1 Introduction

Hurricanes are one of the most devastating natural hazards in the United States, threatening communities, ecosystems, and infrastructures. Hurricanes have caused average normalized damage of US$ 10 billion annually in the continental US over the past century [1]. Economic losses from the 2005 hurricane season exceeded $200 billion, making it the costliest season ever [2]. Over 40% of the US population lives on coastal shorelines prone to hurricanes, and their contribution to the US economy is over US$ 9 trillion [3]. Recently, there have been more destructive and intense hurricanes (e.g., Hurricane Laura in 2020, Michael in 2018, Harvey in 2017, and Irma in 2017) that have ravaged the coastal areas of the US (the Gulf of Mexico in particular), causing extensive damage to infrastructures and loss of lives. For example, Harvey (2017), which mainly affected the coastal regions of Texas and Louisiana, caused a $62.2 Billion in normalized damage and led to 68 direct casualties. Meanwhile, recent hurricane seasons (2021 and 2022) are among the top 5 costliest hurricane seasons. Furthermore, it has been observed that regions that are populated by marginalized populations are significantly more prone to hurricane damage and loss. In other words, differences in the vulnerability of different population groups could result in regions inhabited by marginalized populations being more sensitive to hurricane hazards [4].

Hurricanes present two extreme hazards to coastal communities: intense winds and inundation, mainly due to storm surges. Some of the costliest hurricanes in US history were Category 4 or 5 upon landfall based on the Saffir-Simpson scale [5], and the extensive damage they caused was mainly due to intense winds (e.g., Hurricane Michael in 2018). However, Hurricane Katrina in 2005, which is to date the costliest hurricane in US history, was a Category 3 upon landfall [6]. Unlike some other hurricanes, a significant portion of the damage and loss due to Katrina is directly linked to coastal inundation due to storm surge [7, 8].

Current design guidelines for structures and infrastructure systems vulnerable to hurricanes specify design wind speeds derived based on the statistical analysis of historical hurricane data and simulations. They do not consider any possible changes in the intensity and frequency of hurricanes and the consequent increase in wind loads due to changing climatic conditions in the future. Further, the current flood risk maps from the US Federal Emergency Management Agency (FEMA) do not account for the impact of climate change on future surge and flooding hazards [9]. Meanwhile, the threat of hurricanes is ever-growing due to climate change [10,11,12,13]. Reports from the United Nations Intergovernmental Panel on Climate Change (IPCC) show that the period between 1983 and 2012 was the warmest 30-year period in the last 1400 years. This trend is expected to continue in the future [14]. In fact, the earth’s global mean surface air temperature in 2020 was 1.76 °F (0.98 °C) above the twentieth century average, making 2020 the second warmest year ever, followed by 2019 [15, 16]. The projected climate change scenarios of IPCC show moderate to a significant increase in sea surface temperature (SST) in the future, which would result in more destructive and intense hurricanes [11, 17, 18].

Hurricanes and climate change-related stressors also pose an increasing risk to coastal environments, ecosystems, and related economic activities. Hurricanes are partly responsible for the erosion of valuable barrier island shorelines. For example, the Louisiana barrier island shoreline, which supports a more than US$10 billion per year fishing industry, is eroding at a rate exceeding 20 m per year [19]. Extreme winds from hurricanes can damage ecosystems through the influx of organic material. Hurricane-induced storm surges can lead to scour of estuarine habitats and inundate terrestrial and freshwater habitats with saltwater. Similarly, extreme precipitation due to hurricanes can cause scouring in riverine ecosystems. More intense hurricanes due to climate change are expected to significantly impact the ecosystems in coastal areas. In fact, the IPCC has identified increasing hurricane activity coupled with sea level rise (SLR) as the most significant environmental threat to coastal ecosystems [20].

Many coastal areas of the US and other countries face permanent inundation due to SLR. It is estimated that by 2050, more than $100 billion worth of coastal property in the US is likely to be below sea level if current trends continue [21]. Worldwide, more than 410 million people in coastal areas are expected to be vulnerable to SLR [22]. Rising sea levels and drought induced by climate change can also lead to saltwater intrusion, contaminating freshwater supplies, harming agriculture, and threatening aquatic plants and animals [23]. SLR will also have a tremendous impact on coastal economies. For example, the majority of commercial fisheries depend on coastal marshes, which will potentially disappear due to SLR, affecting the livelihood of coastal communities [2]. The impact of SLR on coastal areas is compounded by land subsidence. For instance, land subsidence in areas of the US Gulf Coast is causing sea level to rise at ten times the world rate [19].

Future climate conditions depend on radiative forcing, defined as the balance between incoming and outgoing radiation [24]. Future radiative forcing depends on several factors, such as policies adopted to control the emission of greenhouse gases, population growth, and economic development. These factors introduce uncertainty in predicting future anthropogenic forcing levels [25]. The impact of future human driving factors of climate is usually modeled using a scenario approach. The IPCC developed four future climate scenarios termed Representative Concentration Pathways (RCPs) to investigate the impact of human-made forcing on climate [14]. The RCPs (RCP 2.6, 4.5, 6.0, and 8.5) represent total radiative forcing by the end of the twenty-first century compared with the year 1750. For example, RCP 8.5 represents an extreme scenario with a high level of radiative forcing (8.5 \(\mathrm{W} {\mathrm{m}}^{2}\)) by the year 2100 that excludes any climate mitigation policies in the future. RCP 2.6 represents a low forcing level, and RCPs 4.5 and 6.0 exhibit medium stabilized future climate scenarios [26].

Some studies have considered the effect of climate change on the frequency and intensity of future hurricanes and their probable impact on the related hazards (wind and storm surge). Mudd [27] compared simulated hurricanes for the year 2100 under the RCP 8.5 scenario with simulated hurricanes for the year 2005 to investigate the effect of changes in SST on hurricane wind hazard. The authors concluded that more intense storms are expected by the end of the twenty-first century, leading to higher structural design wind speeds. Knutson [28] projected an increase of up to 11% in wind speed by the end of the twenty-first century around the globe based on the RCP 4.5 scenario. While these studies use similar methods for hurricane simulation that consider future variations in SST, they do not consider other hazardous aspects such as storm surge and the possible relation between wind and surge hazard. Further, A recent study by Esmaeili and Barbato [29] indicates that the design wind speed corresponding to different mean return intervals considered by ASCE 7 are expected to increase between 14 and 26% on average for US Gulf and Atlantic Coast by the year 2060. A direct statistical approach is used to establish the relation between SST and corresponding hurricane parameters for specific sites. Although this method is computationally efficient, it only focuses on the available historic hurricane parameter statistics of a particular area and does not consider the effect of varying SST on hurricane intensity throughout its path before entering the site proximity. A separate study by Lin, Emanuel [30] shows that the occurrence of 100-year storms in New York may change to an event once every 3 to 20 years, considering climate change. Marsooli [9] found that the 100-year flood will become an annual occurrence for the mid-Atlantic region and a 1–30-year occurrence for the Gulf of Mexico by the end of the twenty-first century due to climate change. Similar studies based on projected future climate have found an increase in hurricane intensity and related hazards with increased average SST [12, 31,32,33,34]. There have also been studies on changes to hurricane induced rainfall and inland flooding due to the impact of climate change [35,36,37]. For example, Knutson [36] found that storm-induced rainfall for U.S. landing hurricanes will increase by 18% on average by the end of the century. Meanwhile, the rainfall rate for major hurricanes (Category 3–5) is predicted to increase even more (+37%). However, the level of this increase varies significantly due to the large uncertainty involved, which requires further studies [38].

However, it is noted that even though some studies have addressed the impact of climate change on hurricane hazards, there is still a necessity for thorough research and assessment of wind and storm surge risk across different regions that considers all the possible future scenarios as there is a certain level of uncertainty on which climate scenario would be occurring by the end of the century. Due to the spatial variation of SST across the Atlantic basin, the impact on hurricane risk across the different regions would vary by various degrees. Considering the variation in SST and the consequent effects on hurricane risk is critical to ensure that currently low-risk areas that may experience a significant increase in hurricane risk in the future understand and prepare for such an increase. Moreover, due to inherent uncertainties involved in hurricane risk assessment, detailed analysis considering these sources of uncertainties in terms of a probabilistic approach could lead to more accurate estimates of threats to coastal communities. Additionally, to account for the effect of climate change on hurricanes, a framework for risk analysis needs to consider two fundamental hazards associated with hurricanes: wind and storm surge. These two hazards are responsible for a significant portion of loss associated with hurricanes across the Atlantic basin. The combined effect of more extreme hurricane winds and increased risk of inundation from storm surges exacerbated by the sea level rise (SLR) will result in severe losses for coastal communities in the future. Finally, another important aspect that requires further attention is the interaction between surge and wind hazard for different locations as there is no guarantee that the most significant increase in surge and wind threat coincide spatially. Hence, depending on different parameters such as location and considered climate scenario, the dominant hurricane hazard could vary. It is crucial to understand the increase in risk level for each hazard for a studied community to provide effective countermeasures against future hurricane threats. Therefore, it is necessary to understand the impact of climate change on hurricane wind and surge hazards and their relationship to make well-informed decisions regarding community and infrastructure design, resilience, risk assessment, disaster preparedness, and post-disaster recovery.

This paper investigates the impact of potential changes in SST on future hurricane wind and storm surge hazards, using possible future scenarios in terms of RCPs. Due to the spatial variation in SST changes across the Atlantic basin, future changes in hazard levels are expected to vary across different regions. Hence, selected locations across the US Atlantic and Gulf coast regions are considered to study the spatial variation in future hazard levels. The objectives of the study are to (i) use a hurricane simulation model to generate hurricane ensembles based on an empirical model considering changes in SST and frequency for present and projected future climate scenarios (RCP 2.6, 4.5, 6.0, and 8.5), (ii) validate the simulation model by comparing simulated hurricanes under present condition and recorded historical hurricane statistical data like frequency, central pressure, heading angle, and translational speed; and (iii) generate the projection of the future hazards (wind and storm surge) for selected locations across the Atlantic and Gulf coast regions followed by a comparison of the hazard profile for current and future climate. The sites located in the Gulf Coast region are Houston, TX; New Orleans, LA; Mobile, AL and Tampa, FL. Meanwhile, the Atlantic Coast sites are Miami, FL; Charleston, SC; Norfolk, VA and New York, NY. These sites, which are amongst the most populated regions on the east side of the U.S., are selected in a way to be relatively scattered across the study region, so the spatial effect of SST variation and parameters such as sea-level rise on the studied hazards across the Atlantic coast and the Gulf area could be captured and observed. The hurricane hazards studied in the selected regions could potentially affect a large number of the population and lead to tremendous loss and damage.

2 Hurricane simulation model

Figure 1 presents the procedure for modeling the full tracks of tropical cyclones based on the Empirical Track Model (ETM), first proposed by Vickery et al. [39]. The model is computationally efficient, especially in cases where a large number of hurricanes need to be simulated. The ETM model is one of the most widely used hurricane simulation models in the literature [18, 30, 40,41,42,43]. Although the model is empirical, it can account for the effect of changes in SST on hurricane intensity and retain the information related to projected SST changes based on climate scenarios for different regions. The model consists of four major modules: Genesis, Track, Intensity, and Decay.

Fig. 1
figure 1

Flowchart for one storm year simulation

Full size image

In the Genesis module, the number of storms to be simulated per year is determined by sampling from a negative binomial distribution [44, 45]. Based on a statistical analysis of historical hurricane records from 1944 to 2020, where reliable historical records are available, the mean value of 11.20 storms/year and a standard deviation of 5.6 storms per year is obtained. Some studies have projected an increase in the frequency of future hurricanes considering the impact of different climate change scenarios with varying degrees [6, 27, 46,47,48,49,50,51]. For example, Emanuel [51] utilized a high-resolution simplified coupled atmosphere–ocean tropical cyclone model to estimate the impact of increasing greenhouse gases on tropical cyclone activity and concluded that the frequency will increase. Conversely, other studies have shown a decrease in the frequency of hurricanes in the future [47, 52,53,54,55]. Knutson [36] employed a two-step dynamical downscaling technique to investigate the activity of U.S. landfalling tropical cyclones in a warming climate. They found a robust basin-wide decrease in frequency (categories 1–5) but no robust increase or decrease in the frequency of U.S. landfalling hurricanes (categories 1–5). However, they did observe an increase in the frequency of landfalling intense hurricanes (categories 4 and 5). Chand [55] used the Twentieth Century Reanalysis (20CR) dataset and a high-resolution climate model to indicate that the frequency of tropical cyclones has decreased in the twentieth century due to anthropogenic greenhouse warming. Knutson [56] reviewed studies modeling the impact of anthropogenic warming on the frequency of tropical cyclones and reported a range of change from -28% to + 22%. Several factors could influence the projections of tropical cyclone frequency, including the choice of moisture variable in the genesis model [57], resolution of global climate models [58], mass flux reduction, and mid-tropospheric saturation deficit, among others [56]. Due to the inconclusive results regarding the effect of climate change on the frequency of hurricanes, three cases are considered: (1) 25% increase, (2) 25% decrease, and (3) no change in frequency of storms by the year 2100 over Atlantic basin compared to historical values. The genesis points, which are the locations of hurricane formation in the ocean, are selected based on available historical data in HURDAT [59] for both current and future climates.

The Track module indicates the location of the storm eye at 6-h time intervals and drives the storm through its path until dissipation. After selecting a genesis point for a storm, the Track module estimates the storm's new position, speed, and heading angle at 6-h intervals until the hurricane dissipates. The Intensity module determines the storm’s central pressure (Pc) in the ocean, indicating the intensity. The central pressure is associated with the SST in the model. Recorded SST data for 2005 and 2020 are used for current conditions, as seen later. The recorded monthly average SST values for 2005 and 2020 on a \(1\times 1^\circ\) grid were obtained from the Hadley Ice and Sea Surface Temperature (HadISST) database [60].

The projected future SSTs are based on a climate model following the protocol of the Coupled Model Intercomparison Project Phase 5 (CMIP5). Using the CMIP5 protocol under the radiative forcing corresponding to each RCP scenario, NOAA’s GFDL—Coupled Model (CM3) has produced monthly average sea surface temperature projections on a grid with a \(1\times 1^\circ\) resolution [61]. These average monthly SST projections are then interpolated for the exact location of the hurricane eye at a specific time during a simulation year and the intensity of the storm is calculated using the intensity module. This procedure repeats each time step as long as the hurricane is located over water. Once a storm makes landfall, its central pressure deficit decreases, weakening the storm over time. Hence, in the Decay module, the storm’s central pressure is calculated using the filling rate (Decay) model after landfall.

Numerous studies have acknowledged the high correlation between hurricane intensity and SST [12, 29, 39, 42, 44, 62,63,64]. For instance, Bhatia [50] used a coupled global climate model to show that tropical cyclone intensity and intensification will increase with warming SST under the RCP4.5 scenario. Meanwhile, some studies noted that other environmental factors such as changes to the environmental wind shear, relative humidity (RH), tropospheric humidity, etc. could also influence the intensity and frequency of hurricanes. Increases in vertical shear have been associated with decreased tropical cyclone activity, while increases in tropospheric humidity, vorticity, and potential intensity (PI) tend to enhance tropical cyclone activity [58, 65, 66]. Lee [57] utilized the Columbia Hazard model (CHAZ) to show that the impact of moisture changes on tropical cyclone activity depends on the choice of the moisture variable (column relative humidity vs. saturation deficit). They also demonstrated that as the climate warms, the saturation deficit decreases, PI increases significantly, while changes in deeplayer vertical wind shear, low-level vorticity, and ambient steering winds are relatively small. Vecchi [58] employed a coupled global climate model (GCM) with three resolutions to show that doubling CO2 will lead to a decrease in wind shear in certain areas (near-equatorial Pacific and northern tropical Pacific) and an increase in others (tropical North Atlantic and South Atlantic). These wind shear changes are expected to make the tropical North Atlantic less conducive to tropical cyclone activity. Additionally, the lowest resolution model shows an increase in mid-tropospheric RH, the highest resolution model shows a drop in the RH, while the mid-resolution model outcome falls in between. Ting [65] examined changes in PI and vertical wind shear due to anthropogenic changes (RCP4.5 and RCP8.5) and concluded that a suppression of wind shear is expected on the US East Coast, while an enhancement is expected in tropical North Atlantic. Emanuel [51] used the outcome of the downscaling of CMIP6 to show that PI and saturation deficit are the dominant factors influencing cyclone activity, while wind shear and vorticity contribute relatively little. In a study by Mudd [67] using a statistical hurricane simulation model, it was found that changes in RH, wind shear, tropopause temperature and heading are statistically insignificant at the 95% confidence level. The focus of this study is on the impact of changes in SST due to climate change on hurricane wind and storm surge hazards. Changes to other environmental parameters such as relative humidity and wind shear are not considered.

The empirical track model provides the full track of a hurricane from its formation in the ocean through the final dissipation while providing key parameters like the central pressure, storm heading, translational speed, and radius to the maximum wind (RMW) at every time step. It should be noted that some parameters, such as RMW and B parameters that describe the shape and size of the hurricane profile, are a function of central pressure, which makes them a function of SST. By using the statistical data from the historical hurricane record to calibrate the ETM and extract the necessary regression coefficients, a Monte Carlo simulation technique is applied to simulate the spatial and temporal evolution of tropical storms from formation to final dissipation over the domain of the Atlantic basin. It should be noted that methods such as ETM used for hurricane simulation are regression-based models and their accuracy and precision improve as more recorded historic hurricanes are used for calibration of the regression coefficients as the time passes. A Monte Carlo simulation method is used to generate a large number of scenarios to sufficiently capture the uncertainties inherent in the hurricane simulation process. The Monte Carlo simulation is performed for numerous years until the results converge. It was found that 20,000 years of hurricanes would suffice to estimate the hurricane risk. Therefore, 20,000 hurricane years were simulated using the method for each considered climate scenario. More details of the hurricane simulation model, including the equations referenced in Fig. 1 and model validation, are available in the supplemental material.

3 Storm surge model

To assess the storm surge hazard from the hurricane ensembles, the two-dimensional finite-difference hydrodynamic model, SLOSH (Sea, Lake, and Overland Surges from Hurricanes) [68], developed by the National Weather Service (NWS), is used. SLOSH is currently used by many agencies, including the Federal Emergency Management Agency (FEMA), National Hurricane Center (NHC), and local emergency agencies for real-time hurricane simulation, prediction, and disaster preparedness [69,70,71]. Compared to other computational storm surge analysis methods, SLOSH is computationally efficient with reasonable accuracy, making it an ideal choice when a large number of storm surge simulations are required. SLOSH can perform storm surge computations with good accuracy even with limited knowledge of storm parameters and intensity. Previous studies have found that where the hurricane is adequately described, the SLOSH model estimates storm surges with an accuracy of ± 20% [71,72,73].

ADCIRC (ADvanced CIRCulation model) is a two-dimensional, finite element hydrodynamic model capable of simulating hurricane wind field and wind forced circulation, i.e., storm surge and inland flooding [74]. The ADCIRC model estimates the surface wind speed using an analytical hurricane wind profile. Meanwhile, the SLOSH model deploys a semi-parametric hurricane model to determine the wind from the pressure field [68]. Storm surge simulation using ADCIRC may be more accurate than SLOSH due to the applied governing equation, grid resolution, and different wind profiles. However, the advantage is lost when considering the uncertainties in the generation of hypothetical hurricanes and associated high computational cost due to the number of cases to be studied. Therefore, SLOSH is used to assess the storm surge from the hurricane ensembles due to its extreme computational efficiency and reasonable accuracy, and a limited number of hurricanes are used to validate the surge analysis results using ADCIRC.

The SLOSH model is available for 36 different coastal basins covering most US coastal regions. The corresponding basins to the selected study sites are used to simulate the hurricane storm surge. The SLOSH meshes covering these basins extend approximately 400 km from the study locations. With the generated synthetic hurricane ensemble for the Atlantic basin for different scenarios, the storms that pass within 400 km of the study sites are used in the SLOSH model to estimate the most significant peak storm surge generated by the storm along the studied county’s coastlines for that hurricane year. Little to no surge height is observed for hurricanes that are further than 400 km from the studied site. The highest storm surge is normally coincident with the RMW since the strongest winds within a tropical storm lie at this distance [33]. As a result, the assumed radius is expected to be sufficient since the maximum observed radius-to-maximum-wind (RMW) from the simulated results is no more than 160 km. Further, to estimate the surge risk profile, only the peak surge height for each simulated year is recorded for each region, and minimal surge heights are not considered. The inputs, such as position, size, intensity, heading angle, and translational velocity for a 100-h window (one-hour intervals), are used to describe a hurricane. These parameters are also used in the wind field model and the driving force vector field on the water surface necessary for surge analysis.

4 Climate change impact

Following the simulation model, the hurricane hazard is assessed for the selected locations for both the current and the projected climate scenarios at the end of the century. For current conditions, the simulation is carried out for the years 2005 and 2020. For future scenarios, the simulation is carried out for RCPs 2.6, 4.5, 6.0, and 8.5 for the year 2100. Hereafter, the presented results are based on 20,000 years of hurricanes simulated for each of the six scenarios (2005, 2020, RCPs 2.6, 4.5, 6.0, and 8.5). Furthermore, four scenarios that study the impact of climate change on hurricane frequency and the resulting effect on hurricane hazards have also been discussed.

4.1 Impact of climate change on wind hazard

To evaluate the impact of climate change on the future design wind speed, assuming only changes to the hurricane intensity, the 3-s gust wind speed under different climate scenarios as a function of the Mean Recurrence Interval (MRI) for the eight different locations are simulated, and the results are shown in Fig. 2. It shows that the impact of climate change will lead to an increase in the intensity of design wind speed for all locations. It also indicates that intense winds are expected to occur more frequently in the future. For example, the wind speed corresponding to the 700-year MRI under current conditions (2020) in New York is predicted to become a 40-year event in 2100 under the 8.5 RCP scenario. On the other hand, Mobile, AL, is expected to experience the least changes in the intensity of future winds. Based on the wind speeds provided by ASCE 7–16, a 69 m/s wind is expected to occur every 700 years in Mobile, AL. However, under the RCP 8.5 climate scenario, the return period for this wind speed will be around 250 years by the end of the century. A similar trend is observed in other locations, especially for intense wind speeds with high MRIs.

Fig. 2.
figure 2

3-s gust wind speed and Mean Recurrence Intervals (MRI) for different climate scenarios

Full size image

A detailed comparison of changes in wind speed under various climate scenarios for all locations is shown in Fig. 3. It is seen that while the wind speeds for all future climate scenarios show an increase for all locations, such an increase is not uniform. For example, in New York, the 3-s gust wind speed for the 700-year MRI increases from 52 m/s under current conditions to 86 m/s at the end of the century under RCP 8.5, a 66% increase. On the other hand, Miami, FL’s 700-year MRI wind speeds are 75.5 m/s and 100 m/s under current and RCP 8.5 scenarios, respectively, a 32% increase. In general, a 32–66% increase in the 700-year MRI wind is observed in the Atlantic Coast regions. Meanwhile, a 15–27% increase in the 700-year MRI wind is predicted for the sites located in the Gulf due to the expected temperature changes under RCP 8.5.

Fig. 3
figure 3

Changes in 3-s gust wind speed from 2020 to 2100 for different Mean Recurrence Intervals (MRI)

Full size image

Figure 3 also shows that a relatively higher increase in wind speed is observed in New York for all climate scenarios. This can be explained by looking at the changes in SST from 2020 to 2100 (Fig. 4), where significant temperature changes in the Atlantic basin are observed towards the northern part, including New York. In Norfolk, VA, only a slight increase in wind speed is observed under RCP 2.6. For example, the 700-year MRI wind speed increased by about 4% under RCP 2.6. However, Norfolk, VA, ended with the second-highest increase in wind speeds under the RCP 8.5 scenario (about 52% for the 700-year MRI). The slight SST increase predictions for the RCP 2.6 scenario have led to a slight increase in future hurricanes in the Atlantic, resulting in a minor increase in the design wind speed. However, under the RCP 8.5 scenario, SST will experience a significant increase in the northern Atlantic, leading to more intense hurricanes. Mobile, AL, shows a relatively lower increase in wind speeds for all climate scenarios than other Gulf Coast regions. This could be because the increase in SST in the areas around Mobile, AL, is lower than in other Gulf Coast areas, as seen in Fig. 4.

Fig. 4
figure 4

Average changes in SST from 2020 to 2100 for RCPs 2.6, 4.5, 6.0, and 8.5 in ℃

Full size image

Figure 3 also shows that the predicted wind speeds under RCP 2.6 led to more significant changes in the design wind speed across all MRI spectrum for Gulf sites compared to the Atlantic Coast. This could be due to the minimal changes in SST in the Atlantic Coast region since the RCP 2.6 scenario predicts the slightest increase in temperature. This would lead to a slight increase in the intensity of hurricanes through their path for the sites located on the Atlantic Coast. Meanwhile, the uniform SST increase across the southern part of the Atlantic basin would gradually lead to a minor increase in the intensity of hurricanes going towards the Gulf of Mexico, leading to less significant changes in wind speeds in this region. When it comes to RCP 6.0 and 8.5 scenarios, the most significant rise in SST is observed in the Northern-Atlantic region. Hence, more intense hurricanes will be observed in this region compared to the Gulf coast. As a result, the rate of wind speed increase gradually shifts from the Gulf of Mexico to the Eastern Atlantic Coast for warmer future climate conditions.

Note that the results for RCP 4.5 and RCP 6.0 are very close. The RCP 4.5 and 6.0 results are similar because the difference in SST over a large portion of the Atlantic basin for the two scenarios is negligible (see Fig. 4). This led to only slight differences in average central pressure for simulated hurricanes under the two scenarios, as shown in Figure S4 in the supplemental document.

4.2 Impact of hurricane frequency changes on wind hazard

The impact of climate change on the intensity and frequency of future hurricanes has been extensively discussed in the available literature [25, 34, 42, 75,76,77,78]. While there is an established relationship between the increasing SST and the intensity of future hurricanes due to climate change, the effect of this phenomenon on the frequency of future storms is still debatable, and there is no consensus in the literature on even the direction of the change. The formation of hurricanes depends on several climatic variables such as sea surface temperature, atmospheric stability, “El Nino” effect, North Atlantic and Southern Oscillations, vertical wind shear, and other factors [6, 79]. Hence, variability in hurricane occurrence is a complex problem that requires multiple aspects to be considered. Table 1 summarizes some of the literature's reported changes in hurricane frequency. It can be observed that the reported changes in the frequency of hurricanes vary over an extensive range, from − 25% to + 55% for RCP 8.5 (and scenario A2 since it follows a similar emission trajectory), for example. The hurricane occurrence frequency is typically modeled using a statistical distribution such as negative binomial or Poisson [11, 80]. The mean hurricane occurrence per year is established by either assuming a constant mean value (no changes in frequency by the end of the century) or an assumed relationship that predicts this value in the future (e.g., linear trend, Linear Moving Average, or oscillating average). In this study, to consider the potential impact of climate change on the frequency of future hurricanes, a 25% increase and a 25% decrease in the storm’s occurrence over the Atlantic basin for RCP 8.5 are considered. This method can be applied to all the other available climate scenarios. However, only the changes for RCP 8.5 are shown herein for brevity.

Table 1 Summary of research on the impact of climate change on hurricane frequency (only studies that provide numerical values under specific emission scenarios are reported)
Full size table

To capture the impact of climate change on future wind hazard for the studied sites, the following scenarios are assumed for storms at the end of the twenty-first century: (i) 25% frequency increase with no intensity change; (2) 25% frequency decrease with no intensity change; (3) Constant frequency with intensity increase (based on the predicted SST of RCP 8.5 for the year 2100); (4) 25% frequency increase and intensity increase; (5) 25% frequency decrease and intensity increase. Note that the intensity of future hurricanes is a function of the SST variations. As a result, where no changes in the intensity compared to the year 2020 are considered, the recorded SST for 2020 is used. The 3-s gust wind speeds for the studied locations have been calculated using the previously explained method. Since the results across all locations follow a similar trend, only the results for Miami, FL, have been presented in Fig. 5.

Fig. 5
figure 5

Percentage difference of 3-s gust wind speed corresponding to different MRIs between the year 2020 and different frequency/intensity change scenarios for the year 2100 for Miami, FL

Full size image

It is observed that the assumed 25% increase/decrease in the hurricane frequency across the Atlantic basin leads to less than a 5% variation in the calculated wind speeds. In the case of the scenarios with only frequency changes where the hurricane intensity is assumed to be similar to the year 2020, the frequency change has a more significant impact on low MRI winds such as 25 and 50 years. This could be attributed to an increase in the occurrence of low-intensity storms as there are no changes in the SST assumed, and hence, no increase in hurricane intensity is observed overall. Furthermore, the wind hazard profile is extracted based on the peak observed wind speed in a simulation year. Hence, while there would be an increase in the hurricane occurrence rate over a region, the maximum recorded wind speed for the year could stay the same. A similar trend can also be observed for other study sites across the Atlantic basin and the Gulf area. The results show that changes in frequency have little impact on the wind hazard in cases where an increase in intensity due to changes in SST is also considered. A similar trend is also observed in the cases of intensity increase with no change in frequency and with ± 25% change in frequency. The difference in wind speeds in the three cases is below 5% for all MRIs. The increase in wind speeds for all sites corresponding to 700-year MRI ranges from 10 to 28 m/s by the end of the century, with New York having the most significant increase. Under the RCP 8.5 scenario and assuming a 25% increase in hurricane occurrence over the Atlantic basin, an approximately 70% increase in 3-s gust wind speed year700 year return period is observed for New York. Should there be a 25% decrease in the frequency of hurricanes, a 63% increase is observed for the design wind speed. Meanwhile, the projected changes in the 700-year MRI wind speed for Mobile, AL, assuming a 25% decrease and 25% increase in frequency, are estimated to be − 0.5% and 0.55%, respectively. Approximately 14% increase is calculated for the three cases that assume an intensity increase in future hurricanes for Mobile, AL. The fluctuations due to frequency changes for Mobile, AL are negligible as this site experiences the least variation due to hurricane intensity increase because of climate change. Based on these results, it can be concluded that the wind speed is dominantly affected by the intensity of the hurricanes and the variation of the SST. Meanwhile, the changes in frequency would result in a minor variation.

4.3 Impact of climate change on storm surge hazard

The storm surge for all simulated hurricanes that get within the vicinity of the study sites (400 km) is modeled using SLOSH. The surge heights are then used to find the storm surge corresponding to different MRIs for the study locations. Note that due to the uncertainty involved in the process for the future climates, the flood heights are determined, assuming the local current mean water level with no tide. Figure 6 presents the predicted surge height levels under present and future climate scenarios. Unlike the wind speed in the previous section, a clear relation is hard to observe between surge height and future hurricanes and the changes in SST. The main reason behind this is that topography, bathymetry, and geographic layout play a considerable role and directly impact the resulting surge height. Furthermore, variations in the hurricane track like heading angle, translational speed, landfall location, and RMW, also play a crucial role in surge heights. However, to this day, there is no clear evidence of the relationship between these parameters and climate change impact. Herein, the hurricane properties unaffected by SST are assumed to experience no changes in the future.

Fig. 6
figure 6

Estimated storm surge height and Mean Recurrence Interval (MRI) for different climate scenarios

Full size image

Figure 6 shows that high surge levels are expected to occur more frequently. For example, in Houston, TX, a 1.25 m surge that is likely to occur on average every 100 years is predicted to occur on average every 40 years under the RCP 8.5 scenario. Meanwhile, the expected surge with 100-year MRI for New York is expected to become a 55-year recurring event. Furthermore, while some study sites like Miami, FL, do not experience a significant surge height increase due to projected rising SST, the probability of exceedance of high surges increases considerably. In fact, Miami, FL’s 100-year surge height will become a 25-year surge incident for the RCP 8.5 scenario, which is the highest increase in frequency among the studied locations. On the other hand, should the RCP 2.6 scenario prevail in the future, Miami, FL, is expected to experience a 100-year surge every 80 years on average by the year 2100.

Changes in surge heights for all locations under RCP 2.6, 6.0, and 8.5 are shown in Fig. 7. As expected, there is a rise in surge level observed across the studied sites due to increased hurricane intensity. In addition, based on the analysis of the average translational speed of the storms for mileposts across the study region, an increase in the number of slow-moving storms is observed for future climate scenarios, with RCP 8.5 having a 1.7 m/s slower translational speed for hurricanes on average compared to current climate scenario. Storms with a slower translational speed allow the wind to blow for a more extended period leading to higher storm surges.

Fig. 7
figure 7

Changes in storm surge hazard from 2020 to 2100 for different Mean Recurrence Intervals (MRI)

Full size image

Figure 7 shows that the Gulf Coast will experience a more significant increase in predicted surge levels by the end of the century compared to the Atlantic coast, especially under RCP 2.6 and 6.0. The Gulf coast locations will experience a surge level increase of 2–9% under RCP 2.6 compared to 0–4% for the Atlantic Coast locations. Under RCP 6.0, the increase will be 7–21% and 4–11% for the Gulf and Atlantic Coast, respectively. Under the RCP 8.5 scenario, the changes are 7–29% and 10–16% in the Gulf and Atlantic Coast, respectively. Except for Houston, TX, the changes under RCP 8.5 are similar for the Gulf and Atlantic Coast locations.

A comparison of Figs. 3 and 7 shows that the impact of rising SST on the predicted surge heights is not as significant as the impact on wind speed, especially for the RCP 8.5 scenario. For example, the maximum increase in surge heights under RCP 8.5 is below 30%, while the maximum increase in wind speed is over 65%. However, as mentioned earlier, the surge heights are significantly affected by other factors related to the characteristics of the locations and the hurricanes. An index such as \({V}_{max}^{2}.RMW\) can be used to measure hurricane intensity and size for every studied site [9], where \({V}_{max}\) is the maximum wind speed of a hurricane, representing the intensity of the hurricane; and RMW which represents the size of the hurricane. The largest increase of this index is observed in the Gulf of Mexico, which leads to a greater increase in projected storm surge height in this area compared to the Atlantic Coast region when comparing the increase due to future climate scenarios such as RCP 8.5. It is also worth noting that the average translational speed of hurricanes in the Gulf of Mexico is relatively lower than the mileposts located across the Atlantic coast, specifically for Northern-Atlantic, as can be observed from Figure S2 in the supplemental material.

4.4 Impact of sea-level rise on storm surge hazard

The storm surge risk along the US Atlantic and Gulf coast is evolving due to climate change's impact on hurricane intensity and frequency. In addition, changes to the current sea level due to sea-level rise (SLR) are expected to affect the flood risk that threatens coastal communities and infrastructures. The future rate of global sea level is primarily controlled by the thermal expansion of ocean water combined with mass loss from glaciers, ice sheets, and ice caps [83]. As the local sea level could differ significantly from the global average sea level, a localized assessment of SLR is crucial for adaptation planning and risk mitigation [84].

To account for the combined effect of storm-induced surge and SLR on coastal flood risk, the probabilistic SLR projections under different RCPs for the end of the twenty-first century developed by Kopp, Horton [85] are used. This method provides projections of SLR at tide gauge stations across the globe by the end of the century. The SLR projections for each climate scenario have been used for the closest station to the counties in this study. The projected SLR for the locations range from 0.7 to 1.2 m for RCP 2.6, 1.0–1.5 m for RCP 4.5 and RCP 6.0, and 1.2–1.9 m for RCP 8.5. It should be noted that SLR and storm surge are assumed independent, and the nonlinear interactions are neglected. Figure 8 presents the relative increase in storm tide (effect of SLR and storm surge) with a 100-year return period for RCP 8.5 compared to the estimated surge values for the current climate scenario based on the year 2020. A comparison of the potential average SLR and the expected surge height with specific MRI (e.g., 100-years) could provide insight into the significance of the hurricane induced surge height and the potential changes of the hazard in the future for each region.

Fig. 8
figure 8

Percentage increase due to storm surge and sea level rise by the end of the century for RCPs 2.6, 4.5 and 8.5 compared to the estimated surge height for the year 2020 corresponding to 100-year MRI

Full size image

An approximately 25–55% increase in the 100-year flood (combined surge and SLR) is observed across all the studied locations for RCP 8.5 compared to the estimated values for the year 2020, with a more considerable impact of storm surge observed for the Gulf region. Amongst these sites, Houston, TX, showed the highest increase, with storm surge causing an increase of 23% and SLR causing a 30.5% increase. Meanwhile, the combined surge and SLR increase are more considerable for the sites across the Atlantic coast than the Gulf Coast locations. Norfolk, VA, experienced the highest predicted increase for RCP 8.5 (average increase of 1.05 m in SLR and 0.23 m in 100-year surge height). Across all studied sites, the highest increase in surge tide is predicted for Houston, TX, and Norfolk, VA, by the end of the century. Tampa, FL, is expected to experience the least increase in both storm surge and SLR, leading to a 30% increase in surge tide for RCP 8.5 by the end of the century. It should be noted that the predicted absolute surge heights for Tampa, FL, are the highest across all sites, which could lead to a reduced impact of increasing surge due to storms and SLR as a result of climate change.

5 Discussion

As mentioned earlier, the IPCC emission scenarios (RCP 2.6, 4.5, 6.0, and 8.5) represent four possible total radiative forcing by the end of the twenty-first century compared with the year 1750 [14, 86]. RCP 2.6 represents a scenario where mitigation actions are taken, leading to a very low forcing level that peaks and declines before 2100. RCPs 4.5 and 6.0 are stabilization scenarios where radiative forcing stabilizes by 2100 for RCP 4.5 but does not peak for RCP 6.0. RCP 8.5 is a scenario with very high greenhouse gas emissions representing a high-risk future. Such a scenario-based approach is warranted because the future climate will depend on policies adopted to control the emission of greenhouse gases, population growth, and economic development. The scenario-based approach does not consider the likelihood of the various scenarios. While such an approach is appropriate for impact assessment, as done in this paper, it may be inappropriate for climate adaptation decisions as it does not capture the associated relative risk of the various scenarios [87]. Many past studies on the impact of climate change on natural hazards tend to consider the most extreme emission scenario, mainly in an attempt to reduce computational effort. However, the most extreme scenario might not be the most plausible [88]. Therefore, all four scenarios were considered in this study.

Table 2 shows the IPCC projected change in global mean surface air temperatures and sea surface temperatures relative to the 1986–2005 period [89]. Note that for sea surface temperature, only projections for RCP 2.6 and 8.5 are provided by the IPCC. The likelihood of the various IPCC emission scenarios has been a subject of much discussion. Rogelj, den Elzen [90] assessed the effect of current Intended Nationally Determined Contributions (INDCs) of countries outlining their post-2020 climate action and concluded that a median surface temperature warming of 2.6–3.1 °C is expected by 2100. Compared to the IPCC projections in Table 2, it indicates that a scenario in between RCP 6.0 and 8.5 is likely. Some researchers have argued that the RCP 8.5 scenario is more likely than initially thought because of factors such as the release of greenhouse gases from thawing permafrost, which are larger than currently estimated [91, 92]. Other researchers have argued that the RCP 8.5 scenario is becoming increasingly implausible. Hausfather and Peters [88] argued that RCP 8.5 is increasingly implausible because (i) it will require a fivefold increase in coal use, which is highly unlikely, and (ii) the cost of clean energy sources will continue its falling trend.

Table 2 IPCC projected change in global mean surface air temperatures and sea surface temperatures relative to the 1986–2005 period
Full size table

There is an increasing call for a risk-based or probabilistic approach to modeling future climate scenarios [87, 88, 92, 93]. Such an approach will consider the likelihood of the various climate scenarios and assign probabilities to them. However, there are several challenges to moving to such an approach. The main challenge is that probabilistic climate scenarios might underestimate the uncertainty because of an inadequate number of global climate model runs due to computational limitations and the use of improper probability distributions in models [94]. Also, the likelihood of the various scenarios will keep changing constantly and will need to be updated as new data is collected and climate models are being updated [87]. Lastly, to assign the probabilities, climate modelers will have to move from focusing on only modeling the physical processes and work with policymakers, social scientists, and industry experts. Such a move is expected to take years of work [88]. In this study, results for all four RCPs are presented and discussed. While estimating the likelihood of the various scenarios is out of the scope of this work, readers and users are encouraged to consider the discussion and references above regarding the current climate trend and the most likely future climate scenarios.

6 Conclusions

This paper studied the potential impact of changes in SST on the intensity of future hurricanes and subsequent major causes of loss and damage, i.e., wind and storm surge. An Empirical Track Model was employed to simulate hurricane ensembles considering the effect of changes in SST on hurricane intensity. Using the four IPCC emission scenarios (RCPs 2.6, 4.5, 6.0, and 8.5), Projected SST was used to simulate future hurricanes. Eight locations across the Atlantic and Gulf coast regions were considered to illustrate the impact of climate change.

The intensity of future hurricanes is expected to increase for all the locations across the Atlantic coast due to climate change. Since the most significant changes in SST are predicted around the Northern-Atlantic coast, more intense hurricanes are expected in this area specifically. As wind speed is a function of hurricane intensity and is sensitive to the central pressure, higher wind speeds with more frequent intervals are predicted in the future. For example, it has been found that the present 700-year wind speed in the Atlantic coast will occur approximately every 30–60 years in 2100 under the most extreme emission scenario (RCP 8.5). Meanwhile, the results show that the 700-year wind will be an event with 140–200-year intervals for the Gulf of Mexico by the end of the century. It is also observed that the frequency of rarer wind speeds, e.g., 1700-year return period winds, will increase significantly compared to more frequent wind speeds with lower MRIs. A ± 25% change in hurricane frequency had a minor effect on the wind speed in all considered locations.

Similarly, an increase in the surge height for future climate scenarios is noted. However, the storm surge height relies on factors other than hurricane intensity like environmental properties, including geographical layout and bathymetry, and hurricane track characteristics like heading angle, translational speed, and landfall location. Hence, the predicted surge results show a slight increase compared to the wind speed. The changes are still significant, however. For example, the 100-year surge height for the present climate is expected to become a 25–50-year occurrence under RCP 8.5 for the studied locations. SLR exacerbated the coastal flood hazard in all locations, with Atlantic Coast areas experiencing a more remarkable impact relative to the Gulf Coast. It should be mentioned that in this study, the spatial variation of the surge return levels and relative impact of SLR and surge height under different climatology is examined.

It should be noted that the purpose of this study is to investigate the impact of increases in sea surface temperature due to climate change on future hurricane characteristics, specifically the intensity and frequency of storms, and to quantify the consequent hazards—extreme wind and storm surge—as a metric for comparison between present and future climate. The goal is to shed some light on the impact of climate change on the wind and surge hazards threatening coastal communities, which will help develop possible hurricane mitigation strategies. Hence, potential changes in other climatic (e.g., wind shear, tropopause temperature, relative humidity) and non-climatic factors such as topological and geographical changes of the studied coastal communities, layout and use of the land across the coastal region, and possible mitigative measures and actions taken in the future have not been considered in this study. Furthermore, it is worth noting that the accuracy of regression-based models such as ETM that are used to predict the activity of future hurricanes is sensitive to the training data, which is from recorded historical hurricanes. As more recordings of historical hurricanes become available with time, the accuracy of the hurricane simulation method, specifically in the Intensity module where the relationship between SST and hurricane intensity is established, will further improve. Another shortcoming of the regression-based ETM model is that it does not incorporate various physical processes such as large scale subsidence that are expected to be affected by climate change. Such omissions might lead to over- or underestimation of the projected changes in hurricane activity. Future research can aim to incorporate other factors that could possibly impact how climate change will affect hurricane activity in the Atlantic.

Another area of future research is the consideration of compound flooding, defined as the occurrence of multiple types of flooding (storm surge, riverine, rainfall) simultaneously or sequentially such that the overall hazard is increased. It has been shown that compound flooding resulting from the co-occurrence of storm surge and heavy precipitation has increased significantly in the U.S. over the past century [95]. The occurrence of such compound events is expected to increase due to climate change, especially due to changing hurricane patterns and sea level rise. The impact of compound flooding and sea level rise will be further exacerbated by increasing urbanization and population growth in coastal areas. The combined effects of riverine flooding, storm surge, sea level rise, and coastal development are not well understood, and modeling such effects is one of the critical research areas identified recently [96]. It is, therefore, essential to consider the co-occurrence of storm surge, heavy precipitation, riverine flooding, and sea level rise in future research.