Coastal flooding at predictable hours

Abstract

Flooding is typically perceived as a sudden and unpredictable hazard. Here, we show that recurrent flooding can occur at highly predictable times in tidally dominated coastal systems. This predictability stems from phase-locking of tidal constituents and constituent pairs with the solar day, causing peak tides to recur at consistent local times set by regional tidal propagation. Using tide-gauge records from the United States and the United Kingdom, we quantify the intraday timing of coastal flood events and show strong clustering at specific hours, particularly where semidiurnal or mixed tides dominate. For example, floods in Boston cluster around noon and midnight, whereas in southern California they occur in the morning. Sites with stronger non-tidal variability show weaker clustering. This temporal predictability extends beyond nuisance flooding to larger consequential events involving inundation, road closures and infrastructural damage, highlighting opportunities for anticipatory risk communication, emergency planning and time-sensitive coastal adaptation.

The study shows that coastal floods in tide-dominated regions often occur at predictable hours, enabling more targeted warnings, emergency planning, and time-sensitive adaptation.

Introduction

Coastal flooding poses a major and increasing threat to communities, infrastructure, and ecosystems in low-lying coastal regions around the world¹. Rising sea levels and intensifying extreme events amplify flood hazards^2,3, leading to increasing social, economic and environmental impacts^4,5. Flood risk is commonly expressed as the product of hazard, exposure, vulnerability and response^6–8:

$$\text{Flood risk}\left(x,t\right)=\text{hazard}\left(x,t\right)\times \text{exposure}\left(x,t\right)\times \text{vulnerability}\left(x,t\right)\times \text{response}\left(x,t\right)$$ 1

where x and t denote the spatial and temporal dimensions. Hazard reflects the probability and severity of flooding, shaped by water level characteristics (e.g., magnitude, duration, frequency, intensity) and local protective features such as embankments or natural barriers. Exposure refers to who and what may be affected and depends on the presence and activity of humans. Vulnerability captures the susceptibility of these exposed elements to harm. Response refers to the capacity and actions taken to anticipate, manage, and recover from flood events, ranging from emergency services and evacuation protocols to long-term adaptation measures. A key but underexplored dimension of flood risk is timing, which directly modulates all four components. For example, minor to moderate flooding during commute hours or business operations can cause greater economic disruption due to increased exposure than physically similar events that occur overnight or during non-business hours^9–12. In contrast, the vulnerability to injury or loss of life may increase during the night or early morning hours, when people are asleep or less alert^13,14. A striking example is the 1953 North Sea flood, which occurred overnight and on a weekend, delaying the emergency response and amplifying its human toll¹⁵. Therefore, any possible alignment of flood events with daily rhythms of exposure and vulnerability could play a critical role in shaping their real-world impacts¹⁶.

Flood hazard varies with both season and time of day. Although the seasonal timing of river and coastal floods has been widely studied^17–19, the intraday timing of floods has received little attention²⁰, reflecting the stochastic nature of meteorological forcing. In tidally influenced coastal systems, however, flooding often arises from the superposition of tidal high waters and non-tidal processes, including meteorological and oceanographic contributions (e.g., storm surge, wave run-up, and steric variability). Tidal dynamics impose strong temporal regularity: the probability of exceeding flood thresholds increases sharply during high tides, particularly during spring tides. In semi-diurnal regimes, these spring tides are the result of the fortnightly alignment of the Sun and Moon, and because the solar constituent (S₂) is locked to the solar day, high spring water recurs at consistent hours. This phase locking has long been recognized in theory and practice. Historical tide predictions used the ‘age of the tide’ or ‘establishment’ to estimate the delay between lunar zenith and spring high water²¹. These empirical observations, rooted in the experience of mariners, linked spring tides to a predictable delay after the full moon reaches its zenith, which itself typically occurs near midnight²². By the mid-19th century, tide prediction methods in the United States formalized this concept through the so-called first and second reductions²³. Although modern tide tables still report the establishment or age of the tide²⁴, the implications for flood timing are often overlooked. Specifically, both minor high-tide floods and compound events involving storm surge may preferentially occur near the time of establishment. At specific sites such as Boston, a semidiurnal preference for flooding has been documented since the early 20th century, near midnight and noon local time^25,26.

Despite this physical regularity, most studies still characterize flooding using coarse temporal metrics such as flood days or monthly event counts^2,27,28. These approaches obscure sub-daily timing patterns that could enhance risk understanding, early warning, and adaptation planning. Moreover, previous literature has cautioned against overstating the role of peak tides^29,30, given the masking effects of atmospheric variability. However, the timing of the flood threshold exceedance remains a physically meaningful and underutilized dimension of flood risk.

In this study, we systematically analyze coastal flood timing using tide gauge records and flood event registrations from the United States (US) and the United Kingdom (UK), selected for their contrasting tidal regimes and surge characteristics, which together span a broad range of coastal forcing conditions and extensive data availability. We assess (see “Methods”) how flood events are distributed on both daily and seasonal timescales and how this temporal predictability varies between regions and with local forcing characteristics, particularly the relative strength of tidal versus non-tidal drivers. Our findings reveal that in many tide-dominated locations, flood events follow highly predictable daily patterns. With sea-level rise, calm-weather flooding has become increasingly common in low-lying urban areas, where elevated spring tides now periodically lead to inundation^27,31,32. This growing prevalence of tide-only flooding further underscores the importance of understanding not just whether floods occur but also when. Such insights allow anticipatory risk communication, infrastructure planning, and policy design to account for the timing of floods.

Results and discussion

Why do coastal floods favor certain hours?

Coastal water levels are strongly shaped by tidal dynamics, particularly by the interaction between the principal lunar (M₂) and solar (S₂) tidal constituents. These constituents oscillate at slightly different frequencies, once per 12.42 h for M₂ and once per 12.00 h for S₂, producing a beat pattern that modulates the tidal amplitude, resulting in the spring-neap tidal cycle (Fig. 1a, b). The S₂ phase lag sets the timing of spring tides, with the highest water levels occurring when M₂ and S₂ are in phase. Because the S₂ constituent completes exactly two cycles per solar day, spring high water levels recur at two specific hours, namely those aligned with the S₂ peaks.

Fig. 1. Tidal dynamics explain the intraday timing of flood peaks at Boston.

a heights of the principal lunar (M2, orange) and solar (S2, black) tidal constituents at Boston, MA (NOAA ID: 8443970) over half a lunar cycle; b combined M2 and S2 signal over two lunar cycles, illustrating spring-neap modulation; c full tidal prediction using all 37 NOAA tidal constituents (relative to NAVD88, the standard North American vertical datum); d timing (in UTC) and height of predicted tidal high waters in the year 2025 (relative to NAVD88); orange and black lines indicate the timing of M2 and S2 peaks, respectively; e timing of threshold exceedances above NOAA's NOS minor flood level (2.078 m+NAVD88); f predicted tidal high water heights (relative to NAVD88) and their intraday timing. Colors in panels (b–d, f) indicate the height of local high water. Lunar phases are indicated in panels (a, b, d) by filled, open and half-filled circles, together with gray vertical grid lines.

Beyond the dominant M₂ and S₂ constituents, additional tidal components modulate the timing and amplitude of high waters (Fig. 1c, d). Diurnal constituents, primarily K₁ and O₁, generate once-daily oscillations that produce inequalities between successive high waters, which can cause the highest tides to occur preferentially at only one particular time of day. In predominantly diurnal regimes (Form factor > 3, see “methods”), this would yield no strong intraday clustering of peaks; however, interactions between K₁ and smaller diurnal constituents such as P₁ introduce amplitude modulations that still favor particular hours of the day.

On longer timescales, (semi-)annual constituents (Sa, Ssa) and phase locking between specific tidal pairs produce seasonal clustering of extreme tides³³, such as S₂-K₂, S₂-T₂, K₁-P₁, or S₂-K₁. Semidiurnal tidal amplitudes typically peak around the equinoxes (March and September), whereas diurnal tides reach their maxima near the solstices (June and December). Non-tidal influences, including daily and seasonal variations in atmospheric pressure and wind patterns, may further enhance or shift this clustering^34,35. At Boston, for instance, the annual highest tides typically occur in May and November, reflecting the combined seasonal modulation of diurnal and semidiurnal constituents (Fig. 1d). Supplementary Note 1 provides a mathematical framework for predicting when such phase-wrapping occurs, how many distinct peaks emerge in the solar reference frame, and over what interval the pattern repeats. This helps explain the emergence of intraday and intrayear clustering in tidal extremes.

The contribution of tidal dynamics to the timing of real-world flood events is exemplified using observational data from Boston, Massachusetts (Fig. 1e). Extreme high waters and threshold exceedance events show a marked concentration around specific hours of the day, consistent with patterns predicted by tidal constituent interactions. Notably, 70.4% of all exceedances above the NOAA National Ocean Service (NOS) minor high-tide flooding (HTF) threshold (hereafter the minor NOS threshold; 2.078 m + NAVD88 at Boston) occur within ±1.5 h of the dominant peak times. These peaks typically occur near 4 am and 4 pm UTC (corresponding to 11 pm and 11 am EST), forming a bimodal daily distribution. The slight skew toward the midday peak (43.2% of all exceedances) coincides with business hours and may amplify societal impacts. This preference for local noon and midnight has been noted historically^25,26, and is consistent with the phase-locking behavior of key tidal constituents.

Spatial patterns in timing

Our analyses proceed from general high-water variability to events more directly associated with flooding. We first examine the timing of high water events with exceedance intervals of one year or less, as these provide a consistent and comparable basis across stations. We then assess whether these timing characteristics persist for subsets of events exceeding site-specific thresholds more directly linked to flooding.

The intraday timing of high water events with exceedance intervals of one year or less across coastal regions closely follows the spatial progression of the solar semidiurnal (S₂) tidal wave (Fig. 2). To investigate the relevance of this spatial control, we classified each tidal station by its dominant exceedance timing (see “Methods”) and compared these classifications with the S₂ phase distribution. Peak timing modes were identified by locating the three-hour window that contains the majority of threshold exceedances (Mode 1). After removing this block, a second three-hour window with the next highest concentration of exceedance peaks was identified as Mode 2. Prominence refers to the fraction of peaks that occur within each mode relative to the total. The sensitivity of peak hour and prominence to the chosen recurrence interval is explored in the Supplementary Fig. 1.

Fig. 2. Spatial patterns in the intra-day timing and prominence of threshold exceedance peaks across the US and UK.

a, b First exceedance peak mode for the mainland US (a) and the UK (b), based on long-term tide gauge records. c, d Second mode for the same regions. Marker color indicates the time of day (in UTC), with round symbols denoting morning (AM) peaks and square symbols denoting afternoon (PM) peaks. Marker size reflects the prominence of the mode, defined as the percentage of annual exceedance events occurring within 1.5 h of the modal time. White-filled markers indicate locations where the distribution of peak times is not significantly different from a uniform distribution (p > 0.01). Inset maps in (a, c) highlight Puget Sound, San Francisco Bay, and Chesapeake Bay. The background shading shows the phase of the S₂ tidal constituent relative to UTC, derived from the EOT20 global tidal model⁶⁶. Coastlines and political boundaries from the Global Self-consistent, Hierarchical, High-resolution Geography database (GSHHG;⁶⁷).

Along the West Coast of the US, the S₂ tidal wave exhibits a clear northward progression, traveling from the Mexican to the Canadian border in approximately 5 hours. This results in a systematic shift in exceedance timing from south to north (Fig. 2a–c). This region displays some of the highest Mode 1 prominence values, particularly in the Pacific Northwest, reflecting a strong and temporally coherent daily peak. Mode 2 is generally not statistically significant, indicating a predominantly unimodal distribution (one prominent daily high water). In contrast, the southern US Pacific coast shows a higher prominence of Mode 2, suggesting a more bimodal pattern in exceedance timing.

On the US East Coast, peaks are more synchronized, and most locations experience threshold exceedances approximately at the same time. The prominence of Mode 1 is moderate (typically 25–50%), and Mode 2 is often not statistically significant, suggesting a more complex interaction between tidal and non-tidal influences. This widespread synchronization results from shelf co-oscillation, where tidal forcing reflects off the continental slope, producing a standing-wave pattern with a nearly uniform phase along the US East Coast coast^36–38. A notable exception is the northern East Coast close to Canada, where local tidal dynamics in the Gulf of Maine introduce a 4–5 h delay in exceedance peak timing and enhance the prominence of both daily modes.

The Gulf Coast in the south of the US exhibits a more diffuse and less predictable pattern. Here, the tidal range is smaller (not shown), and the diurnal constituents dominate (Fig. 3a). Both the small tidal range and the dominance of the diurnal constituents result in weak clustering of the exceedance timing. Both Mode 1 and Mode 2 show low prominence, and the S₂ wave progression is limited. This reduced regularity likely reflects the combined influence of diurnal tidal forcing and the frequent impact of tropical cyclones. These geographic contrasts mirror the regional division in which tides control HTF along the West Coast and in the Gulf of Maine, whereas non-tidal residuals dominate along most of the East Coast and the Gulf coast in the south³⁹.

Fig. 3. Spatial patterns in the intra-year timing and prominence of threshold exceedance peaks across the US and UK.

a, b First exceedance peak mode for the mainland US (a) and the UK (b), based on long-term tide gauge records. Marker color indicates the time of year of the dominant intra-year peak, while marker size reflects its prominence, defined as the percentage of annual exceedance events occurring within 1.5 months of the peak modal time. White-filled markers indicate locations where the distribution of peak dates is not significantly different from uniform (p > 0.01). The background shading shows the tidal form factor (F), calculated as (K₁ + O₁)/(M₂ + S₂), which reflects the relative influence of diurnal versus semidiurnal tidal constituents. Coastlines and political boundaries from the Global Self-consistent, Hierarchical, High-resolution Geography database (GSHHG;⁶⁷).

In the UK, the exceedance timing closely follows the progression of the S₂ tidal wave, although the presence of multiple amphidromic points introduces notable spatial complexity (Fig. 2b–d). Both Mode 1 and Mode 2 are prominent along much of the UK coastline, together indicating a high degree of temporal regularity. Most sites exhibit most exceedance peaks during the afternoon or evening hours (UTC), with a few exceptions in the southeast and southwest of the UK, where morning peaks are more common.

In semi-enclosed basins and estuaries, such as Puget Sound, Chesapeake Bay, and San Francisco Bay in the US, and the Bristol Channel in the UK, exceedance timing often lags behind that of the adjacent open coast. These lags reflect the inertial response of shallow water to the tide-generating forcing under the influence of local bathymetry and geometry and may amount to several hours (e.g., up to five hours at Philadelphia in the Delaware River). As a result, exceedance timing can vary substantially over short spatial scales, and mode prominence tends to be more spatially variable.

Across both the US and the UK, the first intra-year exceedance timing mode (Mode 1) exhibits strong temporal clustering, with slightly higher prominence observed at US tide gauge stations compared to the UK (Fig. 3). Along the West Coast of the US and the northern segment of the East Coast, peak timing is concentrated in December and January, reflecting enhanced diurnal tidal amplitudes around the winter solstice³¹. In contrast, the Gulf Coast and the Southeast US show peak timing between August and October. In the UK, peaks occur most frequently in January and February, although several sites exhibit earlier peaks in October and November. The timing of these annual peaks likely primarily reflects non-tidal influences, such as the regional storm season that dominates in the UK during winter months.

Balancing tides and storm surges

To assess the influence of tidal versus surges on exceedance timing, we use the ratio of the tidal range to the surge height corresponding to a once-per-year exceedance as a proxy for tidal dominance (see “Methods”). We find that regions with high tide-surge ratios, where tidal forcing dominates over non-tidal variability, exhibit stronger clustering and regularity in exceedance timing (Fig. 4a). For gauges with a ratio exceeding 2, the minimum prominence is 58.1%, with a median value of 89.0%.

Fig. 4. Influence of tidal dominance on the prominence of exceedance timing modes.

Subplot a shows the combined prominence of exceedance timing modes 1 and 2 (see “Methods” for details), expressed as the percentage of annual exceedance events occurring within ±1.5 h of the modal times. The prominence values are shown in relation to the tide-surge ratio, which is defined as the tidal range divided by the surge height corresponding to a once-per-year exceedance. Subplots b, c display the individual contributions of mode 1 and mode 2, respectively. Blue markers represent NOAA (US) tide gauge locations; pink markers represent BODC (UK) sites. Crosses indicate locations where the distribution of peak timings is not significantly different from a uniform distribution(p > 0.01). Dashed lines show expected values if peaks were evenly spread across the day (uniform distribution). Subplot d presents a global map of the tide-surge ratio, derived from the GTSM model^40,41. The map was created by interpolating data from 43,119 GTSM output locations. High-latitude regions near the poles were excluded due to expected limitations in model performance. The underlying tidal range and 1-yr surge maps used to compute this ratio are provided in Supplementary Fig. 2. Coastlines and political boundaries in Subplot (d) are from the Global Self-consistent, Hierarchical, High-resolution Geography database (GSHHG;⁶⁷).

The prominence of mode 1 is consistently elevated in these regions, indicating a robust primary timing signal. In contrast, the prominence of mode 2 varies depending on the diurnal inequality (difference between the two daily highs). Where tides exhibit high diurnal inequality, mode 1 tends to dominate, suppressing mode 2 prominence. This distinction is evident when comparing US and UK tide gauge sites. US locations, particularly along the west coast, generally show higher prominences in mode 1 and lower in mode 2 for similar tide-surge ratios, while UK sites often exhibit a substantial prominence in mode 2, approaching the theoretical maximum of 50%.

A global tide-surge ratio map, derived from predictions from the Global Tide and Surge Model (GTSM)^40,41, extends these regional findings to a global context (Fig. 4). High ratios are observed near the equator, where storm surges are relatively low, and along the western coasts of the Americas and Euro-African continents. These regions are characterized by strong tidal dominance, and consequently, exceedance peak timing should be more predictable and temporally clustered. In contrast, areas with low ratios, such as those near amphidromic points or where surge variability is high, are expected to exhibit a weaker clustering of exceedance events. The observed spatial patterns are consistent with earlier results indicating strong clustering of exceedance timing in both the US West Coast and the UK. A limitation of this map is that it does not distinguish between tidal regimes. Diurnal tides show less clustering in the timing of daily peaks than semidiurnal tides, so intraday regularity is slightly reduced in diurnal or mixed regimes with comparable tidal ranges.

Timing patterns in observed coastal floods

Having established that tidal dynamics strongly govern the timing of water-level exceedances occurring once per year or less, we now examine how this regularity extends to actual flood events, those that cause measurable disruption or damage. We use standardized impact-based thresholds from the NOAA NOS⁴² for the US and the SurgeWatch database for the UK⁴³ to distinguish between flood severities. NOS minor floods correspond to “nuisance” or “sunny-day” flooding, localized ponding or street flooding without major damage, while moderate floods cause road closures and structural damage. SurgeWatch moderate events (Category ≥ 3) meet at least three impact criteria, including inundated property, service disruption, extensive flood descriptors, or sea-defense damage. These consistent definitions allow us to assess whether tidal control extends to real, observed floods and to examine how their timing aligns with typical daily human activities (Fig. 5). We note that such thresholds provide a simplified representation of flood impacts; actual flooding can vary depending on local conditions, infrastructure, and processes (e.g., wave run-up, drainage, and flood pathways), and may not be fully captured by a single water-level threshold.

Fig. 5. Timing of flood events per station.

Hourly distribution of flood peaks exceeding impact-relevant thresholds at selected tide gauges along the US and UK coasts in local time. Sites were chosen to provide a representative spatial spread along both coastlines and to include well-known or metropolitan locations. a, b Relative probability of exceedances over NOAA NOS minor (a) and moderate (b) flood thresholds; c Timing of events classified as moderate or worse (Category 3 and higher) in the SurgeWatch database⁴³. Probabilities are normalized per station to account for differing event counts (with the number in brackets specifying the number of floods per station). Time windows associated with varying societal vulnerability (e.g., working hours, rush hour, nighttime) are indicated by yellow bars next to each heatmap.

In the US, exceedances of the NOS minor flood threshold exhibit clear regional patterns (Fig. 5a). Along the Northeast Coast, the timing of floods is mostly clustered with bimodal peaks. From Eastport to Boston, these peaks are very pronounced and typically occur around noon and midnight local time, aligning with business hours and nighttime. South of Boston, the temporal organization remains evident, though slightly reduced. Flood peaks shift to the morning and evening, coinciding with morning commute periods, increasing the potential for transportation-related impacts. The Delaware (Cape May & Philadelphia) and Chesapeake Bay (Sewells Point to Baltimore) regions provide exceptions with weak clustering. Further south, parts of the Carolinas and Florida (Wilmington to Key West) retain moderate regularity, while the timing along the Gulf Coast remains largely unpredictable, shaped by small diurnal tides and frequent tropical cyclones.

The West Coast of the US displays some of the strongest temporal coherence. In Southern California (San Diego to Port San Luis), flood peaks align with the morning rush hour, while further north, timing shifts progressively later, aligning with midday business hours. Within Puget Sound (Port Townsend, Seattle & Cherry Point), flood peaks occur earlier than on the outer coast, again aligning with morning rush hour to early business hours. Hawai’i (Honolulu & Hilo Bay) also exhibits a bimodal pattern, with peaks in the early morning and late afternoon. At the moderate flood threshold (Fig. 5b), exceedances are less frequent, but predictable timing persists in several regions. The Northeast and Northwest Coasts continue to show clustering, albeit with reduced prominence.

In the UK, flood events classified as moderate or higher in the SurgeWatch database (Fig. 5c) show remarkable temporal regularity. Most locations exhibit bimodal peaks, with a clear spatial progression along the coast. Flood timing aligns with business hours, rush hours, or nighttime, depending on location. Notably, in the southeast UK (Harwich to Dover), where the 1953 North Sea flood caused significant loss of life, flood peaks occur in the middle of the night, a factor likely contributing to the high casualty rate¹⁵.

Implications

Our results demonstrate that flood events in many coastal tidally dominated systems occur at consistent times of day and year, driven by the phase-locking of key tidal periodicities with the solar day and annual cycle, such as the semidiurnal S₂ and annual SA constituents. Although this has long been recognized in tidal theory³⁰ and observed at selected sites⁴⁴, we present a systematic investigation of intraday flood timing across a broad range of tidal conditions. The same tidal mechanisms also govern the timing of the lowest low tides, with implications for navigation and ecological exposure⁴⁵. We show that this timing persists even in the presence of pronounced non-tidal residuals, revealing a more widespread and robust pattern than previously recognized. This challenges the prevailing view that atmospheric variability diminishes the relevance of peak astronomical tides^29,30, and instead suggests that tidal timing plays a more critical role in flood occurrence than previously recognized⁴⁶. One reason why this pattern remains poorly known may be that most flood studies report only the day of occurrence, and metrics such as ‘flood days’^27,47 obscure the sub-daily structure that governs real-world impacts.

The timing of floods in tidal systems is shaped by a combination of diurnal inequality, tidal constituent interactions, and seasonal modulation. Bimodal patterns arise when both daily high tides regularly exceed flood thresholds, typically in systems with small diurnal inequalities. In contrast, many locations, such as the Pacific Northwest, exhibit strong diurnal inequality, resulting in a unimodal distribution of flood peaks. This inequality varies seasonally, as the timing of the higher high water gradually shifts between day and night. In mixed tidal regimes, the M₂-S₂ spring-neap cycle interacts with the O₁-K₁ fortnightly tidal cycle, amplifying spring tides twice per year³³. When combined with an underlying annual cycle, be it tidal, meteorological, or oceanographic, this shift can amplify one peak over the other, resulting in consistent threshold exceedance at a specific time of day. While M₂-S₂ phasing governs the dominant timing, additional constituents such as K₂ and T₂, along with long-period cycles such as 4.4 and 18.6-year nodal cycles, modulate flood timing on seasonal and interannual scales^2,48,49. These tidal effects interact with broader meteorological and oceanographic cycles (e.g., storm seasons, discharge regimes, ocean currents) to constrain exceedance timing at tide gauges to narrow windows, often limiting possible exceedance to just a few dozen hours per year, although local land-based factors may modify flood timing on land⁵⁰.

Peak times are not arbitrary, but reflect the basin-scale propagation of the S₂ tidal wave: In the Atlantic, standing-wave behavior causes semidiurnal high waters to align across thousands of kilometers³⁶; along the US West Coast, tides are characterized by a progressive Kelvin wave, producing a northward shift in flood peak timing. This makes flood timing a defining characteristic unique to each coastal site. Within estuaries, shallower depths and river discharge lead to more spatially variable timing (Fig. 2a–c), and these systems are especially sensitive to anthropogenic changes such as dredging, land reclamation, and sea-level rise^51–54. Consequently, secular changes in timing⁵⁴ and exceedance patterns are increasingly observed in estuarine and shallow coastal zones⁵⁵. We note that tide-gauge timing reflects local tidal propagation and may not directly represent flood timing at regional scales. Nonetheless, these local variations can in principle be accounted for, analogous to the way tidal predictions are routinely translated to specific sites. Our two case study regions, the US and the UK, span a broad spectrum of tidal environments, including progressive and standing-wave regimes, estuaries and semi-enclosed basins, and a wide range of form factors and tide-surge ratios.

The strength of timing depends on the ratio of tidal to non-tidal variance: where tides dominate, exceedances cluster at one or two clock times; where the non-tidal forcings dominate, timing is more diffuse. Crucially, the influence of non-tidal residuals is frequency-dependent: low-frequency residuals (e.g. seasonal mean sea level) shift the background state without disrupting intraday clustering, whereas high-frequency residuals (e.g. tropical cyclones) that exceed the spring-neap envelope can break down the degree of clustering. The shape of the tidal envelope and the high water have a further influence on the clustering and duration of flood events^56,57. At the same time, part of the non-tidal signal contains energy at the same solar frequencies that organize the tides (Sa, Ssa, S₁, S₂), arising from radiational heating, sea-breeze cycles, monsoon seasonality, meltwater rhythms, or river discharge periodicity; these processes can reinforce rather than degrade timing coherence when they phase-align with the underlying astronomical cycle^34,35. On shallow continental shelves, where compound tide-surge interactions are strong (e.g., North Sea^58,59), flood peaks remain phase-anchored. Tide-surge interaction can shift the peak relative to the predicted tide, but when that shift is systematic across events, exceedances still align in clock time. Consistent with this, UK North Sea stations exhibit a pronounced clustering of peaks.

As sea level rises, exceedances first appear at the dominant tidal peak times and then expand to other hours (see Supplementary Note 2). Flood predictability is tied to its observed frequency, a relationship that holds under sea-level rise. As moderate and major floods become more common, they inherit the temporal predictability currently associated with minor floods. Conversely, the predictability of minor floods diminishes as they transition into a persistent, chronic background condition. This transition underscores the importance of viewing sea-level rise not just in terms of magnitude, but also timing. Minor shifts in flood timing may occur as sea-level rise alters tidal propagation^52,60,61, but dominant timing modes are expected to persist.

The predictability of flood timing offers clear opportunities for anticipatory adaptation. Targeted flood warnings and infrastructure operations, such as flood barriers⁶², pumps, gates, and transit systems, can be aligned with peak flood hours to reduce disruption and damage. Emergency response and public awareness efforts can also be tailored to high-risk times. More broadly, timing data could be integrated into flood risk maps, insurance models, and urban planning codes to better reflect the temporal dimension of coastal flood risk. The practical implication is direct: flood risk has a schedule, and accounting for it offers a concrete lever for anticipatory adaptation.

Methods

We analyze long tide-gauge records from the United States (US) and United Kingdom (UK) to quantify the temporal organization of flood events. Hourly observations and tidal predictions are used to decompose water levels into tidal and non-tidal components and to identify exceedance events of consistent impact severity. We extract the time-of-day and time-of-year of each event to diagnose dominant timing modes and relate these patterns to tidal range, residual magnitude, and tidal propagation characteristics. Finally, we place local findings in the global context using tide-surge ratios from a global tide-surge model.

Tide gauge data

Hourly water level records were obtained from NOAA (U.S.; https://tidesandcurrents.noaa.gov) and BODC (U.K.; https://www.bodc.ac.uk/data/hosted_data_systems/sea_level/uk_tide_gauge_network/). For US stations, we selected sites with published sea-level trends and retrieved all hourly observations and predictions since 1900. Stations with <20 years of data were excluded to provide sufficiently robust statistics of tidal timing and to ensure that records capture key sources of variability (e.g., the 18.6-year nodal cycle and interannual climate modes). Data were referenced to the station datum, reported in metric units, and standardized to UTC. All records were quality-controlled by the data provider, and no additional cleaning was required. The UK data was accessed via GESLA-3⁶³, filtered for stations with >20 years of data, manually cleaned and resampled at hourly intervals. Tidal predictions were generated using UTide⁶⁴, using the most recent year of observations per site. The residual water level was defined as the observed minus the predicted tide. The tidal range was determined from mean high and low waters in the predicted tide; for US stations, the mean range (MN) was taken from NOAA datums. The strength of the non-tidal component was characterized using the once-per-year exceedance level of the residual water level, based on a peak-over-threshold (POT) approach applied to maxima (≥12-h separation). Return levels were estimated by selecting the ranked peak corresponding to the record length, adjusted for data completeness. We quantified the relative importance of tidal and non-tidal contributions as the ratio of tidal range to this residual level.

Flood event detection and mode analysis

Flood events were identified using annual exceedance thresholds (adjusted for data gaps) of observed water levels, based on the POT approach, and using site-specific impact thresholds. The annual threshold (~1 event per year) was selected to ensure sufficient sample size, maintain relevance to impactful high-water events, and allow comparability between stations (see Supplementary Fig. 1 for sensitivity). As site-specific impact thresholds, we adopt NOS HTF minor/moderate flood thresholds for the US⁴² and SurgeWatch-2 Category 3+ events for the UK⁴³, the latter of which are explicitly impact-based and provide impactful threshold exceedances. NOS thresholds are not strictly impact-based and may over- or underestimate actual flooding⁶⁵. Nevertheless, they provide consistent coverage across stations—whereas NWS thresholds, which are more directly linked to impacts, are not available for all sites—and therefore offer a practical basis to demonstrate the timing characteristics of flooding, capturing the underlying phenomenon even if the link to impacts is more implicit^16,42.

For each exceedance, we extract the time of day and the day of the year. To detect intraday modes, we identify the 3-hour window containing the largest number of events (Mode 1), remove those events, and then repeat to identify a secondary 3-h Mode 2. Mode centers are taken as window midpoints; prominence is the fraction of events within the window. Significance is assessed with a binomial test (insignificant if p > 0.01). The same procedure is applied seasonally with a 3-month window to detect the dominant annual mode(s).

Tidal propagation, form factor and global tide-surge ratio

To test whether intraday timing aligns with tidal propagation, we extract the phase of S₂ relative to UTC from EOT20⁶⁶. The amplitudes of M₂, S₂, K₁, and O₁ were used to compute the tidal form factor, which indicates whether tides are predominantly diurnal (F > 3), semidiurnal (F < 0.25), or mixed (0.25 ≤ F ≤ 3):

$$F=\frac{K_1+O_1}{M_2+S_2}$$ 2

The relative tide-surge strength was defined as above. To assess global prevalence, we used tidal range and once-per-year residual from the Global Tide and Surge Model^40,41 at 43,119 output locations and evaluated the same ratio.

Author contributions

Conceptualization: J.B., S.T., I.H., B.H., and G.L. Methodology: J.B., S.T., I.H., B.H., and G.L. Investigation: J.B. Visualization: J.B. Writing—original draft: J.B. Writing—review & editing: J.B., S.T., I.H., B.H., G.L., D.M., and A.H. Supervision: S.T., I.H., B.H., D.M., and A.H. Funding acquisition: J.B., S.T., A.H.

Peer review

Peer review information

Nature Communications thanks Katherine Anarde and the other anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was funded by the Netherlands Organization for Scientific Research (NWO) within the Vici project “Deltas out of Shape: Regime Changes of Sediment Dynamics in Tide-Influenced Deltas” (Grant NWO-TTW 17062). Joris Beemster acknowledges that this research was supported by the WIMEK PhD Grant 2021. Financial support for this research was also provided by the U.S. Department of Defense’s Strategic Environmental Research and Development Program (SERDP) under Project RC12-3456.

Data availability

Tide-gauge data used in this study are publicly available from NOAA and BODC, and both datasets are also accessible through the GESLA-3 archive⁶³. UK flood data used in this study are available from the SurgeWatch-2 database⁴³ at https://www.surgewatch.org. The global ocean tide model EOT20, used to map the phase of the S₂ tidal constituent and form factor, is available via SEANOE at 10.17882/79489⁶⁶. Global tidal range and once-a-year residual water-level thresholds, used to map the global tide-surge ratio, are available through the Copernicus Climate Change Service Climate Data Store under DOI 10.24381/cds.6edf04e0^40,41. US flood thresholds and sea-level rise scenarios used in this study are available from NOAA at https://tidesandcurrents.noaa.gov. The source data generated in this study and underlying the figures are provided with this paper as Source Data files. No data used in this study are subject to restricted access. Source data are provided with this paper.

Code availability

The Jupyter notebooks used to determine the intraday and intrayear modes and their statistical properties are archived on Zenodo at 10.5281/zenodo.20272888.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-75710-5.