1. Introduction
High temperatures and heat waves can cause numerous problems for railway infrastructure. In the United Kingdom, most railway assets show increased failure rates at even moderately low temperatures, such as 20°C, with incident rates increasing significantly from 26°C (Network Rail 2014a, 2015). These temperatures are well within the climatological norms for the United Kingdom (Barrow and Hulme 2014) and also well within the required operational envelope of the railway infrastructure. Given that high temperatures and heat waves such as those experienced in Europe in 2003 are predicted to become increasingly common (Coumou and Robinson 2013), understanding the nature of these heat-related incidents is of clear importance for Network Rail and for the four million passengers who use the railway network on a daily basis (Network Rail 2016).
Heat affects different asset types in different ways. For example, temperatures above 30°C are often associated with track-buckling incidents (Network Rail 2014a). Network Rail own and operate Britain’s rail infrastructure, incorporating over 20 000 miles of track, more than 2500 railway stations, and a wide variety of signaling, telecommunication, and overhead line equipment. The majority of the track consists of continuously welded steel rails fixed to concrete beams (sleepers or crossties) with stone ballast around and below. In direct sunshine on hot days, the rails can be up to 20°C warmer than ambient temperature (Chapman et al. 2008). When steel warms, it expands, increasing compressive stress and potentially causing buckling and possibly derailment. To reduce the risk of buckling, the tracks are prestressed to 27°C. However, if the temperature is very high, or if the track is in poor condition, track buckles can still occur, often driven by the additional energy provided by a passing train (Dobney et al. 2010). To reduce the potential for track buckling and to minimize the severity of a consequential derailment, speed restrictions are introduced at specific critical rail temperatures, which vary depending upon the nature and condition of the track [a comprehensive list is given in Palin et al. (2013)]. Blanket speed restrictions can also be introduced where there is doubt or insufficient data availability or when ambient temperature exceeds 36°C (Chapman et al. 2008). Blanket speed restrictions are applied ubiquitously over large areas. These speed restrictions put passenger safety first, but unfortunately lead to passenger disruption and delay and cost Network Rail money in the form of compensation payments to the passenger and freight train operating companies that use the infrastructure. Several studies have considered the incidence and cost of track buckles under a future warmer climate (e.g., Baker et al. 2010; Dobney et al. 2009; Palin et al. 2013), and without targeted adaptation and/or mitigation, the costs associated with heat-related delays are projected to double in the future (Dobney et al. 2010; Thornes et al. 2012).
Thermal expansion can also affect the overhead lines that supply electricity to trains via a pantograph. If the overhead lines expand, their tension reduces, which can lead to excessive line sag. This can cause the pantograph to disconnect. Most new overhead lines have a design temperature range up to 38°C and are autotensioned via a pulley system that adjusts the tension with temperature variation (Palin et al. 2013). However, for older equipment that does not have the automatic tensioning of the more modern equipment, or poorly maintained equipment, or at the joins between older and new equipment, heat-related problems can occur. The incidence of overhead line sag is also higher in urban areas (RSSB 2015), because of the urban heat island effect, which can result in urban areas being several degrees warmer than their surroundings (Oke 1973). Cities are often critical transport nodes, and a failure here can be very problematic, quickly propagating through the rest of the railway network (Chapman et al. 2013).
Far less well studied is the impact of heat on signaling and communication assets. These include a broad range of equipment, such as stand-alone continuously manned signal boxes, remote relay and signaling rooms, and lineside location cases. In contrast to overhead line assets, it is modern types of signaling equipment that tend to be more susceptible to heat-related failure because of their increased dependence on electric and electronic components (RSSB 2015). Assets in direct sunlight are most vulnerable to overheating, and in particular the equipment inside location cases can experience more rapid changes in temperatures and higher temperature extremes than those that occur outside (RailCorp 2012). Indeed, little is known about the temperature within location cases and its relationship to ambient air temperature (Thornton et al. 2011). Other heat-related problems affecting the railway network include a reduced opportunity to undertake track maintenance (for this may not take place over an ambient temperature of approximately 32°C); staff exposure to heat stress; and passenger and freight risk from train failure in extreme weather (Palin et al. 2013).
This study examines industry fault data to characterize the nature of heat-related incidents in southeast England between 2006 and 2013. Research into the effects of weather on the railway infrastructure are scarce (Koetse and Rietveld 2009), and there is no clear understanding of how the day-to-day failures of all asset types may be influenced by heat. In particular, the temporal analysis considers the concept of failure harvesting (Chapman et al. 2008). It is hypothesized that a greater number of heat-related incidents will occur earlier in the summer season (summer is defined by the industry as April–September). Once these failures have been harvested, and the failed equipment replaced, the infrastructure system within that particular region will become resilient for the remainder of the year at that particular temperature. Consequently the resilience of the railway network will improve over the course of each summer season, and, theoretically, once the maximum temperature is reached, no more heat-related incidents should occur in the particular location.
An improved understanding of heat-related incidents will improve the evidence base for strategic decision-making for the industry [e.g., for “Weather Resilience and Climate Change Adaptation Plans” produced by Network Rail (2014b)] and, in turn, will facilitate better heat risk management, ultimately to improve the service for passengers and freight alike. It will also provide the robust evidence baseline that is required by climate change impact studies, such as Palin et al. (2013).
2. Method
a. Study area
For management purposes, the Great Britain railway network is divided into several smaller regions, known as routes (Fig. 1). This study focuses on four routes that are located in the relatively warmer, drier, and sunnier southeast of England (Mayes 2013): namely, Anglia, Kent, Sussex, and Wessex. These four selected routes have more than twice the number of days per year where their daily maximum temperature exceeds Network Rail operational thresholds for heat compared to the U. K. average (Network Rail 2014a), and it follows that heat-related failure incidents will occur more frequently here than other parts of the United Kingdom.
The insert shows Great Britain divided into different routes for operational management by Network Rail. The main map shows the extent of the Anglia, Kent, Sussex, and Wessex routes and the weather stations used in this study.
Citation: Weather, Climate, and Society 8, 2; 10.1175/WCAS-D-15-0068.1
The four routes also include key sections of the London railway network servicing the major stations of Waterloo, Kings Cross, St. Pancras, Victoria, and Liverpool Street. These are critical transport hubs, and the interdependent nature of the railway system means that asset failure near or at these nodes can quickly propagate throughout the rail network, causing extensive delays, passenger dissatisfaction, and a disproportionate additional cost compared to that of the original fault itself. For example, Network Rail can potentially repair an electrical failure near Waterloo in six minutes, but this can result in a total delay of three hours (Network Rail 2013). Consequently, understanding the nature of heat-related incidents along such critical sections of track is essential.
b. Temperature data
To explore the failure-harvesting hypothesis, the fault incidents were linked with ambient maximum temperature data recorded by weather stations that are part of the Met Office Integrated Data Archive System (MIDAS) Land and Marine Surface Stations Dataset (BADC 2014; Fig. 1). This dataset was selected after considering data availability and the type of measurement appropriate for the scale of analysis and for decision-making procedures at Network Rail. Ideally the hypothesis would be explored using the asset temperature at point of failure; however, Network Rail do not routinely monitor asset temperature, nor would it be economically feasible given the size of their infrastructure assets.
Ambient temperature data are indicative of the daily weather conditions but not of the temperature of the asset. For example, following a recent overhead line malfunction at Stratford, East London, the asset temperature was recorded at 38°C: 10°C warmer than the ambient temperature observed at the nearest weather station at London City Airport (RSSB 2015). Although, it is possible to estimate track temperature from ambient temperature using the approximation Trail ≈ 3/2 Tair (Hunt 1994), in practice this is too simplistic, as evidenced by Chapman et al. (2006), who measured an actual temperature range of 39°C using thermal imaging along a short section of track. Chapman noted that much of this variation in temperature depended on shading by vegetation or infrastructure, such as buildings or tunnels.
Therefore the temperatures of railway assets, such as track, overhead lines, or communication units can vary significantly depending upon the specific location and type of asset and may well be different from the temperature recorded at the nearest weather station. Therefore, there is no advantage to using weather data from the nearest station or the nearest group of weather stations, as all are unlikely to be representative of the asset temperature. This is particularly true given the range of asset types (e.g., track, signaling, and buildings) considered in this study, which could be located in urban or rural locations and influenced by any number of external factors, such as shading or exposure. Instead, a daily maximum temperature value (Tmax) was calculated for each route by taking an average of the daily temperature maximum recorded at several weather stations across each route (Fig. 1). While not representative of the asset at point of failure, Tmax is indicative of the daily maximum ambient temperature at the route level at which Network Rail operational decisions are made. The maximum distance between the weather station and the asset is 27 km.
c. Network Rail datasets
Network Rail collects and archives information on equipment failures and other incidents on its proprietary rail infrastructure of Great Britain and the consequential delay minutes and/or financial cost that failures or incidents may incur. These databases are generally underutilized for research purposes (Jaroszweski et al. 2015); however, they contain a wealth of temporal and spatial information that can be combined with meteorological datasets to obtain insights into the impact that weather has on rail transport. Here, two databases are used: the Fault Management System (FMS), which includes all failures and incidents logged in real time at route level; and the Train Running System (TRUST), which compares the scheduled and observed train times in order to log delays in terms of minutes, from which financial costs can be derived. The two databases can be linked by a common identifier; however, only those FMS incidents that incur a cost in terms of finance or delay minutes are recorded in TRUST. Unfortunately, this linkage between FMS and TRUST is incomplete because of ambiguities and incomplete information in the free text link fields. This will result in some underreporting of delay minutes/costs for heat-related events. Previous studies that have used these or similar databases for weather/climate research include the following: Dobney et al. (2009, 2010), which quantified the effects of higher summer temperatures as a result of climate change on the railway network; Jaroszweski et al. (2015), which described the impact the 28 June 2012 storms had on the railway network; and Palin et al. (2013), which examined the impact that climate change may have on several temperature-related issues, including track buckling, worker heat stress, sag of overhead lines, and delays to track maintenance. The databases are also used for internal analyses and research undertaken by Network Rail (e.g., Network Rail 2014a, 2015).
In this study, the descriptions of individual faults and incidents in the FMS database (2006–13) were searched using an algorithm specifically designed to identify failures or incidents related to heat. The algorithm searched for key words and phrases related to heat, such as buckle, temp, expansion, hot, thermal, and so on. The incidents selected by the algorithm were carefully examined by eye and also cross-compared with other Network Rail databases holding information on emergency speed restrictions (ESRs) and weather-related performance, in order to remove erroneous or ambiguous data. This was particularly important for track incidents in order to distinguish between those due to faults or failures caused by heat and those due to preventive speed restrictions. It is important to note that the latter are not included in the failure-harvesting analysis, for these do not represent infrastructure failure. The meticulous selection process ultimately produced a high-quality dataset of faults and failures that could be clearly attributable to heat. In reality it is likely that the actual number of heat-related incidents will be significantly higher than calculated by this study, because some heat-related incidents were discarded because they were ambiguously recorded, or heat may not have always been listed as a contributing factor by the Network Rail engineers, not least because heat may not have been the obvious cause of the incident.
3. Results
Over 340 000 faults and incidents that occurred in southeast England between 2006 and 2013 were searched using the algorithm. Of these, approximately 850 could be unambiguously classified as heat related from the description of the incident recorded by the Network Rail engineer. The majority of incidents related to either signaling (57%) or track (20%) assets (Fig. 2a). Figure 2b shows the ambient maximum air temperature Tmax at which subclasses of signaling and track assets failed in southeast England. Note that heat-related incidents for signaling assets were recorded on days where the daily Tmax was as low as 8.2°C. The highest recorded daily Tmax for signaling assets was 31.7°C, and for track assets it was 33.4°C.
(a) The proportion of heat-related incidents by asset type between 2006 and 2012. Incidents from 2013 are not included, as asset type information is not always available. (b) The ambient maximum air temperature Tmax at which track and signaling assets failed in southeast England between 2006 and 2013. It should be noted that the Tmax values were calculated on a daily basis at route level, as outlined in section 2b. The top and bottom of the boxes represent the first and third quartile, respectively, and the band inside the box shows the median. The ends of the whiskers represent minimum and maximum Tmax values for each asset type. The box width represents the incident count; the scale is shown in the top-left corner.
Citation: Weather, Climate, and Society 8, 2; 10.1175/WCAS-D-15-0068.1
a. Temporal analysis (failure harvesting)
Figure 3 shows the total number of heat-related incidents for each month between 2006 and 2013 compared to the monthly maximum Tmax for the Anglia route. Equivalent data for Kent, Sussex, and Wessex are shown in Table 1. For all routes and years, the majority (Anglia 77%; Kent 72%; Sussex 76%; Wessex 79%) of heat-related incidents occur before the date on which the annual Tmax value was recorded. Indeed, in years like 2006 there is a significant reduction in the number of heat-related incidents later in the summer season for all four routes, despite equivalently high temperatures continuing into the summer season. A similar drop-off can be seen in 2009 and 2013 for Anglia, Sussex, and Wessex. Combining the data for all routes and years clearly demonstrates the reduction in incidents from August onward (Fig. 4). This is in line with the failure-harvesting hypothesis, whereby the resilience of the infrastructure system improves incrementally for each temperature increase as the summer season progresses and failed equipment is replaced. Theoretically, once the maximum temperature is reached, no more heat-related incidents should occur. However, in reality, a few late incidents can be seen in September and October (Figs. 3 and 4). The high number of track-related incidents in July 2006 compared to other years is also notable; this month was particularly hot across the Anglian route, with an average daily maximum temperature across the region of 27°C, which is 5°C warmer than the 2006–13 mean and 5.5°–6.5°C above the 1961–90 average daily maximum temperature (Prior and Beswick 2007). This is likely to have contributed to the high number of track incidents.
The number of heat-related incidents each year in Anglia between 2006 and 2013 as compared to the monthly maximum Tmax value; Tmax was calculated on a daily basis for each route by taking an average of the daily maximum temperature from all operational weather stations. Note that results from 2006 begin in April. Equivalent data for Kent, Sussex, and Wessex are shown in Table 1.
Citation: Weather, Climate, and Society 8, 2; 10.1175/WCAS-D-15-0068.1
Summary tables for: (a) Anglia; (b) Kent; (c) Sussex; and (d) Wessex, comparing the total number of monthly incidents to the route-level monthly maximum temperature Tmax (°C). This information is represented graphically for Anglia on Fig. 3.
The total number of heat-related incidents for signaling and track assets between 2006 and 2013 for all routes.
Citation: Weather, Climate, and Society 8, 2; 10.1175/WCAS-D-15-0068.1
Figure 5 shows the heat-related failure rates (incidents per week) before and after the date of the annual maximum Tmax for all routes and years, and Table 2 contains the p values following Welch’s unequal variances t test. With the single exception of Sussex in 2009, the rate of heat-related incidents after annual maximum Tmax is consistently less than before the annual maximum Tmax. This is also in line with the failure-harvesting hypothesis.
The heat-related failure rates (incidents per week) before and after the date of the annual maximum Tmax for the four routes from 2006 to 2013. Mean failure rates and p values following Welch’s unequal variances t-test outcomes are shown in Table 2.
Citation: Weather, Climate, and Society 8, 2; 10.1175/WCAS-D-15-0068.1
The mean failure rates before and after the date of the annual maximum Tmax and the p values following Welch’s unequal variances t test for the four routes. The heat-related failure rates (incidents per week) for the four routes from 2006 to 2013 are shown in Fig. 4.
b. Heat-related costs and delays
Heat-related impact in terms of delay minutes or financial costs can result from heat-related incidents, as discussed in section 3a, or as a consequence of the introduction of ESRs in order to reduce the likelihood and impact of track-buckling events. Figure 6 compares these costs for Anglia from 2006 to 2012, and Table 3 provides equivalent information for Kent, Sussex, and Wessex routes. For all asset types, and in line with the failure-harvesting hypothesis, the vast majority of heat-related delays and costs occur before the annual maximum Tmax. The same is true for track assets in Anglia, Sussex, and Wessex. For Kent, a greater percentage of heat-related delays and costs take place after the annual maximum Tmax, although the sample size for this route is very small. Concerning the ESRs, all routes continue to impose these later into the summer season, when the number of heat-related incidents has reduced as a result of failure harvesting. Indeed, a disproportionate number of delay minutes and delay costs associated with ESRs occur after the annual maximum Tmax. For example, in Wessex, nearly half the delay costs and minutes associated with ESRs took place after all heat-related track incidents had taken place. This mismatch between the occurrence of heat-related track incidents and the ESRs designed to reduce them suggests that there is scope for Network Rail to improve the efficiency of their heat risk management procedures.
The proportional heat-related costs in terms of delay minutes and financial cost for (left) all heat-related incidents, (middle) heat-related incidents for the track assets only, and (right) emergency speed restrictions for Anglia from 2006 to 2012 before, on the day of, and after the maximum Tmax value. The actual number of delay minutes is also given.
Citation: Weather, Climate, and Society 8, 2; 10.1175/WCAS-D-15-0068.1
The proportional heat-related costs in terms of delay minutes (min) and financial cost for all heat-related incidents, heat-related incidents for the track assets only, and emergency speed restrictions for all routes from 2006 to 2013 before, on the day of, and after the maximum Tmax value. The actual number of delay minutes is also given.
Figure 7 compares the number of delay minutes associated with heat-related failures and incidents for signaling assets, track assets, and ESRs for all routes between 2006 and 2013. There are clear differences between the different routes. For example, Anglia and Kent have more delay minutes resulting from ESRs than actual heat-related incidents. In contrast, there are only a small number of delay minutes from ESRs in Sussex. Kent has very few delay minutes associated with track incidents. With the exception of Anglia, the delay minutes from heat-related incidents associated with signaling assets are greater than those associated with track assets.
The heat-related impact in terms of delay minutes associated with heat-related failures and incidents for signaling assets, track assets, and emergency speed restrictions for all routes between 2006 and 2013. Proportional amounts (%) are also shown on the columns.
Citation: Weather, Climate, and Society 8, 2; 10.1175/WCAS-D-15-0068.1
c. Spatial analysis of heat-related delays
Figure 8a shows the total number of delay minutes for each asset location for all heat-related incidents between 2006 and 2013: 53% of locations accumulated less than 1 h delay; 21% accumulated 1–3 h of delay; 19% accumulated 3–12 h of delay; and a remaining 7% of locations accumulated over 12 h of delay up to a maximum of 70 h at a particular location in the Anglia route area. Figure 8b shows the 19 locations that accumulated more than 12 h of delays between 2006 and 2013. Of these, 10 locations are located within Greater London, where the density of infrastructure and passengers is high. Other locations are rural. For example, at a location in northeast Anglia (Fig. 8b, annotated with *) there were 14 heat-related incidents at the same location between 2006 and 2012, accumulating 1390 delay minutes. Inspection of FMS data shows this to be a swing bridge that repeatedly fails as a result of thermal expansion. Repeated failures can also be seen at a location near Woking (Fig. 8b, annotated with **), where signaling equipment failed 13 times between 2006 and 2012. This is a known problem site with respect to heat for Network Rail. Indeed, 11 of the sites recorded more than five heat-related incidents over the 8-yr analysis period, suggesting certain locations are repeatedly vulnerable to heat.
(a) The total number of heat-related delay minutes between 2006 and 2013 in southeast England. (b) The location of heat-related delays lasting over 12 h. These are colored to indicate whether the delay resulted from a single heat-related incident or multiple incidents. Those locations mentioned in the text, one in northeast Anglia and one near Woking, are annotated with an * and ** respectively.
Citation: Weather, Climate, and Society 8, 2; 10.1175/WCAS-D-15-0068.1
In contrast, five of the delays lasting over 12 h were caused by a single heat-related incident. One incident was attributed to a location box overheating; unfortunately, there is insufficient evidence on the exact nature of the other single heat-related incidents.
4. Discussion
a. Insights into heat-related incidents
This analysis provides several new, and potentially important, insights into the nature of heat-related incidents on the railway network. First, Fig. 4 clearly shows that the majority of heat-related incidents happen in the early to midsummer season (summer is defined by the industry as April–September) and then reduce significantly, despite the temperature remaining high. This is further supported by Fig. 5 and Table 2, which show a significant reduction in the weekly incident rate from before the annual maximum Tmax (i.e., early/midsummer) to after the annual maximum Tmax (late summer). This is consistent with the failure-harvesting hypothesis and suggests the infrastructure system is becoming increasingly resilient throughout the summer as faulty equipment is repaired or replaced. Late incidents in September or October may arise from local heterogeneities in temperature caused by localized variations in shading. Chapman et al. (2006) demonstrated that shading from man-made structures or trees, the extent of which can vary with the time of day and time of year, is the major cause of spatial variation in track temperature. The role of diurnal temperature change also requires further investigation; during the first cold spells of autumn, the drop in rail temperatures from daytime highs of up to 25°C to very low temperatures at night can lead to track breakages. Indeed, Network Rail record increased incident rates and costs when the diurnal change is greater than approximately 12°C (Network Rail 2014a). The converse may prove problematic in spring, and such a rise then fall in temperature is the likely cause of the signaling incident that occurred on a March day when the daily Tmax was only 8.2°C (Fig. 2b). Alternatively, the late incidents could also result from locally varying maintenance regimes or external factors, such as people. However, the indisputable skewing of heat-related incidents toward the early/midsummer (i.e., before the annual maximum Tmax) has implications for current operational procedures at Network Rail, as discussed in section 4b.
Second, the analysis has clearly highlighted the vulnerability of signaling assets to heat. Previous research into the current and future impacts of heat on the railway (e.g., Dobney et al. 2009, 2010; Gardiner et al. 2009; Palin et al. 2013; Thornes et al. 2012) have focused almost entirely on the problem of track buckling, but in southeast England 53% of heat-related incident costs and 51% of heat-related incident delay minutes are associated with signaling asset incidents or failures. From discussions, it is clear that Network Rail is aware of the sensitivity of signaling assets to heat (e.g., Network Rail 2015). However, as this asset type tends to fail safely, reducing the number of these incidents is important to improve infrastructure but is not considered to be a key safety risk (unlike track buckling, which could lead to derailment).
Third, this analysis adds detail to existing work (e.g., Network Rail 2014a, 2015) that shows a wide range of assets are failing because of heat-related causes at ambient temperatures well within the operational envelope of Network Rail and also within the current climatic norms of the United Kingdom. While this study has focused on southeast England, routes in Scotland and the northwest are also susceptible to asset failures at ambient temperatures on average approximately 3°C lower than the rest of the United Kingdom (Network Rail 2015). The reasons behind these failures are complex; the railway network has suffered from historical underinvestment from previous management (Department for Transport 2014), and it is likely that old legacy equipment may not be as resilient as modern equipment (Network Rail 2015). Additionally, “maintenance procedures may not always be carried out to the design standard,” and equipment reliability can depend on other external factors, such as people and vegetation (Network Rail 2015). It is of interest to the industry to be aware that signaling assets are more vulnerable in early summer at lower temperatures than the track assets (Fig. 4) and that the median Tmax for signaling assets is 2°–3°C lower than for track assets (Fig. 2b). In general, track incidents consist of a crack or buckle in the rail that occurs following the expansion of rail on hot days. Accordingly, when compared to signaling assets, track assets tend to fail at higher temperatures, which tend to occur later in the summer season.
Finally, this analysis demonstrates the difference in heat risk management practices and/or the consequences that heat risk management practices may have for the different routes. For example, Fig. 7 shows that the number of delay minutes associated with ESRs in Anglia and Kent far exceeds the delay minutes resulting from heat-related track failures. For Anglia, the large number of delay minutes from ESRs is a direct function of its larger geographical extent; imposing an ESR and slowing a train down results in a substantial number of delay minutes in this route, for the length of track between stations is far greater.
b. Implications for heat risk management
The disproportionally higher number of heat-related incidents occurring earlier in the early/midsummer season (failure harvesting) and the consequential increased resilience of the infrastructure system over the course of the summer season could permit an innovative new approach to heat risk management on the railway network. Currently, in order to mitigate the occurrence and impact of track-buckling incidents, ESRs are introduced at static temperature thresholds, which vary depending upon the nature and condition of the track [a comprehensive list is given in Palin et al. (2013)]. However, this analysis suggests that the resilience of the railway infrastructure system increases over the course of the summer season. Accordingly, there is the potential to add regionally specific increments to these temperature thresholds over the course of the summer season as the regional infrastructure becomes resilient to higher temperatures. This dynamic heat risk management would reduce the heat-related disruption from unnecessary ESRs on the railway network and therefore offer a significant improvement in service. It would also reduce some of the costs associated with ESRs. However, some heat-related incidents do occur late in the season, and the pattern of failure harvesting is not consistent for every route and year (Fig. 3 and Table 1). Crucially, this analysis does not include the impact that preventative ESRs may have on the number of track incidents. Dynamic heat risk management relies on the assumption that failure harvesting and the consequential repair/replacement of the asset leads to resilience across the infrastructure system for a specific temperature. However, where ERSs are used, the vulnerability of the track at that specific temperature is not assessed, and therefore removing the ESR or applying it a higher temperature threshold may not be appropriate. This could be addressed by including advice from line watchmen on the resilience of the track. More generally, given that Network Rail is a “safety first” organization, the practicalities and consequences of incorporating dynamic heat risk management into operational procedures requires further exploration before implementation can be possible.
There is also scope for Network Rail to reduce costs by addressing the mismatch in timing between the occurrence of track incidents and the timing of ESRs (Fig. 7). However, as discussed above, the analysis does not include the impact that preventative ESRs may have on the number of track incidents, and therefore without ESRs the number of track incidents, and associated delays and costs later in the summer season could be higher. Estimating the number of track incidents that are prevented by the implementation of ESRs would enable Network Rail to undertake a more accurate cost–benefit analysis of the value of ESRs, with a view to reducing some of the costs and delays associated with mistimed ESRs.
It is also questionable whether the current practice of letting signaling assets fail-safe is sustainable in the longer term. If without targeted adaptation and/or mitigation the incidence and costs associated with track buckling are projected to increase in the future (Dobney et al. 2010), then it seems very likely that the incidence, costs, and delays associated with signaling asset failure will also increase in a future warmer climate. Given that in southeast England 53% of heat-related incident costs and 51% of heat-related incident delay minutes are associated with signaling asset incidents or failures, the implications of this should not be underestimated. Some actions to mitigate against signaling assets overheating are listed in the 2014 “Weather Resilience and Climate Change Adaptation Plans” (WRCCA) produced by Network Rail; for example, Anglia route will “accelerate the completion of cabinet ventilation installations,” and the Kent and Sussex routes will “improve air conditioning.” However, these actions are unlikely to be sufficient given the wide range of assets affected by heat (Fig. 2b). In particular, the WRCCA produced by the Wessex route does not contain any action points to reduce the overheating of signaling assets, despite these accounting for 60% of both heat-related delay minutes and costs between 2006 and 2013.
The spatial analysis of total delay minutes for each location over 2006–13 showed that the delays that last over 12 h fell into two broad categories that require different approaches to manage heat risk: these are, very disruptive one-off events; and locations, such as Woking, where there are repeated problems with a particular asset type. With regards to the latter, local solutions, such as painting at-risk rails or points white so they absorb less heat, or hosing down equipment to lower temperatures (e.g., Network Rail 2014c) may offer site-specific solutions to manage the heat risk. Indeed the value of tacit knowledge (i.e., the knowledge gained through the experience of working) is crucial for dealing with complex interdependent infrastructure systems, such as the rail network (Geldof et al. 2015), particularly given the future challenges of increasing capacity and climatic change.
With regards to the very disruptive one-off events, many of the extensive delays that originate from a single heat-related incident are located within Greater London (Fig. 8b). There, as typical of many urban areas, the railway system is highly interdependent, which means that delays quickly propagate throughout the rail network, causing extensive delays, passenger dissatisfaction, and an additional cost that is disproportionate to the original fault itself (e.g., Jaroszweski et al. 2015). Heat risk management at critical nodes is therefore imperative, particularly in urban areas, where the heat island effect can give rise to temperatures that are significantly warmer than surrounding countryside. Improved condition monitoring, perhaps harnessing the emerging Internet of Things (IoT), could offer an innovative approach. Recently there has been a rapid technological advance in low-cost monitoring solutions that can wirelessly connect to the Internet (Chapman et al. 2016, manuscript submitted to Proc. Inst. Civ. Eng.: Transp.). These devices could be deployed at high resolution alongside (e.g., track) or inside (e.g., location cases) assets in critical nodes to provide an early warning system of extreme temperatures. This would reduce the time needed to respond to a heat-related fault and ultimately the length of disruption to service.
High-resolution temperature networks using the IoT could also be used to develop an evidence base required to scope out the potential for, and ultimately become part of, regional dynamic heat risk management. This localized approach to heat risk management is not limited to temperature, and a whole suite of weather parameters could be monitored and included in localized alert systems [e.g., wind (Easton et al. 2013)]. Such networks of sensors could also be used to verify and improve high-resolution meteorological forecasts for use by the rail sector (e.g., Chapman and Thornes 2006) to predict the temperature of track or other asset types.
5. Conclusions
This analysis has produced a unique evidence base of heat-related incidents of value for both business and research purposes. Moreover, it addresses a knowledge gap highlighted by Network Rail by providing clear and accurate information on the asset types most vulnerable to temperature and those that result in the highest cost. This consequently will support decision-makers by providing the evidence required to review current operational procedures and thresholds; highlight priority areas for resources; and make improved decisions on future strategic investment and development. For researchers, it provides a much needed evidence-based dataset of heat-related failures that can be used as input for climate change impact studies. Although this study is regional in context, the failure-harvesting approach and its potential to innovate current management and adaptation strategies are internationally significant, particularly given the paucity of studies that investigate the effects of weather or climate change on railway infrastructure (Koetse and Rietveld 2009).
Using the algorithm to mine the Network Rail databases in order to identify the heat-related incidents used by this study has also led to the development of a list of clear recommendations describing how Network Rail can less ambiguously record weather-related incidents on the railway network. Adopting these recommendations will add rigor to the Network Rail evidence base used for weather and climate decision-making. Future research in this area will consider two different impacts of heat, including the role of diurnal temperature variation and “heat dosage.” The latter refers to accumulated temperature exposure and whether there is a relationship between how many hours or days an asset may be exposed to heat before it fails. This is particularly important for signaling assets that tend to be more vulnerable to lower temperatures occurring in early summer.
In the future, the railway network will face unprecedented challenges from more frequent high temperatures, and, without targeted adaptation and/or mitigation, the costs and delays associated with heat will increase. As such, improving the climate resistance and resilience of the network is now of paramount importance, particularly considering that railway use is to double by 2050 (UKCIP 2000), placing increasing strain on the railway infrastructure, which is already operating at 100% capacity along certain critical sections. Studies, such as this one, that address regional climatic conditions and infrastructure peculiarities are essential to develop effective adaptation strategies (Oslakovic et al. 2013). Dynamic heat risk management, particularly when coupled with high-resolution condition monitoring, could offer an innovative solution for the future to reduce the heat-related delays and disruption on the railway network nationally and internationally.
Acknowledgments
This research was undertaken in collaboration with Network Rail and funded by the NERC Environmental Risks to Infrastructure Innovation Programme (NE/M008355/1). The authors are grateful for the kind and helpful comments offered by the reviewers, which added depth and clarity to the manuscript.
REFERENCES
BADC, 2014: Met Office Integrated Data Archive System (MIDAS) Land and Marine Surface Stations Data (1853–current). National Centre for Atmospheric Science, British Atmospheric Data Centre, accessed 23 October 2014. [Available from http://badc.nerc.ac.uk/view/badc.nerc.ac.uk__ATOM__dataent_ukmo-midas.]
Baker, C., Chapman L. , Quinn A. , and Dobney K. , 2010: Climate change and the railway industry: A review. Proc. Inst. Mech. Eng. Part C, 224, 519–528, doi:10.1243/09544062JMES1558.
Barrow, E., and Hulme M. , 2014: Climates of the British Isles: Present, Past and Future. Routledge, 454 pp.
Chapman, L., and Thornes J. E. , 2006: A geomatics-based road surface temperature prediction model. Sci. Total Environ., 360, 68–80, doi:10.1016/j.scitotenv.2005.08.025.
Chapman, L., Thornes J. E. , and White S. , 2006: Thermal imaging of railways to identify track sections prone to buckling. Proc. Inst. Mech. Eng. Part F, 220, 317–327, doi:10.1243/09544097JRRT73.
Chapman, L., Thornes J. E. , Huang Y. , Cai X. , Sanderson V. , and White S. , 2008: Modelling of rail surface temperatures: A preliminary study. Theor. Appl. Climatol., 92, 121–131, doi:10.1007/s00704-007-0313-5.
Chapman, L., Azevedo J. A. , and Prieto-Lopez T. , 2013: Urban heat & critical infrastructure networks: A viewpoint. Urban Climate, 3, 7–12, doi:10.1016/j.uclim.2013.04.001.
Coumou, D., and Robinson A. , 2013: Historic and future increase in the global land area affected by monthly heat extremes. Environ. Res. Lett., 8, 034018, doi:10.1088/1748-9326/8/3/034018.
Department for Transport, 2014: Transport resilience review: A review of the resilience of the transport network to extreme weather events. Her Majesty’s Stationary Office Rep. CM8874, 166 pp.
Dobney, K., Baker C. , Quinn A. , and Chapman L. , 2009: Quantifying the effects of high summer temperatures due to climate change on buckling and rail related delays in south-east United Kingdom. Meteor. Appl., 16, 245–251, doi:10.1002/met.114.
Dobney, K., Baker C. , Chapman L. , and Quinn A. , 2010: The future cost to the United Kingdom’s railway network of heat-related delays and buckles caused by the predicted increase in high summer temperatures owing to climate change. Proc. Inst. Mech. Eng. Part F, 224, 25–34, doi:10.1243/09544097JRRT292.
Easton, J., Jaroszweski D. , Quinn A. , Chapman L. , and Baker C. , 2013: The assessment of anemometer-based wind alert systems for implementation in GB: Operational context and requirements. Rail Safety and Standards Board Rep. T1020, 38 pp.
Gardiner, L., and Coauthors, 2009: Climate change effects on the land transport network volume one: Literature review and gap analysis. New Zealand Transport Agency Research Rep. 378, 226 pp.
Geldof, G. D., Cath A. G. , van der Heijden C. M. G. , and Valkman R. , 2015: Future cities, tacit knowledge and new craftsmanship. Accessed 22 January 2016. [Available online at http://www.academia.edu/5016441/Future_Cities_Tacit_Knowledge_and_New_Craftsmanship.]
Hunt, G. A., 1994: An analysis of track buckling risk. Tech. Rep. RR-TM-013, British Railways, 31 pp.
Jaroszweski, D., Hooper E. , Baker C. , Chapman L. , and Quinn A. , 2015: The impacts of the 28 June 2012 storms on UK road and rail transport. Meteor. Appl., 22, 470–476, doi:10.1002/met.1477.
Koetse, M. J., and Rietveld P. , 2009: The impact of climate change and weather on transport: An overview of empirical findings. Transp. Res. D: Transp. Environ., 14, 205–221, doi:10.1016/j.trd.2008.12.004.
Mayes, J., 2013: Regional weather and climates of the British Isles—Part 2: South East England and East Anglia. Weather, 68, 59–65, doi:10.1002/wea.2073.
Network Rail, 2013: A better railway for a better Britain. Network Rail, 36 pp. [Available online at http://www.networkrail.co.uk/WorkArea/DownloadAsset.aspx?id=30064784479.]
Network Rail, 2014a: NR weather analysis report issue 1.0 March 2014. Rep. TS-PR669-00010, Network Rail, 34 pp.
Network Rail, 2014b: Weather and climate change resilience. How our routes are mitigating the impact of weather and climate change to maintain an efficient service. Network Rail, accessed 19 January 2016. [Available online at http://www.networkrail.co.uk/publications/weather-and-climate-change-resilience/.]
Network Rail, 2014c: Route weather resilience and climate change adaptation plans: Anglia. Network Rail, 29 pp. [Available online at http://www.networkrail.co.uk/Anglia-Route-WRCCA-Plan.pdf.]
Network Rail, 2015: NR weather analysis report issue 1.0 March 2015. Rep. TS-PR669-00041, Network Rail, 55 pp.
Network Rail, 2016: Delivering a better railway for a better Britain. Network Rail, accessed 19 January 2016. [Available online at http://betterrailway.co.uk/.]
Oke, T. R., 1973: City size and the urban heat island. Atmos. Environ., 7, 769–779, doi:10.1016/0004-6981(73)90140-6.
Oslakovic, I. S., ter Maat H. , Hartmann A. , and Dewulf G. , 2013: Risk assessment of climate change impacts on railway infrastructure. Proc. Engineering Project Organization Conference, Devil’s Thumb Ranch, CO, Engineering Project Organization Society, 9–11.
Palin, E. J., Thornton H. E. , Mathison C. T. , McCarthy R. E. , Clark R. T. , and Dora J. , 2013: Future projections of temperature-related climate change impacts on the railway network of Great Britain. Climatic Change, 120, 71–93, doi:10.1007/s10584-013-0810-8.
Prior, J., and Beswick M. , 2007: The record-breaking heat and sunshine of July 2006. Weather, 62, 174–182, doi:10.1002/wea.101.
RailCorp, 2012: Signalling equipment temperature control guidelines, version 2.0. Rail Corporation New South Wales Tech. Doc. TMG G1520, 19 pp. [Available online at http://www.asa.transport.nsw.gov.au/sites/default/files/asa/railcorp-legacy/disciplines/signals/tmg-g1520.pdf.]
RSSB, 2015: Tomorrow’s railway and climate change adaptation: Work package 1 raw data. Rail Safety and Standards Board Ltd..
Thornes, J., Rennie M. , Marsden H. , and Chapman L. , 2012: Climate change risk assessment for the transport sector. Department for Environment, Food and Rural Affairs Tech. Rep. EX 6426, 139 pp. [Available online at http://randd.defra.gov.uk/Document.aspx?Document=CCRAfortheTransportSector.pdf.]
Thornton, H., and Coauthors, 2011: The impact of climate change on the GB rail network. Met Office Hadley Centre Consultancy Rep., 143 pp.
UKCIP, 2000: Socio-economic scenarios for climate change impact assessment: A guide to their use in the UK Climate Impacts Programme. U.K. Climate Impacts Programme Rep., 127 pp. [Available online at http://www.ukcip.org.uk/wp-content/PDFs/socioeconomic_tec.pdf.]
