Review on Energy Resilience

Transportation Research Part A 69 (2014) 142–164

Contents lists available at ScienceDirect

Transportation Research Part A

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / t r a

Estimating economic and resilience consequences of potential navigation infrastructure failures: A case study of the Monongahela River

http://dx.doi.org/10.1016/j.tra.2014.08.009 0965-8564/� 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author. Tel.: +1 412 719 6901. E-mail addresses: [email protected] (G.S. DiPietro), [email protected] (H. Scott Matthews), [email protected] (C.T. Hendrickson).

Gwen Shepherd DiPietro ⇑, H. Scott Matthews, Chris T. Hendrickson Civil and Environmental Engineering, Carnegie Mellon University, Pittsburgh, PA 15213, USA

a r t i c l e i n f o

Article history: Received 10 March 2014 Received in revised form 11 August 2014 Accepted 12 August 2014 Available online 16 September 2014

Keywords: Inland waterway Infrastructure failure Coal shipment Vulnerability assessment Network resilience

a b s t r a c t

The paper examines the potential effects of failure of heavily used, outdated locks and dams on the Monongahela River in southwestern Pennsylvania. Catastrophic failure would result in lengthy outage of barge traffic. The displaced volume of coal shipments from mines to power plants is estimated using Energy Information Administration survey data. The resilience of the impacted facilities, the viability of their shipping alternatives, and their ability to re-organize into new markets is assessed. Lost revenues are estimated for facilities that close due to an inability to adapt, as well as the replacement cost of towboats and barges trapped by a catastrophic and sudden failure. The aggregate costs to these facilities as a result of a year-long closure are estimated at $0.56-1.7 billion.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

The motivation for this study is the concern about the reliability of the inland waterway infrastructure. This infrastruc- ture, managed by the U.S. Army Corps of Engineers (USACE), allows for the shipment of 2.3 billion tons of commodities in 2010 (USACE, various). More narrowly, this study is motivated by the extended delay in the completion of a major USACE project to replace the antiquated components of the three lower Monongahela lock and dam facilities with two modern facilities. Replacement of the first dam at Braddock is complete, while work on the third lock at Charleroi (and subsequent removal of the second lock and dam at Elizabeth) is delayed due to Federal funding constraints. These funding constraints are felt throughout the U.S. inland waterways (NRC, 2012). In the interim, river transportation is reliant on the ability of USACE to keep the Elizabeth and Charleroi facilities functioning. Catastrophic failure of the highly compromised Elizabeth dam would likely cascade into loss of the fragile river lock wall at the Charleroi lock, resulting in a complete and prolonged loss of navigation on this stretch of the Monongahela. The risks of this specific failure scenario exist at various other aged components of the Nation’s inland waterways system.

The purpose of this paper is to examine the potential impacts associated with extended loss of navigation due to cata- strophic failure of aging infrastructure on the Monongahela River in southwestern Pennsylvania. The analysis focuses on coal shipments from mines to power plants; coal shipments account for three quarters of the commodity tonnage shipped on the region’s rivers. An assessment is provided of the resilience of the regional ‘‘coal-to-utility network’’ in response to an extended loss of navigation through a key stretch of river. Numerous regional and national studies have previously examined various aspects of commodity transport and congestion using GIS tools and national databases, making general assumptions

G.S. DiPietro et al. / Transportation Research Part A 69 (2014) 142–164 143

about the impact of river failure on commodity movements (Kruse et al., 2007, amended 2009; USACE, 2011b). This analysis looks in more detail at actual transport through the at-risk locks, and assesses local and regional impacts, particularly with respect to potential infrastructure failure and the abilities of regional coal mines and power plants to adjust to a long-term system constraint.

The contributions of this paper lie in extending the current body of work which assesses and models (i) the commodity transportation network, (ii) the integrated energy system (focused on the coal-to-utility network), and (iii) impacts of and vulnerability to catastrophic infrastructure failure. The specific application is to failure within the inland waterways, but the findings are applicable to broader transportation infrastructure analyses. Tools are developed to identify mine and power plant accessibility constraints that may compromise components of the integrated energy system. Consideration of these vulnerabilities allows for a methodology to assess resilience and to quantify losses at facilities that may not be able to remain in business without access to barge transport.

Section 2 reviews key prior studies and concepts relevant to this work. Section 3 provides a description of the river infrastructure in the study area and the commodities that are shipped through the region. Section 4 identifies at-risk coal shipments that were barged through the potential failure zone in 2010. Section 5 characterizes the impacts on the mines and power plants that relied on barge shipment of coal in 2010, and attempts to predict likely responses to an unexpected and prolonged closure of a portion of the Monongahela River. Section 6 explores the likelihood that new markets could successfully emerge above and below the failure zone, allowing the impacted facilities to remain operational. Section 7 quantifies the financial impacts to the mines, power plants, and barge shipping operations due to displacement of their coal shipments. Analysis conclusions are presented in Section 8. Additional data and supporting analyses are available in a Supplementary materialdocument.

2. Prior studies and key concepts

Extensive work precedes this analysis developing increasingly sophisticated approaches to modeling the transport of commodities with differing levels of focus on economic theory (Chulkov, 2012), traffic flow modeling, queuing modeling, agent-based modeling (Reis, 2014), and optimization. These analyses differ in terms of whether they assess normal, unper- turbed conditions, short-term disruptions (Jenelius et al., 2006; MacKenzie et al., 2012; Wang et al., 2006; Pant et al., 2011), or long-term catastrophic disruptions (Tsang et al., 2002). Some assess broad commodity flows using input/output models (Ham et al., 2005; MacKenzie et al., 2012; Pant et al., 2011), while others focus more narrowly on specific commodities of importance to a specific region (Kruse et al., 2011). In all of these analyses, a tension exists between what can be modeled under appropriate simplifying assumptions, and the interest in incorporating a robust set of parameters and potential out- come sets. The ultimate need for robustness in these types of analysis has several drivers. First, efficient infrastructure investment decision making in a resource constrained world requires robust accounting of the costs and benefits of proposed projects. In the specific case of the inland waterways infrastructure, Congress requires USACE to conduct extensive analyses to document the need for all major infrastructure projects, including the rehabilitation of aging locks and dams (Bray et al., 2004). Second, national security concerns have increasingly driven more sophisticated failure and resiliency analyses – both in assessing the potential infrastructure vulnerabilities that could lead to system failures, as well as the resiliency of the broader systems to continue to function in the wake of catastrophic losses (Department of Homeland Security, 2009; Baroud et al., 2014; Folga et al., 2009; Campo et al., 2012). Third, the transportation sector accounts for 32 percent of U.S. greenhouse gas emissions (U.S. Environmental Protection Agency, 2014), and the development of sensible and effective strategies to move this sector to a lower level of negative impacts requires an accurate accounting of costs and benefits (Reis, 2014).

Resilience is generally understood to mean the ‘‘capacity to adapt to changing conditions without catastrophic loss of form or function’’, with a more refined definition suggested by Park et al. to be ‘‘an emergent property of what an engineering system does, rather than a static property the system has’’ (Park et al., 2013). In the context of the commodity distribution system that supplies coal to power plants, resilience (or the lack thereof) applies to the physical infrastructure that facilitates movement (e.g., the locks and dams of the inland waterways), as well as to the entities that represent the supply and demand forces in the system (the mines and power plants), and more broadly the integrated energy system. Park et al. goes on to characterize resilience as the ‘‘persistence of relationships’’ where fundamental basic influence relationships are maintained. Applying these concepts to the coal-to-utility system when subjected to failure of a key infrastructure component, we agree that the generic relationship between fuel supplier and electricity generator will remain intact (the lights will stay on), but the nodes will shift. We will demonstrate the capacity of the coal-to-utility system to utilize a variety of suppliers, re-organizing into new sub-markets (above and below the potential failure zone). We will also explore the possibility that not all nodes will remain in the altered system if the new sub-markets are reorganized in a way that maximizes profit (rather than the retention of facilities in the system). We explore the potential fragility or marginal nature of some of the nodes as an important component of an accurate assessment of the impact of catastrophic failure.

The range of potential responses of these entities to an extended river outage differs as a function of their accessibility to the coal-to-utility market. Jenelius discusses accessibility as a key component in road network vulnerability analysis, and the concept is useful in this waterway network as well (Jenelius et al., 2006). Geurs and van Wee explore different accessibility measures for land-use and transportation strategies, as well as the inaccuracies that can arise in analyses with incomplete

144 G.S. DiPietro et al. / Transportation Research Part A 69 (2014) 142–164

accessibility measures (Geurs and van Wee, 2004). For the mines and power plants, their vulnerability to catastrophic infra- structure failure is a direct function of their accessibility to the coal-to-utility market (i.e., their location, shipping patterns, available shipping alternatives, and alternative raw material sourcing flexibility). Understanding the most likely impact of failure requires consideration of these factors.

A number of proof-of-concept modeling exercises have explored different ways to simulate the movement of commod- ities on the inland waterways. To demonstrate these models, simplifying assumptions are frequently used to allow the devel- opers to validate the underlying equations and theory. For example, models might assume inelastic demand for transportation services, where the movement of the commodity is assumed to be a given; other alternatives, such as waiting for a port to re-open or cancelling the shipment, might be ignored (Jenelius et al., 2006; Ham et al., 2005). Models that focus on understanding the incremental cost associated with shifting transportation modes in response to a port closure might assume that rail or truck is always a viable alternative to disrupted barge shipments, both in terms of accessibility and capac- ity (MacKenzie et al., 2012; Kruse et al., 2007, amended 2009; USACE, 2011b). Broad analyses of the value of the inland waterways might assume closure of an entire waterway, requiring the complete displacement of all barge traffic to alterna- tive modes (Kruse et al., 2007, amended 2009; Bray et al., 2011; USACE, 2011b). These assumptions serve their purpose in allowing for the exploration of modeling advances and screening impacts. However, to characterize impacts for the purposes of infrastructure investment decision making, the modeling needs to continue to evolve to better approach the constraints of reality (NRC, 2001; Gravens et al., 2008). This national priority was clearly stated in the 2013 update to the Principles and Requirements for Federal Investments in Water Resources:

‘‘In consideration of the many competing demands for limited Federal resources, it is intended that Federal investments in water resources as a whole should strive to maximize public benefits, with appropriate consideration of costs. Public benefits encompass environmental, economic, and social goals, include monetary and non-monetary effects and allow for the consideration of both quantified and unquantified measures (U.S. Council on Environmental Quality, 2013).’’

As will be shown below, examination of reported shipments between coal mines and power plants in 2010 shows that simplistic assumptions of blanket switching to alternative transportation modes is not appropriate given the geographical and infrastructure constraints that these entities operate under.

3. Regional inland waterways infrastructure and commodities

Pittsburgh, home of three major rivers, has a proud and enduring industrial history that is profoundly intertwined with its waterways. Hundreds of millions of tons of finished and raw materials move in and out of the region’s mills, powerplants and mines, relying on a sophisticated web of infrastructure in, beside, above and below the rivers. The focus of this paper is a particular layer of infrastructure, the locks and dams used to make the rivers navigable, and their role in the integrated energy system and broader commodity flow system.

The region’s river infrastructure dates to the 18th Century when General Braddock’s engineers built fords across the Monongahela River as part of their disastrous campaign to dislodge the French from Fort Duquesne. In 1791 and 1792, the Pennsylvania Assembly passed legislation funding the first formal improvements to the Pennsylvania waterways, includ- ing removing obstructions, stabilizing shorelines, dredging channels, digging connecting canals and portages. Pittsburgh’s first set of locks and dams were completed in the 1840s to improve navigation on the lower Monongahela (Johnson, 1978).

Over the past 170 years, the river infrastructure has been improved and redesigned repeatedly in response to the city’s growth, heavy use, and the needs of advancing transportation technology. Today, USACE’s Pittsburgh District operates nine navigation locks and dams on the Monongahela River, eight on the Allegheny River, and six on the Ohio River (Fig. 1) (USACE, 2010b), in addition to 16 flood control and multi-purpose reservoirs with a combined capacity of over 3.7 billion cubic meters (3 million acre feet) (USACE, 2013d).

The 23 locks and dams in the Pittsburgh District are of varying sizes. The up-river facilities tend to be smaller, single- chamber locks, while the down-river facilities can accommodate much larger barge tows, with the added capacity and backup of auxiliary chambers (see Table 1).

The dimensions of the lock chambers limit the configuration of tows that can pass through the lock. The 34 meter (110 foot) widths of the Ohio locks, as well as the main chamber at Braddock on the Monongahela, allow for tows of 11 meter (35 foot) wide ‘‘jumbo’’ barges that are three barges abreast. Above Braddock, the Elizabeth and Charleroi locks (as well as all of the locks on the Allegheny) can only accommodate single wide configurations of jumbo barges, or double width configura- tions of the smaller ‘‘standard’’ and ‘‘stumbo’’ barges (8 meter (26 feet) wide). Similarly, the length of the lock chamber serves as a capacity constraint; the three largest locks in Table 1 are 366 meters (1200 feet) long, accommodating tows up to four barges deep. Large tows of jumbo barges must be broken into smaller tows or pushed through in stages (double or triple locking). The constriction of the lock capacities is particularly tight at Charleroi where the auxiliary lock has been closed since 2004 due to the ongoing and delayed construction project. Above Charleroi, six more modern up-river locks can accommodate double widths of jumbo barges.

The Corps’ major project to redesign and update the first three lock and dam facilities on the Monongahela, known as the Lower Mon Project, was first authorized in the 1992 Water Resources Development Act (WRDA), calling for a new dam at Braddock, new, larger locks at Charleroi, and Elizabeth’s elimination. In 1995, detailed design and construction planning

Fig. 1. Locks and dams in the Pittsburgh District (USACE, 2010b).

Table 1 Lock dimensions, Pittsburgh District (USACE, 2013d).

River Lock names (kilometers/miles from confluence) Main chamber (meters, feet) Auxiliary chamber

Ohio Hannibal (203/126), Pike Island (135/84), New Cumberland (87/54)

34 m � 366 m, 110’ � 1,200’ 34 m � 183 m, 110’ � 600’

Montgomery (51/32), Dashields (19/12), Emsworth (10/6)

34 m � 183 m, 10’ � 600’ 17 m � 34 m, 56’ � 110’

Monongahela Braddock (18/11) 34 m � 219 m, 110’ � 720’ 17 m � 110 m, 56’ � 360’ Elizabeth (39/24) 17 m � 219 m, 56’ � 720’ 17 m � 110 m, 56’ � 360’ Charleroi (66/41) 17 m � 219 m, 56’ � 720’ 17 m � 110 m, 56’ � 360’ (closed) Maxwell (98/61) 26 m � 219 m, 84’ � 720’ 26 m � 219 m, 84’ � 720’ Grays Landing (132/82) and Point Marion (146/91) 26 m � 219 m, 84’ � 720’ None Morgantown (174/102), Hildebrand (184/108), Opekiska (196/115)

26 m � 183 m, 84’ � 600’ None

Allegheny All eight locks 17 m � 110 m, 56’ � 360’ None

G.S. DiPietro et al. / Transportation Research Part A 69 (2014) 142–164 145

was completed, resulting in a cost estimate of $750M with a 2004 completion date. Work to replace the nearly 100 year old fixed crest dam at Braddock with a gated dam was completed in 2004. Work at Charleroi began in 2002, but budget con- straints are stretching the project well beyond the initial 2004 completion target. Upon completion of the new 26 m � 219 m (840 � 7200) Charleroi locks, the entire lock and dam at Elizabeth will be removed, creating a 48 km (30 mile) pool of unimpeded navigation between Braddock and Charleroi (USACE, 2013b). As of 2012, USACE estimated it will take $1.5B to complete the work by 2024. The May 2014 revision and reauthorization of WRDA (U.S. Congress, 2014) is expected to ameliorate some of the funding roadblocks, although contracting constraints and appropriation concerns remain. Assum- ing funding continues to be slow, completion could be delayed till the 2030s at a higher overall cost.

3.1. Risk of infrastructure failure

The delay in completing the Lower Mon Project exposes the Monongahela River navigation system to extended risk of failure of the antiquated components still in place. The Corps’ 5-year work plan identifies additional maintenance costs of over $3M/year at both Charleroi and Elizabeth to keep these facilities functional until the project is completed. The 2013 Lower Mon Factsheet, summarized in Table 2 lists a number of risks associated with the viability of these facilities (USACE, 2013b).

While each of these risks threatens the viability of navigation, failure of the Elizabeth dam has particularly severe impli- cations. The Dam Safety Action Classification System (DSAC) is used by USACE to ‘‘to provide consistent and systematic

Table 2 Major risks associated with long-term viability of Elizabeth and Charleroi locks (USACE, 2013b).

Locks and Dam 3 – Elizabeth Locks and Dam 4 – Charleroi

Primary lock’s filling and emptying flume is structurally deficient – both roof and walls have severely deteriorated concrete

Only one operational lock chamber (Age – 80 years)

Auxiliary lock’s components are on poor condition and are subject to frequent failure

Downstream guide wall beyond capacity

DSAC I Dam – repair in 2007–2008 expected to last 5–10 years (Age – 105 years)

DSAC II Dam – Stilling basin inadequate

Lock walls founded on timber piles – wall movement Serviceability – wall anchorages (wall armor, corner protection, etc.) Resource constraints of piecemealing construction (batch plant, work area, etc.)

146 G.S. DiPietro et al. / Transportation Research Part A 69 (2014) 142–164

guidelines for appropriate actions to address the dam safety issues and deficiencies of USACE dams’’ (USACE, 2013a). Elizabeth’s Class I DASC status is used for ‘‘urgent and compelling’’ conditions ‘‘where progression toward failure is confirmed to be taking place under normal operations and the dam is almost certain to fail under normal operations within a time frame from immediately to within a few years without intervention; or, the combination of life or economic consequences with probability of failure is extremely high.’’ The emergency repairs conducted on the Elizabeth dam in 2007–2008 are now approaching the end of their expected lifetime. Charleroi’s Class II DASC status, ‘‘high urgency’’ is assigned to dams ‘‘where failure could begin during normal operations or be initiated as the consequence of an event’’.

Failure of the Elizabeth dam would result in loss of pool below Charleroi. It is not unreasonable to assume that the fragile downstream guide wall and lock walls at Charleroi would then collapse. Locks at both Elizabeth and Charleroi would be out of commission as a result, with repairs taking many months to restore navigation. Main gate failures, a somewhat less traumatic type of failure, can take many months to repair, as evidenced by the eight month repair period at the Columbia River’s John Day lock (Grier, 2004). This failure scenario, cascading dam and lock wall collapse, is unlikely, however the known issues with these particular facilities and ongoing aging and deterioration can only increase this probability unless additional piecemeal repairs or completion of the Lower Mon Project occur.

The surge from a sudden and catastrophic dam failure at Elizabeth might hypothetically cause damage to the Braddock facilities, but for the purposes of this analysis, the Braddock lock and dam is expected to hold and remain functional. The Charleroi dam is also assumed to be likely to hold, and USACE likely would be able to secure the upper lock gates at Charleroi to maintain pool above Charleroi. Thus upper Monongahela navigation above Charleroi is assumed to be maintained.

Under this scenario, navigation is assumed to remain possible below Elizabeth in the vicinity of the mouth of the Youghiogheny River (approximately Mile 15), and above Charleroi (Mile 42). (Assuming no damage to the Braddock dam, the lower pool likely could be restored on a priority basis to protect the industrial and municipal water intakes below the failed Elizabeth dam.) The development of the scope of the failure scenario represents an improvement over prior analyses that assume closure of the entire course of a waterway.

As with all modes of infrastructure, there are many ways, large and small, that locks and dams can fail. USACE is positioned to rapidly address routine issues associated with filling flumes, lock gates, runaway barges, etc. Catastrophic failures such as the loss of a dam, however, cannot be remediated quickly, resulting in cascading impacts over the course of the river closure. The scenario is artificially set to reflect failure for all of 2010, a crude estimate of how long it might take to restore navigation. Setting the scenario length to one year of closure allows for a simplified use of the survey data, as well as the assessment of a closure extended beyond those seen to date (Chulkov, 2012). Impacts from a longer closure would be similar to those presented in this paper while remaining reflective of the extensive work that would needed to restore navigation after catastrophic failure. Obviously, a shorter remediation period would minimize impacts, and perhaps allow entities to avoid closure. This potential extended loss of transit is the motivation for this research.

The extended outage scenario examine here has not happened, and hopefully will be avoided. However, the inland waterway has experienced major unscheduled maintenance and disaster outages:

� 52-day closure of the Greenup Locks and Dams on the Ohio, 2003 (USACE, 2003) � 67-day closure of the Mississippi River between the Upper and Lower St. Anthony Falls Lock and Dams due to the collapse

of the I-35 bridge in Minneapolis (re-opened for limited commercial barge traffic after 37 days) (National Transportation Safety Board, 2008) � 22-day closure of Locks 12–25 on the Upper Mississippi River due to flooding (USACE, 2009).

3.2. Pittsburgh Regional Commodity Flow

USACE compiles extensive data on commodity movements on the inland waterways, available through the Waterborne Commerce in the United States website (USACE, various). These statistics are derived from monthly reports submitted by vessel operators and data collected at locks via USACE’s Lock Monitoring Performance System (LPMS). WCUS shows that

Fig. 2. Barge tonnage through the Port of Pittsburgh over time (U.S. Army Corps of Engineers, 2013c).

Table 3 Port of Pittsburgh commodities barged, 2011 (USACE, 2013e).

Metric tons (million) % of total

Total coal, lignite and coal coke 23.1 75.4 Total crude materials, inedible except fuels 4.71 15.3 Total primary manufactured goods 1.14 3.7 Total petroleum and petroleum products 1.10 3.6 Total chemicals and related products 0.59 1.9 Total food and farm products 0.008 <0.1 Total all manufactured equipment, machinery 0.003 <0.1 All commodities 30.7

G.S. DiPietro et al. / Transportation Research Part A 69 (2014) 142–164 147

the inland waterways system processed 2,200 million metric tons (2,500 million tons) nationwide in 2012. The Port of Pittsburgh, encompassed by the Pittsburgh District, is the 17th busiest port in the Nation, handling 31.9 million metric tons (35.2 million tons) in 2012 (USACE, 2014b). Unlike the coastal ports, the inland Port of Pittsburgh does not handle containers or foreign shipments (USACE, 2010c, 2013e). The Port is defined by the 200 miles of navigable rivers across a 12 county area, and includes about 200 river terminals and barge industry service suppliers (Port of Pittsburgh Commission, 2014). Fig. 2 shows the trends in commodity tonnage handled on Pittsburgh’s rivers since 1996. The downward trend, stabilized over the past four years, indicates that the rivers are capable of carrying a significantly higher volume of commodities.

The predominant product shipped through the Port of Pittsburgh is coal, accounting for over 75 percent of the materials handled on the rivers in 2011 (USACE, 2013e). Table 3 provides an overview of the major commodity categories tracked by USACE1 (a full breakdown is provided in the Supplementary material). The second largest category, ‘‘Crude Materials, Inedible Except Fuels,’’ includes 1.8 million tons of sand and gravel, 1.5 million tons of limestone, and 0.4 million tons of gypsum. Primary iron and steel products account for 0.9 million tons barged.

The Port of Pittsburgh is predominantly a regional port. Within the Ohio River System (ORS), which encompasses USACE’s Pittsburgh, Huntington, and Louisville Districts, 80 percent of the shipments start and end within the region (USACE, various).

Within the Port of Pittsburgh, the locks at Braddock, Elizabeth, and Charleroi handled 13.7, 10.8, and 10.2 million metric tons (15.1, 11.9, and 11.2 million tons) of commodities in 2012, respectively (see Supplementary material for additional details) (USACE, 2014a). Focusing on Braddock, coal and coke represent 81 percent of the commodities locked in 2012, show- ing a higher prevalence of coal shipments than seen across the broader Port. Fig. 3 shows the distribution of commodities over the past two decades that have been locked through the Elizabeth facility. The pattern is similar for Braddock and Char- leroi (see the Supplemental materials for additional figures) (USACE, 2013e). The coal mines of southwestern Pennsylvania and northwestern West Virginia use docks on the upper Monongahela to reach coal-fired power utilities and coking plants. There are a number of power generation and coke production plants located on the rivers of the Pittsburgh District that uti- lize coal barged from both local and distant mines. Given the predominance of coal shipments, the primary focus of this anal- ysis is the movement of thermal coal from coal mines to coal-fired utilities.

Inland waterways shipment data are also presented in the context of the national freight network from the U.S. Depart- ment of Transportation (DOT) Bureau of Transportation Statistics’ (BTS) Commodity Flow Survey (CFS) (U.S. DOT Bureau of Transportation Statistics, 2007) and Oak Ridge National Laboratory’s Freight Analysis Framework (FAF) (Oak Ridge National

1 Note: the Corps uses its own unique commodity codes, the Waterborne Commerce Statistical Codes, which are closely tied to the Harmonized System of commodity codes used by the World Customs Organization, with a focus on commodities likely to be transported on the waterways (USACE, various) (TRB, 2013). The WCSC do not correlate directly to the U.S. Department of Transportation’s commonly used Standard Classification of Transported Goods (SCTG).

Fig. 3. Commodity tonnage barged through Elizabeth Lock and Dam, Monongahela River (U.S. Army Corps of Engineers, 2013).

148 G.S. DiPietro et al. / Transportation Research Part A 69 (2014) 142–164

Laboratory, n.d). The CFS is a sample survey of shippers conducted every five years, designed to serve as a primary source of national and state-level data on domestic freight shipments. The FAF is a tool which compiles a variety of shipping, economic and geographic data for the analysis of commodity movements (including the CFS). Upon examination, the WCUS and LPMS provide much more detailed and complete information on inland waterways shipments. For example, the WCUS shows 2.3 million tons shipped from Ohio to Pennsylvania in 2007 (USACE, 2011a). The FAF only reports 0.001 million tons shipped by water for the same origin and destination (O/D). The CFS, while reporting similar overall tonnage as the FAF, reports no water shipments from Ohio to Pennsylvania. This gap is a good example of the CFS’s inability to provide consistent coverage outside of a ‘‘small number of rather large geographic regions’’ (Cambridge Systematics Inc., 2008; Ludlow, 2011). See Supplemental materials for additional information.

4. Identifying at-risk coal shipments

A river O/D table was developed linking barged coal transactions in 2010 between 211 U.S. mines and 94 U.S. power plants, giving tonnage barged and total tonnage shipped from the mines to each utility. The River O/D table was derived from the Department of Energy’s (DOE’s) Energy Information Administration (EIA) Survey 923 (EIA-923), which includes data about the amount and source of fuel purchased by each utility, and primary and secondary modes of transportation used to move the fuel to the utility; this data set includes coal transactions across all modes of transportation for over 750 mines and 480 utilities in 2010, as well as other transactions for natural gas, diesel fuel oil and other fuel types (EIA, various).

A regional model was developed from the River O/D table to assess the potential impact on the power plants and mines of infrastructure failure at Elizabeth and Charleroi. The River O/D table was combined with other public sources to create a model of thermal coal movements between mines and power plants. Latitude and longitude coordinates for each mine and utility were extracted from data maintained by the Environmental Protection Agency (EPA) (U.S. Environmental Protection Agency, 2009) and the Mine Safety and Health Administration (MSHA) (U.S. Mine Safety and Health Administration, 2013). U.S. Army Corps data was also extracted for the mile number of docks controlled by the utilities and mines (USACE, 2010b).

A key assumption in the model development was the assignment of docks and river miles to the coal mines. Some mines actually own and operate their own docks, which were readily identifiable in USACE dock dataset (e.g., the Cumberland Mine dock at Mile 81.5 on the Monongahela). Other mines are controlled by operating companies that have river loading capacity that is used by multiple mines (e.g., CONSOL’s Alicia transloading facility at Mile 58 on the Monongahela). Some mine oper- ators identify their river loading facilities locations on their websites. One mine operator responded to an email request for clarification regarding the loading docks used for two of their strip mines. Less certain mine river loading assignments were made after examining satellite images for likely loading operations (Google, 2013). Where clear determinations could not be made from these methods, docks assigned in the Velocity Suites (Coal Transaction Analyst) software were used (Ventyx, 2013). The Velocity Suites source was found to be inaccurate for some facilities, predicting rail deliveries for mines and power plants known to have no connection to the national rail system. As a result, this source was used cautiously.

Starting with the four Monongahela River power plants, linkages to each of the mines that supplied these power plants with coal in 2010 were characterized in terms of their viability after failure at Elizabeth/Charleroi. Similarly, linkages to the 15 coal mines that barge their product from the Monongahela were coded, with transits that pass through the failure zone coded as not viable (these linkages included 16 power plants located on the Ohio or further downstream). A final layer of connections was then added to code shipments to the non-Monongahela River power plants from other mines that had not otherwise been included in the model. The resultant regional model thus included 36 power plants linked to 129 mines, with direct impacts predicted at the subset of facilities summarized in Table 4. Fig. 4 shows the distribution of most of these

Table 4 Profile of directly impacted facilities.

Mines Count Total coal production (million tons)

Above failure zone 15 43 Within failure zone 0 0 Below failure zone 11 270 Subtotal 26 320

Power plants Count Total electricity generation capacity (MW)

Above failure zone 2 2,900 Within failure zone 2 880 Below failure zone 16 21,000 Subtotal 20 24,000

G.S. DiPietro et al. / Transportation Research Part A 69 (2014) 142–164 149

facilities, omitting the Powder River Basin (PRB) mines and marginally connected power plants located further down the Ohio and Mississippi River.

The resulting model links power plants and mines, allowing for an assessment of the importance of reliance on barge movements on the Monongahela, as a function of tonnage barged, location of mine and utility docks, and documented availability of alternative transportation modes.

Fig. 5 depicts the impacted mines and power plants in terms of the expected viability of their coal shipments and receipts, as well as their geographic placement with respect to the failure zone. 6.4 million metric tons (7.1 million tons) of coal shipments/receipts would have been displaced by the Elizabeth/Charleroi failure scenario. These displacements represent the direct impacts of river closure. As will be discussed further below, additional secondary impacts, including subsequent mine and power plant closures and losses to towboat and barge owners, are of interest.

This paper focuses on the power plants and coal mines potentially impacted by the Monongahela River failure scenario, and the 6.4 million metric tons (7 million tons) of thermal coal barge shipments that would not be viable in a year-long river outage. The impacts of failure at Charleroi and Elizabeth, however, would also extend to several major steel and coke-making facilities located on the Monongahela, as well as the metallurgical coal mines that barge to these facilities. As shown in Fig. 3, overall shipments through the Elizabeth lock in 2010 totaled 10.2 million metric tons (11.3 million tons), 8.0 million metric tons (8.7 million tons) of which were coal. The 1.6 million metric tons not accounted for by the 6.4 million tons of thermal coal tracked in the EIA 923 survey is likely to be metallurgical coal used for coking and other industrial applications (EIA, 2010). The Port of Pittsburgh Commission identifies ten river terminals on the Monongahela, four of which are located within the failure zone and thus would be directly impacted by a river outage (Port of Pittsburgh Commission, 2014). There are tow- boat operators and barge carriers who would also be directly impacted (five major docks within the impact zone), as well as a variety of other businesses that service the region’s river freight industry or rely on the river pool provided by the Elizabeth

Fig. 4. Potentially impacted power plants and coal mines using barge coal delivery.

Fig. 5. Viability of thermal coal mine and power plant transits, 2010.

150 G.S. DiPietro et al. / Transportation Research Part A 69 (2014) 142–164

dam. Examination of USACE’s navigation maps shows three municipal water intakes and seven industrial water intakes within the failure zone, as well as 13 private marinas, six municipal docks and two state fish commission docks (USACE, 2004). See the Supplemental materials for additional details. Other potential impacts not quantified here include recreational boating, movement of lime and limestone to upriver utilities, as well as the shipment of coal combustion residuals to disposal sites. Kruse et al. (2007, amended 2009) discusses a wide range of potential impacts.

5. Potential responses to infrastructure failure

The impact of an extended outage of navigation on the Monongahela River would have a profound impact on the industries that rely on barge transportation. The range of potential responses of these entities to an extended river outage differs as a function of their accessibility to the coal-to-utility market. Assessing disruptions to the inland waterway has some parallels to assessments of roadway disruptions in that the mines may respond in a variety of ways (Jenelius and Mattsson, 2012): some may be completely thwarted, some may be partially blocked, and others scarcely impacted at all. In road networks, the possible responses to a disruption have been summarized as: (1) cancelling the trip, (2) postponing the trip, (3) choosing another destination, (4) choosing another mode, (5) choosing another route (Nicholson and Dalziell, 2003). This type of framework has been used in a combined travel demand model (CTDM) to estimate long-term equilibrium network conditions due to network disruptions (Chen et al., 2007). USACE’s Navigation Economic Technologies Program (NETS) has conducted a number of surveys and interviews of shippers in the wake of several extended lock closures to assess the impacts of and response functions to closures (USACE, n.d). An extensive interview with a utility impacted by the 2003 closure of the Greenup lock identified a tiered priorities in the face of a river closure: stockpile as much as possible (assuming suppliers can step up production) and wait until closure is over (relevant for planned closures), divert to alternative modes (assuming excess capacity exists), shift coal sources to avoid a closed facility; close plants that cannot receive coal and re-dispatch remaining plants or purchase power off the grid (USACE, 2003). In response to a catastrophic failure of river infrastructure, a barging mine may consider these same responses, as discussed further below, and then applied to the facilities potentially impacted by a closure at Charleroi and Elizabeth.

5.1. Shift transport mode from barge to rail

Some prior studies have made the simplifying assumption that all disrupted barge shipments could be shifted to rail (MacKenzie et al., 2012; Kruse et al., 2007, amended 2009). In the specialized coal-to-utility network, however, this is not always possible. In a study focused on the Illinois waterway, Folga noted that ‘‘shifting to an alternative mode may not be feasible in the short-term because of the characteristics of the cargo or the physical absence of a practical alternative’’ (Folga et al., 2009).

A number of mines and power plants do not have any access to the national rail system. Bray discusses the operating lim- itations of power plants, ‘‘utility companies are set up to operate as they are now because of vested infrastructure, and to orient their businesses more to rail transportation would be very expensive’’ (Bray et al., 2004). Drawing from shipper sur- veys, Bray goes on to describe specific power plant limitations: no access to rail, dependence on bridges with limited weight capacity, favorable barge back haul rates being unavailable for rail. These limitations can be seen in the power plants of inter- est in this study. Fig. 6 depicts the Fort Martin power plant that relies entirely on barge deliveries of its coal (Mapquest, 2014). The map view shows that a railroad bounds the plant site, but provides no service to the site (as evidenced by the lack of a spur). The response scenarios for the barge-only mines and power plants cannot include a shift-to-rail option.

Fig. 6. Satellite and map images of a barge-only power plant.

G.S. DiPietro et al. / Transportation Research Part A 69 (2014) 142–164 151

Several researchers have used USACE’s dock database (USACE, 2010b) to determine whether rail access exists at a site (Campo et al., 2012; Folga et al., 2009). The Corps’ data is primarily derived from on-site surveys which are no longer being conducted (the last survey covered the Houma, LA region, conducted in 2008) (USACE, 2014c). For example, the data set indi- cates that Fort Martin, depicted above in Fig. 6, has ‘‘one surface track serving plant at rear; connects with Norfolk Southern Railway.’’ This no longer appears to be the case.

5.2. Rely more on rail (when available)

Some of the impacted power plants have both rail and barge access. Campo described five physical and operational factors that limit transits that rely on both rail and river: lack of terminals and suitable locations, lack of transshipment capability, excess terminal capacity, limited railroad access to transshipment terminals, and environmental laws (Campo et al., 2012). Wang describes the capacity of transloading facilities as a limitation as well (Wang et al., 2006). Discussions with power plant operators revealed that rail access does not guarantee sufficient rail off-loading capacity to run operations totally from rail deliveries. For example, the geographically constrained riverside location of these plants may allow for a rail spur, but not a rail loop. Thus such a site can accommodate one coal train on site which must be fully off loaded and then backed out onto the main line before another train can be admitted to the plant. If the plant also lacks access to sufficient siding to allow for staging of the ‘‘next’’ train, the logistical challenges of keeping the plant fully fueled increase. Further, an extended outage would cause regional ripple effects, with unexpected increased demand for rail creating shortages (USACE, 2003).

5.3. Shift to truck

Prior analyses have sometimes assumed that truck transport is used if rail or barge are not available (Folga et al., 2009). This may be appropriate for limited shipments (time or volume), however, the logistics associated with a mode shift from barge to truck are staggering and generally not likely to be feasible for a long-term displacement. A single ‘‘stumbo’’ barge holds 1,085 tons (American Waterways Operators, 2006), while a highway tractor-trailer truck’s capacity is 25 tons. As an example, the Hatfield’s Ferry power plant burned 4 million tons of coal in 2010. Supplying this volume by truck instead of barge would require offloading a new truck every three minutes, 24 h per day, 365 days per year to deliver this volume of coal. Thus truck delivery is highly infeasible. The feasibility of this option is even less feasible given that it is unlikely that (1) the plants and mines have the requisite truck off-loading and loading equipment, (2) sufficient excess truck capabilities exist to step into a sudden need, and (3) the roads and bridges between the mines and power plants have the requisite load- bearing ratings and capacity to bear this load.

5.4. Transload and rail around failure zone

Mines that barge through the failure zone may have the option to use rail to avoid the failure zone. Mines that currently use a combination of rail and barge to reach the barge-only power plants might be able to re-route to a transloading dock above the failure zone. Barge-only mines could only physically reach the barge-only power plants by adding transloading to rail around the failure zone, rail past the failure zone, then transloading back to barge for delivery to the plants. The costs associated with these additional transfers and rail shipment are significant and may make this alternative infeasible. Limitations on increased use of these segments of rail might include the transloading capacity, equipment availability, and track ratings (Bray et al., 2004). Transloading from rail to barge costs $1.50/ton and rail switching to short line rails if needed can cost $2 per ton (U.S. Environmental Protection Agency, 2013). The cost of rail itself adds about $15/ton for

152 G.S. DiPietro et al. / Transportation Research Part A 69 (2014) 142–164

shipments within Pennsylvania. For perspective, the average total cost of coal for power plants with impacted barge transits is $57 per ton (see section 7), transloading around the failure zone thus might increase coal costs by a third.

5.5. Plant closure

Having acknowledged the significance of the mode shift constraints, what alternatives remain? If a power plant must receive all or most of its coal by barge, what will happen if barge deliveries are disrupted for an extended (say, year-long) period of time? One alternative not commonly considered in prior studies is the possibility that a facility’s accessibility to its market is so tailored to its geography that it is unable to adapt to a catastrophic constriction of its key infrastructure. Such a plant thus would no longer be a functional participant in the coal-to-utility network and conceivably would face closure.

In 2007, the I-35W bridge over the Mississippi River collapsed. A study of traffic reactions to this catastrophic failure showed that one third of previous trips using the bridge disappeared (Zhu et al., 2010). The study authors postulated that these trip reductions may have been due to changes in destination and trip consolidation, and to a lesser extent changes in mode. We would add to this list the possibility that entities that had trips that were completely tied to a transit across the bridge might close in response to an inability to make that critical transit.

The inability to adapt to catastrophic infrastructure losses is a function of many factors, one of which is proximity to the failure zone. Barge-only facilities within the failure zone would have extremely limited options. Facilities outside the failure zone that partially rely on shipments through the failure zone have many more options.

Another key adaptability factor is the resource base of the facility itself. We postulate here that facilities at the end of their useful lives (as evidenced by low capacity factors, dropping production rates, or closures since 2010 in the absence of infrastructure failures) will simply close sooner rather than expend additional resources needed to cope with a constrained transportation network. We suggest a simple approach to quantifying this loss (aka unmet demand) in Section 6.

(For barge-only facilities with significant remaining useful lives, their approach might be to run down their stockpiles and/or shift to lower levels of production until navigation is restored.)

5.6. Assessing mode shift opportunities

Mode shift opportunities have been modeled in a variety of ways, simplistic and sophisticated. The simplest models assume that shifts will occur when the costs of the alternative mode become lower than the current mode. Bray used the results of an extensive shipper survey to explore the maximum willingness to pay for shipping in the context of mode shifting, concluding that ‘‘[t]ransportation decisions are often highly integrated with choices regarding other inputs. They are also highly integrated with other supply chain costs such as loading and unloading or storage costs.’’ (Bray et al., 2004).

The potential responses (mode shift, closure, supplier shift, etc.) of the power plants and thermal coal mines to failure at Elizabeth and Charleroi is deduced largely from data extracted from the EIA-923 survey (EIA, various), with additional information gleaned from satellite images (Google, 2013; Mapquest, 2014). The EIA-923 survey provides a valuable tool for determining the accessibility of specific mines and power plants to the coal-to-utility network. Power plants report the primary and secondary transportation methods used to move coal from mine to power plant. The primary mode is defined to be ‘‘the method used to transport the fuel over the longest distance from point of origin [the mine] to consumer.’’ The secondary mode is reported when ‘‘more than one method of transportation is used in a single shipment’’ for the ‘‘second longest method used to transport the fuel to consumer.’’ Additional modes are not reported (EIA, 2013b). Nine modes of transport are used in the survey. The top eight combinations of these modes (out of 29 reported combinations) are presented in Table 5. 90 percent of all shipments are limited to single mode of transportation, with about 10 percent being shipped by ‘‘river-only’’.

Table 5 also shows the reported transport codes for the four largest power plants in the immediate study area. These results show the full dependence of the Mon power plants (Fort Martin and Hatfield’s Ferry) on barge transport, and the more varied modes used by the two Ohio power plants (Bruce Mansfield and W.H. Sammis). Thus the EIA-923 provides a useful tool to assess the accessibility of the power plants (as well as the mines) to a range of transportation modes. Additional confirmation can be derived by examining additional years of EIA-923 data, and by examination of satellite images (see Fig. 6).

Note that the FAF and EIA both show a significant use of trucks for coal within Pennsylvania. The EIA-923 allows for the close examination of the specific utilities that rely heavily on truck shipments, most of which are not located near the inland waterways (New Castle, Homer City, Seward).

5.7. Change supply chain

Analysis of the EIA data over the 2009–2012 period shows that power plants have some flexibility in their access to coal sources. The potentially impacted power plants each generally rely on a limited set of mines from year to year for the majority of their fuel, but also access a wide number of additional and different mines in any given year for the balance of their fuel.

Table 5 Major transport modes reported for shipment of coal to power plants, industry-wide and for selected plants in 2010 (EIA, various).

Transport mode Codes Metric tons (millions) % of total Hatfield’s Ferry Fort Martin Bruce Mansfield W.H. Sammis

Railroad only RR 507 62.7 0.5 2.5 Truck only TR 84.5 10.5 0.00009 0.04 River only RV 79.3 9.8 3.5 2.5 5.2 1.3 Tramway or conveyer only TC 47.1 5.8 Railroad/ river RR RV 28.7 3.6 0.03

RV RR 1.41 0.2 0.05 Railroad/ truck RR TR 26.7 3.3

TR RR 1.48 0.2

G.S. DiPietro et al. / Transportation Research Part A 69 (2014) 142–164 153

� Over the course of four years, all of the power plants used at least 10 unique mines, with an average of 21 unique mines (see Table 6). � Each power plant had a smaller subset of base mines that they used at least three of the four years, averaging 27 percent

of the total number of mines they used over the four years. These ‘‘go-to’’ mines provided 73 percent of the coal consumed by these power plants over the four years.

Thus the power plants and mines maintain a complex and shifting web of transactions. The ability of the power plants to supplement their base coal contracts (their go-to mines) with a variety of non-base mines appears to provide resilience and flexibility to their logistics chains, with the non-base mines serving as a spot market. The expectation then is that the loss of access to certain mines as a result of the failure scenario of concern would not be devastating to the power plants in terms of availability to coal and continued access to the coal ‘‘spot market’’. This conclusion coincides with the overall conclusions of the Corp’s study of the impact of closing the Monongahela to barged steam coal shipments (USACE, 2011b).

5.8. Impact and responses for powerplants and mines

The power plants and mines assessed fall into five general classes (see Table 4). The range of potential impacts and failure scenario responses for these five categories are discussed briefly below and in more detail in the Supplementary material.

� Power plants on the Monongahela River above the failure zone � Power plants within the Monongahela River failure zone � Non-Monongahela River power plants burning coal sourced from Monongahela River loading coal mines � Monongahela River loading coal mines � Non-Monongahela coal mines fueling Monongahela River power plants.

Table 7 summarizes the potential responses of the impacted mines and power plants to catastrophic river closure at Elizabeth and Charleroi. Facilities within or above the failure zone are much more likely to bear significant impacts than

Table 6 Impacted power plant and coal mine linkages, 2009–2012.

Total tons (million) of coal used 2009–2012

Number of unique mines used 2009–2012

Percentage of mines used 3–4 times in 4 years

Percentage of coal from mines used 3–4 times in 4 years

Bruce Mansfield 26.5 19 42% 89% Cumberland 24.2 20 20% 79% J. M. Stuart 23.9 15 40% 89% Clifty Creek 19.1 11 27% 56% Hatfield’s Ferry 15.9 26 27% 75% W. H. Sammis 15.5 36 14% 65% Miami Fort 13.9 39 36% 92% Mountaineer 13.7 38 16% 69% W. H. Zimmer 11.9 30 33% 83% Fort Martin 9.78 26 42% 81% Muskingum River 8.81 29 17% 59% East Bend 7.53 26 27% 71% Killen Station 6.86 15 33% 83% Cheswick 4.06 11 18% 79% Tanners Creek 3.99 10 30% 41% Mitchell 1.68 26 12% 55% Average 11.0 21 27% 73%

Table 7 Summary of potential responses of power plants and mines to catastrophic river closure.

Number Dependency on barging

Level of direct disruption Response options

Power plants above failure zone 2 large Very high 30–35% of 2010 coal deliveries would not have been viable

� Replace with local coal � Re-route around failure zone � De-rate � Closure

Power plants within failure zone 2 small Very high 100 � Closure

Power plants below failure zone burning Mon. loaded coal

3 small High >50 � Replace with viable barged coal � Increase rail shipments � Closure

6 medium to large

High <50%, but > 200,000 tons � Replace with viable barged coal � Increase rail shipments

7 Low to moderate

<30%,<150,000 tons � Replace with viable barged coal � Increase rail shipments

Mon. loading mines 1 large High >60% � Increase barge shipments to local power plants � Re-route around failure zone � Closure

2 small High Low � Continue barge shipments to local power plants

2 strip mines

High Moderate � Truck to alternate docks

4 marginal

Low to high None to low � No change for most shipments � Increase rail and truck shipments

2 Low to moderate

None � No change

4 Low to moderate

<10% � No change for most shipments � Shift to rail � Shift to non-impacted power plants

Non-Mon mines fueling Mon power plants

4 PRB Low to moderate

<1% � Insignificant impact

2 Moderate to high

<1% � Insignificant impact

3 High 15–40% � Shift to non-impacted power plants � Closure

2 Moderate to high

Moderate � Shift to rail � Shift to non-impacted power plants

154 G.S. DiPietro et al. / Transportation Research Part A 69 (2014) 142–164

the downriver mines and power plants. Facilities with barge-only access to the supply chain are also much more susceptible to significant impacts. Models that incorrectly assume all facilities can shift to rail shipments significantly underestimate the significance of infrastructure failure to these facilities. The response options listed are subjective projections of the authors; the actual options available to the facilities may be different given the full range of factors that determine a site’s viability and transportation alternatives at any point in time.

6. Market-wide response to potential failure

A probable response to a prolonged outage due to infrastructure failure is the emergence of reorganized coal-to-utility markets above and below the failure zone. A supply and demand analysis was conducted to determine whether the regional market could feasibly re-organize into separate markets above and below the failure zone, while recognizing that many of the mines above the potential failure zone will continue to be able to reach power plants below the failure zone by rail, truck, or alternate docks. The new market above the failure zone would be significantly smaller and more constrained than the new market below the failure zone.

The specific question of interest is whether sufficient supply and demand for coal remains after river failure and subse- quent plant shut downs to ‘‘make’’ new markets. The key assumption driving this analysis is that extended river closure at Charleroi/Elizabeth will be a sufficient disruption to some power plants and mines that they will stop operating, and thus be removed from the coal-to-utility network. Given the interdependency in this system, we explored the potential ripple effects of such postulated closures.

To address this market stability question, the five grouping of mines and power plants discussed above were character- ized in terms of (1) total tonnage shipped or received by all modes of transportation, (2) barged tonnage that would have been displaced by a river outage, and (3) the likelihood of closure in response to the outage (see Table 7). Four scenarios were considered:

G.S. DiPietro et al. / Transportation Research Part A 69 (2014) 142–164 155

� Scenario 1 assumes the direct closures of six power plants, and two mines thought to be ‘‘marginal’’ and likely to close directly in response to the catastrophic loss of key river infrastructure (see prior section’s discussion of Plant Closure). These closures include the two power plants within the failure zone (Elrama, Mitchell) and other facilities that have subsequently closed (AES Beaver Valley, R.E. Burger, Willow Island, and Tanner’s Creek; Yellowbush Mine and Mine 84) � Scenario 2 assumes the baseline direct closures plus the closure of the Hatfield’s Ferry power plant. First Energy has

announced the closure of this facility in late 2013 (First Energy, 2013). Thus closure in response to the additional complications of operating in a constrained river delivery system seems reasonable. � Scenario 3 assumes that the Cumberland Mine closes in addition to the baseline closures. This Monongahela-loading mine

barges all of its output to its customers. Loss of access to its customers below the failure zone, combined with increased competition for contracts with the remaining Monongahela power plants, could conceivably change the financial viability of this mine. � Scenario 4 reflects a cascading scenario, where both Hatfield’s Ferry and the Cumberland Mine close, as well as two other

barge-only Monongahela mines (Crawdad #1 and Prime #1) and the baseline facilities.

For each scenario, the projected excess coal supply from the remaining mines and lost supply from the remaining power plants was calculated and compared for the reorganized markets above and below the failure zone.

A supply and demand analysis of these impacted mines is presented in Table 8. As described earlier, 6.4 million metric tons (7 million tons) of coal shipments from mines to power plants would not have been viable in 2010 as a direct result of closure of the Monongahela at the failure zone. The impacts would have been particularly acute for both of the power plants and the large mine located above the failure zone that are primarily dependent on barge transport and would essen- tially be ‘‘trapped’’ by river closure; the impacted mines are identified as a discrete ‘‘barge only’’ subcategory in Table 8. For the purposes of this analysis, the question is whether the re-organized coal-to-utility markets above and below the failure zone could absorb the portion of the 6.4 million metric tons trapped within those markets.

In addition to the no-longer-viable shipments, Table 8 shows that 0.65 million metric tons of coal would have been with- drawn from the market due to the assumed mine closures, and 3.1 million metric tons of demanded coal would have been canceled due to the assumed power plant closures. This withdrawal of supply and demand would impact linkages between power plants and mines that otherwise would have been viable despite a river closure; the final column in Table 8 describes these secondary impacts (0.6 million metric tons of secondary lost sales to mines, and 0.33 million metric tons of secondary lost deliveries to power plants).

In the scramble to remain operating in the aftermath of a catastrophic and extended river closure, it is reasonable to assume that the mines and power plants would attempt to form new markets to reflect the new challenge of moving coal cost effectively, particularly above the failure zone. Several power plants and the ‘‘barge only’’ mines lack the near-term ability to shift coal transport to rail to circumvent the failure zone; these facilities would need to find a way to continue to ship/receive coal by barge or face closure. Table 9 summarizes the new markets that could form in response to an extended river closure, incorporating the supply/demand losses of the facilities above and below the failure zone. Negative net supplies indicate a shortfall within the new markets. This analysis assumes full fungibility of the coal within this mine-to-utility network, which ignores the power plants’ needs to balance their fuel mixes to match their boiler and envi- ronmental permit constraints. Consideration of these constraints would further complicate the ability of the new markets to adapt to the loss of navigation through Elizabeth and Charleroi. Discussion of these results is provided below.

In Scenario 1, where the marginal power plants and mines with high dependency on barge traffic through the failure zone close, the market above the failure zone has the potential to re-organize adequately to keep the two power plants operating. In other words, the 2.0 million metric tons of unmet demand at those power plants created by the inability of Ohio River coal mines to barge up through the failure zone could be offset by the 4.2 million metric tons of excess supply available at the

Table 8 Viability and closures in response to failure (million metric tons).

Impacted mines and power plants Direct closures Impact from closed but viable links (barged)

Total (all modes)

Total barged Barged, not viable

Total (all modes)

Barged, not viable

Mines: Lost demand Above failure zone (with access to rail) 28 2.2 1.2 0.012 0.012 0.00 Above failure zone (barge only) 6.8 6.6 3.0 0.0 0.0 0.00 Below failure zone 248 42 2.2 0.64 0.11 0.60 Subtotal 283 51 6.4 0.65 0.12 0.60 Power plants: Lost supply Above failure zone 6.15 6.09 2.0 0.00 0.00 0.00 Within failure zone 0.72 0.56 0.56 0.72 0.56 0.00 Below failure zone 43.1 33.1 3.8 2.4 0.77 0.33 Subtotal 50 40 6.4 3.12 1.32 0.33

Table 9 Scenario analysis (million metric tons barged).

Scenarios

1. Baseline closures

2. Baseline + Hatfield’s Ferry closures

3. Baseline + Cumberland closures

4. Baseline + Hatfield’s Ferry + barge-only Mon mine closures

Facilities above failure zone Power plant lost supply �2.03 �0.75 �3.24 �0.80 Mine (with rail access) excess barge supply 1.16 1.33 1.16 1.33 Mine (barge only) excess supply 3.03 5.18 0.05 0.14

Net supply above failure zone 2.16 5.75 �2.03 0.67

Facilities below failure zone Power plant lost supply �3.39 �3.39 �3.39 �3.39 Mine (with rail access) excess barge supply 1.13 1.13 1.13 1.13 Mine (barge only) excess supply 1.57 1.57 1.57 1.57

Net supply �0.68 �0.68 �0.68 �0.68

Net system wide coal 1.48 5.06 �2.71 �0.018

156 G.S. DiPietro et al. / Transportation Research Part A 69 (2014) 142–164

Monongahela mines that could not be barged down through the failure zone. Key observations of the Scenario 1 results include:

� Marginal power plants and mines close (1.9 million metric tons of barged demand and 0.65 million metric tons of barged supply, respectively). � Remaining Monongahela power plants: The lost barged supply (2.0 million metric tons) for these power plants could be

fully powered by the displaced barged coal available above the failure zone (4.2 million metric tons). � Barge-only Monongahela mines: The viability of Scenario 1 is dependent on whether these barge-only mines (with 3.0

million metric tons of displaced coal) can secure contracts for a significant amount of that unmet 2.0 million metric tons of barged demand with the remaining Monongahela power plants and absorb the remainder by reducing output. It is unli- kely these mines can cost-effectively transload the remainder of their excess supply from barge to rail to compete against those Monongahela mines with established rail access for contracts to power plants below the failure zone. The outlook for these mines under Scenario 1 is not encouraging. � Monongahela mines with rail access: These mines, with 1.1 million metric tons of displaced barged coal, are competitors

of the barge-only mines, and would likely compete to supply the lost barged supply for the Monongahela power plants, as well as for the 0.6 million metric ton projected shortfall of supply to power plants below the failure zone. With their established rail and barge alternatives, they are expected to have a cost advantage over the barge-only mines. � Below the failure zone: A slight shortfall in supply (�0.68 million metric tons) would result in the market below

Elizabeth; this shortfall could be readily addressed by excess supply from those Monongahela mines with rail access, or slight increased production below the failure zone. The impact below the failure zone is unchanged for the remaining three scenarios (which remove additional facilities only from above the failure zone). � Overall, more demand is removed from the combined markets than supply as a result of river closure, resulting in an

excess supply of 1.5 million metric tons after if the market reorganizes successfully without additional closures. Much of the excess supply, however, is from barge-only Monongahela mines that would be ‘‘trapped’’ above the failure zone. While this represents less than one percent of the total volume of coal produced by the impacted mines, if the trapped Monongahela barge-only mines are unsuccessful in joining the re-organized markets, their closure may occur, as modeled in Scenario 3 and 4.

Scenario 2 considers the added impact of closure of the Hatfield’s Ferry power plant, which in fact will have already occurred by the time of publication of this paper. As shown in Table 9, closing Hatfield’s Ferry in response to the river failure would have resulted in an excess of 5.7 million metric tons of coal that otherwise would have been barged through the region. Additional Scenario 2 observations:

� Marginal power plants and mines close. � Remaining Monongahela power plant: Fort Martin could be fully powered by the remaining barged coal available above

the failure zone. � Barge-only Monongahela mines: Closure of Hatfield’s Ferry would have a significant additional impact on these ‘‘trapped’’

mines, increasing their surplus barged coal 70 percent to 5.2 million metric tons; 25 percent of Cumberland Mine and half of the Crawdad and Prime No. 1 mines’ customer base would have been lost in this scenario. While transloading to rail to reach customers below the failure zone would be physically possible, it may not be cost-effective for these mines. Per- manent or temporary closure is even more likely under this scenario than in Scenario 1.

G.S. DiPietro et al. / Transportation Research Part A 69 (2014) 142–164 157

� Monongahela mines with rail access: These 11 mines can be expected to cost-effectively shift their excess supply from the river to rail (or truck). Their excess supply would only increase by 0.2 million metric tons by closure of Hatfield’s Ferry. � System wide, closure of Hatfield’s Ferry would create an excess supply of 5.1 million metric tons of coal. This glut would

likely cause additional mine closures in the constrained market above the failure zone.

Scenario 3 considers the impact of closure of the Cumberland Mine, in addition to the baseline closures. This large mine shipped coal over 4.8 million metric tons of coal in 2010 to the 13 power plants below the failure zone and the two power plants on the Monongahela. Impacts include:

� Marginal power plants and mines close. � Remaining Monongahela power plants and mines: These power plants would need to find an additional 2.0 million metric

tons of barged coal to remain fully powered. It is reasonable to assume that the remaining Monongahela mines (with total production of 34 million metric tons in 2010) would step up to this unmet demand with a 6 percent increase in production. � System wide, closure of the Cumberland Mine would result in an additional 4.2 million metric ton barged shortage over

the baseline scenario.

Scenario 4 represents a cascading series of closure in which the Hatfield’s Ferry power plant (as in fact has already occurred without a river outage) and the three barge-only Monongahela-loading mines then close (Cumberland Mine, Crawdad No. 1, and Prime No. 1):

� Marginal power plants and mines close. � Remaining Monongahela power plant: Fort Martin’s 0.80 million metric ton shortfall can be met fully by the surplus

barged coal from the remaining mines, leaving an excess of 0.6 million metric tons of coal. � Barge-only Monongahela mine: One remaining barge-only mine (Washington County Strips) would be competing to meet

Fort Martin’s shortfall. This strip mine, however, has some flexibility to truck its output to alternative docks below the failure zone for barge transloading. � Monongahela mines with rail access: These five mines, with 1.4 million metric tons of displaced barged coal, are assumed

to have crowded out their barge-only competitors in this scenario, and would be competing among themselves to supply the lost barged supply for the Fort Martin, as well as for the 0.6 million metric ton projected shortfall of supply to power plants below the failure zone. Three of the five mines were owned by the same company in 2010. All five mines had robust rail connections to the broader market below the failure zone. � System wide, closure of both Hatfield’s Ferry and the barge-only mines would result in an insignificant barged shortfall

over the baseline scenario.

7. Estimating financial impacts of closure

The previous section explored impacts in terms of logistics and the projected successfulness of reorganized markets above and below the failure zone. This section builds on those scenarios, characterizing the revenue losses and displacements asso- ciated with facility closures, specifically in terms of revenue and employment losses to those power plants and mines that might close, and the revenues lost by the barge shipping industry due to displaced coal movements. The revenue losses assume a one-year long river closure at Elizabeth and Charleroi, after which time revenue streams to the impacted entities could be resumed.

Revenues are used as the metric of interest as an upper bound of the financial impacts of plant closures. A more detailed financial model would yield a more accurate assessment of losses associated with plant closures, differentiating (for example) between cost of materials, wages, and profits. With the exception of the cost of delivered coal, the data used for this analysis do not support this level of detail; development of a detailed financial model is left to future modeling.

This analysis recognizes that these private losses will be offset largely in the broader economy by stepped up electricity generation (and revenues) and fuel production elsewhere in the grid; however the localized impacts where facilities close will be real and painful.

Prior models have estimated the cost of failure in terms of increased travel costs due to delays and mode shifts. Jenelius also describes the cost of ‘‘unsatisfied demand’’ where the perturbed system is not able to move product to customers (Jenelius et al., 2006). The valuation of unsatisfied demand, according to Jenelius, is ‘‘still an open question, however, or a matter of political decisions’’, and quantified unsatisfied demand as a fraction of total demand. In the examination of this particular case for the inland waterways, an alternative valuation is proposed: the value of revenues displaced at facilities that are likely to close in the face of failure of critical infrastructure.

Consideration of the closure potential is a valid means for assessing where the impacts of a disruption may be most severe (Jenelius and Mattsson, 2012). Chulkov’s application of social net present value to port technology investment decision mak- ing also identifies lost revenue as a cost of port closure (Chulkov, 2012).

Table 10 Potential revenue losses by coal mines under modeled year-long failure scenarios.

Total 2010 Production at Impacted Mines (million metric tons)

Baseline Value ($million, FOB, 2010)

Lost 2010 revenue at closed mines ($million, FOB)

Baseline (Scenario 1)

Baseline + Cumberland Mine closure (Scenario 3)

Baseline + closure of Hatfield’s Ferry and barge-only Mon mine closures (Scenario 4)

Mines above failure zone 35.2 1,240 1.1 337 465 With transport alternatives 28.5 772 1.1 1.1 1.1 Barge only 6.83 471 – 336 464 Mines below failure zone 248 2,820 44.2 44.2 44.2 With transport alternatives 228 1,890 – – – Barge only 20.4 924 44.2 44.2 44.2 Mine Subtotal 283 4,060 45.3 382 $ 509 Closed Mines Yellowbush Yellowbush Yellowbush

Mine 84 Mine 84 Mine 84 Cumberland Mine Cumberland Mine

Crawdad No. 1 Prime No. 1

# of employees 19,844 79 619 827(4% of total)

158 G.S. DiPietro et al. / Transportation Research Part A 69 (2014) 142–164

This analysis does not attempt to quantify all of the financial impacts associated with a catastrophic loss of infrastructure, but instead is focused on an impact area not fully explored in prior analyses – what are the financial impacts associated with firms that cannot adapt to the loss of critical infrastructure?

7.1. Revenue losses at closed coal mines

Table 10 summarizes the revenue losses associated with a one-year coal mine closure for each of the four scenarios characterized above, ranging from $45 million to $509 million. In addition, direct job losses might range from 79 to 830 mine employees.

These losses were estimated from the EIA-923 data set by calculating the value of the shipments from specific mines to specific utilities (reported as fuel cost, $/MMBtu, and combined with reported average heat content, MMBtu/ton, and quantity, ton). The publicly reported valuations represent the total cost of the coal, including transportation costs. To isolate the loss of revenue to the mines, these total coal valuations were adjusted to estimate freight-on-board (FOB) costs. (Actual FOB costs are collected in the EIA-923, but are withheld from the public datasets used for this analysis.) Transportation costs, based on state-to-state (barge and truck) or coal basin-to-state (rail) rates ($/ton) reported by EIA were subtracted from the total coal prices to give rough estimates of FOB coal prices (EIA, 2012). In a limited number of cases, fuel costs were not reported in the public data set, and were extrapolated from similar transits (mode and length). Similarly, missing transport costs for specific O/D combinations were extrapolated from similar transits or prior years, as available (see the Supplementary material). Employment losses were taken from MSHA mine data for 2010 (U.S. Mine Safety and Health Administration, 2013).

Note that these estimated revenue losses are direct private costs to the owners of the mines. They do not include the additional indirect costs associated with shuttering a mine or power plant. This paper does not attempt to calculate net costs to society of changes in utility fuel sources due to river infrastructure failure. Also, these estimated revenue losses assume a year-long closure, after which the river is assumed to be re-opened and mine or power plant operations could resume. The actual length of the river closure would affect the scale of the losses, as would the feasibility of the facilities to re-open with the river.

7.2. Revenue losses at closed power plants

Using EIA data for electricity prices by state in 2010 in conjunction with the net electricity generated by each of the impacted power plants (EIA, various), the revenues potentially lost by these owners at the closed power plants under the four modeled year-long Scenarios were calculated, as summarized Table 11. Folga used a similar approach for power plants forced to close due to a loss of water intake for their cooling water supply (Folga et al., 2009).

Table 11 also provides (parenthetically) the cost of delivered coal as reported by these facilities in the EIA-923 survey. Accounting for the cost of coal allows for the subsequent addition of lost revenues of mines and power plants without double-counting the coal.

These lost revenues represent direct private costs to the power plant owners, and are not a quantitative assessment of the net impact to society, as electricity generation (and revenues) will shift to other power generators and other fuel sources. The

Table 11 Potential revenue losses by power plants under modeled year-long failure scenarios.

Total 2010 Production from Impacted Plants (million MWh)

Baseline 2010 Value ($million)

Lost 2010 revenue at closed power plants [cost of delivered coal]($million)

Baseline (Scenarios 1 + 3)

Baseline + additional closures (Scenario 2 + 4)

Power plants above failure zone 15.3 $ 1,400 $ – [–] $925 [$232] Power plants within failure zone 1.49 $153 $153 [40] $153 [40] Power plants below failure zone 115 $ 10,300 $503 [149] $503 [149] Subtotal 132 $ 11,800 $ 656 [189] $ 1,580 [421] Closed power plants Elrama Elrama

Mitchell Mitchell Beaver Valley Beaver Valley R.E. Burger R.E. Burger Willow Island Willow Island Tanner’s Creek Tanner’s Creek

Hatfield’s Ferry

Table 12 Barge shipment year-long revenue losses due to mine closure (million $).

Shipping Origin Total 2010 Shipping Revenue, no failure, all modes

Baseline (Scenario 1)

Baseline + Cumberland Mine closure (Scenario 3)

Baseline + closure of Hatfield’s Ferry and barge-only Mon mine closures (Scenario 4)

Lost revenue - not viable barge shipments ($million) Mines above failure zone 590 31.4 Mines below failure zone 5,670 11.0 Lost revenue – cancelled viable barge shipments from closed

mines ($million) (not including shipments to closed pp) Mines above failure zone 0 14.4 7.71 Mines below failure zone 1.65 1.65 1.65 Lost revenue – cancelled viable barge shipments to closed power

plants ($million) Mines above failure zone 0 0 14.1 Mines below failure zone 2.76 2.76 2.76 Total 6,260 46.8 61.2 68.6

G.S. DiPietro et al. / Transportation Research Part A 69 (2014) 142–164 159

catastrophic infrastructure failure scenario is not expected to cause immediate shutdowns at the directly impacted power plants – stock piled coal will allow for continued operation in the near term and grid-wide transitions to alternative power mixes (Folga et al., 2009). The timing of the failure thus would drive the length of time that the power plant remains oper- ating while drawing down its stockpiles. EIA tracks the utility stockpiles in terms of tonnage and the average number of burn days that the stockpiles can support (EIA, 2013a). Stockpiles peak seasonally in early spring before the cooling season, and again at slightly lower levels in early fall before the heating season. Generally the average burn days fluctuate between 65– 75 days, with higher levels seen in 2012 as the industry adjusted to increasing availability of competing natural gas, and lower levels in the first half of 2014.

7.3. Revenue losses to barge shippers

Shipping costs are substantial for large quantity, high density commodities such as coal. Closure of the river due to infra- structure failure will cause the displacement of many millions of tons of shipment. Assuming the grid remains fully func- tional, which seems reasonable given the availability of coal stockpiles at the utilities and extensive redundancies within the grid, the shipments that are impossible due to river closure would be replaced by other shipments elsewhere. The direct impact to the individual shippers will be a function of their ability to supply those replacement shipments, an analysis which is beyond the scope of this paper and the available data. A worst case scenario, however, simply assesses the lost shipping revenues from the cancellation of the shipments that (1) become impossible because they need to pass through the failure zone, (2) are from mines that close, and (3) are to power plants that close, resulting in estimated lost shipping revenues of $47–69 million (see Table 12). Shipping costs were determined as described above for the lost mine revenue calculations. The largest impacts would be on those shippers who service the mines above the failure zone.

Table 13 Replacement value of vessels and barges trapped in potential failure zone, 2010.

Braddock Elizabeth Charleroi

Total number of lockages in 2010 3355 5299 5455 Barges per lockage, average 6.1 2.9 3 Daily number of lockages, average 9 15 15 Daily number of lockages, maximum 20 31 26 Daily number of unique vessels locked, average 7 8 8 Daily number of unique vessels locked, maximum 13 16 15 Average towboat power (hp) 2261 1684 1532 Replacement value per vessel ($million) 2.9 2.1 1.9 Replacement value of average daily number of vessels locked ($million) 18.8 17.0 16.3 Barge replacement ($million) 19.5 14.6 15.6

160 G.S. DiPietro et al. / Transportation Research Part A 69 (2014) 142–164

As with the revenue losses at impacted power plants, these shipping revenue losses reflect the direct private losses to specific shipping firms over the course of a one-year river closure. Given society’s unrelenting demand for electricity, fuel shipments will continue, albeit not via barge through the failure zone. The net shipping cost to society, not estimated here, would be a function of where the replacement electricity is generated, what specific fuel sources are used, and selecting the most cost-effective shipping modes available between the replacement power plant and fuel source.

For those mines with access to both rail and barge, it is possible to assess the magnitude of the increased shipping costs associated with moving their barge shipments to rail in response to a loss of passage through Elizabeth/Charleroi. Comparing barge shipping rates across the region with rail rates for the same O/D pairings shows that rail rates are $2-20/ton more expensive than comparable barge rates (see Supplemental materials). However, as discussed previously in the mode shifting discussion and as Wang noted, ‘‘the interfaces between the rail and barge networks are limited, and there are real costs to the shipper for making this modal change, which are not all captured in the freight rates (Wang et al., 2006).

7.4. Value of trapped vessels

Barge and towboat owners also are exposed to the risk of being trapped within the failure zone. Such entrapment might result in losses ranging from total losses of the vessels, to temporary grounding until the pool could be re-established at the downstream Braddock dam. The size of such losses can be estimated from the number of vessels and barges that typically move through the potential failure zone on an average day (USACE, 2013c), with replacement values estimated by the Corps of Engineers (USACE, 2000) and CPI-adjusted to 2010 costs. The daily lockage profiles were taken from a dataset of lockages through USACE’s Pittsburgh District for 2010; each record in the dataset identifies the timestamp, lock number, vessel num- ber and number of barges locked. Additional data on vessel size as measured by engine rating were taken from a Corps data- set characterizing registered towboats (USACE, 2010a).

If failure were to occur on a day with average traffic through the failure zone, seven or eight vessels would have locked through the three locks bounding or in the failure zone. On average, the replacement value of these vessels and their barges would be on the order of $32-38 million. On the highest traffic day, there might have been twice as many vessels moving in the region, doubling the potential losses. See Table 13.

This analysis provides a very rough approximation of the potential magnitude of vessel and barge losses; these represent indirect private costs. Some of these vessels would have only moved through one lock (e.g., larger towboats that typically do not travel higher than Braddock) and may have moved out of the failure zone. Others would have been in transit across the entire zone (e.g., tows moving from the Monongahela mines to the Ohio River power plants) and thus would have been more likely to be trapped by a sudden failure. In addition, moored vessels and barges, as well as harbor-duty towboats are not visible in USACE lockage data set used to construct this analysis, yet would also be trapped and possible damaged by river infrastructure failure.

7.5. Aggregated lost revenues associated with thermal coal shipments due to infrastructure failure

In summary, the private cost of extended river closure due to catastrophic infrastructure failure at Elizabeth and Charleroi will be significant, particularly for those mines, power plants, and vessel owners in the near vicinity of the failure zone. The aggregate revenue losses due to the closure of mines and power plants that may not be able to adapt to the loss of navigation are estimated to range from $560 to $1,700 million over a year-long river closure (see Table 14). Additional indirect costs for vessel and barge losses are about $35 million. Of course, there are other significant costs not tallied here for those impacted facilities that continue to operate in the post-failure markets.

From a broader social perspective, these lost revenues will be transferred to other power plants and fuel producers, as the region’s demand for electricity will be unchanged. The transfer of these revenues is of no comfort to the local economies that depend on and benefit from the closed facilities. These estimated aggregate private costs due to facility closures in response

Table 14 Displaced revenues from thermal coal mine and power plant closures due to year-long infrastructure failure (2010 $ million).

Scenarios

1. Baseline closures

2. Baseline + Hatfield’s Ferry closures

3. Baseline + Cumberland Mine closures

4. Baseline + Hatfield’s Ferry + barge-only Mon mine closures

Mines $ 45 $ 45 $ 382 $ 509 Power plants $ 656 $ 1,580 $ 656 $ 1,580 Cost of coal, delivered $ (189) $ (421) $ (189) $ (421) Shipping $ 47 $ 47 $ 61 $ 69 Sum (million) $ 559 $ 1,251 $ 910 $ 1,737

G.S. DiPietro et al. / Transportation Research Part A 69 (2014) 142–164 161

to a year-long river closure are on the same order of magnitude as USACE’s estimate of the total cost of the Lower Mon Project ($1.5 billion).

The costs summarized above focus on the power plants and mines that could close in response to a prolonged loss of nav- igation through Elizabeth and Charleroi. To the extent that these values include revenues from closed mines that supplied closed power plants, these results include some double counting of impacts. The remaining mines and power plants would also incur costs, not quantified here, associated with procuring alternative shipping or shifting to new positions within the regional coal-to-utility network.

Additional costs would be borne by the disruption of shipments of metallurgical coal and other commodities not assessed in this analysis.

� Metallurgical coal may account for another million tons of shipments through Elizabeth, although there is no publicly available data to quantify these shipments or their origins and destinations. To the extent that these shipments are from barge-only mines, additional mine closures could result. � Certainly a significant vulnerability is the U.S. Steel coking facility at Clairton (mile 19–22) which receives significant

feedstock by river and relies on the river for its process water. � There is also a significant movement of limestone and lime through Elizabeth, totaling over 1.4 million metric tons in

2010, much of which is expected to be shipped to the upper Monongahela River coal-fired utilities that rely on lime for the operation of their air pollution control devises. Disruption of these shipments could increase the likelihood that these plants would close in response to catastrophic failure at Elizabeth and Charleroi. These key shipments warrant fur- ther investigation. � The current market equilibrium differs from the modeled 2010 scenarios due to the actual closure of the Mitchell, Elrama

and Hatfield’s Ferry power plants. This demonstrated methodology could be extended readily to different time frames or regions. The Emsworth lock and dam faces similar vulnerabilities.

8. Conclusions

A careful examination of actual transit alternatives in response to a significant river infrastructure failure shows that impacts are highly localized to the failure zone, with particularly significant impacts on the river segments that become isolated from the remainder of the inland waterways. A failure further down river would impact more industries, but would result in more symmetric and larger new submarkets that would, by the nature of their size, be more resilient.

Impacts would be much more significant for those facilities with accessibility limited to only one mode of transporta- tion. Facilities with active access to multiple modes of transportation (barge, rail and truck) are better positioned to cope with the catastrophic loss of major infrastructure such as a lock (or bridge or rail spur). This resilience gives them the flexibility to shift transport modes. Without this flexibility, facilities that are completely reliant on barge transport may be forced to close.

Most power plants and mines participate in a fairly extended supply network, providing resilience in the face of loss of a specific supplier/customer. The ability of power plants to supplement their base contracts with coal from a variety of non- base mines provides resilience and flexibility to the logistics chain. Thus for those power plants that are not within the failure zone nor fully dependent on barge coal delivery, loss of access to certain mines as a result of the failure scenario of concern could be managed by their continued access to the coal ‘‘spot market’’.

The impacts to the power plants and mines that would be stranded above the failure zone, as well as the owners of the towboats and barges servicing these facilities, would be significant and perhaps catastrophic from a business viability perspective. The aggregate private cost of a year-long infrastructure failure to the facilities that cannot adapt and close is comparable in magnitude to USACE’s estimated cost for the ongoing and long delayed project to replace that infrastructure.

162 G.S. DiPietro et al. / Transportation Research Part A 69 (2014) 142–164

The modeled failure scenario assumes that certain power plants and mines would close in response to the disruption caused by a prolonged river closure at Elizabeth and Charleroi. Many of these plants have in fact closed since 2010. The viability of facilities in the face of major infrastructure failure is closely related to their fundamental financial viability. Facilities that are approaching the end of their useful lives, or that face other constraints on their profitability (such as pending environmental regulations), are more likely to close than to absorb additional costs associated with shifting trans- portation modes or with competing in a newly constrained market.

This paper focused on impacts to coal mines and their power plants customers who used the Monongahela River for transport of 6.4 million metric tons of thermal coal in 2010. Other commodities of interest shipped through the potential failure zone include an estimated million tons of metallurgical coal and 1.4 million metric tons of limestone and lime. Insufficient public data exist to assess the potential impact of the displacement of these metallurgical coal shipments, but could be expected to be similar in nature to the thermal coal impacts modeled in this paper. The limestone and lime shipments are likely being delivered to the upper Monongahela power plants and local cement plants, and thus the disruption of these shipments would further complicate the viability of these facilities in the face of a river closure.

In conclusion, the contributions of this paper are several. We have demonstrated the importance of assessing the actual accessibility of impacted plants to multiple modes of transportation, as well as a novel methodology for determining that accessibility. We have added the overall viability of plants as a criterion to whether they would be able to adapt to the loss of critical infrastructure, and suggested a methodology for quantifying their removal from the coal-to-utility network. We have assessed the viability of formation of new sub-markets in response to an extended closure of an important stretch of river. We have suggested a methodology to assess the likely loss of towboats and barges and their valuation as a result of a sudden catastrophic loss of pool. These contributions provide additional tools for the full assessment of the value of maintaining the viability of the inland waterways.

Acknowledgments

This material is based on work supported by the National Science Foundation (Grant No. CBET-1032722). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author and do not necessarily reflect the views of the National Science Foundation.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/ j.tra.2014.08.009.

References

American Waterways Operators, 2006. 2006 Carrier Dues and Assessment Schedule. [Online] Available at: http://www.americanwaterways.com/ membership_meetings/2006Carrierdues.pdf [Accessed 6 September 2013].

Baroud, H., Barker, K., Ramirez-Marquez, J.E., Rocco, S., 2014. Importance measures for inland waterway network resilience. Transp. Res. Part E 62, 55–67, C.M..

Bray, L., Dager, C., Burton, M., 2004. Willingness to pay for water transportation in the Ohio River basin. Transp. Res. Rec. 1871, 5–12. Bray, L.G., Murphree, C.M., Dager, C.A., 2011. Toward A Full Accounting of the Beneficiaries of Navigable Waterways. s.l.: Inland Waterways Users Board. Cambridge Systematics, Inc., 2008. Use of FAF Data for Florida Multimodal Freight Analysis, s.l.: Systems Planning Office, Florida Department of

Transportation. Campo, M., Mayer, H., Rovito, J., 2012. Supporting secure and resilient inland waterways: evaluating off-loading capabilities at terminals during sudden

catastrophic closures. Transp. Res. Rec. 2273, 10–17. Chen, A., Yank, C., Kongsomsaksakul, S., Lee, M., 2007. Network-based accessibility measures for vulnerability. Networks Spatial Econ. 7, 241–256. Chulkov, D.V., 2012. Managing new technology investment for underwater security of ports. J. Transp. Secur. 5, 95–105. Department of Homeland Security, 2009. National Infrastructure Protection Plan: Partnering to Enhance Protection and Resiliency. [Online] Available at:

<http://www.dhs.gov/xlibrary/assets/NIPP_Plan.pdf>. EIA, 2010. Annual Coal Distribution Report, U.S. Energy Information Administration. [Online] Available at: <http://www.eia.gov/coal/distribution/annual/

archive.cfm> [Accessed February 2014]. EIA, 2012. Coal Transportation Rates to the Electric Power Sector, U.S. Energy Information Administration. [Online] Available at: <http://www.eia.gov/coal/

transportationrates/> [Accessed 2013]. EIA, 2013a. Electricity Data Browser, U.S. Energy Information Administration. [Online] Available at: <http://www.eia.gov/electricity/data/browser/#/topic/

7?agg=0,1&geo=0001a001l&endsec=vg&linechart=ELEC.PRICE.PA-ALL.A&columnchart=ELEC.PRICE.PA-ALL.A&map=ELEC.PRICE.PA- ALL.A&freq=A&start=2009&end=2010&ctype=columnchart&ltype=pin&fuel=8&sec=008&pin=&rse=0&ma> [Accessed November 2013].

EIA, 2013b. Form EIA-923 Power Plant Operations Report Instructions, U.S. Energy Information Administration. [Online] Available at: <http://www.eia.gov/ survey/form/eia_923/proposed/2013/instructions.pdf> [Accessed 2014].

EIA, various. Form EIA-923 detailed data, U.S. Energy Information Administration. [Online] Available at: http://www.eia.gov/electricity/data/eia923/ index.html [Accessed 2013].

First Energy, 2013. [Online] Available at: https://www.firstenergycorp.com/content/fecorp/newsroom/news_releases/firstenergy-to-deactivate-two-coal- fired-power-plants-in-pennsyl.html [Accessed July 2013].

Folga, S. et al, 2009. A systems-level methodology for the analysis of inland waterway infrastructure disruptions. J. Transp. Secur. 4, 122–136.

G.S. DiPietro et al. / Transportation Research Part A 69 (2014) 142–164 163

Geurs, K.T., van Wee, B., 2004. Accessibility evaluation of land-use and transport strategies: review and research directions. J. Transp. Geogr. 12 (2), 127–140.

Google, 2013. Map data @2013 Google. [Online] [Accessed various 2013]. Gravens, M.B. et al, 2008. Incorporating complexity and uncertainty in planning models for engineering: economic analysis within the U.S. Army Corps of

Engineers. J. Contemp. Water Res. Educ. Issue 140, 15–23. Grier, D.V., 2004. The Declining Reliability of the U.S. Inland Waterway System. Washington, D.C., s.n. Ham, H., Kim, T.J., Boyce, D., 2005. Assessment of economic impacts from unexpected events with an interregional commodity flow and multimodal

transportation network model. Transp. Res. Part A 39, 849–860. Jenelius, E., Mattsson, L.-G., 2012. Road network vulnerability analysis of area-covering disruptions: a grid-based approach with case study. Transp. Res. Part

A 46, 746–760. Jenelius, E., Peterson, T., Mattsson, L.-G., 2006. Importance and exposure in road network vulnerability analysis. Transp. Res. Part A 40, 537–560. Johnson, L.R., 1978. The Headwaters District: A history of the Pittsburgh District. U.S. Army Corps of Engineers. s.l.:U.S. Army Corps of Engineers, Pittsburgh

District. Kruse, C.J. et al, 2011. America’s Locks & Dams: ‘‘A Ticking Time Bomb for Agriculture?’’. American Soybean Board, Chesterfield, MO. Kruse, C.J., Protopapas, A., Olson, L.E., Bierling, D.H., 2007, amended 2009. A Modal Comparison of Domestic Freight Transportation, College Station, Texas:

Texas Transportation Institute, The Texas A&M University System. Ludlow, D., 2011. Key Observations on Improving the 2007 Commodity Flow Survey. Number E-C158 ed. Washington (D.C.): Transportation Research Board

of the National Academies. MacKenzie, C.A., Barker, K., Grant, F.H., 2012. Evaluating the Consequences of an Inland Waterway Port Closure With a Dynamic Multiregional

Interdependence Model. IEEE Transactions on Systems, Man, and Cybernetics-Part A: Systems and Humans, 42(2), pp. 359–370. Mapquest, 2014. [Online] Available at: www.mapquest.com [Accessed February 2014]. National Transportation Safety Board, 2008. Collapse of I-35W Highway Bridge, Minneapolis, Minnesota, August 1, 2007. [Online] Available at: <http://

www.dot.state.mn.us/i35wbridge/ntsb/finalreport.pdf> [Accessed February 2014]. Nicholson, A. & Dalziell, E., 2003. Risk evaluation and management: a road network reliability study. In: Proc. of the 1st Int. Sym. on Transportation Network

Reliability, pp. 45–59. NRC, 2001. Inland Navigation System Planning: The Upper Mississippi River Illinois Waterway, National Research Council, National Academy of Science,

Washington, D.C. NRC, 2012. Corps of Engineers Water Resources Infrastructure: Deterioration, Investment, or Divestment?, National Research Council, The National

Academies Press, Washington, DC. Oak Ridge National Laboratory, n.d. Freight Analysis Framework Data Tabulation Tool, Version 3. [Online] Available at: <http://faf.ornl.gov/fafweb/

Extraction1.aspx> [Accessed August 2014]. Pant, R., Barker, K., Grant, F.H., Landers, T.L., 2011. Interdependent impacts of inoperability at multi-modal transportation container terminals. Transp. Res.

Part E 47 (5), 722–737. Park, J. et al, 2013. Integrating risk and resilience approaches to catastrophe management in engineering systems. Risk Anal. 33 (3), 356–367. Port of Pittsburgh Commission, 2014. The Port District and the Inland Waterway System. [Online] Available at: <http://www.port.pittsburgh.pa.us/

index.aspx?page=127> [Accessed February 2014]. Reis, V., 2014. Analysis of mode choice variables in short-distance intermodal freight transport using an agent-based model. Transp. Res. Part A: Policy Pract.

61, 100–120. TRB, 2013. Guidebook for Developing Subnational Commodity Flow Data, Transportation Research Board of the National Academies, s.l.: s.n. Tsang, J.L., Lamber, J.H., Patev, R.C., 2002. Extreme event scenarios for planning of infrastructure projects. J. Infrastruct. Syst. 8, 42–48. U.S. Congress, 2014. Water Resources Reform and Development Act of 2014. Washington, D.C.: s.n. U.S. Council on Environmental Quality, 2013. Principles and Requirements for Federal Investments in Water Resources. [Online] Available at: <http://

www.whitehouse.gov/sites/default/files/final_principles_and_requirements_march_2013.pdf> [Accessed February 2014]. U.S. DOT Bureau of Transportation Statistics, 2007. Commodity Flow Survey. [Online] Available at: <http://www.rita.dot.gov/bts/sites/rita.dot.gov.bts/files/

publications/commodity_flow_survey/index.html> [Accessed 2014]. U.S. Environmental Protection Agency, 2009. eGRID2012 Version 1.0 Plant File. [Online] Available at: <http://www.epa.gov/cleanenergy/energy-resources/

egrid/index.html> [Accessed various 2013]. U.S. Environmental Protection Agency, 2013. Power Sector Modeling Platform v. 5.13, s.l.: s.n. U.S. Environmental Protection Agency, 2014. Inventory Of U.S. Greenhouse Gas Emissions And Sinks: 1990-2012. [Online] Available at: <http://

www.epa.gov/climatechange/Downloads/ghgemissions/US-GHG-Inventory-2014-Chapter-Executive-Summary.pdf> [Accessed March 2014]. U.S. Mine Safety and Health Administration, 2013. Mines Data Set. [Online] Available at: <http://catalog.data.gov/dataset/mines-data-set> [Accessed 2013]. USACE, 2000. FY 2000 Planning Guidance Shallow Draft Vessel Costs, s.l.: s.n. USACE, 2003. September-October 2003 Main Chamber Maintenance Closure Greenup Lock and Dam. [Online] Available at: September-October 2003 Main

Chamber Maintenance Closure Greenup Lock and Dam. USACE, 2004. Navigation Charts. [Online] Available at: <http://www.lrp.usace.army.mil/Missions/Navigation/NavigationCharts.aspx> [Accessed 2014]. USACE, 2009. Shipper and Carrier Response to the June–July 2008 Upper Mississippi River Flood Emergency Closure of Locks 12–25. [Online] Available at:

<http://library.water-resources.us/pubsearchS.cfm?series=NET> [Accessed February 2014]. USACE, 2010a. U.S. Waterway Data: Vessel Characterization. [Online] Available at: <http://www.navigationdatacenter.us/data/datavess.htm> [Accessed

2013]. USACE, 2010b. Ports and Waterways Facilities. [Online] Available at: <http://www.navigationdatacenter.us/ports/ports.asp> [Accessed 2013]. USACE, 2010c. U.S. Waterborne Container Traffic by Port/Waterway in 2010. [Online] Available at: <http://www.navigationdatacenter.us/wcsc/

by_portnames10.html> [Accessed February 2014]. USACE, 2011a. 2011 Region to Region Public Domain Data Base by Origin. [Online] Available at: <http://www.navigationdatacenter.us/wcsc/pdf/

pdrgod11.pdf> [Accessed February 2014]. USACE, 2011b. Measuring the Impact of Monongahela Lock Closures on Forecasts of Utility Steam Coal Consumption, Sourcing and Transportation in the

Ohio River Basin. [Online] Available at: <http://www.port.pittsburgh.pa.us/modules/showdocument.aspx?documentid=360> [Accessed 5 August 2014]. USACE, 2013a. Dam Safety Program Activities. [Online] Available at: <http://www.usace.army.mil/Missions/CivilWorks/DamSafetyProgram/

ProgramActivities.aspx> [Accessed 6 September 2013]. USACE, 2013b. Monongahela River Locks and Dams 2, 3, and 4 Project Fact Sheet (Lower Mon). [Online] Available at: <http://www.lrp.usace.army.mil/

Portals/72/docs/HotProjects/LMPJune2013.pdf> [Accessed 6 September 2013]. USACE, 2013c. Navigation Data Center. [Online] Available at: <http://www.navigationdatacenter.us/data/data1.htm> [Accessed 2013]. USACE, 2013d. Pittsburgh District Mission and Vision. [Online] Available at: <http://www.lrp.usace.army.mil/Missions.aspx> [Accessed 5 September 2013]. USACE, 2013e. Part 2 Tons_Comm_Ports Calendar Year 2011. [Online] Available at: <http://www.navigationdatacenter.us/wcsc/webpub11/

Part2_Ports_tonsbycommCY2011.HTM> [Accessed 6 February 2014]. USACE, 2014a. LPMS Summary by Division/District. [Online] Available at: <http://www.navigationdatacenter.us/lpms/pdf/2011%20&%202012.pdf>

[Accessed 29 January 2014]. USACE, 2014b. Navigation Data Center. [Online] Available at: <http://www.navigationdatacenter.us/db/pport/xls/> [Accessed 24 January 2014]. USACE, 2014c. Port Facility spreadsheet. s.l.:s.n.

164 G.S. DiPietro et al. / Transportation Research Part A 69 (2014) 142–164

USACE, n.d. Planning Guidance: Navigation – Inland. [Online] Available at: <http://planning.usace.army.mil/toolbox/guidance.cfm?Option=BL&BL=Inland Nav&Type=None&Sort=Default> [Accessed August 2014].

USACE, various. Waterborne Commerce of the United States. [Online] Available at: <http://www.navigationdatacenter.us/wcsc/wcsc.htm> [Accessed 2014]. Ventyx, 2013. Coal Transaction Analyst. [Online] [Accessed July 2013]. Wang, S., Tao, X., Schonfeld, P., 2006. Modeling shipper response to scheduled waterway lock closures. Transp. Res. Rec. 1963, 33–43. Zhu, S., Levinson, D., Liu, H.X., Harder, K., 2010. The traffic and behavioral effects of the I-35W Mississippi River bridge collapse. Transp. Res. Part A 44,

771–784.

Glossary

FOB: Freight on board NRC: National Research Council TRB: Transportation Research Board of the National Academies USACE: United States Army Corps of Engineers