Extreme heat threatens railroads with connectivity and ridership loss in the United States

“High temperatures may require trains to operate at lower speeds, resulting in delays of up to 60 min between 12:00 p.m. and 7:30 p.m. for the remainder of the week”¹. This announcement, shared on a social media platform on June 20, 2024, came in the wake of a prolonged heat wave that disrupted Amtrak routes and affected millions of Americans over three consecutive days. That same afternoon, as temperatures soared to 96 °F (35.6 °C), Amtrak experienced a power failure near New Jersey. The outage disrupted service across a 150-mile stretch of the nation’s busiest rail corridor for over three hours. The ripple effects of the disruption persisted into the following day, with trains crawling through hours of residual delays.

Extreme heat has been responsible for an average of 50 derailments annually over the past four decades in the United States². Examples include the derailment of an Amtrak train in Kensington, Maryland; a Union Pacific coal train in Wyoming’s Powder River Basin; a Burlington Northern Santa Fe coal train carrying 31 loaded cars in Mesa, Washington; a Union Pacific freight train in California; a Burlington Northern Santa Fe derailment in Minnesota; a Bay Area Rapid Transit derailment in California; and a freight train derailment in Matteson, Illinois. Extreme heat, out of all weather-related rail disruptions, has also been responsible for approximately 30% of late arrivals. This accounted for approximately 1200 h of lost time, twice the total recorded in 2018³.

Railway systems, constrained by limited infrastructure and predetermined timetables, are particularly vulnerable to disruptions propagating across specific nodes and links. Even minor disturbances can ripple through the network, resulting in cascading delays^4,5. This behavior, inherent to railway operations, is exacerbated by safety systems that regulate train movements, often halting or slowing trains to maintain safe distances^6,7. The simplest and most frequent manifestation of disruptions of any size are delays, which reflect the complex dependencies and operational constraints of rail networks. Understanding the connectivity of the network is, therefore, important to identify how disruptions at a single point cascade and affect other parts of the system. Analyzing these interdependencies makes it possible to identify critical nodes and links where interventions can most effectively mitigate broader disruptions. This understanding is essential for designing rail systems that are less vulnerable and more capable of maintaining functionality under extreme heat stressors as (i) global climate change is already driving record-breaking heat, exposing the weaknesses of America’s aging rail infrastructure⁸ and (ii) the system, designed for a climate that no longer exists, struggles to withstand the increasingly severe and persistent extreme weather events of today.

Much of the previous research has advanced knowledge of how extreme heat contributes to track buckling, operational delays, and increased maintenance costs in rail infrastructure^9,10. With a warming climate, extreme heat has become an increasingly threatening stressor for rail systems worldwide¹¹. Exposure to record-breaking temperatures is projected to rise fourfold by the end of the century under high-emissions scenarios¹². Amid the increasing threat of extreme heat^11,12, research is shifting from asset failure to delays in urban rail networks and economic losses in the absence of adaptation strategies, as evidenced by studies from Spain, Norway, the United Kingdom, Sweden, and the United States^{11,13,14,15,16,17,18,19}. The scale of this threat has prompted federal agencies, including the U.S. Federal Railroad Administration (FRA) and the Australian Bureau of Infrastructure and Transport Research Economics, to develop strategic frameworks and guidance for the rail industry, emphasizing the cascading effects of infrastructure failure under future climate scenarios^20,21. Yet, three areas of inquiry remain unexplored. First, little is known about how extreme heat–induced connectivity and ridership loss in railroads emerges across multiple geographic scales (e.g., national, state, county, station), as existing studies are largely confined to either the urban scale (e.g., London metro system)¹⁵ or the national scale (e.g., U.S., Spain)^11,13. Second, research has yet to examine how extreme heat–induced rail disruptions propagate through the broader rail network, as existing studies primarily examine minutes of operating delays^11,15, increased maintenance costs¹⁹, and structural failures¹⁷ at isolated points or across limited segments^11,15. Third, there is limited understanding of how both current and projected extreme heat conditions can be leveraged to inform targeted and proactive adaptation strategies within the rail industry^11,12,13,19.

The overarching goal of this study is to evaluate the susceptibility of the U.S. passenger rail network to extreme heat events, by assessing rail network’s connectivity and ridership loss under moderate to severe disruption scenarios at different geographical resolutions (state, county, and station). Passenger rail network was selected for the analysis due to the transport sector’s significant energy footprint. Transport sector produces nearly 25% of global greenhouse gas emissions²² and shifting more intercity and regional travel to rail can reduce energy use, lower emissions, and ease road congestion. This positions passenger rail as an important component of sustainable urban and interurban mobility. Conducting this analysis in the U.S. is particularly important as: (i) the intercity rail system remains underdeveloped relative to other high-income countries, offering an opportunity to build climate-resilient infrastructure from the outset, and (ii) the nation’s wide climatic and geographic variability presents substantial planning challenges. Extreme heat disruptions may affect not only mobility but also urban productivity, labor access, and business continuity due to rail’s role in facilitating economic connectivity across metropolitan regions. Assessing vulnerability to extreme heat is, therefore, not just a technical concern but a policy-relevant priority for ensuring dependable, low-carbon transport within and between cities.

In pursuit of its overarching goal, this study addresses four specific objectives. The first is to identify where high connectivity losses coincide with high ridership, improving the understanding of the interplay between localized susceptibilities and system-wide connectivity. This analysis is conducted at the state level. The second is to analyzes network susceptibility to extreme heat at the county level by examining how connectivity loss propagates across the broader rail network. This includes identifying impacted counties and evaluating whether a station’s topological position within the network influences the extent of cascading disruptions. The third is to integrate future heat projections at both state and county levels, identifying regions where escalating heat exposure is expected to further destabilize rail operations. This is achieved using mid-century (2040–2060) and late-century (2080–2100) estimates from the Federal Emergency Management Agency (FEMA)’s National Risk Index (NRI). The fourth is to develop a heat exposure-informed railway investment prioritization framework, integrating projected heat risks, current exposure, connectivity, and ridership to classify stations into priority categories. This provides a practical tool to allocate climate adaptation investments. This framework: (i) offers state and county governments, responsible for funding and planning rail infrastructure, a snapshot of how and to what extent railroads under their jurisdiction are affected by current and future extreme heat; (ii) enables the detection of hotspots susceptible to targeted adaptive strategies; and (iii) facilitates improved coordination across governance scales and ensures that adaptive measures align with broader regional and national transport objectives, as railways often span multiple jurisdictions.

The findings advance our understanding of rail network vulnerability by linking theoretical vulnerability frameworks with practical insights into mitigating disruption impacts. Focusing on a multi-day extreme heat event in Summer 2024, during which Amtrak experienced heat-related delays, as well as mid- and late-century heat projections, this study expands the knowledge on how high temperatures have affected and will continue to affect station connectivity and ridership. This assessment highlights the risks posed by extreme heat to rail infrastructure, a threat expected to grow in both intensity and frequency in the coming decades²³, and offers a scalable and spatially granular framework that not only identifies heat-vulnerable segments, but also prioritizes them in relation to their operational importance.

The remainder of the article is structured as follows. First, the article discusses findings on how connectivity in the Amtrak rail network is impacted by extreme heat conditions at both the state and county levels, under current and projected future climate scenarios. Second, it provides an in-depth discussion of the implications of these findings for planning and policy practices. Third, the article concludes with a description of the data sources and an outline of the methodological approach used in the analysis.

This study examines the U.S. nationwide passenger rail network operated by Amtrak, which spans a total rail length of 20,698.9 miles and includes 540 stations (excluding Thruway stations). The analysis focuses on rail segments exposed to extreme heat, defined as surface temperatures exceeding 40 °C (104 °F)²⁴. Rail segments are classified into three percentile groups based on surface temperature data: (i) temperatures between 54 °C and 60.8 °C (129.2 °F and 141.4 °F) (the 99^th percentile), covering 1.043 miles (0.005% of the total network length); (ii) temperatures between 44 °C and 60.8 °C (111.2 °F and 141.4 °F) (the 95^th percentile), covering 15.283 miles (0.07% of the total network length); and (iii) temperatures between 40.8 °C and 60.8 °C (105.4 °F and 141.4 °F) (the 90^th percentile), covering a more substantial length of 67.74 miles (0.33% of the total rail network). The results include discussions on the spatial disparity of surface temperatures across the Amtrak network and connectivity loss and network susceptibility under each percentile-based disruption scenario.

Rail Temperature and High-Risk Buckling Segments

Figure 1 illustrates the spatial variation in surface thermal exposure along the Amtrak network, shaped by regional land surface characteristics and atmospheric phenomena, including heat domes. Certain states and regions contain clusters of high-risk rail segments, indicating localized susceptibilities rather than uniform risks across the system. A visual examination reveals three regional patterns.

Hot spots are identified based on the average exposed temperature. Rail segments are color-coded based on average land surface temperatures, with darker tones indicating higher thermal exposure. The map highlights areas of high heat susceptibility, particularly around major hubs such as Chicago, St. Louis, and the Northeast Corridor. Insets provide greater detail on selected high-density regions.

Full size image

The first pattern emerges in the Midwest, which comprises 5415.3 miles (26.1% of the Amtrak network) and has the highest concentration of high-risk rail segments, particularly in Illinois and Indiana. Illinois stands out, with 1149.6 miles of rail, of which 21.2 miles (1.9%) fall within the 90^th percentile of surface temperatures (40.8–60.8 °C). This elevated risk is most pronounced around Chicago, where the urban heat island effect and dense infrastructure contribute to higher temperatures along critical rail corridors. Conversely, states like Iowa and Kansas show minimal representation in high-risk categories, likely due to cooler rural environments and lower levels of urbanization. The contrast between these states underscores the role of urbanization and localized land use in shaping thermal exposure risks.

The second pattern is evident in the South, which covers 6476.4 miles (31.3% of the Amtrak network) and exhibits a more dispersed distribution of high-risk rail segments. Texas stands out as an exception, with a critical segment near Austin spanning 5.385 miles that falls within the 90^th percentile. This isolated vulnerability could be an indication of a significant risk, as disruptions in this segment could lead to localized but severe impacts, particularly in areas with limited network redundancy. Compared to the Midwest, the South’s relatively lower concentration of high-risk segments reflects reduced exposure to extreme thermal stress, likely due to less urbanization and the region’s overall network structure.

The third pattern appears in the West and Northeast, which together encompass 7364.5 miles (35.6%) and 1378.5 miles (6.66%) of Amtrak rail, respectively. High-risk rail segments in these regions are relatively limited. The West region includes 14.05 miles (0.2%) of high-risk rail, primarily in California, near Santa Barbara and Santa Maria. Despite the moderating effects of coastal climates, these areas face localized susceptibilities due to the interplay of land surface conditions and operational demands. The Northeast, with 2.7 miles (0.2%) of high-risk segments, shows scattered exposure along the heavily trafficked Northeast Corridor, including New York, Pennsylvania, and Massachusetts. This pattern reflects a combination of dense infrastructure, operational pressures, and limited network flexibility, amplifying localized susceptibilities despite generally cooler regional temperatures.

Loss of Connectivity and Station Ridership

Figure 2 illustrates the association between baseline connectivity (attachment to stations) and connectivity loss (loss of attachment to stations) across three percentiles (i.e., removing segments within the 99^th, 95^th, and 90^th percentiles of surface temperatures) while also considering stations’ ridership levels. Three observations are noticed.

The figure consists of two sets of plots. The three full-scale plots on the left illustrate the relationship between baseline connectivity and connectivity loss for each state under different percentile removal scenarios (99^th, 95^th, and 90^th percentiles of surface temperatures). The three zoomed-in plots on the right provide a detailed view of the left plots, showing variations that may not be visible in the full-scale plots. In all plots, bubbles are color-coded by region, with the x-axis representing baseline connectivity and the y-axis indicating connectivity loss. The size of each bubble corresponds to ridership. Together, the plots reveal how disruptions from removing the hottest segments impact connectivity. Average state level ridership and connectivity values are reported in Supplementary Table 1.

Full size image

The first observation indicates that regions with higher baseline connectivity and larger ridership correlate with more substantial connectivity losses across all disruption scenarios. The analysis reveals an average connectivity loss of 1.14 stations for the 99^th percentile, 1.56 stations for the 95^th percentile, and 2.33 stations for the 90^th percentile, indicating a progressive increase in network susceptibility as more segments are exposed to high heat. This trend is evident in states like Illinois (Cook County), where the highest ridership station (2,359,084 passengers) also exhibited one of the largest connectivity losses across all percentiles (e.g., a loss of 6 connections in the 99^th percentile, 11 connections in the 90^th percentile). Similarly, stations in New Jersey (Essex County) and Maryland (Baltimore County) consistently rank among the highest in both baseline connectivity and observed connectivity losses. The Pearson correlation analysis (0.47, p < 0.001) confirms that higher baseline connectivity and ridership are associated with greater susceptibility to disruption. This aligns with the concept of betweenness centrality²⁵, where disruptions to key connectors lead to cascading failures propagating across the broader network.

The second observation indicates that connectivity losses are greatest across all percentiles of segment removals in the Northeast. Stations in New York (New York County), Pennsylvania (Philadelphia County), and Washington, D.C., demonstrate high ridership volumes (e.g., 8,008,700 in New York County, 3,631,677 in Washington, D.C., 3,058,329 in Philadelphia County). These hubs also experience notable connectivity losses, (e.g., 4 to 9 connections in the 95^th and 90^th percentiles), indicating their role as critical nodes in the Amtrak network. In contrast, the South and parts of the Midwest show lower connectivity losses due to their more linear and decentralized network structure. Stations in Missouri (Jackson County) and Michigan (Calhoun County) have lower connectivity losses (e.g., a loss of of 4 to 5 connections across segment removal scenarios) despite their baseline connectivity levels. This pattern indicates that less interconnected networks are less susceptible to cascading disruptions²⁶.

The third observation suggests that the network topology in the West exhibits a hybrid structure, characterized by linear routes interspersed with nodal clusters. Stations in California (Los Angeles County) display moderate ridership (928,558 passengers) and relatively minor connectivity drops (a loss of one connection across scenarios). However, localized susceptibilities emerge in smaller nodes (e.g., Santa Barbara County, Santa Maria County) where isolated disruptions can have disproportionate impacts on local connectivity. Observations indicate that while the West exhibits fewer high-risk segments, the impacts of localized disruptions can still be significant, particularly in isolated nodes with limited redundancy (e.g., Texas).

Loss of Connectivity and Network Impact

Figure 3 illustrates the baseline connectivity and connectivity losses under the three disruption scenarios. Two patterns are observed.

The maps show the spatial distribution of Amtrak station connectivity across the United States, averaged at the county level, under baseline and disruption scenarios (99th, 95th, and 90th percentiles). The graphics reveal regional disparities in connectivity losses, with the Northeast and Mid-Atlantic regions experiencing greater declines under disruption scenarios.

Full size image

First, greater declines in connectivity are observed in counties with high connectivity and high heat exposure. Rail segments exposed to higher temperatures are prone to thermal stress, leading to sun kinks, track deformities, and operational failures. When these segments coincide with critical nodes or central corridors, disruptions disproportionately impact overall connectivity. This pattern is evident in the counties of Maryland, Montana, and Maine. In Montana, with 12 stations, the average baseline connectivity ranges from 4 to 6. Under the 90^th percentile disruption scenario, many stations drop to 0–2 connections. The linear configuration, particularly in counties such as Flathead and Glacier, creates dependencies on single links, limiting redundancy. Similarly, Maryland, with 14 stations and a baseline connectivity of 11–12, experiences a decline to 5–6 connections under the 90^th percentile. The reliance on urban hubs such as Baltimore County means that disruptions in these areas impact connectivity across the state. In Maine, with six stations, the network starts with a baseline connectivity of 5–6. However, under extreme conditions (90^th percentile), it loses all connectivity, reflecting its reliance on isolated nodes in York and Cumberland counties. The absence of alternate routes in Maine leaves the state’s rail network unable to maintain service during disruptions. These patterns demonstrate distinct vulnerabilities: Montana’s dependence on linear configuration, Maryland’s reliance on central hubs, and Maine’s uniquely fragile network that collapses under extreme conditions.

Second, gradual declines in connectivity are observed in counties with moderate temperature exposure and more clustered stations, allowing disruptions to spread evenly and reducing cascading failures. Cook County and Guilford County reflect this pattern. Cook County serves as the largest rail hub in Illinois, with a baseline connectivity of 23 stations. Under the 99^th percentile removal scenario, connectivity declines to 17, reflecting a loss of six connections. However, Cook County absorbs network stress due to its extensive links, preventing immediate large-scale disruptions. The surrounding counties further contribute to this gradual decline, with Sangamon County experiencing a reduction from 7 to 5 connections, and Knox County dropping from 10 to 6. The ability of these nearby counties to share the burden of connectivity loss helps stabilize the regional network. Smaller stations, such as those in Adams and Logan Counties, retain limited but stable connections, ensuring that local service remains operational even as higher-order disruptions occur. In North Carolina, where the rail network consists of 18 stations distributed more evenly, a similar pattern emerges. Guilford County, with baseline connectivity of 9, drops to 8 under disruption, while Cabarrus County declines from 8 to 7. These counties function as stabilizing hubs, mitigating severe disruptions by allowing stress to be redistributed across a decentralized network.

Projected Heat and Railway Susceptibility

Figure 4 overlays heat risk projections, derived from LOCA-downscaled climate data, with Amtrak station connectivity and ridership. This comparison helps evaluate how projected mid-century (2040–2060) and late-century (2080–2100) heat risks could affect regions with potential connectivity and ridership losses, as rising temperatures and prolonged periods of extreme heat are expected to impact broader areas, leading to more frequent and severe heat events²⁷. Analysis revealed three patterns in the association between the risk of extreme heat and rail connectivity: (i) exacerbation risk of heat in already vulnerable counties, (ii) escalation heat risk in currently less vulnerable counties, and (iii) continuation of low risk in counties with minimal susceptibility.

This map illustrates Amtrak station connectivity in relation to future heat risk and ridership. Bubble size represents projected heat risk, while color intensity indicates ridership volume, with darker shades denoting higher ridership. Data is sourced from FEMA’s LOCA-downscaled climate projections for mid-century (2040–2060) scenarios. The map highlights regional disparities, with high-ridership corridors, particularly in the Northeast, facing increasing climate threats. Insets provide detailed views of urban centers (e.g., Chicago, St. Louis, Pittsburgh). Detailed current and projected NRI ratings for each county are reported in Supplementary Table 2.

Full size image

Counties already facing high susceptibilities, characterized by high ridership (5 to 8 million) and high connectivity loss (5 to 10 connections), are projected to experience exacerbated heat risks under future scenarios. New York County and Washington, D.C., which currently have “relatively high” risk, are projected to face “very high” risk based on the future projections. Other counties such as Philadelphia, Baltimore, and Anne Arundel, are currently categorized as high risk and are expected to remain so in future projections. These counties, concentrated in the Northeast, indicate the compounded challenges posed by current susceptibilities and future climate stressors, necessitating urgent mitigation efforts.

Counties in the South and Midwest that are presently less vulnerable (moderate baseline connectivity of 6 to 10 connections and low to zero connectivity loss) are expected to see escalating heat risks. Examples include Henrico, Mecklenburg, Sangamon, Durham, and Richmond counties, which currently have “moderate heat” risk but are projected to experience “relatively high” risk in the future. These regions may face growing disparities in rail access as extreme heat events increase in frequency and intensity, suggesting the need for targeted adaptive strategies. Measures such as (i) network expansion, (ii) route redundancy, and (iii) infrastructure enhancements are useful in addressing these emerging risks.

Counties with low heat risk under the current climate are projected to maintain this status in the future. These counties, which display diverse ridership levels (ranging from 50,000 to 650,000), moderate baseline connectivity (4 to 8 connections), and low connectivity loss (0 to 4 connections), are geographically dispersed but include important examples: Cook, Rensselaer, Norfolk, and St. Louis counties. The relatively stable heat risk in these areas presents opportunities to sustain current service levels and address localized challenges without the added burden of intensifying climate impacts.

Heat Exposure-Informed Railway Investment Prioritization

Figure 5 illustrates a framework for heat exposure-informed railway investment prioritization by integrating projected heat risk, current heat exposure, station connectivity, and ridership. Stations are classified into four priority zones: (i) high priority, (ii) medium priority, (iii) low priority, and (iv) no priority for tailored investment strategies. This approach is particularly necessary, as most climate-sensitive infrastructure decisions in the coming century will occur at the local level, often involving projects with budgets in the tens or hundreds of thousands of dollars rather than large-scale investments worth millions²⁸.

The figure consists of two panels. a Illustrates prioritization of rail stations for adaptive planning based on projected heat risk, current temperature exposure, ridership levels, and connectivity. The matrix categorizes stations into different priority levels for adaptation measures, ranging from no priority (gray) to high priority (marron). Stations with high ridership, high connectivity, and high exposure to both current and future extreme heat are assigned the highest priority, as disruptions in these locations could propagate system-wide. Conversely, stations with low ridership, low connectivity, and low exposure to both current and future extreme heat are designated as low-priority areas. b Illustrates classification of Amtrak rail stations based on projected heat risk, current temperature exposure, ridership levels, and connectivity. The scatter plot categorizes stations into four primary quadrants using two key variables: future heat risk (based on FEMA’s mid-century projected heat risk) and current temperature exposure (with 40 °C as the threshold). Stations in the upper quadrants face higher projected future heat risks, while those in the right quadrants experience higher current temperatures. Within each quadrant, stations are further mapped based on ridership and connectivity, using their respective average values across all stations as thresholds (represented by the dashed lines). Average baseline connectivity is 5 connections and average ridership is 89,457. This secondary classification results in four categories within each quadrant: low ridership-low connectivity, high ridership-low connectivity, low ridership-high connectivity, and high ridership-high connectivity. The color gradient corresponds to station priority, with marron indicating high-priority stations that warrant immediate adaptation measures, and gray indicating low-priority stations. Data associated with each station is reported in Supplementary Table 3.

Full size image

Stations classified as high priority account for 2% of total stations (11 stations), located in cities of Alexandria (VA), Ann Arbor (MI), Bloomington-Normal (IL), Battle Creek (MI), Fredericksburg (VA), Glenview (IL), Laurel (MS), Picayune (MS), Quantico (VA), Redding (CA), and St. Louis (MO). These stations collectively maintain 94 connections, with Alexandria (VA) as the most connected (23 connections) and Battle Creek (MI) as the least (4 connections). Annual ridership across these stations reaches 3.5 million, with Alexandria alone accounting for 2.3 million passengers. Alexandria, Ann Arbor, Bloomington-Normal, Battle Creek, Redding, and St. Louis are classified as urbanized²⁹. Urbanization is relevant as stations in densely populated areas are more susceptible to cascading disruptions. Heat-related delays in these locations could affect larger commuter patterns, regional economies, and multimodal transport integration. Urban heat islands may also exacerbate temperature extremes, intensifying the urgency of adaptation measures. Among these high-priority stations, five are experiencing both high current heat exposure and projected high future heat risk, while the remaining six are primarily at risk due to existing extreme heat conditions. Adaptation strategies for these stations should consider both present vulnerabilities and long-term climate projections, ensuring that mitigation efforts address immediate threats while reducing vulnerability for future climate stressors.

Stations classified as medium priority account for 25% of total stations (132 stations). Notable stations in this category include Jacksonville (FL), Ontario (CA), New Haven (CT), and Albuquerque (NM). Collectively, these stations maintain 831 connections, with Staunton (VA) as the most connected (23 connections) and Bellows Falls (VT) as the least (2 connections). Annual ridership across this group reaches 34 million, with Croton-on-Hudson (NY) alone accounting for 8 million passengers. Among these, only 24 stations are located in urban areas, indicating that many medium-priority stations serve smaller or less densely populated regions. These stations are categorized as medium priority for two primary reasons: (i) they exhibit high temperatures but have low ridership and connectivity, meaning their disruption would have limited system-wide impact, and (ii) they are not currently exposed to extreme heat but are projected to face high risk in the mid-century climate scenario, requiring long-term adaptive planning rather than immediate intervention. Adaptive strategies for these stations should be guided by their urban context and future growth trajectories. For urban stations like Jacksonville and New Haven, planning and adoptive strategies should integrate with broader regional resilience efforts (e.g., transit-oriented developments). Stations in less urbanized areas may require infrastructure reinforcements (e.g., superstructure modifications, development of nature-based solutions around tracks) to increase heat resistance and prevent future disruptions. Given that many of these stations do not currently experience extreme heat, but will in the future, early-stage interventions (e.g., improved station shading, cooling infrastructure, operational flexibility) can help preemptively mitigate risks before they become critical.

Stations classified as low to no priority constitute 73% (43% low priority and 30% no priority) of the total network, encompassing 392 stations. Notable cities within this category include Houston (Texas), Providence (Rhode Island), New Orleans (Louisiana), Washington, D.C., and Boston (Massachusetts). Collectively, these stations facilitate 1717 connections, with Ashland (Virginia) being the most connected (18 connections) and Erie (Pennsylvania) the least (2 connections). Annual ridership among this group exceeds 10 million, with stations in Cary (NC) and Ashland (VA) collectively serving over 6 million riders. Despite the large number of stations falling in this category, only 27 of them are situated in urban areas, indicating their predominance in smaller cities or less densely populated regions. These stations are designated as low priority because they are neither currently at risk from extreme heat nor projected to face significant risk in mid-century climate scenarios. Those with minimal impact on network connectivity and ridership are categorized as no priority.

Of course, these classifications are not definitive, as network connectivity, ridership, and urban development patterns evolve over time. Cities may expand, leading to increased commuter traffic, greater infrastructure dependencies, and shifts in station susceptibility. Climate models and transport planning decisions may also alter the future risk landscape, necessitating adjustments in adaptation and prevention strategies. This requires the ongoing evaluation of classifications to ensure that both preventive and adaptive strategies are routinely aligned with emerging transport and climate challenges.

Climate projections for the 21^st century indicate that rising global temperatures will drive more frequent and intense extreme heat events, with the most pronounced increases expected in the South and Southwest^{30,31,32,33,34}. These regions are simultaneously undergoing rapid urbanization, as seen in cities like Phoenix (AZ) and Austin (TX), where population growth is accelerating demand for efficient rail service. If network connectivity remains fragmented or insufficient, this expansion could introduce new vulnerabilities, compounding the risks posed by extreme heat.

The intersection of rising temperatures and urban growth is not confined to the South and Southwest. Many counties with expanding urban footprints (e.g., Henrico County, VA; Sangamon County, IL) experience similar risks, where disruptions in already limited rail networks could propagate delays across broader corridors. Even in high-ridership regions with dense rail interdependencies (e.g., New York County, NY; Philadelphia County, PA; Washington, DC), extreme heat events can trigger cascading failures, amplifying system-wide disruptions. This growing link between heat stress, network connectivity, and ridership underscores the escalating risks to rail infrastructure, particularly in densely populated megaregions like the Northeast. Addressing these emerging challenges requires proactive adaptation strategies to reduce infrastructure vulnerability, ensuring that urban rail systems can sustain reliable service amid worsening climate conditions.

The United States has massively increased funding for new and existing passenger rail through the passing of the Inflation Reduction Act and the Bipartisan Infrastructure Law, which plans to direct over $260 billion to rail. These investments, if continued, are expected to help build new transit corridors in underserved regions, and many of these are in hot regions. Important examples include the under-construction Los Angeles-San Francisco high speed rail and Inland Empire-Las Vegas line in California, as well as the proposed Phoenix-Tucson Amtrak line. These new projects, if and when completed, would increase base connectivity of the region in one of the hottest and most rapidly warming parts of the country. It is, however, noted that extreme heat is not a unique challenge to the South and Southwest. Large, or synoptic, scale processes that lead to extreme heat, such as persistent high-pressure systems (i.e., heat domes) can occur in any region of the country. Recent important examples include the 2021 Pacific Northwest heat wave, which saw record-breaking temperatures for several days and led to reduced speeds and service cancelation of rail service³⁵. This suggests that as these investments increase the total passenger rail capacity in the United States, planners and engineers must account for temperature extremes in nearly every part of the country.

The findings provide a structured foundation for climate-resilient passenger rail planning across federal, state, and regional scales. In the short term, federal agencies (e.g., FRA, FEMA) can operationalize the station prioritization framework to direct adaptation funding toward high-risk, high-connectivity nodes. This entails retrofitting heat-exposed segments, reinforcing critical links, and introducing operational redundancies in highly interconnected urban counties like Alexandria (VA) and St. Louis (MO), where a single failure can compromise regional mobility. In the long term, the importance of embedding climate exposure and network vulnerability into corridor planning and design processes must be considered, particularly as FRA’s Long-Distance Service Study³⁶ charts new rail expansions across the Midwest and South. Many of these proposed corridors intersect with counties such as Sangamon (IL) and Henrico (VA), which are currently low-risk but projected to face increasing heat threats.

Integrating forward-looking climate data into route alignment, materials specification, and station siting decisions is imperative, as rising susceptibility to extreme heat threatens not just passenger rail resilience but also broader greenhouse gas reduction efforts in the transport sector, the largest source of emissions in the United States as of 2022^37,38. Expanding intercity rail and other forms of public transit is widely recognized as an important decarbonization strategy³⁹. Yet, this transition is threatened by the intensifying frequency and severity of heat extremes. New rail infrastructure must be designed for the climate it will face over its service life, while existing systems will require targeted retrofits and adaptive upgrades. With passenger rail being central to modal shift and emissions reduction goals, climate adaptation must advance in parallel with mitigation. Without heat-resilient infrastructure, network expansion risks reinforcing vulnerabilities and compromising the reliability needed to sustain low-carbon mobility.

This study is not without caveats. First, it relies on the Moderate Resolution Imaging Spectroradiometer land surface temperature data at a 1 km resolution and interpolates missing values due to cloud cover, which may introduce spatial uncertainty in estimating the precise thermal exposure of rail segments. Second, the analysis assumes that the hottest segments are removed concurrently and that the resulting changes in network connectivity is due to this simultaneous elimination of segments. While this assumption simplifies the modeling of disruption, it may not fully capture the temporal dynamics or operational responses of the rail system. Future studies could incorporate sequential failure modeling or simulate adaptive management strategies, such as rerouting or phased shutdowns, to more accurately reflect real-world conditions. Third, while extreme heat disruptions are likely to affect urban productivity and economic activity through impaired labor access and business travel, such impacts could not be quantified in this study due to the lack of spatially resolved economic and ridership data. Future work should investigate these economic consequences more directly to support integrated infrastructure and economic disruptions.

Data

Amtrak passenger rail network is the primary provider of intercity passenger rail service across the United States. It spans across 46 states and the District of Colombia, connecting over 500 cities, towns, and villages. In total it includes 654 segments and 540 rail stations. Amtrak operates most extensively along the Northeast Corridor, which experiences significant intercity travel demand. Amtrak also maintains key routes extending into the Midwest, the South, and the West, serving as intercity connections. The railroad network intersects with regions susceptible to extreme temperatures that can jeopardize rail infrastructure connectivity. These regions are prone to rail buckling that could lead to connectivity and ridership loss throughout the network.

Four data sources are used to assess the connectivity and ridership of Amtrak network and its susceptibility to extreme heat exposure. First, land surface temperature data were obtained from the Moderate Resolution Imaging Spectroradiometer (MODIS) daily 1 km L3 product (MOD11A1), from daytime swath-level imagery covering the period between July 3 and 12, 2024. Second, Amtrak ridership data was obtained from the Bureau of Transportation Statistics, providing annual boarding and alighting figures for each station. The analysis uses the most recent data available, which is from 2022. Third, detailed transit information for Amtrak, including stop locations, trip schedules, directional information, and route details, was sourced from the latest General Transit Feed Specification (GTFS) dataset provided by Amtrak. Fourth, projected heat risk data were drawn from the Federal Emergency Management Agency (FEMA) National Risk Index (NRI)²⁷, which provides both current and future risk scenarios for extreme heat at the county level. Future projections include mid-century (2040–2060) and late-century (2080–2100) scenarios under both lower and higher greenhouse gas concentration pathways.

Susceptibility Assessment

Network susceptibility is its tendency to experience reductions in serviceability during disruptive events^40,41. Such disruptions often degrade performance and propagate across the network⁴². This propagation and its impacts are often assessed by measuring physics of the network (e.g., degree, betweenness, centrality)^{43,44,45,46,47,48,49,50,51,52,53,54,55}. Here, the connectivity of the Amtrak network is quantified through a systematic approach. Connectivity, in the context of Amtrak stations, refers to the extent to which each station is linked to other stations within the rail network⁵⁶. This measure captures the potential for travel between stations, representing the station’s role as a node within the larger transport network. The metric used for quantifying connectivity is degree, which counts the total number of direct connections (routes) that a node (station) has within the network. Consider Station A, which is directly connected to three other stations, receives trains from Station B and Station C, while sends trains to Station D. Station A has an in-degree of 2 (connections from Station B and Station C), an out-degree of 1 (connection to Station D), and a total degree of 3, representing the sum of its in-degree and out-degree. This indicates that Station A has three direct connections within the network. Connectivity metrics help understand the structure and functioning of the rail network. Stations with high degrees often act as critical hubs, serving as transfer points or major destinations, while those with lower degrees may signal susceptibilities or areas where the network could be strengthened. Station connectivity also influences travel accessibility, as higher-degree stations generally support more efficient movement across the network.

Analysis considers the bidirectional flow of trips. The methodology was grounded in extracting and analyzing data on trips, timetables, stop locations, and flow directions, using Amtrak’s GTFS data. The analysis followed four steps. First, relevant rail routes and associated trips within the network were identified. This step involved filtering for passenger rail services, excluding non-rail or non-passenger transport routes to focus exclusively on Amtrak’s passenger network. Second, the sequencing of stop events was retrieved to determine the specific station stops that each train made along its route. This information was used for creating the links between consecutive stations and for maintaining the correct order of travel and distinguishing between inbound and outbound directions. Directional flow data was integrated at this stage to differentiate the routes effectively and maintain a high level of granularity. Third, edges were created to represent the direct linkages between consecutive stations. These edges were defined as unique, directed pairs of stations, ensuring that each link represented a meaningful, unidirectional flow between two stations in a specific order of travel. A directed network graph was then constructed from these edges, with nodes representing individual stations and edges representing the flow between these nodes. The network graph consisted of 654 segments and 540 stations. Fourth, the baseline connectivity of each station in the network was quantified using the degree metric. Figure 6 shows the baseline connectivity of Amtrak stations. From the visual examination of Fig. 6, two observations are made.

The size of the bubbles in the figure represents the connectivity of each station, while the color gradient indicates ridership levels, with darker shades reflecting higher ridership.

Full size image

One observation is that stations in the Midwest (e.g., Chicago) and Northeast (e.g., Washington, D.C., New York City) exhibit the highest connectivity, serving as critical junctions that link multiple rail routes. These regions form the backbone of the Amtrak network, characterized by densely interconnected nodes (i.e., hubs) that enhance network resilience by offering multiple redundant paths. Disruptions at these hubs are likely to have cascading effects, impacting both regional and national rail operations. In contrast, the West and South regions display a more fragmented network structure, with stations exhibiting lower connectivity levels. While higher connectivity is observed in select areas such as California and Florida, large portions of the network in states like Texas, Oklahoma, and New Mexico remain sparsely connected. This dispersed connectivity in the South and West results in a more vulnerable network topology, as fewer redundant pathways increase the likelihood of localized disruptions severing regional connectivity. The lack of integration in these areas limits their role in the broader national rail network, positioning them as peripheral nodes in the network hierarchy.

Another observation is that ridership patterns reveal regional disparities across the Amtrak network, closely aligned with connectivity and demographic factors. The Northeast Corridor, spanning Washington, D.C. to Boston, exhibits the highest ridership volumes, driven by strong intercity travel demand, dense population centers, and a well-established rail infrastructure⁵⁷. This region represents a core segment of the network where high ridership is reinforced by the density of nodes and efficient linkages, reflecting the interplay between connectivity and passenger flow dynamics. The Northeast Corridor functions as a “central cluster,” where high network centrality coincides with elevated usage, amplifying its criticality to the overall system. Chicago in the Midwest and Los Angeles in the West also exhibit moderate to high ridership levels, despite lower passenger volumes compared to the Northeast Corridor. These stations act as regional hubs, demonstrating how nodal importance in the network can drive substantial passenger flows even outside the densest parts of the system. Ridership in these areas, however, reflects a balance between regional demand and the more limited infrastructure compared to the Northeast. In contrast, stations in less populated regions, particularly in the South and West, are characterized by lower ridership. This pattern is likely influenced by (i) sparse population density, (ii) reduced rail service frequency, and (iii) a diminished reliance on rail for intercity travel. These areas represent peripheral nodes, where low connectivity and limited service diminish their role in sustaining significant ridership volumes.

Baseline connectivity is vulnerable to extreme heat, as heat-induced buckling leads to connectivity and ridership loss. Three scenarios are defined based on average surface temperature to exclude segments exposed to high thermal stress points:

• 99^th Percentile Exclusion: The top 1% of segments, corresponding to 6 segments, were removed. These segments have average temperatures ranging from 54 °C to 60.8 °C (129.2 °F and 141.4 °F).

• 95^th Percentile Exclusion: The top 5% of segments, corresponding to 45 segments, were removed. These segments have average temperatures ranging from 44 °C to 60.8 °C (111.2 °F and 141.4 °F).

• 90^th Percentile Exclusion: The top 10% of segments, corresponding to 100 segments, were removed. These segments have average temperatures ranging from 40.8 °C to 60.8 °C (105.4 °F and 141.4 °F).

The selection of temperature thresholds is based on the increased risk of rail buckling when temperatures exceed 39 °C (102.2 °F)^58,59. At this threshold steel tracks are more prone to deformation and failure. After removing segments, the connectivity of the network was re-evaluated by recalculating the degree for each station. The differences between the baseline connectivity and the new connectivity measures represented the extent of connectivity loss for each station, indicating how vulnerable certain stations were to disruptions in the network.

Land Surface Temperature

Land surface temperature (LST) data were sourced from the MODIS 1 km daily L3 product (MOD11A1) daytime swath-level imagery for July 3–12 2024. Original data is organized by granule, which are spatiotemporal snapshots of MODIS instrument imagery. Granules that contain the Continental United States (CONUS) were used, combining them into a single image and clipped over the region’s boundary. One limitation of using individual snapshots at 1 km resolution is the interference of observations from cloud cover and other atmospheric species. To obtain a spatially complete LST composite over the CONUS, a time average was taken of the imagery retrieved from July 3 to 12, which contains the hottest days of the year. After averaging data over the 10-day period, 0.57% of the pixels within the CONUS boundary had invalid values due to persistent cloud cover. These remaining pixels were gap-filled using regularized spline interpolation with tension with a smoothing parameter of 40, a tension parameter of 0.1, and a regularization parameter of 0.01. Figure 7 displays the resulting gap-filled land surface temperature data.

The map illustrates spatial variation in surface heating, with the highest temperatures observed across the Southwest and parts of the Central U.S., reaching above 60 °C in some areas. Data reflect average daytime land surface temperatures over the 10-day period, highlighting regions most exposed to extreme surface heat.

Full size image

To quantify exposure of passenger rail infrastructure to surface temperature, data from the Bureau of Transportation Statistics (BTS) National Transportation Atlas Database (NTAD) were used. This dataset describes railway as vector line geometries and includes attributes for both passenger rail lines as well as all freight railways. Passenger rail was extracted (i.e., all Amtrak routes) and clipped to the CONUS boundary. To assign land surface temperature raster data to rail line geometries, the latter were converted into a raster at 100 m horizontal resolution. The gap-filled land surface temperature data were then interpolated to this raster using nearest-neighbor methods, allowing for consistent spatial matching between rail infrastructure and thermal exposure. This approach provides the spatial foundation for the analysis presented in this study and supports future efforts to integrate high-resolution climate data with transport infrastructure planning.