Freight Transportation Resilience in Response to Supply Chain Disruptions (2019)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

56 replace navigational buoys that were displaced in a flood or accident. Similarly, it would be helpful to increase the number of dredges and supporting equipment so that channels could be readily repaired and maintained. Shoaling often occurs after floods. If the shoaling is extensive, it can make the channel unsafe for navigation, prolonging the closure even after flooding is over. With respect to dredging, users of the inland waterway and USACE representatives alike reported frustration in the funding mechanism for dredging projects. Stakeholders were unanimous in wanting guaranteed multi-year funding for projects so that they can be scheduled far in advance with stakeholder input on the least-intrusive timing for each element of the project. Funding currently covers only a single year and is not guaranteed in subsequent years. This piecemeal approach is an inefficient and costlier way to conduct projects as it incurs multiple mobilization and demobilization costs and adds contracting complexity. 4.7 REGULATORY FACTORS Interviewees emphasized the need to revisit certain provisions outlined in the Jones Act and Stafford Act that, if temporarily waived or revised, could assist in increasing resiliency in freight transportation during and after a supply chain disruption. 4.7.1 Jones Act The Jones Act refers to the section of the 1920 Merchant Marine Act mandating that any waterborne cargo carried between two U.S. ports must be carried on a U.S. flagged ship, built in the U.S., owned by U.S. citizens, and crewed by U.S. citizens or permanent residents. In the event of a necessary diversion, such as Hurricane Sandy preventing ships from calling in New York and New Jersey, ocean carrier’s vessels that do not meet Jones Act requirements must call at a different port outside the disaster zone. When that happens, the Jones Act prevents them from subsequently moving that cargo back from the diversion port to the original port by ship. Their only options are to move the cargo inland by truck or train, which means they may encounter problems of weight restrictions on the highways or insufficient rail capacity or access at the secondary port. They must be very careful about which port they use so that cargo does not get stranded, unable to move either on land or by sea. The carriers interviewed reported that Jones Act waivers are exceedingly difficult to obtain but would be helpful in supporting cargo diversions and increasing supply chain resiliency. 4.7.2 Stafford Act The Stafford Act refers to the Disaster Relief and Emergency Assistance Act passed into law in 1988. It dictates how the federal government disburses aid to states and local governments to help recover from a disaster. Financial and physical assistance are delivered primarily via FEMA once the president declares an emergency. One common criticism of the mitigation portion of the Stafford Act is that it only provides mitigation funding after a disaster has occurred. Many resiliency experts contend that mitigation funding could be more beneficial before an emergency. Interviewees also noted that the provision in the Stafford Act requiring infrastructure be built back to pre-disaster conditions impedes resiliency. Critics question why these vital assets could not be built back with improvements that make it able to withstand another event. For example, relief funds could not be used to rebuild previously above ground utility lines underground where they would be more resilient to storm damage. Other examples include a stretch of road or railway that is vulnerable to accidents due to poor visibility or a tight turning radius, or a bridge with abutments vulnerable to tree or boulder strike during flooding. Any infrastructure destroyed in a disaster and rebuilt using FEMA assistance, must be rebuilt the same way it was prior to the disruptive event. In these instances, there is a lost opportunity to improve the infrastructure and make it more resilient. The FHWA has an emergency relief (ER) funding program to help rebuild highways damaged during disasters. In all cases, the repair must bring the infrastructure element back to the standards it was built with. In some circumstances, the project can include betterments such as realigning a roadway further from a river that floods

57 frequently, incorporating larger culverts, using grouted riprap, or rebuilding a bridge higher or with a wider span. The FHWA relief money can be used for a betterment if it can be shown that the project will save the agency money over time. Another regulatory issue that came up in multiple interviews was the allowance of overweight trucks on roadways during, and immediately after, disruptions. This could be in the context of ships diverting to a different state with lower allowable highway loads, or in the context of barge or train disruptions requiring a modal shift to trucks. Despite the different situations, part of the solution was the same --- having waivers to allow high, wide, and heavy trucks on certain roads during and immediately after a disruption. Shippers and distributers should communicate with state DOTs in advance to understand the regulations and the waiver process. 4.8 BARRIERS TO QUICK RECOVERY In addition to regulatory issues, interviewees identified potential barriers to response and recovery, some of which echoed literature on supply chain resiliency and others that were unique to their industry. The following barriers to implementing effective mitigation and response strategies are presented in three categories: organizational culture, collaboration and resources. 4.8.1 Organizational Culture The following factors pertaining to organizational culture affect disruption recovery:  An organizational mission that is not geared towards disaster resiliency and disruption recovery. For instance, an organization with a mission geared towards maximizing profit may find it difficult to recovery quickly from a disruption if funds are not allocated to bring service back or to invest in improvements to minimize future disruptions.  Organizational motivation in terms of responding quickly and cost-effectively.  Standardized practices or operating procedures that are difficult to work around.  Organizational inertia in implementing responses.  Professional mindset of dominant organizational groups.  Language barriers that govern how easy it is to coordinate disruption response within and outside of the organization. These barriers are most often related to the different terminology used by the participants in a disruption response. 4.8.2 Collaboration Since institutional relationships, processes and procedures, and the "mindset" of those responsible are different from one locale to another, collaboration is essential to a successful response. Interviewees noted that the following barriers hinder disruption recovery:  A major barrier to collaboration is often a reluctance to accept risks by both public- and private-sector organizations.  Lack of leadership at critical times and loss of leadership after the process has been underway for a period can represent serious setbacks to successful collaboration.  More complicated efforts at collaboration which represent more formalized interaction can often be difficult to attain. However, crises or the threat of crises can be an important motivator for putting in place collaborative frameworks that can serve as the foundation for further joint activities.  In many cases, collaborative efforts work well until initiatives are taken to institutionalize relationships in formal agreements. Such agreements often commit agencies to certain courses of action, which are often reviewed for legal, institutional, and political consequences.

58  Fostering public awareness of a collaboration and of what it can produce is a good strategy to weather changes in leadership. If the public has come to expect collaborative undertakings from a group of organizations, it will be difficult for new leadership to disband this group.  The external environment for collaboration is often characterized as some players having more influence than others. 4.8.3 Resources Resources involved in a supply chain disruption can broadly be split into several categories. The following factors affect quick supply chain recovery after a disruption. Commodity  Perishability can be a hindrance to supply chain recovery. Perishable products can deteriorate during a long supply chain disruption, making it tougher for the supply chain to recover quickly.  The seasonality of certain resources can affect supply chain recovery after a disruption.  Resource visibility within a supply chain can affect recovery after a disruption. Knowledge of where resources are will help a supply chain recover quickly. Not all supply chains are able to perfectly disseminate information; this serves as a barrier to quick recovery. Some of the interviewees were in the process of implementing cargo tracking technologies, such as Radio Frequency Identification (RFID), to bolster resource tracking and visibility. Financial FEMA funding is expressly intended to help communities recover after a disaster. Issues involved with FEMA funding that were expressed during the interviews included a lack of enough funds, difficulties in the application process, timely receipt of funds, and restrictions from the Stafford Act as discussed in the 4.7 Regulatory Factors section. Currently FEMA does not have the capacity or the amount of funding required to fully fund every recovery effort. Given that extreme weather events are becoming more intense and frequent, there are proposals to increase local and state funding allocations by 25%, 35%, or 50%, restructuring the assistance as loans, or having states pay a deductible before federal assistance starts. The filing and bookkeeping requirements for getting reimbursed by FEMA are extensive and detailed, requiring lengthy processes that often delay the delivery of funding when it is most needed. Many organizations are not adequately prepared, staffed or trained to successfully complete the proper documentation. For example, a single street cleanup recovery may have to be divided into many individual projects, such as one for each street (in case FEMA reimburses for one portion of the cleanup work but not another). For each project, separate records must be kept detailing personnel, materials and supplies, equipment, rented equipment, and contract work. This type of meticulous record keeping can be a barrier during a recovery effort. Information  Unwillingness or inability of participants to share lessons learned or share data to help with analysis. Such information includes the location and state of cargo in transit.  Lack of response coordination between the public and private sector, and individuals within a competitive industry.  Delays in the transmission of information among participants in disruption recovery.

59  Insufficient number of staff needed for the normalization of operations. This can be caused by internal constraints (e.g., inability to hire extra staff); or external constraints (e.g., staff stranded by a natural disaster).  Staff is too inexperienced to quickly implement disruption recovery procedures. This can be due to low staff retention that results in the loss of employees with valuable experience. This could also be due to insufficient disruption recovery training.  Inability to incentivize existing staff and potential temporary hires with financial rewards due to binding collective bargaining agreements. Capital Resources and Tools  Inadequate analysis tools that can be used to assess the most cost-effective action in disruption recovery.  Inability of the organization to invest in technologies that streamline operations and disruption recovery.  Compatibility issues between information systems used by different organizations.  Systems dependencies in which the resumption of one system depends on the resumption of another.  Breakdown of proprietary tools that can only be repaired by specialists.  Inadequate storage and equipment replacement. Disruptions can cause product build ups and backlogs at supply chain nodes; there needs to be in place contingency plans to handle excess cargo. Airports need to make sure they have enough skids and pallets while seaports need to ensure they have enough containers. 4.9 CONCLUSION The case studies and subsequent interviews were valuable methods for understanding how to make the nation’s freight transportation systems more resilient. For the most part, the resiliency themes that emerged from the interviews transcend geography, commodity, and disruption type, but tended to focus on individual incidents. The exact type of disruption, whether natural or manmade, is not as important for recovery as the duration of disruption and the amount of lead time available before the disruption occurs. Some resilience strategies are specific to mode and type of infrastructure. Notably, river transport by barge relying on a system of locks and dams has a specific set of strategies that cannot be applied to roads, rail, airports, or pipelines. Sharing information is crucial, before, during, and after the incident. Keeping communication lines open is critical to all parties in their recovery. The types of disruptions that are most challenging from a communication standpoint are instances where power, phones, or computers are down. Organizations should have backup communication plans in place, such as radios, satellite phones, and possibly the latest types of applications such as chats and private messaging services such as WhatsApp. Keeping paper copies of contact lists can be invaluable when electronic systems shut down. Another form of sharing information is when groups within the same freight transportation market form coalitions or alliances. Even shippers who are competitors can realize benefits from sharing information and possibly resources during emergencies. These groups all have a common interest in resolving the disruption and returning to normal operations. The best form of logistical resilience is planning. This takes the form of having backup routes along corridors with known vulnerabilities and possibly contracts in place in case cargo needs to switch modes, for example truck-to-rail or barge-to-truck, or if cargo needs to change ports of entry. Planning also includes locating places to stockpile cargo or supplies in a safe place prior to known closures. Labor

60 From an institutional standpoint, a message that was repeated multiple times is that experience is a great teacher. Organizations should take advantage of smaller disruptions to incrementally improve the system and generate lessons learned. In the same vain, it is important to retain individuals who have been through disruptions and have experienced recoveries, both good and bad. These are the people who can guide an organization through the next disruption, constantly building on their connections and contacts with other stakeholders. Hand in hand with this is the concept of fostering a company culture of taking care of employees, which, in return, encourages an “all hands- on deck” attitude during disruption. One of the best ways to make freight transportation more resilient is to build redundancy into the infrastructure of every system. However, this, along with adding extra capacity, is very expensive and not always practical. Another strategy is to design and build infrastructure to be as durable as possible. This will not avoid every disruption, of course, but having alternate bridge crossings, redundant power stations, high durability pipelines, for example, will help avoid total shut down during most incidents. Regional closures, typically caused by weather events, are better mitigated by building durable structures that can withstand the most common events for that region, such as flooding, blizzards, or high winds. If this is the strategy, the organization must then look at potential future types of disruptions to determine how to best position the organization for those types of disruptions. Regulatory barriers to supply chain recovery can be mitigated by working with regulatory authorities. Organizations should liaise with regulatory authorities to determine whether some key regulations can be waived or revised to aid supply chain recovery. Supply chain stakeholders should try to do so prior to a supply chain disruption. There is no single way to make a supply chain more resilient. It is a question of balancing planning initiatives, organizing resources, and sharing information within each market and region. Governments can do their part by helping fund and build durable infrastructure and creating redundancy wherever possible. Individuals and organizations who have lived through disruptions and experienced the road to recovery can be the most valuable resources in making systems more resilient in the future.

61 CHAPTER 5: ANALYSIS TOOLS AND MODELS FOR SUPPLY CHAIN RESILIENCE 5.1 INTRODUCTION An important component of anticipating and addressing the characteristics of supply chain disruptions is understanding freight flows and how they might respond to changes in network availability, and to quantify the costs versus benefits of proposed remedial actions. This is especially relevant where damage to the nation’s transportation infrastructure is widespread and the loss of connectivity is protracted. This chapter reviews the application of analysis tools and models to derive a deeper understanding of transportation network disruptions and their influence on freight network resilience. In particular, the use of supply chain modeling of freight movements is investigated. While to date such supply chain models have only been developed for a small number of planning agencies, they offer a significant improvement in the modeling of transportation choices, including the sort of choices available to freight carriers faced with significant disruptions to the nation’s transportation networks. To orient the reader at the outset, supply chain freight models assume the origin of the freight is the shipper and the destination is the receiver of that freight thus including all the individual movements that are part of the overall trip. Supply chain models are used to model freight trips by each mode (e.g., truck, rail, barge, aircraft) and link all these trips into an origin-destination (O-D) supply chain. When a change of mode occurs, this may involve costly cargo transfer activities via truck, rail, water, or air transportation via intermodal terminals, or in the case of trucking between long-haul and local freight consolidation and break-bulk facilities. Choosing a supply chain in these models means allocating freight based on the utility of each supply chain option available to a specific O-D pair. Hence a supply chain might include the shipment of cargo by truck to a rail terminal, then long haul by rail to another rail terminal, followed by a truck trip to deliver that cargo to its destination. The choice of the supply chain is determined by 1) the utility of each modal link, or trip, in the supply chain, and 2) the utilities of handling (e.g., time and cost) at the transfer points, or nodes between modes along the supply chain. What is maintained throughout the entirety of the supply chain is the nature of the cargo being transported. The choice of supply chain may also require certain attributes of the freight cargo (e.g., the effect of shipment size, shipment frequency, cargo fragility, etc.) be known. This approach is different from trip-based models, which usually maintain a single mode for the long-distance freight movement and treating the transfer between modes as the origin of a new trip, typically as a special freight generator.

62 5.2 TRADITIONAL TRIP BASED FREIGHT PLANNING MODELS For most regional planning authorities, the modeling of freight movements means developing a long-range (e.g., a 30-year) plan using a well-established “four-step” modeling approach geared to the estimation of 1) location-specific freight production and consumption, and its translation into 2) O-D commodity flow volumes, 3) mode selections, and 4) mode-specific route choices. Over the past quarter century, several regional planning agencies have used one or more of the above modeling steps to assess the impacts of disruptions to freight movements, usually by considering both supply and demand aspects of disruption modeling within a specific corridor or class of commodity. Most applications have been for disruptions to the nation’s highway system (for example Chang et al, 2010; Unnikrishnan and Figliozzi, 2011; the papers in Taylor, 2012; Chen et al, 2017; Du et al, 2017; Ashrafi et al, 2017; Mesa-Arango et al, 2017), with a few papers also considering disruptions to inter-regional and multi-modal ground (i.e., road-rail- waterway) networks (for example, Kim et al, 2002 and Ham et al, 2005). Dong et al (2015) use a multimodal freight transportation network model to simulate commodity movements and evaluate the impacts of hypothetical Midwestern highway, waterway and rail transport disruptions. For rail, the models used a fluid-based dynamic queuing approximation to estimate the delays at classification yards and waterway locks caused by network disruptions). See also Jalic (2015) for a review of large-scale disruptions to air transport networks (where a good deal of cargo is carried as “belly freight” in passenger aircraft), and a post event example of modeling the impacts of the 2012 Superstorm Sandy. In nearly all cases, the emphasis to date has been on simulating the physical responses of freight movements to infrastructure-damaged networks, notably road networks. Most studies have also tended to focus on assessments of specific aspects of resilience, examining such characteristics as transportation network redundancy (the provision of excess network carrying capacity), diversity (alternative routes and modes), and responsiveness/adaptability to changing conditions (the substitution, temporarily or on a more protracted basis, of alternative sources, markets, and in theory at least, the substitution of one good for another). Most of these modeling efforts are “what-if” scenario generation studies. Reggiani et al (2015) call this analysis focus non-tested phenomena, i.e. pre-disruption simulation of events that can potentially occur, as opposed to after-disruption event case studies in which resilience or vulnerability measures are directly applied to real-world data. Mattsson and Jelenius (2015) also find that past network disruption studies have tended to focus on how to anticipate and plan/design for more robust-to-disruption networks, with less attention given to modeling the response and recovery phases after a disaster. Recognizing that widespread and prolonged disruptions to a region’s transportation networks can have large indirect economic impacts that can spread far beyond the region and the industries directly involved, several network disruption studies have combined or linked a freight planning model with an inter-industry economic impact model. For example, Kim et al (2002) and Ham et al (2004) used an inter-industry, inter-regional input-output (I-O) modeling approach to simulate the response to a hypothesized earthquake in the New Madrid Seismic Zone in the center of the U.S. Ivanov et al (2008) also used the IMPLAN I-O model to capture the economic impacts of storm-related damage to the I-5 and I-90 Washington State highway corridors in the winter of 2007-08, but combined in this instance with ex post data from a statewide survey of the disruption experience by some 2,758 affected trucking firms and freight-dependent businesses. Tatano and Tsuchiya (2008) provide an example of integrating a spatial computable general equilibrium (SCGE) model with a transportation model that can estimate the traffic volumes of freight (and passengers). The model, applied to the large Niigata-Chuetsu earthquake of 2004 in Japan, considered the damage to transportation infrastructure and indicated the extent of regional economic losses arising from the earthquake because of disruptions in both intra- and inter-regional trade.

63 5.3 BEHAVIORALLY BASED, SUPPLY CHAIN MODELING OF NETWORK DISRUPTIONS While important for infrastructure utilization, land use planning, and traffic congestion studies, the traditional flow- aggregated and trip-based approach to freight modeling has received extensive criticism for its lack of behavioral realism. Comparatively little effort has been given to incorporating the responses of the different freight actors involved in goods movement, including the shippers, carriers, receivers, warehouse operators and brokers of goods contracts. This weakness affects our ability to model freight agency responses to sudden, catastrophic, and protracted disruptions to the nation’s transportation infrastructure. As numerous authors have pointed out, the increased use of global supply chains can spread the impacts of major network disruptions quite broadly, both geographically, economically, and ecologically. These supply chains can also involve a significant number of different freight agents. A long-distance goods shipment may require more than one shipment leg, making use of a cargo consolidation or break-bulk and transfer facility, possibly involving a cross-docking activity or transferring cargo between two different modes of transport (e.g., via truck-rail, rail-water, or truck-air terminals). A disruption to one of the modes might affect the use of another, and planning models need to be able to recognize which situation is likely to exist and how best to respond to it. As more modes, transfer facilities, and freight agents become involved in producer-to-final customer deliveries, there is more opportunity for different supply chain solutions to emerge. Understanding both the movement options offered and the impacts of network disruptions brings planners closer to understanding both the resilience (or lack thereof) that currently exists in our transportation networks, and the likely impacts of taking specific actions towards system recovery. An extensive enterprise-centric supply chain literature on the topic of disruptions exists, including a wide range of concepts and methods for building resilience into cargo delivery systems (for useful recent reviews see Klibi and Martell, 2010; Snyder et al, 2014; Mattsson and Jelenius, 2015; Elleuch et al, 2016; Ivanov, et al, 2017; and Rosyida et al, 2018). However, as Mattsson and Jelenius (2015) point out, this sort of modeling involves the collection and use of a great deal of data. This not inconsiderable challenge, from a technical perspective, means finding ways to capture disaggregate, microeconomic activity data within necessarily aggregate, macro-analytic freight planning models. In working towards such a goal, a still limited but growing number of planning agencies have begun to explicitly incorporate behavioral elements of supply chain management into their agency models. However, agencies are still far away from explicitly incorporating appropriate financial/ transactional, informational, logistical and physical factors of supply chain management as indicated earlier. When combined with improved reporting and tracking of the physical aspects of natural and man-made network disruption events (causality, geographic extent, timing and duration), the following three modeling directions promise a more comprehensive approach to the modeling of planning for, responding to, and recovering from major disruption events. 1. The use of agent-based modeling to capture the major factors considered in the decision making of different supply chain participants (agents) to bring some freight business-based realism into the modeling process that leads up to the freight activity patterns we see in practice. This includes how these decisions translate into how (mode), when (schedule), where from (origin) and where to (market) goods are shipped, and which supply chain options are used to bring these decisions together. 2. The use of micro-simulation techniques to simulate and subsequently aggregate individual shipments to an inter-zonal (corridor or O-D channel) level of geography for network (congestion impacting) assignment. This allows for freight agent decision making to influence shipment load size, vehicle type, mode and route choice, including both single mode, direct dyadic connections as well as multi-leg and multimodal use of intermediate, distribution center/warehouse/terminal/port-based supply chains.

64 3. The fusion of a variety of data sources, running the gamut from remotely-collected digital datasets to expert interviews. The data collected includes characteristics of goods and freight traffic movements and costs (from federal, state and local surveys), private sector-purchased transactions, digitally based “big” datasets (including GPS, RFID, cellular, bar-code, and traffic counter datasets), and the use of Delphi focus group responses with shipper, carrier, customer, and (as yet largely untapped) freight broker interviews. All these efforts aim at understanding business responses to both past and potential future disruptions to freight networks. European modelers led early developments in this type of modeling. Roogra et al (2010) described some of these modeling efforts, while more recent model applications are described in, for example, (Cavalcante and Roodra ,2013; Holmgren and Ranstedt, 2017; Ben-Akiva and de Jong, 2013; Burgholzer et al, 2013; and Reis, 2014). Indeed, the state-of-the-art in freight modeling is still undergoing development even at the enterprise level. An early attempt at enterprise-level modeling was conduct by Nagurney et al (2002) who used a multi-level network formulation based on the solution of a variational inequality. Subsequent effort by Xu et al (2003) integrated the physical components of the transportation network with appropriate components of the broader business logistics network, financial network, and information network using micro simulation and GIS tools to address freight transportation problems within a much broader decision-making and policy sensitive environment than current models. Bringing similar ideas into regional or network-wide planning models, and finding ways to access the necessary data sources, represents a task for future research. Southworth (2018) provides a brief summary of similar model developments in the U.S., pointing out the relatively recent emergence of actual modeling applications by planning agencies There are only a handful of such model applications in use by metropolitan planning organizations (MPOs) and state DOTs, including: 1) the Chicago Metropolitan Agency for Planning (see RSG Inc. Cambridge Systematics, et al, 2015, also Gliebe et al. 2013, and Outwater et al. 2013); 2) Mariposa Association of Governments (MAG) for freight movements within the Phoenix and Tucson megaregion (see Cambridge Systematics, Livshits et al, 2017; Stinson et al, 2016, and Hong et al, 2017); 3) Puget Sound Regional Council (PSRC) in Seattle and 4) Florida, Maryland and Wisconsin DOTs (See Cambridge Systematics, 2016a). It is important to understand that all these U.S. MPO and DOT supply chain freight models rely on the origin- destination commodity tables from the FHWA Freight Analysis Framework (FAF), which is based on the National Commodity Flow Study (CFS). Both the FAF and CFS are based on national surveys and therefore have limitations and complexities that will impact model analysis. However, they can be used to model supply chain disruptions, as demonstrated in this study. Another important consideration was identifying candidate supply chains as part of this research that would be reflected in the FAF data from origin to destination. For example, the FAF does not recognize seaports for international outbound or inbound flows since the FAF is a compilation of domestic transportation movements. Therefore, supply chains connected to ports were not selected for modeling. In addition, the study team evaluated the other candidate supply chains to be sure FAF data were documented for that movement. Taking into consideration these limitations, the two existing models were chosen for illustrating how freight models could be used for understanding freight resiliency. They included the (MPO level) Chicago Mesoscale Freight Model developed for the Chicago Metropolitan Agency for Planning (CMAP) and FreightSIM developed by RSG, Inc. for the Florida Department of Transportation (FDOT).

65 5.4 CASE STUDY: THE PHARMACEUTICAL SUPPLY CHAIN FROM FLORIDA TO TEXAS This case study focused on the pharmaceutical supply chain from Miami, FL to Houston, TX (see Figure 4). The pharmaceutical industry is highly consumer-driven as manufacturers strive to match production volumes with consumer demand. Small profit margins, operational costs, and market fluctuations have encouraged many pharmaceutical suppliers to adopt a “just-in-time” delivery strategy. Because the availability of drugs to patients heavily relies on the supply chain, catastrophic consequences could follow if the supply chain fails. Pharmaceutical stakeholders need to be proactive regarding both anticipating a disruption event and developing protocols to minimize the negative effects of a disruption. Theoretically, pharmaceutical distributors have the option of changing the transportation mode, changing the transportation route, or changing the business partner in the event of a disruption. The assumption in this analysis is that the distributor can only change the transportation mode and not the route or business partner. The FreightSIM model was manipulated to simulate a disruption in the flow of pharmaceuticals between Miami and Houston. Outputs from the program were analyzed and tested against information gathered through industry outreach and internet searches. Pharmaceuticals are shipped between Miami and Houston primarily by air and truck. Outputs from the model confirmed this mode split for the majority of the tons before the disruption was simulated. The model estimated that an interruption in the air mode would shift most of the tonnage to truck with a minimal amount likely to be shifted to rail. 5.4.1 About FreightSim and the Florida Statewide Model “FreightSim” is a travel demand model component integrated into the Florida Statewide Model (FLSWM). The model simulates the transport of freight between supplier and buyer businesses in the U.S., focusing on movements that include movement in Florida. FreightSim produces a list of commodity shipments by mode and converts these to daily vehicle freight truck trip tables that can be assigned to national and statewide networks in the FLSWM along with trip tables from the passenger model.

66 Figure 4: Origin-Destination Pair for Pharmaceutical Supply Chain Scenario FreightSim is designed to be a policy-sensitive freight model that can be used to:  Inform infrastructure investment decisions  Evaluate congestion on Florida highways  Test the effectiveness of statewide transportation policies on mobility and the economy  Produce multimodal system performance measures for freight  Evaluate the impacts of private sector decisions on the state transportation system  Provide regional agencies with intercity freight travel options for regional planning purpose The model uses FAF Version 3 (FAF3) data along with other data such as Florida employment data as inputs into freight flow forecasting. FAF3 provides origin-destination commodity flow data by 2-digit Standard Classification of Transported Goods (SCTG2) for all the U.S. Forty-four (44) commodity groups in FAF3 data are represented by SCTG2 codes. These commodities range from Live Animal/Fish (SCTG2 #01) to Mixed Freight (SCTG2 #43). Pharmaceuticals are represented by SCTG2 #21 in the FAF3 dataset. FreightSIM is built in the “R” programming language, a computer programming platform with robust statistical analysis capabilities. The model consists of seven major modules that work sequentially. These modules, called “steps” in the model (based on their order of application), include: 1) firm synthesis, 2) supplier selection, 3) FAF flow apportionment, 4) distribution channel, 5) shipment size, 6) mode path selection, and 7) trip table preparation. In order to run the entire model, the steps must be performed in sequence.

67 The model utilizes pre-defined mode combinations to determine what modes are used to transport freight between each OD pair. These mode combinations are called “mode paths.” For example, a "truck-air-truck mode path" means cargo is shipped by truck from its origin to a nearby airport, shipped by air to another airport near the destination, and shipped by truck from that airport to the destination. Similarly, a "direct truck mode path" means that the cargo was shipped by only truck from origin to destination. There are 18 mode paths built into the model including: direct truck, truck-air, truck-rail, truck-intermodal (e.g., an intermodal container movement by truck, either connecting to an intermodal yard associated with a railroad, port or inland port) and others. FreightSIM produces many intermediary and final outputs. Intermediary output files are produced as a result of running each step, while final output files are produced after all steps are finished. Some of the major outputs of the model include tons and values by mode and by direction, tons and values by shipment size and direction, and number of truck trips between each OD pair. Analyzing a disruption scenario requires changes to one or more of the intermediary files. 5.4.2 Base Case The base case in this analysis produces the outputs of the model when supply chains are not disrupted. The base base tonnages and values were from the 2007 FAF3 database. Table 3 illustrates the base case for pharmaceuticals transported between Miami and Houston. The tonnage and value estimates in this table include both directions of the commodity flow (i.e., Miami to Houston and vice versa). As seen, truck is the dominant mode, by both tonnage and value, accounting for 63 and 58 percent of the total tonnage and value moved, respectively. Air is the second most common transportation mode with a 33 and 39 percent share of tonnage and value, respectively. Only a small fraction of the pharmaceutical flow is transported by rail. Table 3: Total Pharmaceutical Tonnage and Value between Miami and Houston Origin-Destination Pair, Base Case (2007) Mode Tonnage (K-tons) Tonnage % Value ($M) Value % Air (truck-air-truck) 99 33% 7,139,305 39% Rail (truck-rail-truck) 11 4% 601,631 3% Truck 189 63% 10,578,271 58% Total 299 100% 18,319,206 100% Source: FreightSIM model Table 4 shows the flow of pharmaceuticals from Miami to Houston. Comparing this table with Table 3 reveals that the overwhelming majority of the pharmaceutical flows between Miami and Houston are in the Miami-to-Houston direction. Only nine thousand tons of pharmaceuticals, valued at $2 million, were going from Houston to Miami. The mode split, however, is very similar to that of Table 3. Trucks dominate the tonnage and value, accounting for 65 percent of the total for both, while air accounts for 31 percent of the total in both tonnage and value. Table 4: Pharmaceutical Tonnage and Value from Miami to Houston, Base Case (2007) Mode Tonnage (K-tons) Tonnage % Value ($M) Value % Air (truck-air) 90 31% 5,049,924 31% Rail (truck-rail) 11 4% 601,631 4% Truck 189 65% 10,578,271 65% Total 291 100% 16,229,826 100% Source: FreightSIM model

68 Interviews with industry stakeholders suggested that that just over half of the pharmaceuticals transported to Houston from Miami are transported by air. At first glance this seems to contradict the findings from the model base case. A closer look, however, reveals that the share of truck in both tonnage and value is twice that of air. Considering that all air cargo requires truck transport for the first and last mile, and given that the shippers responding to the US CFS that underpins the FAF do not always know the line-haul mode used, it seems likely that a good deal of the value and tonnage reported by truck-only is actually delivered using a combination of truck first mile-last mile and airline-haul transport. If this is indeed the case, the model outputs are consistent with the findings from what was learned from talking to industry representatives (and shows the value of a supply chain-based approach). Choosing and defining a mode disruption proved to be a challenging proposition, involving both removing each mode one at a time. Although this might seem reasonable at first, there are legitimate reasons not to simulate a disruption for all the modes and thereby simplify the process. First, the truck mode cannot be removed from the model because both air and rail use truck for the first and last mile of the trip. FreightSIM uses mode paths to simulate the transportation of the goods, and at the end aggregates the tonnage and value of each mode path by mode. If “truck- air” and “direct truck” are considered as the available mode paths, the amount of tonnage and value transported via truck-air and direct truck are part of the estimated output; total tonnage and value are aggregated on truck and air. This results in adding the tonnage and value of truck associated with "truck-air" to those of the direct truck shipment. Even if "direct truck" was removed from the model, the output would still show values for truck which would belong to "truck-air." Second, simulating rail disruption is not worthwhile given that only a tiny fraction of the pharmaceutical flows is moved by rail. This leaves air as the only viable option for disruption modelling. 5.4.3 Disruption Case In order to model a disruption in the supply chain, changes were made in the mode path selection step of FreightSIM. These changes included removing a mode path option for the pharmaceuticals shipped between the Miami and Houston areas. Supply chain disruptions can have different implications for a freight carrier based on factors such as the commodity type and nature of the business. For a minor disruption such as a blocked roadway, the carrier could use an alternative route. For a more serious disruption, the carrier may have to use another mode. For instance, truck transport could be a viable option to air if the airport system itself is shut down due to a terrorist attack (a single airport shutdown might cause shippers to truck the product to a nearby airport if the value of the product is such that fast delivery is necessary).8 A much more severe disruption could force the carrier to change its destination or consumer. For example, a shipment of pharmaceuticals can be transported to Miami instead of Houston if Texas is expecting a hurricane to make landfall. This occurred in 2017 for a pharmaceutical third-party logistics provider during Hurricanes Harvey and Irma. For the purposes of this study, a second disruption scenario was modeled to supplement the analysis. Changing the route or destination was not considered feasible because: 1) modeling a route change requires trip assignment, choice set generation, and route choice modelling modules which were beyond the scope of this task, and 2) the origins and destinations were fixed in the scope of the analysis. FreightSIM was manipulated to reflect the changes in pharmaceutical mode split if a transportation mode was disrupted. This corresponds to the second scenario above that assumes the air mode is not available due to an airport shutdown. 8 As reported in FAF, all the flows were either by truck-air-truck or totally by truck. While an alternative movement could conceivably use truck-rail-truck or truck-water-truck, these were supply chains that were never selected in the database. All supply chains require some truck usage. If any truck usage is allowed, and when the airline haul is disrupted, then "all truck" is the only option available (according to the model).

69 In freight modelling, a mode path is selected based on the logistics cost associated with the use of that link. Intuitively, mode paths with low logistics costs carry more volume compared to mode paths with higher costs. FreightSIM uses logistics cost functions to predict the path and mode of long-haul movements of freight into, within, and out of Florida. The logistics cost function has the form (De Jong and Ben-Akiva, 2007): Total Logistics Costs = constant value + (ordering + transport + damage + inventory in transit + carrying + safety stock) costs The constant in the formula above represents all the hidden costs that cannot be accounted for in the model. The FreightSIM mode path selection step provides the change in mode split after a disruption is simulated. In order to discourage the use of a mode path in the model, a high number was assigned to this constant value for the truck-air mode, causing the “direct truck” mode to become the more cost-effective shipment option. Table 5 illustrates the mode split after a disruption in the "truck-air" mode was simulated between Miami and Houston in both directions. As observed, the direct truck mode is dominant, accounting for 87 and 86 percent of the tonnage and value, respectively. Rail carries the remaining flows, 13 and 14 percent of tonnage and value, respectively. Table 5: Total Pharmaceutical Tonnage and Value between Miami and Houston Origin-Destination Pair, with Disruption (2007) Mode Tonnage (K-tons) Tonnage % Value ($M) Value % Rail 39 13% 2,539,830 14% Truck 260 87% 15,779,376 86% Total 299 100% 18,319,206 100% Source: FreightSIM model Table 6 shows the mode split of pharmaceutical flows from Miami to Houston after disrupting "truck-air." Similar to the table above, truck is the dominant mode carrying 87 percent of both tonnage and value while rail accounts for the remaining 13 percent. Table 6: Pharmaceutical Tonnage and Value from Miami to Houston, with Disruption (2007) Mode Tonnage (K-tons) Tonnage % Value ($M) Value % Rail 37 13% 2,067,908  13% Truck 254 87% 14,161,917 87% Total 291 100% 16,229,826 100% Source: FreightSIM model 5.4.4 Results of the Analysis Overall, the results from the model analysis are intuitive. Pharmaceuticals are primarily transported by air and truck. Once air is not an option due to a disruption, truck is the best alternative since it is the most cost feasible means of transportation compared to rail and water. The model illustrates this in Table 5 and Table 6. These tables also show some volume and value increase for rail. This is expected since FreightSIM will not allocate all the air cargo to truck and assigns small portions of it to rail as well. Table 7 shows the mode split results for the base case (i.e., before the disruption) and the disruption scenario (i.e., after the disruption in "truck-air") between Miami and Houston area for both directions. Pharmaceutical distributors favor truck as the main mode of transportation for pharmaceuticals over other modes when air is disrupted. The truck share of tonnage and value increases by 24 and 29 percent, respectively, after the disruption. The share of "truck- rail," on the other hand, increases much more moderately with only 9 and 10 percent increase in tonnage and value, showing the inability of "truck-rail" to compete with "truck."