Comment #1 The truck crash rates play a significant role in determining the overall risk to the public posed by each potential hazardous material route considered in the report. MassDOT recognizes that it is often difficult to collect information on truck crashes and the number of trucks traveling along the roadways along the proposed routes – both the key factors in calculating truck crash rates. Therefore, it is acceptable that Battelle and the City of Boston have employed techniques to extrapolate a full data set from the limited information available. However, it is not possible to have an open and transparent discussion of these issues in the public without the City supplying additional information on the techniques that were used to calculate the truck crash rates; specifically we request that the City provide: A detailed discussion on how the University of Massachusetts calculated the number of truck crashes based on the limited data available; A discussion of how the annual truck miles of travel were calculated when the sample size available through the Freight Analysis Framework (FAF) was so small for some classifications of roadways (5% coverage for Urban Minor Arterials and less that 1% coverage for Urban Collectors and Local Roads), and the impacts of this uncertainty on the calculated truck crashes; and A discussion of what parts of Massachusetts were included in the “urban area” definition and why efforts were not undertaken to calculate the specific crash rates for the Boston urbanized area where there is increased congestion and commuter traffic. The truck crash rate does play a central role in determining the overall risk to the public posed by each potential hazmat transportation route. In the routing guidelines document (FHWA 1996), route risk is defined as the product of the truck accident rate times the number of people adjacent to the route that are potentially affected should an accident occur. We supplement and clarify, below, the discussion contained in the Battelle report concerning the methodology used by Battelle and UMASS on these issues. Discussion in Report In the Battelle Hazmat Route Evaluation report, the Accident Rate discussion begins on Page 48. The following paragraphs in the report describe the process used to estimate truck accident rates for the various roadway functional classifications. An essential part of the analysis was provided by Battelle. Battelle maintains a GISbased compilation of average annual daily traffic (AADT) flows for trucks as part of the Freight Analysis Framework (FAF) program funded by the FHWA. These truck flows were given to UMass in a format compatible with MassDOT’s GIS system to facilitate merging of accident and truck flow data. There was not always a perfect match between the route segments used in assembling the accident data and the route segments used in estimating flows. Where gaps existed, for each road classification the fraction of the routes with flow data was estimated, the total annual truck miles for those route segments was estimated, and then that number was divided by the fraction of the routes with flow data to get the total estimated annual miles traveled by truck on all the route segments by road class. The accident rate was then obtained by road class by dividing the total 1 number of accidents recorded on that road class by the total annual truck miles traveled on that road class. As mentioned above, the UMass analysis considered accident rates in three population zones –urban areas, urban clusters and rural areas. Discussions with the UMass personnel verified that all the alternative routes being evaluated in this analysis were urban. The results for an urban area in the Commonwealth of Massachusetts, (UMass, 2010), are shown in Table 1. The total road miles by functional class are based on the functional classification assigned by MassDOT. MassDOT maintains a GIS link, http://services.massdot.state.ma.us/maptemplate/RoadInventory that can be used to determine the functional classification assigned to any route segment. The total miles for which truck flow estimates are available for each functional class of highway is based on data from the FAF GIS network maintained by Battelle for the FHWA. The estimated truck percentage is from the FAF. The total crashes are taken from a database maintained by UMass for the State Police. The total estimated truck miles traveled is a calculated number that is obtained by extrapolating the annual truck miles estimated from the FAF data and then prorating it to the entire functional class. To obtain the estimate, the annual truck miles from the FAF was multiplied by the total road miles and divided by the total miles with truck flow data, column 2 divided by column 3. The accident rate was obtained by dividing the total crashes by the annual truck miles traveled, column 5 divided by column 6. As expected, the accident rates are lowest on expressways and increase as the level of access control decreases. The uncertainty in the accident rate decreases as the number of accidents increases. Specifically, the variance is inversely proportional to the number of accidents. Thus if there are 1,000 accidents, the variance is 0.001 and the 95 percent confidence level in the accident rate is two times the square root of 0.001 or ± 6 percent. When just the number of accidents is considered, all the differences shown in Table 1 are statistically significant. A greater source of uncertainty is associated with the fraction of the road classification for which there is truck flows from the FAF GIS network. For the Interstate, Principal Arterial and Urban Principal Arterial classes, the fraction of the roads for which there is truck flow data is above 85 percent. However, it is less than 4 percent for the urban minor arterial and less than one percent for the local and urban collector road functional classifications. If any Local or Urban Collector road classes are used in the routing analysis, the analysis will use sensitivity analyses to verify that the data from those road functional classifications do not play an important role in any routing recommendation.1 1 To clarify upon the original report text, please note that no Local or Urban Collector road classes were considered by the City of Boston for use as candidate alternative routes and therefore, none of these roadway classes with poor or limited FAF coverage, were used in the routing analysis. See the discussion in the response below for additional information. 2 Table 1: Estimated Annual Truck Accident Rates by Functional Class Applicable to the City of Boston, Massachusetts Total Road Miles Total Miles with Truck Flow Data Estimated Truck Percentage Total Truck Crashes (2007 thru 2009) 24,619 1.1 5.0 1,347 239,756,951 1.87 982 932 7.1 1,881 2,085,665,880 0.30 Principal Arterial 1091 986 7.5 1,096 1,237,903,630 0.30 Urban Principal Arterial 1,816 1,580 6.5 1,893 613,415,784 1.03 Urban Minor Arterial 3,962 182 6.6 2,200 603,623.948 1.21 Urban Collector 2,876 0.6 6.3 847 170,704,340 1.65 35,345 3,682 6.9 9,269 6,009,991,603 0.51 Urban Accident Rates by Functional Class Local Interstate All Annual Truck Miles Traveled Accident Rate / 106 Miles Response: The first bullet asks how the University of Massachusetts (UMass) transportation research group identified the truck accidents for the various functional roadway classes. MassDOT has been using Geographical Information Systems (GIS) to display highway features for many years. Thus, there is a layer in the GIS showing the functional classification assigned to each roadway. http://services.massdot.state.ma.us/maptemplate/RoadInventory We are also fortunate that the UMass transportation research staff have been identifying the coordinates, latitude and longitude, and then identifying the accident location on a layer of the statewide GIS system. A map showing the serious truck accident locations in the Boston metropolitan area between 2007 and 2009 is shown in Figure 1. UMass was able to tabulate the accidents that have occurred on each roadway functional class because both the roadway functional classes and accident data are in the same GIS. The resultant truck crash totals, for each roadway functional class, are shown in the fifth column of Table 1. This is a robust, reviewed, verified and geocoded truck crash data set based on recent (2007-2009) commercial motor vehicle (CMV) accident information collected and reported in the Commonwealth of Massachusetts. The second bullet addresses the denominator in the accident rate calculation. The concern expressed in the second bullet focuses on the difficulty of estimating total annual vehicle miles by roadway functional class when there are gaps in the truck accident data for certain classifications of roadway types or the accident sample size for certain roadway classification types is small. First, it is important to clarify that the accident sample size in the data set used in conjunction with the coincident Freight Analysis Framework (FAF) network truck flow data was suitablet for all of the relevant roadway classification types being considered for potential designation as alternative hazmat routes examined in the Battelle Routing Evaluation report. The route segments being considered for proposed hazmat routing fall into two roadway functional classes, Interstate and Urban Principal Arterial. For these two classes of roadway, the vehicle counts are available for 932 of the 982 miles classed as Interstate (95 percent coverage), and for 1,580 of the 1,816 miles classed as Urban Principal Arterial, (87 percent coverage). Therefore, it is not relevant and is of no significance to the routing analysis that CMV vehicle flow coverage 3 for certain roadway classes, e.g., 5 percent coverage for Urban Minor Arterials and less than 1 percent coverage for Urban Collectors and Local Roads, is limited. This limitation in the data set is of no consequence to the routing analysis because no proposed hazmat route alternatives or segments falling under those roadway classifications are being proposed. Second, the equation used for estimating the total annual truck miles by road functional classification is: Figure 1. Reportable Truck Crash Locations in the Boston Metropolitan Area In the above equation, AVMT is the annual vehicle miles traveled for that functional road class, L is the total miles of roadway of that functional class, AADTTi is the average annual daily truck counts for segment i of length li. The issue raised in the second bullet concerned extrapolation given that truck counts (i.e., truck flow volumes) may not be available on all roadway segments or for all roadway classification types. To look at the effect of missing data it is useful to slightly rearrange the AVMT equation as shown below 4 The question of the uncertainty regarding the missing data for 50 miles of interstate and 236 miles (13%) f the urban primary arterial is difficult to quantify. It would be expected that some of the missing links would have higher vehicle counts and some lower so the net effect of the missing data would be minor. Assume that the data on the missing links was in more remote urban areas and therefore less traveled than a typical route segment and therefore the traffic count was half of the average. If this were the case, the total vehicle miles traveled would decrease 5 percent for Interstates and the accident rate on the interstate would increase 5 percent. If the AADTT value for the 236 miles of Urban Primary Arterial was half the average for the other route segments, then the total vehicle miles traveled would decrease 13 percent and the accident rate would increase 13 percent. This is maximizing the effect of the missing data. In reality, the AADTT values for some of the missing route segments would be above average and others lower, so the effect of the missing data would be much less than 5 percent on the Interstates and 13 percent on the Urban Primary Arterial. The important factor is that even raising either of these accident rates would not change the route selection recommendation. The third bullet relates to the definition of an urbanized area. The report stated that all the route alternatives being considered were in an urbanized area. Once again, taking advantage of the MassDOT GIS, Figure 2 shows the roads in the Boston area of the Commonwealth that are in an urbanized area. 5 Figure 2. Roadway Inventory by Urbanized Area All the roadways being evaluated for potential designation as hazmat routes are in the Urbanized Area. The final component of the comment raised in the third bullet, concerning whether efforts were made to calculate specific accident rates in the Boston urban area, is addressed in the response to comment #2, below. 6 Comment #2 Although there is insufficient crash data along many of the roadways in Massachusetts to establish specific crash rates, data is available along the interstates and some arterial roadways. We encourage Battelle to utilize crash rates for specific roadways such as I-95 and I-93, where sufficient data is available. This would help better reflect the risk of crashes along these routes, many of which have exit ramps and other design aspects that do not meet current design standards and can affect overall truck crash rates. Response: The risk associated with hazardous materials movements over each alternative route is the primary criterion for routing designation. Accident probability taken together with population potentially exposed to an accident, are the key factors defining that risk. The probability of an accident is the likelihood or chance that a vehicle carrying hazardous materials will be involved in a roadway accident. However, the estimation of accident probabilities or crash rates, particularly when evaluating different roadway types and on different route segments, is complex and can present difficulties related to available data. First, crashes are not common events, occurring less than once every million miles traveled. The availability of accident data will influence the choice of analytical techniques and simplifying assumptions used for estimating the hazardous materials accident probabilities. Risk analysts need to distinguish among the routes. This means we need sufficient accidents to assure ourselves that the accident rate for the two data sets are statistically different. If this is not the case, the data sets should be combined. Within the truck accident data set developed by UMass and used by Battelle for the routing evaluation, the total number of accidents on the Urban Interstates was 1,881 and on the Urban Principal Arterial 1,893, and the accident rates on the two roadway functional classes differ by a factor of three. If a statistical test is performed, the two data sets would be shown to be statistically different with a high level of confidence. On the other hand, if we zoom in, and attempt to narrow the field of focus to look at accident rates on specific routes and specific roadway segments along those routes as suggested, we can introduce problems related to data insufficiency, creating additional uncertainty in any resultant estimate of accident rates. Moreover, whatever the source of data on accident rates, it is extremely important that the same type of data be used for alternative routes examined, and if sufficient accident data is not available for each specific segment of each of the alternative routes, it is then most appropriate to use available aggregate data based on each roadway functional classification to estimate the accident rate. The route evaluated through downtown Boston consists of two segments. Southbound from Everett to Quincy the route consists of an interstate segment 4.8 miles long and an urban principal arterial segment 5.8 miles long. Northbound the interstate segment is 4.85 miles long. When considering the use of I-95 and I-93 around Boston, there are also are multiple segments and while they are slightly different northbound and southbound they consist of approximately 45.6 miles of interstate and 1.3 miles of urban principal arterial. Getting accident rates on short route segments is a very challenging task because the denominator is vehicle miles per year and the shorter the segment the more time, i.e., multiple years of data, is required to obtain a sufficient number of accident incidents to obtain an accurate estimate of the corresponding 7 accident rate. Based on discussions with the team at UMass that manages the truck accident database for Massachusetts that was used in the analysis, and based on Battelle’s experience in conducting similar accident rate estimates where statistical analysis methods have been used to assess the level of uncertainty in the data used for estimating the resultant risk values, it was determined that there were an insufficient number of accidents in the database on the specific roadway segments of interest to develop specific accident rates looking only at the I-93/I-95 proposed alternative route segments. Furthermore, the use of statewide data based on roadway functional class was judged to be more appropriate and significantly more meaningful from a statistical standpoint. A quick analysis of the accidents file provided by the University of Massachusetts shows that from 2004 to 2009 there were 144 accidents on RA1, the through route and 366 accidents on RA3, the beltway around Boston. Thus, over that period of time, there are about 14 accidents per mile on the downtown route and 8 accidents per mile the beltway. The volume of trucks is certainly much greater on the beltway which would suggest that the accidents rate ratio of the downtown route is greater than the ratio of 14/8, or 1.76. The accident rate on the Urban Principal Arterial functional roadway class developed from the UMass truck accident data set used in the routing evaluation study is three times the rate on the Interstate (1.03 versus 0.30, respectively). The quick analysis just discussed which estimated the accident rate ratio comparing the downtown route (RA1) and the beltway around Boston (RA3) using values of 14 accidents per mile versus 8 accidents per mile, respectively, is consistent with these findings when compared to the UMass estimate of a three times higher accident rate on the Urban Principal Arterial route (RA1 through Boston) than on the Interstate route (RA3), based on the estimated accident rates by functional roadway class. Thus, even if an accident rate estimate could be developed just for the specific route segments being considered in the analysis (as opposed to using statewide truck accident rates developed by UMass and based on roadway functional classes for urbanized areas (as was done in the report), it is likely that two observations would be true: (1) using more specific accident rate estimates for the I-93/I-95 segments would likely affect the resulting risk ratios in such a way as to increase the favorable risk ratio for selecting RA3 over RA1, indicating that alternative route RA1 through the City presents even more risk than previously calculated in the routing risk analysis; and, (2) if these route segment-specific accident rate estimates could be developed and were to be used, the degree of confidence from a statistical-based uncertainty perspective, would be much lower in such narrowly focused accident rate estimates (where the available sample size of representative truck crash accidents on each segment is small). It would take a great deal more time and effort to determine if an accident rate that is statistically different from the accident rate currently being used could be developed for just the route segments being considered in the analysis. Battelle does not believe that such an undertaking is necessary or warranted, particularly given the clear results from the original (and rigorous) routing analysis. The suggested modification to the methodology and additional work suggested certainly are not required by either the regulations or the federal routing guidance, and would likely introduce more uncertainty, not less, into the resultant alternative route risk ratio estimates. 8 Comment #3 Battelle utilized a mail survey for their data collection efforts on trucks using I-93 and the surface roadways in Boston for NRHM deliveries. This effort produced only limited results with 150 surveys out of 1,500 being returned. Given the low response level to the survey, additional data collection should be pursued such as gate surveys at the Everett terminals and traffic counts along the existing Hazardous Materials route in Boston. This information is critical for determining the impact on commerce, as the burden from any change in the route may impact each shipper differently. This added data collection will help provide a stronger case for the overall impact on commerce than the one provided in the report which relied on a number of assumptions. The additional data collection can also help to determine if the potential new routes impede or unnecessarily delay the shipment of goods. Response: "Comment 3" suggests that additional data collection should be pursued (gates surveys/truck counts) to effectively assess “burden on commerce.” Battelle believes that collecting the additional data as suggested would be burdensome and unproductive. Additional quantitative data on commodity flow is not needed to analyze the burden on commerce and shipment delay issues because Battelle has collected sufficient hazmat flow data from a variety of sources to estimate the major types of hazmat moving in the Boston area. With respect to the hazmat carrier and shipper survey Battelle conducted as part of this routing evaluation report, in Battelle’s experience on similar projects and research conducted for the FMCSA in other contexts, a 12 percent response rate for these kind of surveys is not atypical and useful information can still be learned from such responses, even if only a small percentage of the questionnaires are responded to and returned. In addition, the regulatory requirement to examine the potential that a routing designation may affect an unreasonable “burden on commerce” can be met using a qualitative evaluation and the routing evaluation process itself as recommended in the federal guidance is expressly designed to satisfy this burden on commerce requirement as an intrinsic component, without any additional steps or actions. Routing guidance, p. 33 (FHWA 1996). Battelle went beyond that level of qualitative analysis by utilizing ATRI data on operating costs to approximate cumulative economic costs on operators and estimating the potential burden on commerce quantitatively. Moreover, motor carriers and transporters, themselves, are in the best position to understand their operating costs and can submit comments and information with respect to any potential or perceived burden on commerce issues during the public comment period being conducted by MassDOT should they wish to do so. MassDOT and the City of Boston will review and consider all such comments during that process. Because no formal hazmat commodity flow survey existed or had been specifically performed for the Boston Metropolitan Area, Battelle collected hazmat commodity flow data from a number of diverse sources. These included the following: 1. The U.S. Census Bureau commodity flow survey data. In the past, the Census Bureau has issued reports documenting the hazmat flows at the national level and at the State level. A year ago all the State level hazmat commodity flow survey data were available for download from the Census Bureau website using 2003 9 2. 3. 4. 5. 6. census data (U.S. Census 2003) but are no longer available there.. A formal request to the Census Bureau was made for both the Commonwealth of Massachusetts and Boston Metropolitan Regions hazmat flow survey results (U.S. Census 2011). These data provided a valuable source of information regarding hazmat flows in the Commonwealth and in the Boston region. Survey of carriers and shippers. In order to estimate the number of through hazmat shipments in downtown Boston Battelle sent out a questionnaire to over 1,200 carriers and shippers. Battelle received 150 responses. This level of response, over 12%, conforms to the results of this sort of questionnaire in other Battelle projects. Although the questionnaire provided valuable data on hazmat shipments in the region, only 1/3 of these responses said that they transported HM through Boston. We sent the questionnaire to every PHMSA registered carrier or shipper within 75 miles of Boston so the 1/3 that state they shipped through Boston would not be an unexpected answer. The survey results for downtown Boston are based on about 50 carriers and some of those left some key fields blank, making it impossible to estimate the total number of monthly shipments from this data. Follow-up calls were made but few answers were obtained. Several truckers stated information on the number of daily shipments was proprietary information. Thus, it was not possible to get a good count of the number of daily HM shipments through downtown Boston from the 150 respondents. Examination of video tape recordings of vehicle movements. Battelle obtained video tape records from Boston traffic cams but it was difficult to identify placarded trucks using the surface streets in downtown Boston. In addition, we were unable to differentiate between through trucks and those making deliveries within the City. State videos taken on Route 128 proved to be unsuitable for determining hazmat movements. Boston’s hazmat permits and applications. Boston permits and applications were tallied to identify hazmat shipments in the city. The applications provided information on number of shipments, commodity type and origins and destinations. Inspections of hazmat vehicles conducted by Boston. Although the inspections are not random, they provide some indication of the distribution of hazardous classes/divisions that are using Boston thoroughfares and consequently a distribution of hazmat classes/divisions can be developed from that data. Note that MassDOT and the State Police were unable to provide hazmat inspection data outside the City of Boston. Lists of hazmat spills reported to the U.S. Department of Transportation Pipeline and Hazardous Material Administration (PHMSA). These spills are found in the Hazardous Materials Information Reporting System (HMIRS). While the data are limited to accidents that result in hazmat spills, the HMIRS database were used to identify any carrier-reported spills that occurred in Boston, at another location but originated from a shipper located in Boston or, was to be delivered within Boston. While hazmat spills are an extremely small fraction of the hazmat shipments, they provide a distribution of hazmat classes/divisions for which a carrier has reported spills. 10 The finding from these surveys was that Class 3 flammable and combustible liquids are the dominant hazardous material using the downtown Boston Streets, by some estimates as high as 95 percent of all the hazardous material shipments. The responses to the carrier and shipper surveys showed a lot of variability, ranging from less than one hundred to several hundred hazardous material shipments per day, probably depending on the day. For that reason the burden on commerce analysis Battelle conducted was based on the potential economic effect on a single vehicle, looking at estimated increased costs per truck, based on travel time/distance differentials between alternative routes and industry reported operating costs on a per mile and cost per hour basis. As such, the analysis is scalable to look at larger macro effects on commerce. However, contrary to the sentiment expressed in the comment above, the additional data collection being suggested (e.g., gate surveys at terminals and truck counts along existing hazmat routes) to obtain additional or alternative estimates of hazmat commodity flows, or even simply the number of trucks affected, would not yield better data for estimating any potential effect on commerce on a per truck basis. From the existing available data sources used and the per truck estimated increased operating costs evaluating potential burden on commerce in the analysis, as was shown in the report, it is quite possible to extrapolate from this to look at largerscale effects on a quantitative or distributed cost basis, even though such analyses are not required. Battelle’s analysis, based on adequate data, demonstrated that any such effects do not present an unreasonable burden on commerce, and further, that alternative hazmat routes to the current through routing through Downtown Boston will significantly decrease public risk. “The primary criterion for a routing designation is that the designated route significantly reduces risk.” Routing guidance, p. 11 (FHWA 1996). 11 Comment #4 The report states that it is focused primarily on the transport of Class 3 flammable and combustible liquids which have a relatively small (100 meter) impact radius for fire or vapor cloud explosions. However, the impact radius used in the risk analysis is half a mile based on the guidance provided by the FMCSA. The reasoning for this discrepancy should be stated more clearly, along with the chances that an incident involving Class 3 flammable and combustible liquids would impact people located a half mile from the incident. Response: When an accident occurs, it is typically the people who are in the vehicles involved in the accident that are most at risk. The next most vulnerable group is the individuals who are close to the accident at the time of the accident. These include, pedestrians, occupants of other vehicles on the roads but not directly involved in the accident, residents in occupied dwellings and others such as employees in businesses located on the side of the roadway adjacent to the crash can occasionally be affected. If hazardous materials are present, impact distance can become much greater. Flying debris from the rupture of a cargo tank can injure individuals several hundred meters from the accident, the pressure wave generated from the rupture can break glass and individuals can be injured from the flying glass. The flammable liquid in the cargo tank can travel some distance and involve structures in the subsequent fire. While debris, flying glass and secondary fires are seldom generated in an accident involving Class 3 flammable liquids, the Emergency Response Guidebook prepared for emergency responders specifies two isolation distances for a Class 3, flammable liquid spill. In the event of a spill but no fire, the recommended isolation distance is 25 to 50 meters (80 to 160 Feet). In the event of a fire, the recommended isolation distance is 800 meters (1/2 mile). The conservative prudence of using this assumed ½ mile impact radius can be demonstrated by examining real world incidents. Recent experience in the Boston region with tragic gasoline tanker accidents, on Route 1 in Saugus, MA, for example, provides a cautionary tale not to underestimate potential for distant impacts, where not only did the tanker itself explode killing the driver and involving other vehicles and their occupants, but where spilled fuel on the roadway traveled quite some distance through a storm drain, into a stream, catching on fire and igniting several buildings and damaging homes in a neighborhood at a considerable distance beyond the original accident location, resulting in the evacuation of a large impact area and completely shutting down Route 1 in both directions for more than 10 hours. For contemporaneous news accounts of this incident, see for example, “1 dead, fires break out after tanker rolls over in Saugus,” available at: http://www.boston.com/news/local/massachusetts/articles/2011/07/23/1_dies_fir es_break_out_after_tanker_rolls_over_in_saugus/?s_campaign=8315 and “Tanker truck carrying gasoline crashes in explosion on Route 1,” available at: http://www.myfoxboston.com/dpp/news/local/deadly-tanker-truck-crash-insaugus-closes-part-of-route-1-25-apx-20110723. The analyses of the through routes through Downtown Boston follow the guide prepared by the Federal Highway Administration titled: Highway Routing of Hazardous Materials, Guidelines for Applying Criteria. Section II titled: Method for Determining Risk. The section is further 12 divided into three sections, Route Characteristics, Accident Probability and Accident Consequences. Comment 4 is best addressed by looking at the guidance provided in the section on Accident Consequences. The following quotes from the guide are applicable; … the consequences of a hazardous material release are assumed to be proportional to the number of people within a release impact area along the route. The following methods can be used to determine the impact areas for releases of NRHM. Method 1: A simple but extremely conservative option, is to use a large fixed potential impact distance (e.g. five miles on either side of the route. This option is applicable to routes on which poisonous gases etc are being shipped. … Method 2: A somewhat more detailed treatment of the impact area is to assume that the size of the impact area is based on the type of material released or its hazardous material class. … The distances taken in this guide are taken directly from the DOT Emergency Response Guidebook … Method 3: For a more site-specific or refined determination of impact areas, computer models can effectively incorporate site specific climatic conditions, release quantities, hazardous material types, and topology into modeling the release, explosion, or dispersion of hazardous materials. Chemical and site-specific computer modeling can help differentiate the impact areas for chemicals in the same hazard class by accounting for the range of chemical properties and storage conditions, by accounting for the hazardous materials that behave as dense gases versus neutrally buoyant gases, and by accounting for various levels of toxicity for chemicals in the same hazard class. … Method 2 was selected for the analysis of the Boston HM routes for several reasons. First, every determination of hazardous materials currently using the routes through Downtown Boston show that the vast majority of these shipments, most show in excess of 90 percent, are Class 3 Flammable or Combustible Liquids. Thus there is no reason to use the third method to enable differentiation among the hazardous materials being shipped. Method 1 can be dismissed for the same reason, there is no evidence that any Poisonous Gases are routinely being shipped on the downtown streets of Boston. Another reason for selecting Method 2 is its simplicity. The tools are readily available for determining the number of individuals that might be present within a ½ mile of the hazmat routes being evaluated. As pointed out in the introductory paragraph, for many Class 3 releases the impact area is relatively small, 25 to 50 meters (80 to 160 feet). Even in the event of a fire, many consequence models show an impact radius of less than 150 meters, i.e. a fireball generated when a burning cargo tank splits open. For the vast majority of accidents involving Class 3 flammable and combustible liquids, the actual impact distance is much less than the guidelines distance of 800 meter (1/2 mile) distance being used in Method 2. Clearly, even though it is conservative, there is justification for using the ½ mile distance on either side of the roadway. Using the ½ distance specified for Class 3 flammable and combustible liquids using Method 2, by selecting a ½ mile wide impact zone, eliminates a major modeling uncertainty. As the zone narrows, the uncertainty in the number of people present increases rapidly. Close to the roadway, the number of individuals present can vary greatly over time, with a large number of people present during rush hour to a greatly reduced number present in the middle of the night. Trying 13 to estimate that number with any accuracy would border on the impossible. In effect, by using Method 2, the assumption being made is that the number of people that are likely to be seriously injured as a result of an accident will increase in direct proportion to the number of individuals within a ½ mile of the accident. 14 Comment #5 As required in the federal regulations, the report goes to great lengths to estimate the potential population exposed to risk along each route, figuring the number of people in schools, hotels, hospitals, nursing homes, and visiting national parks. However, the report does not provide any estimate of the number of shoppers potentially put at risk along each route. There are a number of major shopping malls in close proximity to the proposed routes including the Faneuil Hall Marketplace and Burlington Mall. We request that the City add shoppers to the population figures, being sure not to double count shoppers who have already been accounted for. Response: To address this comment and MassDOT’s request for additional analysis examining potential exposure of shopping populations, Battelle conducted a sensitivity analysis of shoppers at major malls along alternative routes1 and 3, the route through downtown Boston and the preferred route using I-93 and I-95 around the city. Note that on alternative route 1 (RA-1), Faneuil Hall Marketplace, was excluded from the analysis because the shoppers to Faneuil Hall were included as visitors related to tourism in the RA-1 evaluation. Along RA-3 and RA-1 four and two malls were identified respectively for this sensitivity analysis. In order to estimate the peak number of visitors to these malls Battelle estimated the number of parking places at each mall and assumed that the peak number of visitors would occur when all of the parking places were filled. Each vehicle was estimated to carry an average of two people. Therefore, the number of parking places multiplied by two represents the great majority of peak visitors. In addition, an assumption was made that some visitors could arrive by transit and that these people would represent an added population. If rail and bus are available we used a number equal to 15% of the total peak visitors to the mall. If only bus transit was available we added 5% to the total number of visitors. The total number of peak visitors for each of the malls is shown in Tables 1 and 2. Because shopping centers are open both during the day and into the nighttime hours, peak visitor numbers would occur in both the day and night time periods Table 1: Estimated Added Peak Mall Population for Route Alternative 3 Name Route Gross Leasable Area Burlington Mall® Rt3 1,318,000 5272 10,544 Bus 528. 11,071 Bus Orange line ,Bus Commuter Rail, Bus 636 13,339 375 2,875 360 2,760 1899 30,045 Parking No Peak Time Visitors South Shore Plaza® Meadow Glen Mall Rt3 1,588,000 6352 12,704 Rt3 260,000 1250 2,500 Woburn Mall Rt3 237,000 1200 2,400 3,403,000 14074 28,148 Total 15 Public Transit Visitors By Mass Transit Total Visitors Table 2: Estimated Added Peak Mall Population for Route Alternative 1 Name South Bay Center Cambridge side Galleria Totals Gross Leasable Area Parking No Peak Time Visitors Rt1 445,626 1800 Rt1 900,000 1,345,626 Route Public Transit Visitors By Mass Transit Total Visitors 3,600 Bus 180 3,780 3600 7,200 Green Line, Bus 1080 8,280 5400 10,800 1260 12,060 The current risk calculations based on population estimates along RA-1 and RA-3 did not include these mall visitors in the Battelle report. The effect of adding these visitors to the malls to the RA-1 and RA-3 populations on the risk ratios shown is in Table 3. Although the relative daytime and nighttime risk ratios comparing alternatives RA-1 and RA-3 are lower with the added population represented by the peak mall visitors, the ratios still remain favorable and even at night exceed the 1.5 cutoff satisfying the through routing criteria, and indicating that RA-1 presents at least 50 percent more risk to the public than the proposed alternative RA-3, when taking into account these additional peak shopping populations. Table 3: Sensitivity Case: Risk Ratios if Maximum Number of Visitors at Malls Located Within One Half Mile of Alternative Routes RA1 and RA3 are Added to Day and Night Populations Ratio of Risk Between Routes Risk Ratios in Hazmat Route Evaluation Report Route RA1 / Route RA3 Risk Ratios if Maximum Number of Mall Shoppers Added to Population Route RA1 / Route RA3 Northbound Day Southbound Night Day Night 4.02 2.45 3.96 2.23 3.67 1.82 3.72 1.93 Note that the estimate used for number of mall visitors is likely to be lower if a more detailed analysis were to be conducted. For example, part of the footprint of some malls is located outside of the half mile threshold. Thus, a more detailed analysis might find that some of the peak visitors to a particular mall should not be tabulated as part of the analysis area. Also, assuming all the parking spaces are always filled is clearly conservative. The satellite photographs used to count parking spaces indicated much less than 50 percent of the parking spaces to be filled. Thus if an unbiased estimate of the number of visitors were used, the risk ratios would be less affected by the mall population. Clearly, incorporating the mall population into the base population exposure and risk calculation would not and does not change the routing selection recommendation. 16