Microsoft Word - 6laurent.docx CHEMICAL ENGINEERING TRANSACTIONS VOL. 36, 2014 A publication of The Italian Association of Chemical Engineering www.aidic.it/cet Guest Editors: Valerio Cozzani, Eddy de Rademaeker Copyright © 2014, AIDIC Servizi S.r.l., ISBN 978-88-95608-27-3; ISSN 2283-9216 Cost Action Es1006. Evaluation, Improvement and Guidance of Local-Scale Emergency Prediction and Response Tools for Airborne Hazards in Built Environments: Ongoing Activities, Experiments and Recent Results Claudio Gariazzo*a, Bernd Leitlb, SilviaTrini Castellic, Kathrin Baumann- Stanzerd, Tamir G.Reisine, Fotios Barmpasf, Gianni Tinarellig, Claire M. Milliezh, Elisabetta Bemporada, Patrick Armandi a INAIL Research Center via Fontana Candida 1 Monteporzio Catone (RM) Italy. b Meteorological Institute University of Hamburg Bundesstrasse 55 D-20146 Hamburg Germany. c Istituto di Scienze dell’Atmosfera e del Clima Consiglio Nazionale delle Ricerche Corso Fiume 4 10133 Torino Italy. d Central Institute for Meteorology and Geodynamics Hohe Warte 38 1190 Wien Austria. e SOREQ NRC Yavne Israel. f Aristotle University of Thessaloniki Laboratory of Heat Transfer & Environmental Engineering Thessaloniki Greece. g ARIANET S.r.l. Via Gilino 9 20128 Milan Italy. h CEREA Teaching and Research Center in Atmospheric Environment Joint Laboratory Ecole des Ponts ParisTech - EDF R&D Universite Paris Est Electricite de France R&D 6 quai Watier 78400 Chatou Cedex France. I CEA, DAM, DIF, F-91297 Arpajon, France c.gariazzo@inail.it The COST Action ES1006 is mainly aimed at evaluating and improving the reliability of neighbourhood- scale emergency response tools on the basis of a comprehensive, concerted and harmonized cross- national approach. The main focus is the evaluation of the air dispersion models, when used in urban or industrial environments with complex building structures, and their integration in emergency response systems. In the frame of this research, three working groups are carrying out several activities such as: to inventory models, tools and methodologies currently applied in the context of emergency management and local-scale threat reduction; to categorize and characterize local-scale threat scenarios; to identify the main gaps, deficiencies and limitations in presently available knowledge and models and to determine the directions for the development of the next generation of models; to test and evaluate available models by model inter-comparison and by comparison against test data from qualified field and laboratory experiments. The ongoing activities in the three WGs and the latest available results will be presented. 1. Introduction COST Action ES1006 (2013) “Evaluation, improvement and guidance for the use of local-scale emergency prediction and response tools for airborne hazards in built up environments” aims at assuring the quality of the airborne hazardous material modelling tools used for emergency response and providing guidance for their use and improvement, within an harmonized European approach. The Action focuses on urban and industrial areas, and pays a particular attention to the specific needs of stakeholders and decision makers. Identifying gaps in knowledge related to local-scale emergency response modelling is one of the main aims, supported by the development of an evaluation strategy designed to consider requirements of airborne hazard modelling. In this context, not only the validity and accuracy of dispersion models is of concern, but also their demands regarding input data, their operational performance as well as their robustness considering the uncertainty of input data. The main tasks of the Action are summarized below: DOI: 10.3303/CET1436089 Please cite this article as: Gariazzo C., Leitl B., Trini Castelli S., Baumann-Stanzer K., Reisin T.G., Barmpas F., Tinarelli G., Milliez C.M., Armand P., Bemporad E., 2014, Cost action es1006. evaluation, improvement and guidance of local-scale emergency prediction and response tools for airborne hazards in built environments: ongoing activities, experiments and recent results, Chemical Engineering Transactions, 36, 529-534 DOI: 10.3303/CET1436089 529 • To elaborate a complete inventory of local threat scenarios and related modelling systems presently used, and to establish a scientific and methodical reference for local-scale airborne hazards modelling. Possible sources and release situations have to be characterized and categorized considering specific model requirements and assessing the fitness for purpose of different modelling approaches. • To setup a dedicated comprehensive inventory of models applicable to local-scale accidental releases. A complete and consistent European catalogue of tools and models is not yet available. • To test and evaluate available models by model inter-comparison and by comparison against test data from qualified field and laboratory experiments, extending the existing model evaluation and validation strategies towards task- and application-specific measures for accidental release scenarios. • To identify the main gaps, deficiencies and limitations in presently available knowledge and models and to determine the directions for the development of the next generation of models. • To address the integration of airborne hazards modelling tools in existing and/or evolving information systems for urban/industrial emergency management, considering both the output results of local-scale airborne hazards modelling and the possible quality improvement of the input information. • To classify existing test data with respect to completeness and usefulness for the present purpose. The uncertainty in the test data has to be assessed and possibly quantified. Identification of desirable test scenarios for which data may be collected during field and/or laboratory experiments in the future. 2. Scientific work plan methods and means The scientific work of the Action is carried out in three structured task-specific Working Groups (WGs). Working Group 1 - Threats, Models and Data Requirements - is characterizing and categorizing existing models as well as typical release scenarios. Main task is to evaluate, complete and uniformly document existing test data. Working Group 2 - Test, Evaluation and Further Development - is defining open and blind test scenarios, testing and assessing different modelling approaches and working on scientific strategies for improving the implementation of corresponding tools. In this context, the Action’s scientific interest is not to rank individual modelling approaches but to identify specific reasons for diverging model results and possible ways for improving modelling quality. The strengths and weaknesses of particular modelling approaches are identified, quantified, and documented in order to stimulate further improvement of model quality. A first version of a best practice manual for the application of neighbourhood-scale airborne hazards models will be compiled and released in order to immediately improve the quality of model results. Working Group 3 - Applicability, Implementation and Practical Guidance – is dealing with the practical constraints in the use of local-scale emergency response models. The specific needs of first responders and authorities in charge of neighbourhood-scale emergency response management have to be taken into account in order to successfully implement scientific improvements. From a clear user's point of view, the work covers tasks such as: the collection of requests and demands of the emergency-response experts for improving the practical applicability of the modelling systems; the provision of guidance regarding the suitability of different types of models and methodologies for specific problems at different stages of an incident or the identification, characterization, visualization and quantification of the uncertainties of emergency response modelling facilitating the proper interpretation by decision makers. 3. The ongoing activities and latest results 3.1 The Inventory of Available Datasets. A first database (Tsiouri and Trini Castelli, 2012) was elaborated in order to classify existing test data with respect to completeness and usefulness for the purpose of validating dispersion models specifically for emergency response systems. Since specific datasets suited for emergency response models are rare, datasets originally gathered in atmospheric dispersion models are mainly described in this document. For each dataset the possible limitations, related to their use when validating models in the frame of emergency response assessment, are discussed. A few features for a qualified dataset for model testing and demonstration of model performance in emergency response in built environments were identified, such as: being representative of an accidental release; addressing some important aspect or issue affecting the dispersion modelling reliability; availability of concurrent meteorological and concentration measured data of possible high quality; describing a variety of meteorological conditions. 3.2 The Inventory of Emergency Modelling Tools. A summary of the state-of-the-art of emergency response tools for airborne hazards from accidental/deliberate releases in complex urban and industrial areas was compiled (Tavares and Baumann-Stanzer, 2012) and a dedicated model inventory was established. A pilot version of the Model 530 Inventory Database Tool (MIDT) was prepared, with the intention to catalogue information of available emergency response models, tools and methodologies developed for local-scale airborne hazards and incidents scenarios. Available information are: computational approaches and models integrated, aspects of hazards and incident scenarios addressed, physical background, input data demands, computational demands and information on model application/use, verification or related performance measures. The MIDT was developed in the form of an Excel workbook comprising a set of eight information worksheets. It comprises specific information concerning emergency response computational tools and models used individually or incorporated in these tools, applicable to individual or multiple interdependent aspects of local-scale airborne hazards and incident events. The following elements are considered in the MIDT: Emergency Response Computational Tools Database (ERCTD); Meteorological Models Database (MMD); Source Term Models Database (STMD); Dispersion Models Database (DMD); Consequence Analysis Models Database (CMD); Risk analysis Models Database (RAMD). This inventory will allow for model- specific guidance regarding an efficient and reliable use of different model tools. 3.3 The Model Evaluation Procedures for Emergency Response Applications. A thorough review (Barmpas and Trini Castelli, 2013) of all recent developments in model evaluation procedures for the validation of dispersion models and that can potentially be applied in cases of accidental or deliberate releases of airborne hazards in urban areas was performed. In order to measure the quality of model results and to improve their implementation, a task-specific validation and application procedure was adopted. According to the latter, the necessary steps that should form the proposed evaluation modelling procedure is the following: 1) definition of the variables which should be compared; 2) definition of how the model is run and the results interpreted; 3) Processing of the experimental data; 4) Exploratory data analysis; 5) Metrics for Model Validation; 6) definition of the quality acceptance criteria. This procedure will be further developed and it will form the basis for the model evaluation exercises planned within the frame of Action. 3.4 The end-users and stakeholders questionnaires/interviews. Questionnaires dedicated to survey the use of Emergency Prediction and Response Tools (EPRT) by stakeholders and their needs and requirements related to the modelling suites was elaborated and distributed to end-users and stakeholders. Compiled questionnaires (61) were received from the main European Countries (17). The results presented here refer to Italy, from where the highest response rate was obtained (received/sent ratio = 25/61): however, according to the overall results, they are consistent with those obtained in other Countries. Table 1 lists the kind of Institutions who responded from Italy. Table 1: List of Italian institutions participating at the ES1006 questionnaires Institutions Received questionnaire Local authorities 1 Fire brigade 17 University 1 Consultants 4 National authority 1 industry 1 TOTAL 25 Questions were related to different aspects on the use of modelling tools in emergency planning and response, such as: type of model used, preferred data fed, operators involved, context and aims for modelling applications. The main results are summarized in Tables 2a,b,c. Table 2a: Summary of answers provided by the Italian stakeholders on the use of modelling tools in emergency. Multiple options selected avoid rate to sum up 100. Use of Emergency Tools rate (%) Model operator rate (%) Model used rate (%) emergency planning 88 risk analyst 75 Gaussian 60 training 32 tech. Emerg. manager 60 Eulerian 0 civil protection 36 technical consultant 10 Lagrangian 20 other 12 modelling expert 15 CFD 20 specialised technician 15 short-cut 70 HSEQ tech. manager 5 PUFF 10 531 Table 2b: Summary of answers provided by the Italian stakeholders on the use of modelling tools in emergency. Input data. Multiple options selected avoid rate to sum up 100. Meteorological data rate (%) Turbulence data rate (%) Emission data rate (%) local measurements 23 Pasquill classification 84 estimated 27 measurements in the nearest station 23 bulk ABL similarity theory 5 modelled 55 classified from max. impact ( D5, F2) 73 k-epsilon theory 11 measurements 23 meteorological simulations 9 statistical fluctuations 5 short-cut assess. 36 local climatological models 23 other 5 classified from extreme conditions 9 Emission characteristics rate (%) Emission characteristics rate (%) stationary 79 multiphase 21 42 53 53 47 16 non stationary 37 single-phase mechanical or thermal driven 16 light gases inert 32 heavy gases reactive 16 pure chemical mixture Table 2c: Summary of answers provided by the Italian stakeholders on the use of modelling tools in emergency. Output data and scenario and operator involved. Multiple options avoid rate to sum up 100. Preferred results rate (%) Main interest scenarios rate (%) Operator involved in ER rate (%) concentration values 44 traffic accident 73 plant emergency team 23 confidence intervals 4 industrial accident 100 plant emergency manager 27 hazardous distances 76 arsons 31 local authorities 32 hazardous area depicted on a map 100 ship accident 35 provincial or national authorities 50 other 0 terrorist attacks 35 other 14 natural fire 38 The questionnaires highlighted that EPRT are mainly used by risk analysts and technical emergency managers in the planning phase of industrial and traffic accidents using either Gaussian or simple shortcut parametric models driven by pre-classified meteorological conditions, simple Pasquill atmospheric stability classes and emissions provided by estimates, models or short cut assessments. Hazardous distances or areas are the preferred output. Advanced and more sophisticated models are rarely used as well as combinations with mesoscale meteorological models and consideration of buildings and structures. Questionnaires were integrated with interviews using open answers. Twenty-eight fire brigade operators and emergency responsible (20 Italian, 8 European countries) participated. The general opinion about the usefulness of EPRT ranges from good to essential in their application for emergency planning /management and environmental /health impact assessment. A number of different aspects were considered to limit the use of complex EPTR in emergency, such as: inaccuracy and limited reliability in input data; inaccuracy in model description of relevant phenomena at their spatial/temporal scales; lack of both trained personnel and standardizes operating procedures; model and training costs as well as the long response time. Source term were found another critical point. Therefore models results have to be checked with measurements or damage observations. Consequently, measurements and personal experience, supported by simple and fast modelling approaches, dominate the emergency management procedures. EPRT are not used for post-crisis analysis. Important aspects in using EPRT were: accuracy; availability of validated models in international or national guidelines; standardization of procedures for EPRT applications and rules for risk assessment; needs of training. Simple and robust systems, integrated in a web GIS systems are needed-. Based on these needs, it is important to document the limitations of different local-scale emergency response methodologies by assessing the actual uncertainty of model results. Consequently reference datasets for testing the model performance have to be made and used. 3.5 The Michelstadt Dataset. A first reference dataset for testing the models was processed, the ‘Michelstadt’ case (Fischer et al., 2010). It was designed to include potential in-homogeneities, characterising the neighbourhood-scale urban areas across Europe with the goal of providing observed data for the validation of local scale emergency response models. Data were gathered during a wind-tunnel flow and dispersion experiment, carried out at the Hamburg University, where a typical European urban site was reproduced. Figure 1a 532 shows the layout of urban structure built for this dataset. In the frame of the COST action two modelling phases were carried out. They consisted in numerically simulating two cases of continuous releases for different source locations (Figure1b), using two wind directions. a) b) Figure 1: Real layout of Michelstadt wind-tunnel experiment (a). Locations of continuous sources releases Sx (dot circles) and sampling points (dot squared) (b). In the first phase all the dynamical and concentration measurements were made available to the modellers (non blind test), whereas for the second phase, only one wind inflow point and the source characteristics were given (blind test). Sampling points were located within the urban structure to test and verify the spatial extension of the studied accident. The models involved in the modelling exercise range from a simple parametric approach to the large-eddy simulation CFD models. Different numerical models were involved, which were sorted in three categories: type 1 models not resolving the flow, type 2 models resolving the flow between the buildings with simplified equations and type 3 models resolving the flow between the buildings with full equations. Table 3 summarizes the involved models. Table 3: models involved in Michelstadt dataset Modelling approach Number of models Remarks Computational time Type 1 7 Gaussian – 2 with building parameterization 1 -5 min Type 2 5 Lagrangian 2 min – 5 hrs Type 3 10 CFD (8 RANS–3 LES; 1 RANS –Lagrangian) 2 hrs – 4 days The analyses and comparisons with the experimental results were made according to these three categories rather than specific models. In order to compare the different modelling approach, a dedicated post processing tool was developed in the frame of the Action. The focus is kept on the pollutant concentrations but the tool also allows detailed analyses of the flow field. Grouping the different models according to their categories, the tool produces scatter plots, profiles and 2D projections as well as statistical metrics. Figure 2 shows the scatter plots of simulated vs observed concentrations for each model approach including all results obtained from the available sources in the non blind test. Table 4 shows the values of two indexes of performance (FAC2, FB) obtained for each model approach as an overall evaluation. As expected, results show discrepancies among the model approaches and between model approaches and observations. In general, the better is the description of dispersion phenomena (as in model type 3 which resolves the flow with full equations) the greater are the performances. High discrepancies among the different approaches are also observed in terms of spatial characteristics of concentration maps (not shown), due to the different approaches in reconstructing flow and dispersion. Table 4: Indexes of performance obtained for the different model types in Michelstadt dataset non blind test. FAC2 (fractions of predictions within a factor 2 of observations). FB (Fractional Bias). FAC2 FB Modelling approach min max average min max Type 1 14 % 33 % 25 % -0.24 0.36 Type 2 25 % 46 % 40 % 0.32 0.7 Type 3 25 % 81 % 59 % -0.43 0.54 533 Figure mode Discre autho 3.6 Th WG1 toxics apply results usabil 4. Co The C hazar mana curren to-dat respo bench severa the m an acc wide r Refer Barmp ES COST Fische da atm (C Gariaz mo Tavar Ac 1s Tsiou Re ww ac Viane Co Tr a) ty e 2: Scatter pl ls (a), type 2 m epancies amo rs when obsta he catalogue is collecting, c s in populated emergency re s. This docum lity, the pros a onclusions COST Action rdous materia gement and p ntly in use. Th te inventory nse, a com hmarking loca al documents models compar cident in a typ range of mode rences pas F., Trini S1006 – Short T Action ES10 er R., Bastigk atasets for the mospheric bo CWE2010), Ch zzo C., Pellicc odeling appro res R., Bauma ction ES10 stYear/Drafting ri V., Trini Cas eport. CO ww.elizas.eu/i ccessed 21.10 ello C., Masch onsequences ransactions, 24 ype 1 lots of simulate models (b) and ong results pro acles and struc of Threats an characterizing areas. At the esponse dispe ment is expecte and cons as w ES1006 was als transform policy issues e major outco reviewing the prehensive d l-scale emerg and results h rison study ca pical European els. Castelli S., 2 t Term Scienti 006, 2013, ww keit I., Leitl B e validation of oundary layer. hapel Hill, Nort cioni A., Brag aches, Chemi ann-Stanzer K 006 – S g/stsm%20rich stelli S., 2012 OST Actio mages/STSM 0.2013 hio G., Alban through CFD 4, 1117-1122 ed vs measure d type 3 mode ovided by diff ctures are con nd Challenge g and documen e same time W ersion models ed to supplem well as challeng presented. Fl ation process have to be co omes expected e current mo database of gency respons here briefly pr arried out usin n urban struct 2013, The M ific Mission CO w.elizas.eu B.,Schatzmann f LES-models Proc. 5th Inte th Carolina, U atto P., 2012 ical Engineeri K., 2012, Eme Short Term hard%20tavar , Drafting inve on ES10 Ms/1stYear/stsm nese P., 201 Modeling and b) type red mean conc els (c). ferent modellin nsidered (Gari es and the Be nting typical a WG3 is prepa s in order to lo ment the user m ges and limita low and dispe ses during d onsidered whe d from the Act odelling tools experiments, se models. Th esented. Of p ng a special d ure. They form Model Evaluat OST-STSM-E n M., 2010, G simulating flo ernational Sym SA, TS 15-1. , Simulation o ng Transactio ergency Mode Scientific res%20scientif entory of refere 006 – m%20vasiliki% 1, Chlorine G d Comparison 2 centrations in ng approache iazzo et.al, 20 est Practice G and relevant lo aring a docum ower the unav manual of a ty tions of differe ersion modelli dispersion as en evaluating tion are best-p employed in , scientifically he Action, alth particular relev developed win ms a unique p ion Protocol. S1006-13802 Generation of ow and disper mposium on of accidental r on, 26, 555-560 elling tools Inv Mission. fic%20report.p ence data ava Short T %20tsiouri%20 Gas Releases with Standard Michelstadt da es were also o 012; Vianello e Guideline. ocal-scale thre ent providing voidable unce ypical model b ent modelling ng, source te s well as em and improvin practice recom n emergency y and practi hough in prog vance are the nd tunnel expe platform to eva Scientific Re . f spatio-tempo sion phenome Computationa release by me 0, DOI:10.330 ventory. Scie www.elizas pdf.pdf, acces ailable for mod Term Sci 0scientific%20 s in Urban A d Software, Ch c) type 3 ataset for type observed by s et al., 2011). eats from relea guidance in h ertainty in sim by information approaches. rm characteriz mergency res ng tools and m mmendations, preparednes ically qualifie gress, has pro results provid eriment repro aluate and vali eport. COST orally high re ena within the al Wind Engin eans of two di 03/CET122609 ntific Report. s.eu/images/S ssed 21.10.20 del testing. Sc ientific M 0report.pdf, Area: Calculat hemical Engin e 1 several ases of how to ulation on the zation, sponse models an up- ss and ed, for oduced ded by ducing idate a Action solved e lower neering ifferent 93. COST TSMs/ 13 cientific Mission. tion of neering 534