv
Modelling the propagation of smoke from a tanker fire in a built-up area Lucyna Brzozowska University of Bielsko-Biala, Faculty of Management and Computer Science, Willowa 2, 43-309 Bielsko-Biała, Poland
H I G H L I G H T S • • • • •
The application of a custom-developed model of pollutants dispersion is presented. The application of model is connected with safety in road transport. Numerical simulations are performed for a case of smoke emission from a tanker fire. The results of simulation of smoke propagation in a built-up area are presented. The GIS is used to provide data for pre- and post-processing purposes.
a r t i c l e
i n f o
Article history: Received 2 September 2013 Received in revised form 13 November 2013 Accepted 25 November 2013 Available online 15 December 2013 Keywords: Numerical modelling Safety in road transport Environmental protection Tanker fire Lagrangian particle model Diagnostic wind field model
a b s t r a c t The paper presents the application of a Lagrangian particle model to problems connected with safety in road transport. Numerical simulations were performed for a hypothetical case of smoke emission from a tanker fire in a built-up area. Propagation of smoke was analysed for three wind directions. A diagnostic model was used to determine the air velocity field, whereas the dispersion of pollutants was analysed by means of a Lagrangian particle model (Brzozowska, 2013). The Idrisi Andes geographic information system was used to provide data on landforms and on their aerodynamic roughness. The presented results of computations and their analysis exemplify a possible application of the Lagrangian particle model: evaluation of mean (averaged over time) concentrations of pollutants and their distribution in the considered area (especially important due to the protection of people, animals and plants) and simulation of the propagation of harmful compounds in time as well as performing computations for cases of the potential effects of road incidents. © 2013 Elsevier B.V. All rights reserved.
1. Introduction Modelling the dispersion of pollutants belongs to a group of fundamental problems that are connected with evaluating the impact transport has on the environment. Although evaluation for longer periods of time and on larger scales (meso- and macro-) is relatively easy, in individual road accidents assessment of their effects proves to be difficult. This includes incidents resulting in the sudden release of pollutants. A tanker fire in a tunnel is a particularly important problem that has been discussed in a number of scientific papers (Beard and Carvel, 2005; Caliendo et al., 2012; Colella et al., 2010; Hu et al., 2008; Miles and Smithies, 2006; Xiaojun, 2008; PIARC, 1999). According to PIARC (1999), in this case models of the CFD class are predominantly used, as in: Caliendo et al. (2012), Beard and Carvel (2005), Hu et al. (2008), Miles and Smithies (2006), and Xiaojun, (2008); less often zone models (Hua et al., 2005; Jain et al., 2008; Xiaojun, 2008; Yao et al., 1999) and one-dimensional models play an auxiliary role (Colella et al., 2010).
E-mail address: [email protected]. 0048-9697/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.scitotenv.2013.11.130
It should be noted that road accidents and vehicle fires, including vehicles that transport fuel, do not only take place in tunnels, but also elsewhere. Fires occurring in close proximity to buildings are especially dangerous — this is the case that this article aims to analyse. Basically, there are two types of models that are used to model dispersion of smoke originating from a fire: microscale ones — such as the aforementioned CFD models (Novozhilov, 2001; Viskanta, 2008), which are predominantly used in problems related to fires in tunnels and closed spaces (e.g. a garage or an atrium (Qin et al., 2009)), and also in other cases when phenomena that spatially occur in the microscale (such as the fire of a fuel tank (Argyropoulos et al., 2010; Markatos et al., 2009)) must be taken into account; and mesoscale as well as macroscale ones, which are used most often to analyse forest fires (Lavrov et al., 2006; Mell et al., 2010). The following types may be distinguished in the group of mesoscale and macroscale models: empirical and theoretical (a comparison of models for the case of modelling dispersion of a smoke plume originating from a fire is presented in Fisher et al. (2001)), statistical (Yang et al., 2004) and numerical models in Euler (Vautard et al., 2007) or Lagrange (Stohl et al., 2005) coordinates. Another division of models of fires is possible: into probabilistic models (event trees, statistical models, Monte Carlo simulations)
902
L. Brzozowska / Science of the Total Environment 472 (2014) 901–911
Fig. 1. Example of approximation of land surface with buildings by using the continuous function s (Brzozowska, 2013).
and deterministic models (zone models, discrete ones (field models, CFD)). A taxonomy of models used in simulations of forest fires and their analysis can also be found in Pastor et al. (2003). Modelling propagation of smoke originating from fires in urban canyons and in close proximity to buildings is a problem that is rather seldom discussed. Nevertheless, there are papers (Hu et al., 2009, 2011; Liu et al., 2011) worth mentioning in which the problem of smoke propagation from a building on fire into the street is analysed
by using either (Hu et al., 2009, 2011) LES or experimental data (Liu et al., 2011) for computer simulation. Geographic Information Systems (GIS) are used in modelling the dispersion of pollutants in the mesoscale and macroscale (Pastor et al., 2003). They may prove useful in preprocessing or postprocessing. The current paper uses a Lagrangian numerical model combined with a diagnostic model of the air velocity field (Brzozowska, 2013). Problems related to transfer of heat were neglected in modelling the dispersion of smoke caused by an accident and the ignition of a tanker on a
Fig. 2. Map of the modelled area and the location of the accident (marked with a cross).
L. Brzozowska / Science of the Total Environment 472 (2014) 901–911
903
2. Numerical models
measurement data are interpolated at a given altitude above the ground; and, next, the initial vertical profile of air velocity is determined at node points of the discretisation net. Brzozowska (2013) contains a basic description of the model.
2.1. Diagnostic model of an air velocity field
2.2. Lagrangian model of dispersion of pollutants
Modelling the phenomenon of dispersion in the atmosphere is directly linked with the need to know the air velocity field. Prognostic models may be used for this purpose. These use complex physical phenomena while requiring huge computing power. Simpler diagnostic models (Ehrhard et al., 2000; Seaman, 2000) are also used. This paper employs a diagnostic model of the air velocity field for conditions involving complex topography. A variational algorithm (Homicz, 2002) was used to satisfy the criterion of mass balance. An equation of the elliptic type was formulated and then solved by using the method of conformal maps and finite differences for a nonequidistant discretisation net. The task was reduced to the problem of solving a system of algebraic linear equations. The iterative conjugate gradient method was subsequently used to solve it. The initial velocity field was determined by adjusting the air velocity field to the measurement data. The method that was used comprises two stages: first,
Particle models, besides their usefulness in analysing problems on larger spatial scales (macro and meso), may be used to predict concentrations of pollutants around areas of direct emission (roads, car parkings), as well as to assess the influence of proposed modifications of road infrastructure on the environment. These models also continually evaluate the impact of existing emission sources (Brzozowska, 2013; Clarke et al., 2004; Grašič et al., 2011; Stohl et al., 2005; Souto et al., 2001; Wilson and Sawford, 1996; Wang et al., 2008). Their role may also be significant in predicting and evaluating the propagation of pollutants emitted in sudden releases Kim et al. (2009) or in malfunctions occurring in road accidents. Location r of a particle representing a conventional mass of pollutants in space is determined by its velocity vector, which is the sum of the mean air velocity u and the fluctuation velocity u′ : u ¼ u þ u′ . The fluctuation velocity can be described with the Monte Carlo method
road. Idrisi Andes GIS (geographic information system) was used to provide data on buildings, landforms and their aerodynamic roughness.
Fig. 3. The modelled area: a) landforms, b) assumed directions of the wind vector and location of the source of smoke emission.
904
L. Brzozowska / Science of the Total Environment 472 (2014) 901–911
in which the behaviour of a particle fulfils one possibility out of an infinite set. Both the location r of a particle in space and its fluctuation velocity u′ are assumed to be driven by the Markov process, which is discrete in time (Mayer et al., 2008; Souto et al., 2001): h i ′ rnþ1 ¼ rn þ τ uðrn Þ þ unþ1 ;
ð1Þ
′
′
unþ1 ¼ ψðr n Þ un þ ω;
ð2Þ
where rn = r(tn), rn + 1 = r(tn + 1), un ¼ uðt n Þ, un′ = u′(tn), un′ + 1 = u′(tn + 1), tn + 1 = tn + τ, τ is the time step, ψ is the matrix defining the autoregressive summand ψ ⋅ un′ of the fluctuation velocity u′, and ω is the vector defining the partially random summand of the fluctuation velocity.
Fig. 4. Air velocity field in the modelled area at the 3rd layer of the discretisation net (at an altitude of about 2 m) for: Case 1 — a), Case 2 — b), Case 3 — c) of the initial air velocity vector and for the fragment of the area around the emission source.
L. Brzozowska / Science of the Total Environment 472 (2014) 901–911
905
the velocity field), the concentration of a pollutant in volume V ið1Þ ;ið2Þ ;ið3Þ is computed from the formula: −1
C ið1Þ ;ið2Þ ;ið3Þ ðt Þ ¼ V ið1Þ ;ið2Þ ;ið3Þ
N X
mp δp ;
ð3Þ
p¼1
where mp is the mass of particle number p, δp equals 1 for rp ∈V ið1Þ ;ið2Þ ;ið3Þ and 0 otherwise, and N denotes the total number of particles in the considered area. 3. Simulation of smoke propagation Fig. 5. Change of emission intensity in time.
Determining the matrix ψ and calculating its elements by using Cholesky decomposition (Janicke, 2000) was performed according to the procedure proposed in Mayer et al. (2008). A model of the motion of particles also requires a suitable algorithm that takes into account the influence of reflection and permeation at the upper boundary of the mixing layer and reflection on the land surface, as analysed in detail in Brzozowska (2013). Both for the diagnostic model and for the Lagrangian particle model, the terrain (including buildings) was assumed to be represented by an approximating function s = s(x(1)x(2)) (Brzozowska, 2013). This approach distinguishes the model that is used from the other models of this type, e.g. the QUIC-Plume model as described in Williams et al. (2004). Fig. 1 shows the way the surface is approximated along with the buildings. By applying function s, which can describe the domain with the buildings inside it, the presented model allows to model the dispersion of pollutants on an urban scale. The model can be especially useful in cases when it is necessary to obtain calculation results in a short period of time. Such urgent calculation results might be necessary when modelling the sudden release of dangerous gases in car accidents or in a fire. The dispersion model was described in detail in Brzozowska (2013). In Lagrangian particle model's concentration of a pollutant is computed by counting particles in isolated volumes of the atmosphere or by using so-called smoothing functions (Vitali et al., 2006). With the method of counting particles in isolated volumes (the volumes may be determined by the discretisation net used to model
Predicting the impacts of emissions for both hypothetical and real sources in a given area requires each time the selection of an appropriate model, whose accuracy has been determined under comparable conditions. For example, the location of an emissions source in a street canyon, or the application of the model used in this paper for development conditions in a city centre area with close high-rise buildings forming street canyons, is not allowable without a prior validation process. The analysed case is the ignition of a fuel-filled tanker due to a road accident on a bypass in Bielsko-Biała (Fig. 2), which is a medium-sized city (about 170,000 inhabitants) situated in southern Poland, in the Silesian foothills. The city features complex landforms and varied buildings. The considered area has low- and medium-sized building heights. The location of potential emission has been situated outside the immediate developed area in an open space comprising a sharp road curve (a route turn angle of over 120 degrees), constituting a ringroad. The road is a dual carriageway with two lanes for each direction. The closest buildings are located at a distance greater than 100 m from the adopted source location. At the same time, this location of the emission source relative to the area, for which calculations will be carried out, corresponds to the conditions for which the dispersion model has been validated. On these grounds it could be expected that the average concentration levels to be determined will be calculated with an accuracy comparable to that obtained in the validation process presented in another study (Brzozowska, 2013). In addition, the mapping of the surfaces limiting the air flow and smoke dissipation employed in the model using a single continuous function additionally restricts its application to areas of less dense urban development. These conditions are met by the area under examination, which comprises the outer areas of the city, through which the ring-road part under consideration passes.
Fig. 6. Change in time of the concentration of pollutants along a road with a distance from the source for Case 3.
906
L. Brzozowska / Science of the Total Environment 472 (2014) 901–911
Fig. 7. Field of a 5-minute mean concentration for Case 1: a) in the entire simulation area for t = 1500 s and in the marked fragment of the area for: b) t = 1500 s, c) t = 2400 s, d) t = 3600 s.
The total area of the analysed region was 9 km2, of which the builtup area constituted 17%. 3.1. Determining the air velocity field Three variants of the initial air velocity field were considered in the analysed case of smoke dispersion. In the first variant the air flow direction was assumed to be α = 140°, in the second variant it was α = 219° and in the third variant it was α = 180°, which was relative to the horizontal axis (x(1)); in all variants the wind speed was u = 2.06 m/s. A terrain map from GIS (Idrisi Andes) was imported into the model along with s(x,y) for buildings and aerodynamic roughness coefficients of the terrain. Fig. 3 shows the modelled area: landforms and both wind directions. Sensors are assumed to be placed at 9 points of the area (at an altitude of 8 m above the terrain), thus giving the initial profile of air velocity at the ends of the edges, midpoints of the edges and the centre of the area. The nodes of the discretisation net were assumed to be nonequidistant in the vertical direction (32 nodes up to an altitude of about 630 m) — according to the Chebyshev polynomial, whereas for the horizontal directions equidistant nodes were used with a 15 m step (200 nodes). In total, the analysed area was divided into 1,280,000 cells of different heights. The air velocity field in the x(1)x(2) plane at the altitude of about 2 m (the 3rd layer) used in computation above the ground is presented in Fig. 4. This altitude is variable depending on the landforms due to the
non-equidistant discretisation net that is used and varies as indicated with a grayscale from 280 m to 400 m. Wind speed was calculated after assuming atmospheric conditions corresponding to the neutral equilibrium of the atmosphere. 3.2. Simulation of the sudden release of smoke particles When modelling the phenomenon of the sudden release of particles, it was assumed that it was caused by the fire of a fuel-filled tanker. According to the Eureka HGV fire test, the peak power of the fire of a fuel-filled tanker reaches the range of 100, 120 MW — but only for a short period of time (PIARC, 1999). Smoke emission intensity is thereby estimated at 100–300 m3/s at a temperature of 300 °C (PIARC, 1999). This analysis was done with an assumed value of 200 m3/s, which for a smoke density of 300 mg/m3 gives a maximum value of smoke emission intensity that is equal to 60 g/s. Propagation of smoke is assumed to last 60 minutes, reaching its maximum after 1200 s and progressing in accordance with the curve as shown in Fig. 5. The subsequent figure presents the change of concentration in time at an altitude of 1 m above the road surface at different distances from the emission source for Case 3 (Fig. 6). Figs. 7–9 present momentary means over 5-minute periods of concentrations of pollutants in the entire simulation area for a simulation duration of 1500 s and in an area where x(1)∈b 2000;3000 N and x(2)∈b1000;2000N for selected moments. These concentrations occur at 1.5 m above the land surface.
L. Brzozowska / Science of the Total Environment 472 (2014) 901–911
907
Fig. 8. Field of a 5-minute mean concentration for Case 2: a) in the entire simulation area for t = 1500 s and in the marked fragment of the area for: b) t = 1500 s, c) t = 2400 s, d) t = 3600 s.
The width of the smoke cloud in which 5-minute mean concentrations exceeded 2 mg/m3 at the time of the most intense smoke emission (t = 1500 s) was 160 m, whereas an hour after the fire had started it was 50 m. Fig. 10. shows concentrations of pollutants in a selected crosssection at a distance of 200 m from the emission source for different moments of the simulation for Case 3. It is important, in terms of the point of view of safety, to examine the way in which a fire influences its surroundings. An example of plume propagation in selected cross-sections that are parallel to axes x(2) at time t = 1500 s for the analysed cases are shown in Figs. 11–13. The diagrams allow for the land topography and development. The built-up area covered by the smoke cloud in cases where the wind direction was α = 140° (Case 1), α = 219° (Case 2), and α = m2, 31,050 m2, and 13,730 m2, respectively, and 1 hour after the fire had started it was 0 for Case 1, 1350 m2 for Case 2 and 225 m2 for Case 3. In the first case no concentrations exceeding 5 mg/m3 were found in the built-up area at the time of peak emission, in the second case such concentrations occurred in areas of 11,900 m2 in total, whereas in the third case they occurred in areas of 3800 m2. Also, concentrations with values N 10 mg/m3 in the built-up area were registered in the second case, in areas of 4950 m2 in total and in areas of about 900 m2 in Case 3.
An hour into the simulation (fire) the determined maxima of concentration for the built-up area were 1 mg/m3 in Case 1, 9 mg/m3 in Case 2 and 3 mg/m3 in Case 3. As can be seen, the largest number of buildings is under the smoke cloud when the wind direction is α = 219°, i.e. northwest wind. The smoke cloud covers a smaller total built-up area in Case 3; however, the road itself was not taken into account as the built-up area in the computations. On the other hand, in the first case mostly forest areas, orchards and scrubland were exposed to smoke concentrations more then 2 mg/m3. In the third case the axis of the smoke plume coincides with the centre line of the road on which the accident happened. Fig. 14 shows a map of the smoke concentration for Case 3 at an altitude of 1 m above the land surface for t = 1500°s. In this area concentrations with values over 2 mg/m3 were found at distances of more than 1 km from the emission source, both at the fire's peak intensity and at the moment of 1 hour into the simulation. Higher concentrations, with values above 5 mg/m3 and 10 mg/m3, occur respectively in 800 m and 600 m strips at time t = 1500 s, and 700 m and 550 m at time t = 3600 s. Modelling of the propagation of the smoke plume formed as a result of the fire enables the identification of the areas, where the time of exposure to various chemicals contained in the smoke can jeopardize the human health. The currently applicable standards of air quality,
908
L. Brzozowska / Science of the Total Environment 472 (2014) 901–911
Fig. 9. Field of 5-minute mean concentration for Case 3: a) in the entire simulation area for t = 1500 s and in the marked fragment of the area for: b) t = 1500 s, c) t = 2400 s, d) t = 3600 s.
depending on the exposure time, define the relevant allowable levels for annual and 24 hours' mean values. In the case of intermittent emissions, limits defined for shorter averaging times are practically applied for the assessment of the level of risk to human health. In further calculations, a single indicator defined for total suspended particulate matter was decided to be adopted as the threshold value defining the allowable exposure level. Thus, it is not necessary to take the assumption concerning chemical composition. In such cases, a 30-minute averaging period has been used for calculation purposes in Poland, with taking the value of 350 mg/m3 for total suspended particles (TSP) as the allowable average 30-minute concentration limit (Official Gazette, 1998). The use of this indicator enables the identification of the zones of the area under analysis, in which the calculated average concentrations exceed the adopted limit. Fig. 15 shows the overall range of the smoke plume for each of the considered cases, and the identified areas where the concentration exceeds the threshold value. The presented data applies to the values calculated for the altitude of 1.5 m above ground level. Taking into account the type of development, the number of inhabitants in the area under examination (about 52,760 people) and in individual zones can be estimated based on the statistical data for the city. Depending on the wind direction, the maximum range of impact the smoke plume covers the examined area to a varying extent. In Case 1, the number of people who will potentially fall into the smoke plume is approximately 6.7% of the total population of the examined area; in Case 2 it is 23.5%, and in Case 3, 14.8%.
In turn, the number of people in zones, where the adopted allowable limit values are exceeded is, respectively, 2.5%, 5% and 3% of the area inhabitants for the analysed cases. Based on the performed calculations, the worst case in terms of the effects of the potential impact of the smoke plume is considered to be Case 2. For this wind direction, the number of people exposed to smoke concentrations above the threshold value is approximately 2640, while in the most favourable instance — Case 1, it is about 1320 people. And for Case 3, where the smoke plume axis largely coincides with the road axis, the number of people in the zone with the exceeded threshold value is about 1580. These differences are due to not only the density of development, but also its type. The smoke plume in Case 2 covers predominantly multi-family residential areas, while the development on the right side of the road (Case 1) is made up exclusively by single-family houses in a detached and row housing. 4. Summary The presented results of computations and their analysis exemplify a possible application of the Lagrangian particle model. In addition to standard simulations, whose subjects of evaluation are mean (averaged over time) concentrations of pollutants and their distribution in the considered area and especially taking into account areas that are important due to the protection of people, animals and plants, such models also lend themselves to other tasks. One of these is the simulation of the propagation of harmful compounds in time as well as performing
L. Brzozowska / Science of the Total Environment 472 (2014) 901–911
Fig. 10. Profile of concentration in cross-section for x(1) = 2640 m, for selected moments in Case 3.
Fig. 11. Cross-section parallel to axes x(2) at time t = 1500 s for Case 1.
Fig. 12. Cross-section parallel to axes x(2) at time t = 1500 s for Case 2.
909
910
L. Brzozowska / Science of the Total Environment 472 (2014) 901–911
Fig. 13. Cross-section parallel to the axes x(2) at time t = 1500 s for Case 3.
Fig. 14. Concentration of smoke above the road at an altitude of 1 m, t = 1500 s for Case 3.
Fig. 15. The overall range of the smoke plume and the zones for which the average 30-minute smoke concentrations exceed the allowable values (the squared area).
L. Brzozowska / Science of the Total Environment 472 (2014) 901–911
computations for cases of the potential effects of road incidents that take place. The numeric analyses that were performed allow us to conclude that obtaining prognostic results of computations is possible in a time period that is shorter than 1 hour. This is the case for dispersion of smoke or other harmful substances emitted into the air in selected vertical or horizontal profiles (the duration of a simulation for 10,000 cells of a discretisation net and a time period of 3600 s with step = 1 s is about 45 minutes on standard PC). To further accelerate the computations, a coarser discretisation net should be used or the analysed area should be further limited. In the analysed case (200 × 200 × 32 cells of the discretisation net), computations of the velocity field took about 20 minutes. Once computed, the air velocity field may be used for further simulations of propagation of pollutants without having to redo the computations. Employing parallel computing and an efficient computing unit, while using a non-equidistance computation grid computing in all directions, could significantly reduce the required computation time. An important feature of the model is its integration with a GIS (Idrisi Andes), thus enabling both preparation of initial data on landforms and aerodynamic roughness (preprocessing), and performing analyses related to the influence of the effects of a road incident on the area of simulation (postprocessing).
References Argyropoulos CD, Sideris GM, Christolis MN, Nivolianitou Z, Markatos NC. Modelling pollutants dispersion and plume rise from large hydrocarbon tank fires in neutrally stratified atmosphere. Atmos Environ 2010;44:803–13. http://dx.doi.org/10.1016/ j.atmosenv.2009. 11.034. Beard A, Carvel R. The handbook of tunnel fire safetyIn: Telford Thomas, editor. ; 2005. Brzozowska L. Validation of a Lagrangian particle model. Atmos Environ 2013;70:218–26. http://dx.doi.org/10.1016/j.atmosenv.2013.01.015. Caliendo C, Ciambelli P, De Guglielmo ML, Meo MG, Russo P. Numerical simulation of different HGV fire scenarios in curved bi-directional road tunnels and safety evaluation. Tunn Undergr Space Technol 2012;31:33–50. http://dx.doi.org/10.1016/j.tust.2012. 0.004. Clarke AG, Robertson LA, Hamilton RS, Gorbunov B. A Lagrangian model of the evolution of the particulate size distribution of vehicular emissions. Sci Total Environ 2004;334–335:197–206. http://dx.doi.org/10.1016/j.scitotenv.2004.04.038. Colella F, Rein G, Verda V, Borchiellini R, Carvel R, Steinhaus T, et al. Design of tunnel ventilations systems for fire emergencies using multiscale modelling. Fourth International Symposium on Tunnel Safety and Security. Germany: Frankfurt am Main; 2010. [March 17-19]. Ehrhard J, Khatib IA, Winkler C, Kunz R, Moussiopoulos N, Ernst G. The microscale model MIMO: development and assessment. J Wind Eng Ind Aerodyn 2000;85:163–76. http://dx.doi.org/10.1016/S0167-6105(99)00137-3. Fisher BEA, Metcalfe E, Vince I, Yates A. Modelling plume rise and dispersion from pool fires. Atmos Environ 2001;35:2101–10. http://dx.doi.org/10.1016/S1352-2310(00) 00495-7. Grašič B, Mlakar P, Zlata Božnar M. Method for validation of Lagrangian particle air pollution dispersion model based on experimental field data set from complex terrain. In: Nejadkoorki Farhad, editor. Advanced Air Pollution 2011; 2011. p. 535–56. http://dx.doi.org/10.5772/17286. [Chapter 27]. Homicz GF. Three-dimensional wind field modeling: a review. Sandia National Laboratories, Albuquerque, SAND Report; 20022002–597 [Source:http://prod.sandia.gov/techlib/ access-control.cgi/2002/022597.pdf]. Hu LH, Peng W, Huo R. Critical wind velocity for arresting upwind gas and smoke dispersion induced by bear-wall fire in road tunnel. J Hazard Mater 2008;150: 68–75. http://dx.doi.org/10.1016/j.jhazmat.2007.04.094. Hu LH, Huo R, Yang D. Large eddy simulation of fire-induced buoyancy driven plume dispersion in an urban street canyon under perpendicular wind flow. J Hazard Mater 2009;166:394–406. http://dx.doi.org/10.1016/j.jhazmat.2008.11.105. Hu LH, Xua Y, Zhub W, Wua L, Tanga F, LuaLarge KH. Large eddy simulation of pollutant gas dispersion with buoyancy ejected from building into an urban street canyon. J Hazard Mater 2011;192:940–8. http://dx.doi.org/10.1016/j.jhazmat. 2010.12.063.
911
Hua J, Wang J, Kumar K. Development of a hybrid field and zone model for fire smoke propagation simulation in buildings. Fire Safety J 2005;40:99–119. http://dx.doi.org/ 10.1016/j.firesaf.2004.09.005. Jain S, Kumar S, Kumar S, Sharma TP. Numerical simulation of fire in a tunnel: comparative study of CFAST and CFX predictions. Tunn Undergr Space Technol 2008;23:160–70. http://dx.doi.org/10.1016/j.tust.2007.04.004. Janicke L. A random walk model for turbulent diffusion. Reports on Environmental Physics, 1. ; 2000. p. 1–10. [Source:http://www.janicke.de/data/bzu/bzu-001-01.pdf]. Kim BG, Kim JS, Lee C. Lagrangian stochastic model for buoyant gas dispersion in a simple geometric chamber. J Loss Prev Process Ind 2009;22:995–1002. http://dx.doi.org/ 10.1016/j.jlp.2008.11.007. Lavrov A, Utkin AB, Vilar R, Fernandes A. Evaluation of smoke dispersion from forest fire plumes using lidar experiments and modelling. Int J Therm Sci 2006;45: 848–59. http://dx.doi.org/10.1016/j.ijthermalsci.2006.01.003. Liu XP, Niu JL, Kwok KCS. Analysis of concentration fluctuations in gas dispersion around high-rise building for different incident wind directions. J Hazard Mater 2011;192: 1623–32. http://dx.doi.org/10.1016/j.jhazmat.2011.06.090. Markatos NC, Christolis C, Argyropoulos C. Mathematical modeling of toxic pollutants dispersion from large tank fires and assessment of acute effects for fire fighters. Int J Heat Mass Transf 2009;52:4021–30. http://dx.doi.org/10.1016/j.ijheatmasstransfer. 2009.03.039. Mayer D, Reiczigel J, Rubel F. A Lagrangian particle model to predict the airborne spread of foot-and-mouth disease virus. Atmos Environ 2008;42:466–79. http://dx.doi.org/ 10.1016/j.atmosenv.2007.09.069. Mell WE, Manzello SL, Maranghides A, Butry D, Rehm RG. The wildland–urban interface fire problem — current approaches and research needs. Int J Wildland Fire 2010;19: 238–51. http://dx.doi.org/10.1071/WF07131. Miles S, Smithies N. CFD modelling of tunnel fires for UPTUN WP2. UPTUN and BRE Trust; 2006 [July]. Novozhilov V. Computational fluid dynamics modeling of compartment fires. Prog Energy Combust Sci 2001;27:611–66. http://dx.doi.org/10.1016/S0360-1285(01)00005-3. Pastor E, Zarate L, Planas E, Arnaldos J. Mathematical models and calculation systems for the study of wildland fire behaviour. Prog Energy Combust Sci 2003;29:139–53. http://dx.doi.org/10.1016/S0360-1285(03)00017-0. PIARC Committee on Road Tunnels (C5). Fire and smoke control in road tunnels, PIARC, 05.05.B. World Road Association; 1999. Qin TX, Guo YC, Chan CK, Lin WY. Numerical simulation of the spread of smoke in an atrium under fire scenario. Build Environ 2009;44:56–65. http://dx.doi.org/10.1016/ j.buildenv.2008.01.014. Seaman N. Meteorological modeling for air-quality assessments. Atmos Environ 2000;34: 2231–59. http://dx.doi.org/10.1016/S1352-2310(99)00466-5. Souto MJ, Souto JA, Pérez-Muñuzuri V, Casares JJ, Bermúdez JL. A comparison of operational Lagrangian particle and adaptive puff models for plume dispersion forecasting. Atmos Environ 2001;35:2349–60. http://dx.doi.org/10.1016/S1352-2310(00) 00537-9. Stohl A, Forster C, Frank A, Seibert P, Wotawa G. Technical note: the Lagrangian particle dispersion model FLEXPART version 6.2. Atmos Chem Phys 2005;5:2461–74. The Regulation of the Minister of the Environment. Natural Resources and Forestry of 28 April 1998 on the allowable concentrations limits of pollutants in the air. Official Gazette of 6 May 1998; 1998. Vautard R, Ciais P, Fisher R, Lowry D, Breon FM, Vogel F, et al. The dispersion of the Buncefield oil fire plume: an extreme accident without air quality consequences. Atmos Environ 2007;41:9506–17. http://dx.doi.org/10.1016/j.atmosenv. 2007.08.055. Viskanta R. Overview of some radiative transfer issues in simulation of unwanted fires. Int J Therm Sci 2008;47:1563–70. http://dx.doi.org/10.1016/j.ijthermalsci. 2008.01.008. Vitali L, Monforti F, Bellasio R, Bianconi R, Sachero V, Mosca S, et al. Validation of a Lagrangian dispersion model implementing different kernel methods for density reconstruction. Atmos Environ 2006;40:8020–33. http://dx.doi.org/10.1016/j.atmosenv. 2006.06.056. Wang J, Hiscox AL, Miller DR, Meyer TH, Sammis TW. A dynamic Lagrangian, field-scale model of dust dispersion from agriculture tilling operations. Trans ASABE 2008;51(5): 1763–74. Williams MD, Brown MJ, Singh B, Boswell D. QUIC-PLUME theory guide. Los Alamos National Laboratory; 2004 [LA-UR-04–0561(Source: http://lanl.gov/projects/quic/ open_files/QUICPLUME_theory.pdf)]. Wilson JD, Sawford BL. Review of Lagrangian stochastic models for trajectories in the turbulent atmosphere. Bound-Layer Meteorol 1996;78:191–210. http://dx.doi.org/ 10.1007/BF00122492. Xiaojun C. Simulation of temperature and smoke distribution of a tunnel fire based on modifications of multi-layer zone model. Tunn Undergr Space Technol 2008;23: 75–9. http://dx.doi.org/10.1016/j.tust.2006.10.005. Yang J, He HS, Gustafson EJ. A hierarchical fire frequency model to simulate temporal patterns of fire regimes in LANDIS. Ecol Model 2004;180:119–33. http://dx.doi.org/ 10.1016/j.ecolmodel.2004.03.017. Yao J, Fan WC, Kohyu S, Daisuke K. Verification and application of field–zone–network model in building fire. Fire Safety J 1999;33:35–44.
