A drainage structure designed for the 2 percent annual chance flood, the one usually called the 50-year event, will spend its service under a bridge with the 75-year design life that AASHTO’s bridge specifications assume. FHWA works the arithmetic in Hydraulic Engineering Circular 17 and prints the answer: “There is a 78 percent chance that the 50-year flood or greater will occur at least once over a 75 year period.” The circular’s own table extends the pattern. A 100-year event has a 0.53 probability of occurring at least once in 75 years, and a 500-year event a 0.14 probability.
Nothing in those figures is a criticism of the standards. They are what the standards mean. A design event is a stated probability rather than a ceiling, and the guidance says so directly: exceedance “is expected under conditions that are defined, and therefore exceedance does not necessarily constitute ‘failure,’ although it may result in consequences ranging from inconvenience, to damage, to complete structural failure.”
The return period is an average, not a schedule
HEC-17’s glossary is worth reading before any conversation about the 100-year flood. A return period is “The average length of time between occurrences in which the value of a random variable (e.g. flood magnitude) is equaled or exceeded,” and the definition adds the clause that public argument usually drops: “Actual times between occurrences may be longer or shorter, but the return period represents the average interval.” It is the inverse of the annual exceedance probability, so a 0.01 AEP is a 100-year return period and nothing more.
The circular prefers probability to years for exactly this reason. Design events for highway drainage are specified as exceedance probabilities, and an example table drawn from AASHTO practice assigns 2 percent to Interstates, freeways, principal arterials and minor arterials above 3,000 vehicles a day, 4 percent to lighter minor arterials and to collectors above that threshold, 10 percent to collectors below it, and a band from 20 percent down to 10 percent for local roads. Traffic volume is a proxy for consequence, though the circular cautions that it is a partial one, since “a lack of convenient alternate routes, or a condition of ‘landlocking,’ where no alternative routes exist” can put a low-volume road in a higher consequence class than its count suggests.
The federal minimum covers less than most people assume
Federal floodplain rules for highways live in 23 CFR 650 subpart A, codified in 1974 and updated on 26 November 1979, whose full text FHWA reproduces as an appendix to HEC-17. The design standard at section 650.115 is short, and it carries exactly one numeric floor. Through lanes of Interstate highways must be designed to a flood no smaller than the 2 percent annual chance event. The following sentence exempts everything else: “No minimum design flood is specified for Interstate highway ramps and frontage roads or for other highways.”
That is the whole of the federal minimum. Ramps, frontage roads, arterials, collectors and local roads carry no federally specified design flood at all, and what the regulation asks of them instead is a documented comparison of alternatives plus a risk analysis. The mechanics of that comparison are worked through in highway drainage systems; what matters here is which floods it has to name.
The pairing of floods it has to name is the interesting part. A design flood is the event “associated with the probability of exceedance selected for the design of a highway encroachment,” and the regulation then asserts, as a matter of definition rather than of evidence, that a road will not be inundated at the stage of its own design flood. An overtopping flood is separately defined as the event whose water surface elevation puts flow “over the highway, over the watershed divide, or through structure(s) provided for emergency relief.” The required risk analysis has to cover whichever of the overtopping flood and the base flood is greater. So a compliant design names the flood the road survives and, in the same set of calculations, the flood that runs over it.
Freeboard is discretionary, and defined twice
Section 650.115 handles freeboard in a single sentence with a hinge in it: “Freeboard shall be provided, where practicable, to protect bridge structures from debris- and scour-related failure.” HEC-17 reads the qualifier as substantive rather than decorative, noting that practicability “acknowledges that designing with freeboard, especially for retrofit projects, may not optimize cost, safety, and other considerations.” The circular goes further than permission. “In some cases, allowing short duration overtopping and designing the embankment to withstand overtopping may present a better risk profile, and thus better serve the public, than designs that avoid overtopping at the design discharge.” A perched bridge is the built form of that judgment, and it anticipates that its approach embankments will be overtopped.
The word also carries two definitions inside the same document. The regulation reproduced in the appendix defines freeboard as “the vertical clearance of the lowest structural member of the bridge superstructure above the water surface elevation of the overtopping flood.” HEC-17’s own glossary defines it as the “Vertical distance above a design water-surface elevation that provides a safety factor for waves, surges, drift, uncertainty in hydrologic estimates, and other contingencies.” One is measured from the overtopping flood, the other from the design flood, and the second is explicitly a bucket for uncertainty. A freeboard figure quoted without its definition is two different clearances at once. What debris and scour then do to the structure is the subject of bridge inspection and rating.
Culvert sizing runs on the same logic, in smaller pieces
For a culvert, the design intent is that the design event “should pass through the facility without significant loss of service or damage,” so that at the design discharge water does not overtop the road. Design return periods across transportation infrastructure “range from 10-year to 500-year depending on the type and purpose of the infrastructure,” and the circular shows how quickly ambition costs money: to hold the probability of exceeding the design criterion to 5 percent over 25 years, the design event has to be the 500-year storm.
Multiply that across a corridor and the exposure grows rather than averaging out. Extending the probability of exceedance across many features “increases the probability that managing the consequences of exceedances will be required over the design life of a project,” which argues for treating a route as a system rather than a list of structures. On tidal rivers the same logic compounds, because a rising ocean raises the downstream boundary condition, a mechanism traced in sea-level rise and coastal highways.
A worked case where the answer depended on which story was told
HEC-17 carries a case study from the second phase of the US Department of Transportation’s Gulf Coast study, at the Airport Boulevard crossing of Montlimar Creek in Mobile, Alabama. The site drains 3.3 square miles into four concrete box culvert cells, each 8 feet in rise and 12 feet in span, under a six-lane arterial linking downtown Mobile with its western suburbs and the airport. The applicable standard, a City of Mobile criterion derived from Alabama DOT standards, is to pass a 25-year flood with no less than 2 feet of freeboard measured from the roadway edge of pavement.
Two changes were tested against the observed 1980 to 2009 precipitation record. Future land use raised peak flows by 3 to 5 percent depending on return period, which the analysis judged not a significant contributor. The climate projections split. Under a wetter narrative, peak flows ran “72 to 80 percent greater” than the observed-precipitation baseline, while “the equally likely drier narrative suggests no change in peak flow.” The existing culvert met the criteria under existing conditions and failed some future 25-year scenarios, and the options priced were six cells instead of four, cells enlarged from 12 by 8 feet to 21 by 9 feet, or watershed measures to reduce runoff.
That is what an honest resilience analysis often looks like. The engineering was not in doubt. The input was two defensible futures pointing in opposite directions.
What the water does after the flood recedes
The structural question does not end when the level drops. FHWA’s 2023 survey of resilience practice across the pavement field reports that studies “tend to find that flooding saturates pavement sublayers rather quickly, weakening pavement structural components that are susceptible to increases in moisture content,” and that some work concludes thicker structures resist inundation better. At FHWA’s 2020 peer exchanges, participants said current design “does not take inundation and loss of support into account,” and one state had built a nondestructive testing protocol to judge when a flooded pavement could carry traffic again, finding “that it can take weeks before full structural capacity is restored.” Reopening a road that looks dry is therefore a structural decision, and it is one of the harder unresolved items in climate resilience coverage.
The loss figures are real and cannot be narrowed to roads
NOAA’s National Centers for Environmental Information counted 403 billion-dollar weather and climate disasters from 1980 to 2024, with inland flooding accounting for 45 of them at $203.0 billion in costs adjusted to 2024 dollars. The dataset counts public infrastructure such as roads, bridges and buildings among direct costs and does not report the highway share separately, so it is not recoverable from the published figures. NCEI also names its exclusions, among them natural capital, healthcare-related losses and the value of life, and draws the conclusion itself: that its estimates should be read as conservative relative to what is truly lost.
The federal repair channel is narrower still. FHWA’s Emergency Relief program is authorized at “$100 million annually” under 23 U.S.C. 125, pays a 100 percent federal share for emergency repair work in the first 270 days, then 90 percent on Interstates and 80 percent elsewhere for permanent repairs. An annual authorization at that level, set against flood losses in the hundreds of billions, is the clearest available measure of how much of this bill arrives outside the program built for it.
The regulation, in the end, makes an agency name two floods, and that is more revealing than either number. A standard that requires a design flood the road survives and separately requires analysis of the flood that runs over it has already conceded that the road will be overtopped inside its service life. The concession is the resilient part. What agencies have generally not written down is what the road is supposed to do on the day it happens.