ORI GIN AL PA PER

Extreme weather impacts on freight railways in Europe

Johanna Ludvigsen • Ronny Klæboe

Received: 22 October 2012 / Accepted: 26 August 2013 / Published online: 8 September 2013� Springer Science+Business Media Dordrecht 2013

Abstract Four cases are studied in this assessment of how the harsh 2010 winter weather

affected rail freight operations in Norway, Sweden, Switzerland and Poland and also of the

reactive behaviour rail managers mobilised to reduce the adverse outcomes. The results are

utilised in a fifth case assessing the proportion of freight train delays in Finland during

2008–2010 by modelling the odds for freight train delays as a function of changes in met-

states on the Finnish network and weather-induced infrastructure damage. The results show

that rail operators were totally unprepared to deal with the powerful and cascading effects

of three harsh weather elements—long spells of low temperatures, heavy snowfalls and

strong winds—which affected them concurrently and shut down large swathes of European

rail infrastructure and train operations. Rail traffic disruptions spread to downstream and

upstream segments of logistics channels, causing shippers and logistics operators to move

freight away from rail to road transfer. As a result, railways lost market share for high-

value container cargo, revenues and long-term business prospects for international freight

movement. Analyses of measures employed to mitigate the immediate damage show that

managers improvised their ways of handling crises rather than drew on a priori contin-

gency, i.e. fight-back programmes and crisis management skills. Modelling the co-varia-

tion between extreme weather and freight train delays in Finland during 2008–2010

revealed that 60 % of late arrivals were related to winter weather. Furthermore, the

combined effect of temperatures below -7 �C and 10–20 cm changes in snow depth

coverage from 1 month to the next explained 62 % of the variation in log odds for freight

train delays. Also, it has been shown that changes in the number of days with 10–20 cm

snow depth coverage explained 66 % of the variation in late train arrivals, contributing to

626 min or 10.5 additional hours’ delay. Changes in the number of days with snowfalls

over 5 mm accounted for 77 % variation in late train arrivals, implying that each additional

J. Ludvigsen (&) � R. KlæboeInstitute of Transport Economics, Oslo, Norwaye-mail: jlu@toi.noURL: www.toi.no

R. Klæboee-mail: rk@toi.no

123

Nat Hazards (2014) 70:767–787DOI 10.1007/s11069-013-0851-3

day with this snowfall could contribute to 19.5 h’ delay. Finally, the combination of

increased mean number of days with 5 mm snowfall and temperature below -20 �C

explained 79 % of the variation in late arrivals, contributing to 193 min or 3.25 h’ delay.

All results were significant (p = 0.00).

Keywords Extreme weather impacts � Preparedness level � Traffic and

supply disruptions � Crisis management � Reputational and business damage �Extreme weather indicators

1 Introduction

Extreme weather events are a threat to individual companies, their personnel and collab-

orative arrangements such as supply chains. The literature on management indicates that

there is no one best way of overcoming the negative impacts of these occurrences and one

reason is that they fall within the high-impact/low probability risk category and that there

is therefore a scarcity of the historical data needed for devising universally effective

prevention, containment and mitigation tools. Another reason is that low-frequency inci-

dents are hard to predict and it is thus difficult to allocate resources to proactively manage

the risk. If the risk never materialises, the costs incurred are hard to justify to the company

leadership and/or shareholders (Zsidisin et al. 2004).1

Yet empirical evidence indicates that weather-induced disasters tend to be occurring

more frequently and with increased severity. The report entitled ‘‘Performance Measures

for Freight Transportation’’ (Transportation Research Board 2011) clearly states that this is

reminiscent of ‘‘sub-optimisation’’ in managerial decision-making, where the focus is

inordinately upon achieving narrow, easily and immediately justified sub-goals to the

detriment of broader business objectives such as long-term operational continuity through

resistance development.

So why do some organizations cope so much better than others with the prospects and

impacts of weather-induced adversity?

Organizations do not have a common secret formula or even many of the same pro-

cesses for dealing with weather-generated risk, but they do share one critical trait: resil-

ience. Conceptually, resilience is the antithesis of vulnerability, which Svensson (2002)

defined as ‘‘… unexpected deviations from the norm and their negative consequences’’.

Mathematically, vulnerability may be measured in terms of ‘‘risk’’, a combination of the

likelihood of an event and its potential severity (Sheffi 2001, 2005). The notion of

organisational resilience entails functional and structural preparedness. Functional resil-

ience implies that a given entity is capable of efficiently and effectively dealing with

adversity and may recover unscathed through drawing on internal resources. On the other

hand, structural resilience is the organizational ability to absorb and/or withstand external

risks and/or perturbations thanks to built-in robustness and internal reserves (Bundschuh

et al. 2003; Holmgren 2007; Lai et al. 2002).

There is no doubt that supply chain disruptions are costly. In order thus to prevent,

mitigate and neutralise negative consequences of chain ruptures, it has to be understood

1 As Qiang et al. (2009) have shown in numerical modelling of changes in supply chain risk level invokedby transport disruptions, this statement indicates that manufacturers, retailers and transport carriers within agiven supply network place zero weights on disruption risks (page 108).

768 Nat Hazards (2014) 70:767–787

123

how an abrupt cessation of the movement of goods or stoppage of material flows may

affect not only the focal transport operator but also other supply chain segments (Hendricks

and Singhal 2005). A recent example is the earthquake in Japan, which in March 2011

damaged several plants producing microchips and other electronic components for

equipment manufacturers in the US and Taiwan. This contagion has spread to Europe,

causing transient shortages of smart phones, tablets and other high-tech consumer elec-

tronics (Financial Times, 23 April 2011). As summarised by Sheffi (2005; 74), one of the

main characteristics of disruption in large-scale supply networks is the ‘‘high-level

transmission between vulnerabilities stemming from the large systems’ inter-

connectivity’’.

Yet, the risks caused by weather-related disruption in Europe are, surprisingly, rarely

addressed in the supply chain management literature (Kleindorfer and Saad 2005). Even

worse, the consequences that extreme weather events exerted on freight transport opera-

tions in Europe received hardly any mention from researchers in this field (Wilson 2007).

One reason could be that, compared to disruptions paralysing manufacturing plants and/

or warehouses which result in large supply shortages, a rupture in movement of goods

within a supply pipeline may be potentially less contagious because it halts only the

transfer of merchandise and/or materials within a given conduit. The uniqueness of

transportation disruption is that although the goods in transit have been stopped, the

remaining supply network operations may still function undisturbed.2

However, as observed by Gunipero and Eltantawy (2004) and by Adegoke and Go-

palakrishnan (2009), this is very far from the truth. Transport interruption is a risk that can

quickly cripple an entire supply chain because, in addition to halting the flow movement,

the stoppages in materials and/or goods transfer spread quickly to the upstream and/or

downstream supply chain segments, causing stock outs, inventory depletion, production

downtimes, unfulfilled customer orders, information distortion and/or a backlog of goods

in transit.

2 Purpose statement

Against this backdrop, our article sought to assess:

1. how the different extreme weather events affected European rail freight systems, and

2. the action the affected parties mobilised to mitigate and neutralise the resulting

impacts.

We anticipated that the results from this enquiry would help us identify the most

severely vulnerable areas within the entire European freight rail system and also the

managerial and physical assets capable of improving the sector’s overall preparedness.

Unfortunately, multiple searches for literature on measures proficient at managing the

impacts of hazardous weather on the European freight transport industry, and particularly

the railways, did not produce tangible results.

2 Although a disruption in transportation will certainly delay the arrival of goods at destination, a distinctionis made here between a transportation disruption and a transportation delay which fall into two different riskcategories. Because of larger element of surprise and lower preparedness level, Wilson (2007) maintainedthat risk drivers for a delay are much smaller than those of disruptions, which may last longer and hit severalsupply segments simultaneously. This distinction was also useful for determining the conditions of supplynetwork robustness and strategies for dealing with disruptions caused by natural hazards.

Nat Hazards (2014) 70:767–787 769

123

Consequently, our work had to be revised and broken down into five more specific

explorations, as follows:

1. The impacts the extremely bad weather inflicted on the rail infrastructure and

operations in Poland, Sweden, Norway and Switzerland during the harsh winter 2010.

2. The reactive behaviour managers in the affected companies mobilised to counteract

and/or contain the ensuing outcomes.

3. The proportion of weather-induced delays in all freight train arrivals in Finland during

2008–2010.

4. The effects the harsh weather events had on the odds for freight train delays in the

Finnish network during the above period.

5. How the duration of freight train delays in Finland co-varied with the harsh winter

weather during 2008–2010.

3 Methodology

3.1 Research approach and design

This study focused on natural disasters, the tangible impacts of which on rail freight

companies and their clients in Europe had not yet been adequately recorded, analysed and

assessed. Since these phenomena invoked different material and temporal damage and

since researchers could manipulate neither the independent nor the dependent variables, a

case study method was chosen for this exploration (Yin 1994, 23).

This research approach was supported by Snyder and Swann (1978), who maintained

that the case study method was an appropriate form of empirical enquiry: when impacts of

the determinants studied varied between the targets; when researchers could not manipu-

late either the causes or the specific outcomes; and when the current body of knowledge did

not allow for cross-contextual predictions.

Procedures foreseen for scientifically sound case studies recommend that, first, a

descriptive definition of the phenomena studied is developed. Subsequently, actors with

pertinent experience and data are identified and interviewed, and finally, an understanding

of the phenomena in focus is constructed that facilitates statistical tests of more specific

causal relationships (Yin 1994, 29).

Pursuant to the above, four case studies on how the rail freight operators in Norway,

Sweden, Switzerland and Poland reacted to the harsh winter in 2010 were carried out and

the results used in a Finnish rail modelling case assessing: (1) the proportion of delays in

freight train arrivals attributed to extreme weather and/or weather-inflicted technical

damage on Finland’s network during 2008–2010, (2) the odds for train arrival delays

inflicted by harsh weather and (3) the duration of arrival lateness attributed to bad weather.

3.2 Data sources

3.2.1 European cases of harsh weather

Interviews with managers of rail cargo companies in Poland, Norway and the Netherlands

provided data for the first three case studies; one of the interviewees was a state-owned

incumbent from Norway, while two others were private rail undertakings. Data for the

fourth case came from an interview with a director of the Association of Swedish Train

770 Nat Hazards (2014) 70:767–787

123

Operators, which represents seventeen rail carriers in Sweden. Inclusion of private rail

undertakings was justified by the fact that these entities are relatively small compared to

national incumbents and therefore vulnerable to all types of hazard. Yet, experience shows

that despite small size and relative resource scarcity, private operators are capable of

rapidly adapting to environmental dynamics. These two factors made them interesting for

studying managerial reactions to extreme weather adversity.

3.2.2 Finnish case of harsh weather

The VR Group Ltd., the Finnish Transport Agency and the Finnish Meteorological Institute

provided data which included two sets of aggregated indicators. The first was a register of

eighty different causes of freight train delays and their duration, including thirty-seven

related to bad weather. The second comprised monthly registrations of maximum and

minimum temperatures, precipitation and the numbers of days in each month in the period

2008–2010 with specific temperature and precipitation levels. These monthly temperatures

and precipitation indicators were measured at a number of meteorological stations

throughout Finland and then averaged monthly.

3.3 Data collection

3.3.1 European railway cases

Face-to-face interviews with the executives responsible for management of rail operations

provided data for the European cases. During each interview lasting between one and two

hours, the following were recorded: (1) self-reports on how targets were exposed to harsh

weather and the impact-triggering mechanisms that affected rail operations, (2) types of

adversity experienced and measures mobilised during and immediately after a given

extreme weather instance in order to cope, neutralise and/or reduce the resultant conse-

quences and (3) strategic adjustments that the targets have undertaken and/or planned to

introduce in order to improve overall preparedness for weather-induced damage at com-

pany and/or supply channel levels.

3.3.2 Analyses of European case data

The interview audio-records were transcribed verbatim into an interview protocol, the

content of which was then analysed with regard to the types of adversity that the affected

parties experienced, the actions mobilised and resources employed to counteract the most

immediate impacts and the long-term repercussions.

4 Findings from European cases

4.1 Vulnerability of rail operators to winter disruption

The four European cases revealed the instantaneous impacts that the extremely harsh

winter weather in 2010 inflicted on rail disruption, the aftershocks to other logistics seg-

ments and the long-term consequences for railway business.

Nat Hazards (2014) 70:767–787 771

123

An executive at the Swedish Association of Rail Operators described the different

stages in the weather-triggered crisis as follows:

During winter 2010 the south-western Sweden was affected by unusually low tem-

peratures, winds and heavy snowfalls which begun in December 2009 and lasted

until March 2010. We have been taken aback by a combination of very low tem-

peratures and heavy snow storms. Strong winds formed clouds of light snow which

stuck to wagon undercarriages, immobilized vehicles, and blocked track lines

between Halsberg intermodal terminal and the main Swedish harbors. Snow accu-

mulated under undercarriages has dramatically amplified the vehicle weights causing

the wheel axles to break. We lacked the buffer stocks of spare wheels and personnel

capable of replacing the broken units on short notice. So, we had to sign new

agreements with repair workshops. The smallest rail operators have of course suf-

fered the most because of considerable resource scarcity. In addition, wagon brakes

lost the grip on slippery track surface and many trains had to be stopped. This

reduced the network traffic. Furthermore, many tonnes of temperature-sensitive

goods have frozen while still in wagons because trans-shipment to trucks was

delayed.

This shows that targets were unable to prevent or to neutralise weather-imposed losses.

Efforts used to contain the impacts of infrastructure shutdowns and traffic breakdowns

reveal that both the operators and the infrastructure managers improvised their way out of

crisis rather than drew on the a priori available preparedness measures and/or crisis

management skills. Another excerpt from this interview reveals the dramatic scope of

negative consequences for the entire Swedish rail freight industry.

During this (winter) period the volume of rail cargo in Sweden was reduced by entire

20 percent. The Halsberg marshalling yard which is a centre of Sweden’s rail freight

operations was closed for 14 days. This shutdown alone has cost between 200 and

250 million SEK. This amount has been further attenuated by phasing out of at least

20 rail shuttles between Halsberg and Gothenburg and re-location of large cargo

volumes to road haulage

One interesting finding was that although all the rail freight companies studied function

in northern Europe, where harsh winters are common, not one of them anticipated the

combined impacts of the extraordinarily long spells of low temperatures, heavy snowfalls

and strong winds which in 2010 brought their operations and infrastructure to a standstill.

The Norwegian rail freight manager’s encounter with the unusually harsh winter in 2010 is

reported below.

The 2010 winter weather brought about a rare combination of unusually heavy

snowfalls, low temperatures and strong winds. This resulted in a range of infra-

structure shutdowns and rolling stock breakages. First, Infrastructure Managers

lacked enough snow ploughs to keep all tracks and interchanges snow-free. Second,

low temperature and heavy snow deposits caused that wheels on some of our flatcars

went to pieces. These disruptions reduced our flatcar fleet and supply reliability to 60

per cent (from standard 90 per cent) meaning that considerable number of containers

was not delivered on time. Our customers were aghast; they had to shift goods supply

to road haulage. Further, we had to renew our stock of spare wheels immediately and

that showed difficult. As a consequence of this but also due to the accumulation of

ice deposits on tracks, the deceleration time and braking distance for wagons

772 Nat Hazards (2014) 70:767–787

123

increased considerably. As a result, we were forced to run fewer and shorter trains.

Yet despite fewer trains the manpower at terminals has to be increased to fight

technical emergencies. More people had to step in so that we did not breach the

working time regulation. As a consequence we faced two adversary impacts at the

same time: our operational costs skyrocketed while our cargo volumes along with

customers’ trust plummeted

A Dutch rail manager whose company suffered from 2010 winter traffic breakdown

provided the following account.

In Switzerland we had to stop all our rail container traffic for one week in January

2010 because the track was blocked by unusually heavy snowfalls. However, infor-

mation about infrastructure shutdown reached us in advance so we were able to re-

position our locos, flatcars and containers, and reduce the costs of the standstill. In

Sweden the situation was different because there several of our trains were trapped by

snowfalls blocking connections between the feeder and the trunk lines and could not

return to the main operations depot after discharge of container loads at customers

sidings. Besides persistent low temperature of -20 �C have damaged the rubber

linings on our flatcar brake’s cables which stopped all vehicles. This forced us to stay

put until the cables were replaced. This operation took quite many man-hours .3

The harm inflicted on the Polish rail operator was equally dramatic, although for dif-

ferent reasons.

Heavy falls of wet snow was a nuisance because they broke catenaries and fell trees

along the track lines that blocked the network pathways. In addition, the freezing fog

glazed catenaries and broke pantographs on several locos. Surprisingly, the cold in

the range of -15 �C and below did not inflict much harm on locos’ technical fitness,

but temperature in the range of -1/? 1 �C combined with high humidity caused

shortcuts in locos’ electrical wiring. We protected our locos from freezing by

keeping engines on empty runs before and after each journey. As a result, our loco

drivers had to put extra working hours in emergency shifts.

These excerpts indicate that at least two circumstances amplified the severity of the

attack and the scope of damage. The first was a combination of three different harsh

weather components: low temperature, snowfalls and strong winds. As these elements

coincided in space and time, they produced cascading effects that immobilised long

stretches of the European rail infrastructure and brought all freight transfer to a halt

(Delmonaco 2006). The second factor derived from the unusually large scope and power of

the knock-on impacts, which caught the targets unprepared and rendered them virtually

helpless within a very short time.

Because of the specific position of the railways in the logistics supply system, dis-

ruptions in rail cargo operations produced a chain of contagion that quickly spread to other

segments of logistics channels. However, the railways bore the brunt of weather-inflicted

damage because they could not substitute the rail freight transfer with alternative modes as

could shippers, forwarders and logistics network integrators.

3 Operations of freight trains by Dutch company’ on Swiss and Swedish networks were possible due to theFirst Infrastructure Package, and particularly directive 2001/14/EC, which defined rules for allocation ofinfrastructure capacity on the third countries’ networks, levying of infrastructure usage charges and safetycertification for private rail undertakings.

Nat Hazards (2014) 70:767–787 773

123

Adding to the complexity was the fact that being government utilities, rail infrastructure

administrators were not liable for direct business losses or other disutilities that infra-

structure shutdowns imposed on rail undertakings, cargo owners and logistics companies.

As a consequence, and in addition to sharp spikes in operations and manpower costs, the

railways have suffered through loss of their customers’ reputation and have jeopardised

business prospects in the European freight market. The Norwegian operator summarised

this problem succinctly.

In order to reduce the risks of stock out and the amount of unfilled orders at dis-

tribution centers, we had to re-position our resources. This required higher opera-

tional back-up and closer collaboration with Infrastructure Managers. However, our

hands were tied: our trains were stopped by infrastructure shutdowns. Our customers

demanded compensations for unfilled supply orders. Therefore, we started discus-

sions with the Ministry of Transport and Communications who owns rail infra-

structure in Norway to grant us the rights to charge infrastructure provider with

penalties for track closures and/or pay lower user charges after several track lines

were out of operations which delayed our freight trains’ arrivals. Still, paying con-

siderable delay compensations was not the biggest harm to us. The loss of traffic

which our clients re-located to road transport and the customer trust in our ability to

deliver on time were considered as much more serious setbacks because they seri-

ously threatened our business’ future.

The Dutch rail manager summarised the losses that his company suffered due to

infrastructure shutdown in the following manner:

Recently, we have discussed with the Swedish Rail Infrastructure Administration

what harms the network shutdowns and reduction of network serviceability inflicted

on our operations. We have presented them with bills for losses imposed by infra-

structure closures. The Swedish people have launched a full-blown investigation into

factors causing infrastructure downtime which, we hope may improve infrastructure

security next winter. Still, we expect more harsh winters to come, and with that more

infrastructure closures. We will simply need to live with that and be better prepared.

The cases studied show that the infrastructure closures rendered the railways’ crisis

management efforts futile and have underscored that continuity of rail operations and

punctuality of cargo train arrivals are heavily dependent on infrastructure functionality. In

order to avert the risk of stock outs and/or supply shortages at wholesale and retailer

outlets, logistics integrators moved cargo transfer away from rail to road. The Swedish

operators applied the following measures to contain the damage inflicted by infrastructure

shutdown.

Shippers, rail operators and Trafikverket (The Swedish Transport Infrastructure

Administration) formed task forces to jointly combat these damages. To be effective,

our decisions had to be based on real-time information. Trafikverket fed us with

information about impending and/or already imposed infrastructure closures, lines

open for detour and serviceability conditions on the remaining network segments. To

speed up the most critical consignments, our clients handed us a list of most urgent

shipments and the goods’ physical conditions (i.e. tolerance for cold and longer

transit time). These data helped Trafikverket to re-assign traffic to a considerably

downsized network using three priority rules 1) trains that had to be given green light

immediately, 2) trains that had to be re-scheduled to new time windows and new

774 Nat Hazards (2014) 70:767–787

123

track paths within the next 12 h, and 3) trains that could be kept at sidings and/or

marshalling yards longer than 12 h. Eventually we got the most critical tonnage of

cargo traffic out and moving.

These examples are instructive. On the one hand, they reveal that mobilisation of ad hoc

damage containment is effective in dealing with the unfolding course of disaster events. On

the other hand, however, the lack of backup systems and preventive skills magnifies the

scope of damage and the costs of adversity abatement. The patterns of crisis-fighting

behaviour revealed that all managers strove to alleviate the most immediate impacts on

their operations domains without efforts being made to increase the operational robustness

of the entire supply system. Constrained by the damage to rolling stock, infrastructure

shutdowns and shortages of vital components destroyed by extremely harsh exploitation

conditions, the targets turned to in-house human resources because this type of asset is

readily available and effective at quelling the most immediate harm. This statement by the

Polish manager underscores the need for an internal ‘‘flexibility’’ without, however, it

being specified how this could be achieved.

Our operational flexibility was the main asset that helped us to absorb the conse-

quences of and to deal with these (traffic) disruptions. Today, all these happenings

seem as if they have taken place in a distance past. Now-a-day we are facing and

dealing with quite new and different challenges. We have realized however, that we

need more operational flexibility in our system to be able to withstand the similar

adversities in future.

Although all railway operators deployed extraordinary resources, they still could not

stop aftershocks spreading to the upstream and downstream chain segments. However,

none of these targets acknowledged the need for improving resilience at all channel tiers.

Neither a necessity for systematic risk assessment nor specialised crisis management skills

capable of addressing the different hazards affecting the different supply chain segments

was recognised. Nor was the building of long-term strategic preparedness for level-headed

handling of future crises considered as a long-term risk-containing investment.

Only the manager at the Norwegian state-owned cargo carrier recognised the need for a

strategic overhaul of his company’s command and control system if the carrier was to

withstand the negative ‘‘domino effect’’ between the infrastructure shutdown and the

operations breakdown.

This experience has humbled us. We have to regain the customers’ trust by making

our freight dispatch system more robust. That means that several elements of our

command and control system have to be re-engineered while collaboration with

infrastructure provider reinforced. However, before we re-launch a more reliable

container dispatch system we need a guarantee from Infrastructure Managers as

regards higher standards of network reliability. And that’s the critical area on which

we are working right now.

These results provided background for two questions:

1. Why were all the affected parties so badly prepared to tackle the extreme weather

impacts despite being well accustomed to the harsh north European climate?

2. Why did the executives studied not recognise a priori the high risks that harsh winter

weather might inflict on their operations, personnel, infrastructure, and subsequently

brand reputation and long-term business prospects?

Nat Hazards (2014) 70:767–787 775

123

4.2 Conclusions from European cases

One reason for the inability of rail operators to pre-empt and avert extreme weather

damage was the element of surprise that caught them unprepared for unusually powerful

and commensurate knock-on effects that brought the entire rail system to temporary

paralysis. This ‘‘cascading effect’’ triggered ‘‘failure in a system of interconnected ele-

ments whose entire service provision depends on functionality of the preceding segments,

and whose preceding segments can exert failure on the successive parts’’ (Kappes et al.

2012).

However, the above did not explain the low general level of strategic preparedness and

the absence of weather risk awareness on the part of the managers studied. In search for

more in-depth explanation, we turned to literature on managerial attitudes towards risk in

general and risk perceptions in particular. A classic study carried out by March and Shapira

(1987) provided some indication in showing that:

1. Managers were insensitive to estimates of probabilities of possible outcomes.

2. Managers tended to focus on critical performance targets which affected the way they

managed the risk.

3. Managers made a sharp distinction between risk-taking and gambling.

The first finding could be explained by the fact that managers do not trust, do not

understand or simply do not use probability estimates when making tactical or strategic

decisions (Kunreuther 1976; Fishhoff et al. 1981).

Since managers were insensitive to probability estimates, they were most likely to

define the risk in terms of magnitude of losses such as ‘‘maximum exposure’’ or ‘‘worst

case’’ instead of a broader scale of compound damage. The second conclusion is based on

an observation that the quality of managerial accomplishments is measured by a set of

performance targets. These metrics cause managers to become more risk averse (or risk

prone) when their performance is above (or below) a desired level. Finally, the third

conclusion is based on the fact that managerial rewards are tied to attainment of ‘‘good

outcomes’’, but not to making ‘‘good decisions’’. The more fragmented and specialised a

given service provision chain becomes, the more focus there is on each actor’s slot in a

value chain and less on system-wide functionality. Consequently, the patterns of mana-

gerial contracts and incentive structures follow this line.

Case studies by Closs and McGarrel (2004), Rice and Caniato (2003) and Zsidisin et al.

(2004) show that the pervasiveness of attitudes undermining the need to deal with the risk of

supply chain disruptions prevented managers from carrying out a risk tolerance appraisal

and assessing a risk tolerance threshold. Yet, a few companies did recognise the importance

of risk assessment and used different methods to measure supply chain risk through formal

quantitative models and/or informal qualitative plans. However, these companies appor-

tioned very little time and few resources to mitigation of all supply risks, not to mention the

risk induced by natural hazards. Several factors have underlain this behaviour.

1. Owing to few data points, good estimates of the probability of occurrence of any

particular disruption were difficult to obtain. This hindered performance of cost/benefit

analyses and/or realistic estimation of losses on returns on the assets damaged and

needed honing of risk reduction skills, holding contingency-reducing assets and

reserve capacity.

2. In the absence of an accurate supply chain risk assessment, firms have generally

underestimated the risk of sequential disruptions. As a consequence, many managers

776 Nat Hazards (2014) 70:767–787

123

ignored the impacts of unlikely events and removed natural hazards from their

strategic decision agenda (Tang 2006). This may explain why so few firms took

commensurate action to mitigate the risk of disruption in a proactive manner. Finally,

as aptly summarised by Repenning and Sterman (2001), firms seldom invest in

proactive programmes because ‘‘nobody gets credit for fixing a problem that never

occurred’’.

Still another explanation could be competitive pressures, the ubiquitous search for

higher operational efficiency and lower capital costs that companies all over the world

pursue with great vigour. Well-known managerial terms such as ‘‘lean production’’ and

‘‘tightly coupled’’ supply chain systems with high intra-channel interconnectivity and

‘‘just-in-time’’ manufacturing and supply regimes reveal that there is not much room for

operational slack, ‘‘wasteful reserves’’ or doubling of sourcing and/or manufacturing

outlets.

In dealing with considerable technical and market uncertainties, both private and state-

owned freight railways adopted lean production techniques which prevent them from

keeping a large stock of locomotives, wagons, spare parts and reserve components as

company possessions.4

As the cost of multisystem locomotives in Europe these days reaches €3 million, while

purchase of a multifunctional rail wagon would require at least €0.5 million, hardly any

small rail undertaking or even a state-owned incumbent has the financial capacity to keep

proprietary equipment and spare part buffers on its balance sheet. Market uncertainty

evidenced by seasonal and corridor-dependent spikes and slopes in demand for freight

transfer means that many new entrants own neither locomotives nor wagons, basing their

entire service provision on time-limited lease contracts with specialty rolling stock and

traction companies. This is fortified by many countries’ depreciation rules in tax code

legislation, where lower tax rates are charged on asset-free service providers.

As a rule, service, maintenance and repairs of rolling stock, traction and IT equipment

are also outsourced to external contractors. As widely admitted in managerial literature and

practice, outsourcing may reduce the current costs of service provision but in return will

also reduce the levels of operational robustness and output security due to delayed

responses and/or longer waiting times for emergency deliveries. As evidenced here, this

way of doing business becomes quite expensive under weather-induced crises.

Pursuit of financial effectiveness means that reserves in production capacity, capital

assets and sourcing duplications—which constitute the core of functional and strategic

preparedness—are deliberately avoided as they show on companies’ balance sheets

reducing operating margins and returns on capital assets. They therefore become hard to

justify to board members and/or shareholders.

Finally, business leaders have many other and equally compelling challenges to attend

to in addition to natural disasters. This was very aptly summarised in the 21st Supply Chain

Digest’s June 2011 edition: ‘‘natural disasters are not the only risks in town’’.

In addition, the Virtual Strategy Magazine (http://www.virtual-strategy.com/2011/05/

10), which published results from the BDO study (http://www.bdo.com) of risk factors

most frequently cited in tax filling reports by the one hundred largest publicly traded US

4 None of the interviewees mentioned a possibility of receiving help from an inter-rail aid arrangementwhere several operators deposit equipment and spare parts for use in emergency situations. This differsevidently from a relatively common practice in the US where railways create aid banks from which spareparts, components and other not-so-often used types of equipment could be leased for swift dealing withemergencies and/or other urgent needs.

Nat Hazards (2014) 70:767–787 777

123

technology companies,5 revealed that the risk of natural hazards ranked as number 12

among the most frequently cited and feared business threat categories. However, the study

also showed that this risk type increased in prominence between 2010 and 2011.

5 Finnish case: linkages between extreme weather, freight train delays and theirduration during 2008–2010

5.1 Analytical model

In order to estimate the proportion of weather-induced delays in all freight train arrivals in

Finland during 2008–2010, the odds for late arrival and duration of delivery lateness, an

analytical regression model was developed (Fig. 1). This utilised results from the four

European cases in assessing the strength of covariation between the key weather param-

eters and the punctuality losses in Finnish rail freight traffic. However, since this modelling

exercise also sought to make more disaggregated assessments of how the different ele-

ments of bad weather, and/or combinations thereof, contributed to train arrival delays and

their duration, the aggregated data indicators representing time series with information on

meteorological conditions on the Finnish rail network and train arrival delays had to be

transformed to attain these objectives.

5.2 Problems with train delay data and weather indicators

Data provided by the Finnish Meteorological Institute (FMI) and VR (the national Finnish

rail operator) as monthly aggregated weather and delay indicators posed a considerable

challenge regarding statistical linking of train arrival delays with extreme weather events.

The first issue was that the monthly weather indicators depicted Finnish weather as

national averages and thus confounded meteorological states on the different days of a

month with weather conditions prevailing over Finland’s northern, coastal, inland and

southern regions. Second, since some of the indicators were specific to areas surrounding

the measurement stations, there was the chance that weather indicators could simply reflect

weather parameters at a given station, and/or a change in number of measurement stations

needed for satisfying a given selection criterion and not the actual meteorological

conditions.

Furthermore, the average temperatures in Finland differ sharply not only between the

southern and northern regions, but also within a given time period. In addition, seasonal

changes in the north and south also occur in a time-lagged fashion. Therefore, the average

records of monthly temperature and precipitation did not provide information on when and

where the weather was the most severe.

Likewise, the freight train delays during 2008–2010 were recorded as minute averages

per monthly periods. This hindered assignment of delays to the different rail corridors,

train trips, days of train operations and/or peak traffic hours. Furthermore, inspection of

freight traffic density on the Finnish rail network in 2010 supported an assumption that the

5 BDO Seidman, LLP is the US professional service firm providing assurance, tax, financial advisory andaccounting services to a wide range of publicly traded and privately held companies. The company’sinternational arm, BDO International Limited, serves multinational clients through a global network of1,138 offices in 115 countries.

778 Nat Hazards (2014) 70:767–787

123

likelihood of freight train delays and their duration were greater on lines with high traffic

volumes compared to those with lesser and/or sparse train movements.

5.3 Addition of data file with lagged weather indicators

Faced with these challenges, the results from European cases helped us recognise that

adaptation to sudden bad weather events may be much more challenging than adaptation to

periodically stable conditions. As a result, we conjectured that the odds for arrival delays

might increase when the weather shifted rapidly and when this shift triggered a chain of

follow-on consequences.

Therefore, to capture the shifts in the levels of weather indicators, one-month-lagged

values were calculated, added to the data representing independent variables and defined as

changes in met-states. Consequently, we hypothesised that a rapid accumulation of snow

on the infrastructure network would probably delay train arrivals more than would snow

cover lying over several days. Therefore, changes in the number of days with a given snow

depth from 1 month to the next were also calculated and served as a proxy for snow

accumulation during the period analysed.6 Analyses of data on changes in snow depth

included only 35 observations, as the snow cover depth record for December 2008 was

missing.

5.4 Data transformation

5.4.1 Dependent variable

We assumed that punctuality would be inversely related to adverse weather. To facilitate

the interpretation of regression results, we decided to study delays defined as a proportion

of delayed trips in all train trips during 2008–2010. A worsening of weather conditions and

an increase in harsh weather indicators thus became related to a positive increase in the

Fig. 1 The study’s analytical model

6 The number of days with a stable snow depth cover could also indicate that measurement was simplyundertaken at the end of the winter season and not the actual depth of snow changing between the differenttime periods.

Nat Hazards (2014) 70:767–787 779

123

proportion of delays in freight train arrivals, while the negative parameter estimates would

imply the opposite.

Generally, it is not a good idea to apply linear regression on a dependent variable

defined as a proportion. Studies of disaggregate data show that these relationships are often

nonlinear and S-shaped and may therefore render negative probabilities (below 0 % delay)

or more than 100 % delay occurrence, rendering the results meaningless.

A standard procedure for analysis of a dependent variable in the form of probability or

proportion is to perform a nonlinear logit transformation by taking a natural logarithm of

the odds for a delay (Hosmer and Lemeshow 1989), which is equivalent to running a

grouped regression model (Agresti and Natarajan 2001; Long 1997).

lndelays

1� delays

� �¼ C + b1X1 þ . . .bnXn

Using the logit transformation of dependent variable means that by powering the value

of an estimated parameter, an estimate is obtained of the odds ratio associated with the unit

increase in the independent variable. This implies that eb1 indicates how much the odds for

a delay would increase when a determinant variable representing a given met-state X1

increases by one unit, producing Xnew = X1 ? 1, while the other variables remain

unchanged. Based on the above, we established a linear regression model that linked the

transformed dependent variable with the independent ones.

5.4.2 Independent variables

The temperature indicators (or the number of days below a given temperature) were

overlapping. Therefore, their values were transformed so that it would be possible to assess

the probability of delay and delay duration as a function of each separate low temperature

interval. Hence, one non-overlapping variable category defined as the change in snow

cover was added to the modelling procedure. Consequently, the depth of snow exceeding

10 cm but not more than 20 cm was used rather than the depth of snow exceeding 10 cm.

A simple bar chart juxtaposing proportions of delayed freight trains against 5 �C

temperature intervals from -15 to ?20 centigrade revealed that delays were not a linear

function of temperature changes despite the fact that the proportion of trains delayed

increased when it became colder (Fig. 2).7

Inclusion of mean temperature indicators in a regression model, in addition to those

showing low temperatures, could easily prove counterproductive, because these two were

strongly inter-correlated. The danger was that this combination could explain random

variation rather than systematic co-variation between low temperatures and train delays.

Therefore, the monthly average temperature over Finland during 2008–2010 was removed

from the equations and, instead, indicators of cold weather, i.e., the number of days below

-7 �C and -20 �C, were used.

7 Given that our case studies assessed the impacts of extreme harsh winter weather only, the impacts ofextremely high summer temperatures and/or seasonal flooding were excluded from model analyses. Thisdecision was supported by a finding that values of time lost (a product of valuations assigned to on-timearrivals and a proportion of train cargo arriving late) were highest during the late autumn and winter seasons,although delays occurred all year long.

780 Nat Hazards (2014) 70:767–787

123

5.4.3 Extraordinary delays from March to May 2010

After having included the number of days with cold temperatures, the amount of snowfall

and the depth of snow cover as independent variables, the fit statistics indicated that the

model’s explanatory factors did not account satisfactorily for the variation in freight train

delays during the 35-month period.

A clustered bar chart showing delays per month over 2008–2010 has indicated that the

reason for the model’s ill-fitting could lie in the unexpected long delays during the rela-

tively warm period from March to May 2010 (Fig. 3).

Fig. 2 Delayed freight train arrivals by monthly average temperatures in Finland 2008–2010

Fig. 3 Delayed freight trains by month, Finland 2010

Nat Hazards (2014) 70:767–787 781

123

When asked, VR explained that delays during these months were caused by the lagged

effects of damage that the harsh winter weather inflicted on rail infrastructure, leading to

imposition of train speed restrictions from March to May 2010.

In addition, it has been established that industrial action in the Finnish transport sector

took place in March 2010, but it was not clear how this event affected late train arrivals,

since during the strike the entire train traffic was suspended. Yet, one could expect that

delays caused by the post-strike accumulation of cargo backlogs imposed the needs for

catch-up with late consignment deliveries. Since the strike started at the beginning of

March and lasted for 16 days, it could still have had a time-lagged effect damaging train

arrival punctuality in April and May 2010.

So, to capture the effects of the time-lagged delays between March and May 2010, a

dummy variable ‘‘Lagged/other’’ denoting ‘‘1’’ for delays during these 3 months, and ‘‘0’’

otherwise, was constructed to neutralise the delaying effects of lower train speed limits in

the aftermath of winter infrastructure damage and impacts of non-weather-related events

(industrial action).

5.5 Modelling results

5.5.1 Proportion of weather-related delays in all train arrivals

The proportion of delays induced by bad weather in all freight train delays in Finland was

determined by deriving delays recorded under thirty-seven freight train delay categories

attributed to bad weather and/or technical damage imposed by these events from an

inventory of eighty different delay categories registered by VR during 2008–2010. Sub-

sequently, an Anova regression analysis was run, which showed that weather-related

delays accounted for 60 % of variation in all train delays. The model’s goodness of fit was

not very high (9.6 %), implying that many other causes contributed to train delays in

addition to bad weather. However, all results were statistically significant (p = 0.00).

5.5.2 Freight train delays and delay duration as a function of bad weather

5.5.2.1 Odds for freight train delays Since several independent variables contained

temperature measurements in one form or another—plus data on snowfalls and snow depth

which were a function of below-zero temperatures—we had to decide which indicators to

use. Given the small data set and the exploratory nature of the study, this decision could not

be based solely on statistical records, but also on the experience of rail freight managers

from Poland, Sweden, Norway and the Netherlands who combated weather disasters and

inputs from VR and FMI professionals. Taking stock of the above, three explanatory

weather variables were included in a regression model assessing impacts of weather

conditions on train punctuality: (1) a dummy variable composed of fixed values of delays

during March–May 2010, (2) the mean number of days with temperatures below -7 �C,

and (3) the change from 1 month to the next in number of days with a snow depth of

10–20 cm. These variables explained 62 % of the variation in reported train delays.

The odds for freight train delays were obtained by calculating eC = 13 %, which

translated into an approximately 12 % probability of delay. The compound lagged effects

of infrastructure damage caused by the 2010 harsh winter and (possibly) industrial action

in March 2010 increased the odds of delays by 75 % and were calculated by multiplying

the odds associated with the fixed values of these variables by e0.557 = 1.77 or approx.

175 %. A unit increase in the number of days with 10–20 cm of snow cover from 1 month

782 Nat Hazards (2014) 70:767–787

123

to the next has raised the odds for freight train delays by about 8 %. Similarly, each

additional day of temperature below -7 �C increased the odds for train delay by about

3 %. The increases in odds were calculated ‘‘ceteris paribus’’, i.e. under the assumption

that the values of the remaining variables remained unchanged.

5.5.2.2 Duration of weather-attributed delays Subsequent models analysed the duration

of delays associated with the different delay causes coded by VR. Several weather-related

causes were initially inserted in regression models, but only results from models testing the

relationships between delay determinants and delay durations, which were both significant

and made good sense, are reported here. Good sense meant that they were in accord with

managerial assessments of freight train delay-causing factors and their combined impacts

on service punctuality. As a consequence, only the three linear ordinary regression models

presented below assessed duration of train delays as a function of three independent

variables (and their interactions) representing changes in weather conditions.

Delays attributed to fog, cold weather and leaves on the track

One additional mean day in the number of days with 10–20 cm snow cover compared to

the previous month emerged as a significant explanatory factor. The model including

variables denoted as ‘‘change in the snow cover’’ explained 66 % of the variation in freight

train delays. A unit increase in the average number of days with snow cover between 10

and 20 cm from 1 month to the next might contribute to 629 min or 10.5 h’ duration of

train delays attributed to this weather condition.

Delays attributed to snow barriers

A unit increase in the number of days with snowfalls over 5 mm explained 77 % of the

variation in train delays attributed to snow barriers. This implies that each additional mean

day with this snowfall may contribute to an additional train delay of 19.5 h.

Delays attributed to faults at switch stations

A unit increase in the mean number of days with more than 5 mm snowfall might

contribute to an additional 342 min in freight train delays or nearly 6 h delayed train

arrival. Each additional day with the temperature -20 �C or below might increase the

duration of train delays by 193 min or roughly 3 h and 15 min. This regression model

explained 79 % of the variation in duration of train delays attributed to faults in track

switches.

5.6 Conclusions from the Finnish modelling case

The runs of univariate ANOVA regression model revealed that about 60 % of all delayed

arrivals could be attributed to bad weather and/or weather-related technical damage to the

network.

Assessment of the statistically valid relationship between bad weather and the occur-

rence and duration of freight train delays was difficult because of the small data set and

aggregation of weather and delay records as monthly averages. For establishing the odds

for delays, a dependent variable was converted into the log odds through nonlinear logit

transformation. This allowed the odds for the occurrence of a delay to be calculated. As

regards the independent variable, the monthly changes in number of days with a given

snow depth and snowfall combined with the occurrence of negative temperatures were

used as model parameters. This explained 62 % of the variation in occurrence of the

reported train delays. Afterwards, three linear regression models assessed the strength of

the statistical co-variation between the monthly changes in the number of days with snow

Nat Hazards (2014) 70:767–787 783

123

depth of 10–20 cm, the different categories of bad weather, the weather-related-infra-

structure and/or rolling stock damage, and duration of freight train delays.

The three models explained, respectively, 65, 77 and 79 % of the variation in duration

of freight train delays. It also appeared that train delays attributed to snow barriers were

the most severe impediments to punctuality as these might have contributed to an addi-

tional 19.5 h of late arrival.

These results indicate that, statistically, it was quite difficult to establish clear causal

relationships between bad weather and occurrences of freight train delays and/or bad-

weather-induced technical problems affecting delay duration. However, this does not mean

that such a relationship could not be detected with better quality of met-data and train delay

counts and generally a better understanding of interactions between changes in weather

conditions and freight train arrival dynamics.

6 Suggestions for future work

Based on the results presented here, we admit that obtaining empirically sound results took

more than just a simple analysis of raw meteorological weather indicators and rail delay

data. Without prior knowledge of managerial experience from coping with bad weather

disasters, this task could not have been accomplished meaningfully. In order thus to reduce

the risks of future delays and improve the punctuality of freight trains, a better match

between weather indicators and data on rail operations is needed. This in turn would

require a continuous dialogue between the met-professionals, rail infrastructure managers

and traffic supervisors. Unfortunately, this collaboration was not feasible within the current

study.

Therefore, to improve our understanding of how the different classes of adverse weather

may affect the punctuality of freight train operations, five suggestions are presented below.

To assess the impact that various weather conditions exert on train punctuality and

devise models discerning the effects of low temperatures, wind, icing, snowfalls and snow

accumulation, more detailed data are needed. To attain this goal, rail practitioners need to

assess in detail how the different categories/combinations of bad weather affect the

punctuality of train traffic. In research on weather impacts on humans, a compound

indicator ‘‘chill factor’’ is often used which measures how strong winds, low temperatures

and precipitation conjointly affect the human body. Could a similar ‘‘chill factor’’ be

constructed to measure how low temperatures, precipitation and infrastructure exposure to

bad weather affect the punctuality of rail operations? Could track clogging by leaves, icing

of switch stations, snow accumulation on rail wagon undercarriages or other traffic

impediments be captured through more comprehensive indicators? Would a change in

snow depth from 1 h to the next, 1 day to the next or 1 month to the next be a better time

frame for defining how the different components of adverse weather may threaten freight

train arrivals?

Furthermore, in order to establish which combinations of snowfall and low temperature

critically damage infrastructure and rail operations, these insights need to be discussed with

met-experts and people with knowledge of rail infrastructure topography and topology

across different regions and/or countries.

A second step would involve design and calculation of multi-hazard scorecards

incorporating conjoint impacts of the snowfalls, strong winds, hail and low temperatures

which collectively immobilise rail infrastructure and operations. By adding built-in weights

to multiple hazard components, these indices could be tailored to the different geographical

784 Nat Hazards (2014) 70:767–787

123

areas, i.e. enabling how the shifts in weather conditions might affect traffic on particular

networks, corridors or lines crossing the exposed territories.

The third step would encompass coding the different delay causes more precisely so that

they could be better linked to specific weather conditions. Data on the different classes of

bad weather developed at the first stage could be used here. Since the analyses reported

above indicate the immediate/acute, mediated, lateral and time-lagged types of weather

impacts on infrastructure and train operations, indicators of short-term and lasting damage

need to be developed and connected with the different bad weather categories.

The fourth step would involve making sure that sufficiently large data sets were

available for determining the relationships between more detailed model parameters.

Ideally, data on varying weather conditions on networks cutting through the different

geographical regions would be desired. In this manner, the need for long time observation

series might be reduced, but not entirely. Inter-temporal differences are not the same as

inter-regional differences during the same time. However, with data on both the inter-

temporal and inter-regional differences available, the partial impacts of bad weather

components and their interactions could be established to determine whether the composite

weather indicators from the primary steps were more useful for explaining causes of train

delays compared to single factors.

The fifth and final step would be application of multivariate regression and/or structural

equation models assessing how train arrival delays could be causally linked to multi-hazard

weather impacts and forecast for different settings.

As suggested, a dialogue between the met-scientists, behavioural researchers and rail

operations managers whose daily experience of coping with weather-related and other

threats to train punctuality is needed. This collaboration would provide guidance for data

definitions, data processing, designing of analytical models and formulation of empirically

validated findings.

7 Discussion

Conclusions from railway cases from Poland, Finland, Switzerland, Sweden and Norway

are relatively clear. Managers from the logistics and rail freight industry are well aware of

the impact of extreme events on their business and reputation, but the tools and means

mitigating them are limited. Industry leaders and managers give priority to other issues

than weather, which in the end is an ‘‘act of God’’ and not something they can influence

through better business skills. They expect and hope, however, that infrastructure managers

improve their performance by having pathways clear during, or at least soon after, an

incidence of extreme weather.

However, it would be wrong to conclude that rail and logistics business managers do not

care or are not prepared for extreme events. The simple fact is that these issues are not too

high on their agenda. One obvious reason could be that such preparations are not included

in management contracts—it is shareholder value that matters, not necessarily smooth and

uninterrupted operations. And, with good reason, it may be questioned whether service

functionality should not be included in management and performance contracts of public

sector executives, namely infrastructure managers. If neither side assumes responsibility,

the reliability of rail-based logistic operations may continue to falter.

The modern lean production thinking that shuns reserves in capital, people, equipment

and inventories provides little room for preparedness-building. The higher the efficiency

requirements on business and public organisations, the less room there will be to prepare for

Nat Hazards (2014) 70:767–787 785

123

extreme weather, at least at operational level. However, on a strategic level, the situation

could be otherwise. Production locations, topography and topology of global logistics

channels and gateway hubs, contractual guarantees on supply chains effectiveness, inventory

availability and similar decisions are all examples of strategic choices into which long-term

resilience-building could be incorporated. However, as these judgments have profound long-

term financial implications, they need to be taken by the highest corporate echelons.

The results of the modelling exercise were also instructive. Cold days (below -20 �C)

and snowfalls (C5 mm) explained the majority of winter time freight train delays in

Finland, and when these phenomena are known with reasonable certainty a few days

beforehand—as they usually are—worthwhile preparations can take place. However, even

when the knowledge is there, this does not necessarily mean that it will be taken into

practice unless the benefits are explicit, measurable and long-term. Infrastructure managers

may need to understand that rail operators will lose freight carriage to road hauliers

because shippers’ competitiveness depends on reliability of the goods supply. Market

losses of rail companies had time-lagged scope and duration which by far exceeded the bad

weather occurrence. This will not only affect rail infrastructure usage but also the entire

socio-environmental profile of European freight transportation.

Modelling of the relationship between changes in met-states and freight train delays

revealed a large mismatch between the type and quality of the available data and the need

for rail operators to assess probability of weather-related train arrival delays. Paradoxi-

cally, it appeared that icing of traction lines, heavy snowfalls and low temperatures may

damage the market position of environmentally friendly rail transport much more than

other service shortages. Hence, to ascertain that weather and/or other natural hazards do

not harm rail freight competitiveness, it is necessary to understand how these adversities

affect transport and logistics operations as well as how the managers in these industries

deal with natural hazard risks. Therefore, in order to prevent immediate and long-term

damage, the met-information needs not only to be on time, but also in a format that will

help business people incorporate weather risk forecasts in building up long-term resilience.

To this end, the current modelling exercise was quite useful. We hope that it may

provide guidance when transportation impacts of weather phenomena from other clima-

tological zones are being assessed. Winter phenomena in Finland, Norway, Poland,

Sweden and Switzerland are obvious but heat waves are not. In southern Italy, draught

conditions could pose similar challenge.

Needless to say, both transport and meteorological/climatological scientists need to join

forces to successfully accomplish these tasks.

Acknowledgments Our gratitude goes to Dr Pekka Leviakangas from VTT, Finland, who scientificallycoordinated the EU-co-funded EWENT project (Extreme Weather Impacts on European Networks onTransport), which provided funding and opportunity to perform studies reported in this article. We alsorecognise and acknowledge professional help of Marko Nokkala and Anna-Maija Hietajarvi from VTT asregards facilitation of data provision by VR Group Ltd, the Finnish Transport Agency and the FinnishMeteorological Institute. We thank the Research Council of Norway for funding the Infra-Risk project fromwhich our work also benefitted. Finally, we thank an anonymous reviewer who helped us to clarifyimportant issues related to the modes how the European rail freight rail system functions.

References

Adegoke O, Gopalakrishnan M (2009) Managing disruptions in supply chains: a case study of a retail supplychain. Int J Prod Econ 118:168–174

786 Nat Hazards (2014) 70:767–787

123

Agresti A, Natarajan R (2001) Modelling clustered ordered categorical data: a survey. Int Stat Rev69(3):345–371

BDO Study (2011) Disclosure and management of climate impacts. Virtual Strategy Magazine, October 5(http://www.virtual-strategy.com)

Bundschuh M, Klabjan D, Thurston DL (2003) Modelling of robust and reliable supply chains. OrganizationOnline e-print. htt://www.optimization-online.org. Department of Mechanical and Industrial Engi-neering, University of Illinois at Urbana-Champaign, Urbana, IL

Closs D, McGarrel E (2004) Enhancing security through the supply chain. IBM Center for the Business ofGovernment. Special Report Series

Delmonaco G, Margottini C, Spizzichino D (2006) ARMONIA methodology for mulit-risk assessment andthe harmonization of different natural risk maps. Deliverable 3.11., ARMONIA

Fishhoff B, Lichtenstein S, Slovik P, Derby S, Keeney R (1981) Acceptable risk. Cambridge UniversityPress, New York

Gunipero LC, Eltantawy RA (2004) Securing the upstream supply chain: a risk management approach. Int JPhys Distrib Logist Manag 9(34):698–713

Hendricks KB, Singhal VR (2005) An empirical analysis of the effect of supply chain disruptions on long-term stock price performance and risk of the firm. Prod Oper Manag 14:35–52

Holmgren AJ (2007) A framework for vulnerability assessment in electric power systems. In: Murray A,Grubesic T (eds) Critical infrastructure: reliability and vulnerability. Springer, New York

Hosmer DW, Lemeshow S (1989) Applied logistic regression. Wiley, New YorkKappes MS, Keiler M, von Elverfeldt K, Glade T (2012) Challenges of analyzing multi-hazard risk: a

review. Nat Hazards. doi:10.1007/s11069-012-0294-2Kleindorfer PR, Saad GH (2005) Managing disruption risks in supply chains. Prod Oper Manag 14(1):53–68Kunreuther H (1976) Limited knowledge and insurance protection. Public Policy 24:227–261Lai KH, Ngai EWT, Cheng TCE (2002) Measures for evaluating supply chain performance in transport

logistics. Matekon 13:35–49Long SJ (1997) Regression models for categorical and limited dependent variables. Sage Publications,

LondonMarch J, Shapira Z (1987) Managerial perspectives on risk and risk taking. Manage Sci 33:1404–1418National Cooperative Freight Research Program (2011) Performance measures for freight transportation,

transportation research board, report 10Qiang Q, Nagurnet A, Dong J (2009) Modelling of supply chain risk under disruption with performance

measurement and robustness analysis. In: Wu T, Blackhurst J (eds) Managing supply chain risk andvulnerability. Springer, New York

Repenning N, Sterman J (2001) Nobody ever gets credit for fixing problems that never happened. CalifManage Rev 43:64–88

Rice B, Caniato F (2003) Supply chain response to terrorism: creating resilient and secure supply chains.Supply chain response to terrorism project interim report. MIT Center for Transportation and Logistics.MIT, Massachusetts

Sheffi Y (2001) Supply chain management under the threat of international terrorism. Int J Logist Manag12(2):1–11

Sheffi Y (2005) The resilient enterprise. Overcoming vulnerability for competitive advantage. MIT Press,Cambridge

Snyder M, Swann WB (1978) Hypothesis-testing processes in social interactions. J Pers Soc Psychol36(11):1202–1212

Supply Chain Digest (2011) Managing risk in a multi-tier supply chain. Demand Video cast, June 30th athttp://www.sctvchannel.com/webinars/videocast3.php?cid

Svensson G (2002) A typology of vulnerability scenarios towards suppliers and customers in supply chainsbased upon perceived time and relationships dependencies. Int J Phys Distrib Logist Manag32(3):168–187

Tang CS (2006) Perspectives on supply chain risk management. Int J Prod Econ 103:451–488Wilson MC (2007) The impact of transportation disruptions on supply chain performance. Transp Res E

43:295–320Yin RK (1994) Case study research: design and methods. Sage Publications, Thousand OaksZsidisin G, Elram L, Carter J, Cavinato J (2004) An analysis of risk assessment techniques. Int J Phys

Distrib Logist Manag 34(5):397–413

Nat Hazards (2014) 70:767–787 787

123