Mere Nuisance or Growing Threat? The Physical and Economic Impact of High Tide Flooding on US Road Networks

Rising sea levels threaten to change the nature of coastal flooding events by shifting vulnerable areas further inland, exposing infrastructure that was believed to be resilient to these hazards when it was first designed and constructed. The shift is observable now in many regions of the world (IPCC 2012; Emanuel 2013; Hinkel et al. 2014; Sundermann et al. 2014) and has challenged the way we think about infrastructure design and planning, especially for civil infrastructure (e.g., roads, structures, and pipelines), which are exposed to the environment and are built with the expectation of multiple decades of useful life (Qiao et al. 2020). Considering these expected changes, observed environmental conditions alone are often not sufficient for design and planning—future projections of these conditions are essential in many cases. Planning around flood events is no different, especially for coastal infrastructure in which rising sea levels will exacerbate storm surge and fluvial flooding by raising the flood depth and expanding threats to newly vulnerable areas (Dinan 2017; Neumann et al. 2015b; Hallegatte et al. 2013).

Many studies have evaluated low probability, high damage events (e.g., around 10- to 100-year return periods) for coastal infrastructures, such as storm surges (Hallegatte et al. 2011; Klima et al. 2012; Lin and Shullman 2017; Neumann et al. 2015b; Diaz 2016; Wahl et al. 2017; Dinan 2017), but there has been less focus on high probability, low damage high tide flood (HTF) events (sometimes referred to as minor, nuisance, or sunny day flooding). Historically, HTF events have not caused great damage to infrastructure or posed risks to public safety, at least not to the extent of more extreme events. But due to sea-level rise, HTF is increasingly putting more infrastructure at risk of flooding, with cumulative impacts mounting. Impacts include delays in traffic (Jacobs et al. 2018a; Suarez et al. 2005), declining real estate values (Nabangchang et al. 2015), or damage to low-lying or underground infrastructure, such as storm sewers (Cherqui et al. 2015). Although HTF causes less damage per event than extreme events, expected annual damage may be larger over time than extreme events in some cases because they happen more often and in more locations (Moftakhari et al. 2017). However, damage associated with HTF has been the subject of less monitoring and research to date, and links between specific event characteristics, such as depth and the associated impacts, are not well understood (Jacobs et al. 2018a; Moftakhari et al. 2018).

The frequency of HTF varies through time with some noticeable intraannual-to-decadal signals, driven by a variety of ocean-atmosphere dynamics and physical phenomena, such as the 18.6-year lunar nodal cycle or modes of the El Nino Southern Oscillation (ENSO) (Sweet and Park 2014; Sweet et al. 2018; Thompson et al. 2019). Aside from these oscillations, HTF occurrence and depth are also shifting over time as a result of regional sea-level dynamics, subsidence, and global sea-level rise (SLR) (Sweet and Park 2014; Sweet et al. 2018). High tide flooding has already caused noticeably increased impacts in coastal cities such as Annapolis, MD, Norfolk, VA, Miami Beach, FL (Sweet et al. 2019), and, to a lesser extent, San Francisco (Moftakhari et al. 2017). Analysis of hourly tide gauge measurements between 2000 and 2015 indicates that annual frequencies of HTF have increased by about 125% in the Southeast Atlantic and 75% for both the Northeast Atlantic and Western Gulf (Sweet et al. 2018). Using the intermediate sea level rise projection (global rise of 1 m by 2100), these events are expected to increase from $1.3\mathrm{days}/\mathrm{year}$ in 2000 to $85\mathrm{days}/\mathrm{year}$ in the Southeast, 3.4 to $130\mathrm{days}/\mathrm{year}$ in the Northeast, and 1.4 to $185\mathrm{days}/\mathrm{year}$ in the Western Gulf by 2050 (Sweet et al. 2018). The Eastern Gulf and Pacific Coast are expected to experience less dramatic shifts in tidal flooding events.

Increases in tidal flooding events are likely to have long-lasting impacts on traffic and roadway maintenance in coastal regions. Roads are already vulnerable to changes in temperature and precipitation, causing degradation, more frequent maintenance, or the road surface to be redesigned to be resilient to the projected climatic conditions (Jacobs et al. 2018b; Underwood et al. 2017; Neumann et al. 2015a; Mallick et al. 2018). Neumann et al. (2021) estimate that direct and indirect costs to noncoastal road infrastructure caused by climate change would be roughly $\$90–\$140\mathrm{billion}/\mathrm{year}$ by 2050 in the contiguous United States (CONUS) under a no-adaptation scenario but could be reduced by over 90%, to $\$8\mathrm{billion}/\mathrm{year}$, with proactive adaptation. The study points out that most of the avoidable costs—i.e., the difference between no adaptation and proactive adaptation—are from traffic delays and damage to vehicles, placing the cost burden on drivers. Batouli and Mostavi (2018) developed a model of multiagent adaptive responses to changes in roadway maintenance costs from HTF driven by sea-level rise and precipitation projections and applied the model to a road network in Miami. Their study finds that with adaptation, average network lifecycle costs would be roughly $3.7–$4.8 million per year compared to $15–$17 million without adaptation.

To date, little is known about the impacts of HTF on CONUS road infrastructure over a long planning horizon (through 2100) when considering regional sea level rise scenarios (Sweet et al. 2017). Jacobs et al. (2018a) combined hourly historical water levels, a public road infrastructure database, and fine-scale flood mapping to estimate HTF impacts on traffic delays (in vehicle-hours) for east coast states. Their study introduced a unique procedure that defines road segments at intersections for the purpose of estimating traffic delay risks to HTF. Jacobs et al. (2018a) found that delays would increase substantially along the East Coast, reaching 1.2 billion vehicle-hours by 2060 and 3.4 billion by 2100, whereas all delays combined in the US currently are roughly 100 million vehicle-hours. While Jacobs et al. (2018a) indicate HTF will result in significant regional impacts, to date, no national-scale study has quantified the economic impact of HTF and sea-level rise on the coastal road network nor considered multiple levels of adaptation potential that could be pursued to mitigate these effects.

In this study, we build on Jacobs et al.’s (2018a) methods by expanding the evaluation region to include the Gulf and West Coasts, estimating the economic costs of traffic delays, and exploring various adaptation options that could make the road network more resilient. Specifically, the contributions of this study to the literature are the following. We evaluate indirect adaptation options to alleviate these traffic delay costs by (1) considering road network redundancy by using a detour effectiveness measure; and (2) accounting for roads that would be protected by actions to protect property from storm surge and/or SLR given a least-cost decision by property owners. In addition, we explore options to protect roads directly by applying a screening-level least cost decision and cost evaluation for two adaptation options: raising the road profile and building a sea wall. Finally, we consider changes to maintenance costs that arise from more frequent inundation for a single road segment to illustrate how adaptation costs to prevent direct damage compare to delay costs.

This analysis is being conducted as part of the ongoing Climate Change Impacts and Risk Analysis (CIRA) project, which uses a consistent analytical framework of socioeconomic scenarios and climate projections to estimate and compare economic impacts across multiple sectors and regions (Martinich and Crimmins 2019; EPA 2017).

Following the framework outlined by Jacobs et al. (2018a), risks of traffic delays are estimated for CONUS following the first four steps shown in Fig. 1. In summary, the hourly empirical cumulative density function (CDF) of tide gauge water levels is determined (#1). The road network is segmented by intersections or ramps, and traffic data are assigned to each segment (#2). These datasets are used with the flood plain to identify vulnerable roads and flood duration (#3). Roadway risk, i.e., vehicle hours of delay, is calculated as the product of the flood duration and traffic. The risk is then monetized using hourly rates for passenger and freight truck traffic delays, and costs of direct damage to the road surface are also explored in a sensitivity analysis (#4). Two types of adaptations are simulated (#5): reasonably anticipated adaptation, which includes driver-initiated rerouting and ancillary protection from actions to protect property, and direct adaptation, in which, in addition to reasonably anticipated adaptation, actions are taken to alleviate delays by raising the road profile or building hard structures, such as sea walls.

We now go into further detail. A full list of the data used in this study and the sources can be found in Table S1 of the Supplemental Materials.

The direct adaptation method considers the alleviation of HTF-induced traffic delays through the implementation of adaptation strategies. While there are many conceivable ways to adapt to HTF, we model two well-established options: (1) build a sea wall to hold back the floodwater; and (2) raise the road profile above the effective threshold.

For each road, the decision to adapt using one of these two options depends on the ratio of benefits to costs. Lorie et al. (2020), among others, point out that in many cases, the benefits ($B$) need to be significantly higher than the costs ($C$) to trigger action. We follow Lorie et al. (2020) by running multiple scenarios varying this $B/C$ threshold, $S$, where adaptation is built when the following (Eq. 2) is true

We develop scenarios using $S$-values of 1, 2, 4, and 10. Benefits, $B$, are the avoided traffic delay costs for 10 years, an assumed planning horizon to trigger protection, discounted by 3%. The cost of a sea wall is related to the square of its height, essentially a triangle with a wide base, following previous studies (Yohe et al. 1999; Mendelsohn et al. 2019; among others). The cost of raising the road profile is composed of the material and labor costs of adding base thickness and embankment. Elevating roads requires additional widths for the embankment, and obstacles or development near roads often make this infeasible. We account for this using available information in the HPMS dataset, which excludes roughly 13.7% of the road segments but 48.5% of the total traffic because these are often on high-trafficked roads in urban areas. Other constraints such as soil types that lack sufficient structural stability are not addressed. Details of this and the adaptation costs are provided in Section S5 of the Supplemental Materials.

While the adaptation costs include estimates of material, labor, and construction delays, actual costs will include additional factors—for example, costs associated with management, design, easements, or land acquisition (see Table S2 of the Supplemental Materials for a longer list). In addition, our framework implies that protection will be built without design or construction errors or schedule delays due to sociopolitical or budgetary issues. While construction of either of these protection types would divert floodwaters away from the road, flooding may occur elsewhere as a result, and additional costs would be incurred. Furthermore, these protections may have environmental impacts, such as preventing wetland migration.

Our method results in six sequential categories of results, which build upon the prior step:

Jacobs et al. (2018a) laid the groundwork for this study by developing a method to identify the HTF risks to traffic. This work has extended Jacobs et al. (2018a) from the East Coast states to the Gulf and Pacific coasts, adding economic valuation, as well as the evaluation of both reasonably anticipated and direct adaptation. HTF may currently be correctly classified as a nuisance in many coastal areas of the US, but rising seas will accelerate this threat into a major impact on the coastal road network. National-scale traffic delays are expected to grow steadily over time as sea level rises, but reasonably anticipated adaptation to this threat greatly reduces the impact of HTF events. Nearly half of the impacts occur in the Gulf Coast region, owing to a high percentage of low-lying roads along with subsidence and accelerating sea-level rise. Another 25% occur in the North Atlantic region, owing to a high concentration of affected traffic volume near major coastal cities. Direct damage repair costs for a road segment in South Carolina indicate that repair costs are much less than traffic delay costs; however, these costs are, in a sense, a direct adaptation. If the road is not properly maintained, costs will be higher.

Compared to other CIRA categories of impacts, these costs are significant. In this study, the annual end-century impacts with reasonably anticipated adaptation are estimated to be $280 billion under RCP 8.5 and $220 under RCP 4.5. By comparison, the EPA (2017) indicates that annual costs in 2090 under RCP 8.5 for the coastal property sector are $120 billion [based on the model by Neumann et al. (2015b)], $20 billion for the roads sector [based on the model by Neumann et al. (2015a)], and $8.1 billion for inland flooding [based on the model by Wobus et al. (2017)].

Differences between RCP 4.5 and 8.5 simulations indicate costs are reduced by action to mitigate climate change [in the case of HTF, by roughly 25%, similar to the 22% difference noted by the EPA (2017) for the coastal property sector]. But owing mostly to the already committed levels of SLR this century, the impact of greenhouse gas mitigation on HTF is relatively small, consistent with other coastal impact analyses running through 2100 (Dinan 2017; Neumann et al. 2015b; among others).

While HTFs are fluctuating events, they are frequent and somewhat predictable compared to more extreme flooding events, such as hurricanes or riverine flooding. This allows planners and decision-makers to experience HTF impacts in their communities and to be able to anticipate not only similar but also more frequent impacts in the future. There is likely to be significant technological potential to adapt physically and socially in response to such a large threat. Our simulations with two options for direct adaptation—sea walls and raising the road surface—indicate significant savings compared to costs without direct adaptation: costs are potentially reduced to $4 billion annually in 2090 under the $S=4$ assumption. Our analysis shows that more roads need to be protected from HTF in the coming years, a challenge that will escalate quickly through the end of the century.

It may be tempting to conclude from our results that cost-effective adaptation, including the rerouting of traffic, the timely raising of roads, and construction of seawalls, can be taken as given, and that as a result, HTF impacts may not be an important concern. In practice, there remain many barriers to implementing cost-effective adaptation in all settings. In a general sense, technological, behavioral, and financial barriers stand in the way of achieving an economically optimal adaptation outcome [see a thorough discussion by Chambwera et al. (2014) and more specific examples for the US coastal zone by Moser et al. 2014]. Our modeling in this work assumes that at least three layers of adaptive response will work efficiently together to support an effective response to HTF vulnerability:

Many of these project-level considerations are currently infeasible to model in our national-scale approach, poorly understood in their implications, or lack data. As a result, we caution against using only those results from our work that assume all these elements of adaptive action will operate on time, in a coordinated fashion, and with enough efficacy to realize all the potential benefits of adapting that we model.

In addition, the adoption of the limited benefit-cost method we use to identify likely locations for cost-effective adaptation is not meant to endorse that approach alone in making local adaptation decisions. Ideally, our results will spur local transport authorities and coastal communities to recognize the potentially growing nature of the problem—and then conduct a more nuanced and broader analysis at the project scale to justify HTF adaptation actions. For example, there are numerous additional benefits of adapting, such as an avoided business disruption, property value loss, and local fiscal impact, which can be considered best in a local context. We recognize that expensive capital adaptations will likely need to be justified by some economic analysis. Moss et al. (2019), for example, notes that financial oversight bodies often require benefit/cost analysis for adaptation investments, and we strongly suggest going beyond just the costs of transport delay in characterizing the benefits of adapting for those analyses. Ideally, HTF requires the type of iterative risk management approach recommended in the Fourth US National Climate Assessment (USGCRP 2018) and associated literature (Moss et al. 2019), which considers and then reconsiders climate risks, social and economic context, and technological options both now and as we learn more in the future as uncertainty in these factors resolves, narrows, or increases.

There are other limitations and caveats to note. While the analysis uses projections to the end of the century, the transportation sector may see significant changes over that time that would make some of our assumptions invalid, such as a transition to driverless vehicles or a significant increase in mass transit. Because we rely on a consistent cost per hour of delay throughout the century, this unit cost is representative of the current transportation sector, which is likely to change in a future system significantly. Without direct adaptation, costs scale linearly with the per hour delay costs, i.e., a 50% increase in per hour delay costs increases national costs by 50%, but costs with direct adaptation have a more nuanced sensitivity to per hour delay cost, in which an increase of 50% increases total century discounted costs (at 3%) to 48% but a decrease in per-hour delay cost only reduces total national discounted costs by 6%. This suggests costs with direct adaptation are more sensitive to increases in per-hour delay costs than decreases. We also recognize that the chosen discount rate is important, which reflects the trade-off between consumption today and consumption tomorrow. This paper uses 3%, a commonly employed rate in the climate impacts literature (e.g., Goulder and Williams 2012). There is more on this in Fig. S7 in the Supplemental Materials.

The analysis is also limited to road segments within the flood extent for the current minor flood level. Sea level rise will expand to this extent, and additional roads will become vulnerable, likely causing more delays and road damage than determined in this study. Based on our approximation, total vulnerable traffic could increase by about half in 2050 and more than double by 2100. Absent of road surface elevations, actual flood depth is uncertain, and vehicles may be able to traverse shallow water with only minor speed reductions. Costs are sensitive to depth by about 0.9%/cm. In addition, these flood events are limited to high tide events only. Additional flooding as a result of rainfall or riverine flooding, which are not estimated in this study, may exacerbate the flood extent and/or flood duration if they occur simultaneously with a high tide event. Underground roads are not considered, where flooding may cause significant damage. This study was limited in the site-specific information available, which constrained our ability to investigate effects, such as direct damage to the road infrastructure, so we discourage the use of these results at the project level. Future research should make use of finer resolution topography data, road conditions, drainage, and traffic/routing information. High-resolution inundation maps at various levels are likely to provide a better approximation of the traffic delay risks. Also, sea level and high tide durations and frequencies vary along the coast (these variations are not captured in this study) and result in a lower accuracy for areas further from tide gauges. For smaller-scale studies, it is essential to identify thresholds of impacts important to road managers and users, which might trigger more aggressive and/or permanent responses to this risk over time (e.g., de facto or planned abandonment of certain high-risk roadways. For example, the authors are aware of situations that have already arisen in some coastal communities that have triggered current decisions to abandon roads that currently experience low levels of HTF effectively, but it is clear that problems will worsen).

There are many direct adaptation options to alleviate HTF impacts. We have only considered two protection methods of the many possible adaptation approaches. For example, hydrologic infrastructure, such as dams or pumps, may be used to alleviate high tide flood events. It may be possible in the near term to build temporary protection or use portable pumps if an HTF event is likely to occur. Furthermore, hard-engineered adaptation options, like those evaluated in this study, may not be the most efficient; for example, a viable option may involve altered city planning to direct traffic away from low-elevation roads. These options should be considered in future research. In addition, a systems modeling approach may indicate interactions and cobenefits across sectors. For example, these include merging models of impacts and adaptation of risks to transport, such as this, with property risk models and other types of infrastructure to understand better interactions among at-risk stocks and flows of economic value and productivity, including potentially changing patterns of locations for working, living, and moving around.

All data used in this analysis are presented in detail in the Supplemental Materials. Some or all data, models, or code generated or used during the study are available in a repository online (indecon.com/iec-climate-change-high-tide-flooding-traffic/) in accordance with funder data retention policies. Some or all data, models, or code used during the study were provided by a third party. Direct requests for these materials may be made to the provider as indicated in the Acknowledgments.

We would like to thank Lia Cattaneo for providing assistance with the road segmentation process, as well as providing comments on the study and manuscript. Components of this work were funded under EPA Contract #EP-D-14-032. The views expressed in this paper are solely those of the authors and do not necessarily represent the views or policies of their employers.