Climate Change Impacts on the Risk Assessment of Concrete Civil Infrastructures

Strong winds and tropical cyclones cause substantial damage worldwide every year. Atlantic hurricanes resulted in $150 billion in losses from 2004 to 2005 (Pielke et al. 2008). With the growth of population and economy, coastal regions are increasingly susceptible to more severe damage from hurricanes and tropical cyclones. Additionally, global warming may lead to variations in hurricane wind speed, frequency, and hurricane-triggered storm surges (Bjarnadottir et al. 2014). Although the scientific community is still studying how the relative change of sea surface temperature affects the occurrence and overall intensity of tropical cyclones (Wu et al. 2022; Wahiduzzaman et al. 2022), numerous studies have demonstrated that global warming increases the frequency or/and intensity of strong wind and tropical cyclones.

Bender et al. (2010) investigated the influence of global warming on future Atlantic hurricanes using a hurricane prediction model. Their findings indicated that despite the potential decrease in the overall frequency of tropical cyclones, there would be an almost doubling of the frequency of Category 4 and 5 storms by the end of the 21st century. The most substantial growth was predicted in the Western Atlantic north of 20°N. In the Pacific region, research results have indicated that the frequency of future typhoons will decrease compared to the current climate, but extreme typhoons will be more likely to occur and shift to the northeast (Chen et al. 2021b). A model was proposed by Esmaeili and Barbato (2021) via historical data to aid in predicting the wind speeds of hurricanes under different climate conditions, as shown in Fig. 2. The design wind speeds along the US Gulf and Atlantic coasts would increase by approximately 14% and 26% by 2060 under RCP 2.6 and RCP 8.5 scenarios, respectively.

It becomes evident that the traditional design wind speeds are inadequate for future nonstationarity climates, and this subject has attracted significant attention from researchers. Mudd et al. (2014) evaluated the most severe potential import of the RCP 8.5 scenario on hurricanes based on the historical hurricane events database. According to the simulation results, the design wind speed in the current codes requires to be increased to guarantee the safety and performance of structures. Xu et al. (2020) investigated the import of the tropical cyclone on the building code for two coastal cities, Shanghai and Hangzhou in China, from 1979 to 2098. They pointed out that when adapted to the future climate conditions, the return periods of tropical cyclones would decrease and the design wind speeds recommended by current codes would be relatively low.

The aforementioned literature primarily focuses on the influence of climate change on the frequency and intensity of strong winds, and the validation of design wind speeds in structure codes. In addition, strong wind is a complex climatic phenomenon associated with heavy precipitation and flooding, constituting a typical multihazard scenario that will be further discussed in the following sections.

Snow pressure is one of the main environmental loads considered in civil infrastructure design, generally determined by the ground snow load. The representative value of ground snow load corresponding to a certain annual exceedance probability can be determined based on probability distributions (e.g., the Gumbel distribution being most widely adopted) obtained from the statistical analysis of extreme ground snow load over a 40–50-year period (CEN 1991). To analyze the trend of annual extreme ground snow load, it is critical to consider not only the change in the mean value but also the change in variance, i.e., extreme snowfall conditions. On the one hand, global warming induces an increase in rainfall rather than snowfall (i.e., the average snowfall decreases) and accelerates the rate of snow melting (O’Gorman 2014). On the other hand, an increased extreme snowfall intensity under future climate scenarios will result in larger extreme ground snow loads (Raisanen 2008).

Croce et al. (2019, 2021) investigated snow load models in Europe considering climate change and established a correlation between snowfall, precipitation, and temperature by analyzing available observational data. A decreasing trend was observed in the representative values of ground snow load due to global warming, which decreases the maximum annual mean value of ground snow loads. Conversely, the reduction in the mean value was followed by an increase in the variance so that larger extreme load values could appear in some scenarios. Jeong and Sushama (2018) studied the trend of snow loads in Canada by adopting the snow water equivalent to quantify ground snow loads and predicted the ground snow load values for a 50-year return period via the global climate models (GCMs). The results demonstrated that the ground snow load values generally decrease in southern Canada but increase in northern Canada. The Canadian design codes were subsequently calibrated by Hong et al. (2021) based on historical data, considering the influence of global warming. Fig. 3 illustrates the effect of global warming on the annual extremes of snow loads in Canada. It is evident that the impact of climate warming is not consistent for different regions but overall shows a trend of decreasing mean and increasing variance. Hence, they suggested that different safety factors should be used for the representative values of snow loads in northern and southern Canada, which is consistent with the conclusions drawn by Jeong and Sushama (2018). ASCE7-22 (ASCE 2022) revises the reliability-targeted ground snow loads based on nearly 30 years of additional snow load data. This revision accounts for climate change and site-specific variations in annual ground snow loads across the United States.

In summary, snow loads in Europe and Canada show increased variability under climate change and the effects of various uncertainties need to be comprehensively considered in climate adaptation. It should be added that although the impact of climate change on snow loads in other regions of the world has been rarely studied, the need to update snow load provisions in current design codes is well recognized (Mo et al. 2016; Al-Rubaye et al. 2022).

Changes in river discharges are also one of the consequences of climate change. For example, the melting of more snow caused by rising temperatures and the increase in local precipitation will lead to higher river discharges and increased flood frequency. The consequence is that the global population affected by 100-year floods under 2°C and 4°C warming above preindustrial levels exceeds 200 and 500 million, respectively (Kundzewicz et al. 2010).

According to a report by the Joint Research Centre (Faggian 2018), river floods exhibit greater spatial variability and more significant fluctuations in the frequency of extreme values compared to other hazards under future climate conditions. In western Europe, both average and extreme precipitation are predicted to increase, leading to an increase in the frequency of floods. In the case of southern and east-central Europe, the frequency of floods tends to go down as a result of the decrease in snowmelt-induced floods, which offsets the influence of the increased average and extreme precipitation. Veijalainen et al. (2010) quantified the impact of climate change on floods in Finland and found that the magnitude of the impact varies by season and region. Tabari (2020) illustrates that 30-year flood intensity is expected to increase over most areas of the globe by the end of the 21st century, but there is great uncertainty in different regions. Yang and Frangopol (2019) predicted and evaluated the flow rate of the Lehigh River and the flood scour risk of bridges. The results indicated that the Lehigh River will experience an increase in both flow rate and frequency of flooding. The 20-year flood return period in 2020 is expected to decrease to 13 years by 2099 under RCP 4.5 and RCP 6.0 emissions. In the extreme climate scenario, i.e., RCP 8.5 emission, the frequency of floods is expected to decrease due to the rise in temperatures and the reduction in river discharges. Furthermore, there is considerable variability in the results predicted by different GCMs. Yin et al. (2018) investigated the flood peak discharge and flood volume of the Ganjiang River in China under the RCP 8.5 emission scenario. It was illustrated that for floods with return periods > 50 years, the flood peak and 7-day flood volume are projected to increase by 12.1%–42.4% and 11.6%–37.4%, respectively, by 2080. Generally, climate change has a greater impact on flood frequency than flood intensity.

Based on the preceding analysis, it can be seen that the impact of climate change on river discharges and flooding is difficult to define coherently, and this impact will correspondingly vary with the hydrological conditions of each region. Furthermore, significant variability in flooding occurs under various emission scenarios and climate models. To accurately investigate future river flooding for a particular region, it is crucial to have a deeper understanding of the local hydrometeorological conditions, the influence of precipitation and temperatures on floods, and the conversion of climate scenarios into hydrological conditions (Veijalainen et al. 2010).

The increase in extreme floods is directly related to scour damage to bridges. Scouring endangers the bearing capacity of the bridge foundation and consequently reduces the safety of the bridge. According to Wardhana and Hadipriono (2003), 503 bridges collapsed in the United States between 1989 and 2000, with roughly half of these collapses attributed to flood scouring. Per the US Federal Highway Administration (Richardson and Davis 2001), scour is defined as “the erosion or removal of a streambed or riparian material from bridge foundations due to the river discharges” and is generally classified as (1) long-term riverbed degradation; (2) riverbed shrinkage; and (3) localized scour. The most significant damage is caused by localized scouring around piers or abutments, which can be quantified by scour depth. The calculation of scour depth needs to consider many factors, including riverbed types, river discharges, and bridge abutment shapes, of which river discharges are one of the key factors. Several scholars have studied the calculation of the scour depth and scour risk of bridges considering the influence of climate change on extreme river discharges and flood frequency (Imam 2019; Kallias and Imam 2016; Khelifa et al. 2013; Liu et al. 2020). It is worth noting that river discharges, which are closely related to scour damage of bridges, may vary significantly among different rivers or even different locations within the same river. When considering climate change, it is essential to account for the impact of this uncertainty.

Rising temperatures, increasing water acidity, and sea-level rise are the major impacts of global warming on coastal systems (Nazarnia et al. 2020). Among these impacts, the first two significantly accelerate structural resistance degradation, while the latter augments the structural load. Sea-level rise will elevate the frequency and intensity of natural hazards, such as floods and tsunamis, affecting infrastructure in offshore areas and resulting in substantial economic losses. With every 50-cm sea-level rise, the tsunami risk in certain coastal areas may increase by as much as two times (Li et al. 2018).

To quantify the threat of sea-level rise, it is first necessary to predict the extent of sea-level rise, which can be done using either the semiempirical formula or the GCM-based approaches. In the semiempirical formula approach, the relationship between temperature rise and sea-level rise is fitted based on previous observation data to predict future sea-level rise (Rahmstorf 2007; Vermeer and Rahmstorf 2009). In the GCM-based approach, GCMs are established based on the physical mechanisms, which are used to predict future climate and sea-level rise by assuming future emission scenarios (Church et al. 2013). Compared to the semiempirical formula approach, the GCM-based approach can more effectively account for regional variability, making it more commonly used. Noteworthy, the results predicted by different GCMs vary significantly due to the different assumptions and modeling approaches. Alhamid et al. (2022b) categorized sea-level rise into three components: (1) stereo dynamic sea-level rise; (2) glacier sea-level rise; and (3) ice sheet sea-level rise. A probabilistic model, which considers various uncertainties, including emission scenarios, GCMs, and observational data, was proposed for predicting sea-level rise. The prediction results from Alhamid et al. (2022b) are presented in Fig. 4. Sea-level rise exhibits regional variability and is relatively low in the areas near the Indian Ocean and the Pacific Ocean, such as India, New Zealand, and the west coast of the United States. Nevertheless, coastal cities near the North Atlantic and the Arctic Ocean, such as some in England and Norway, are subject to a greater threat from sea-level rise.

One consequence of sea-level rise is the increase in coastal flooding, which is also related to storm surges (already discussed in the previous sections). This poses a significant threat to critical offshore infrastructure (Dismukes and Narra 2018). Moreover, sea-level rise will greatly increase the frequency of coastal flooding according to Kopp et al. (2014). Buchanan et al. (2017) observed variations in the amplification of flood frequencies for different magnitudes. For example, the probability of flooding in Seattle increased 108, 335, and 814 times for the 10-, 100-, and 500-year return periods, respectively, under a 50-cm sea-level rise scenario, whereas the same scenario in Charleston led to increases of 148, 16, and 4 times for the corresponding return periods. The climate-hydrodynamic modeling was adopted by Marsooli et al. (2019) to quantify the effects of sea-level rise and tropical cyclones (under RCP 8.5 emissions scenario) on flood hazards along the Atlantic and Gulf coasts of the United States in the late 21st century. The study illustrated that the relative influence of tropical cyclones will continue to increase from New England, the mid-Atlantic, and the southeast Atlantic to the Gulf of Mexico, and this influence will be larger than that of the sea-level rise in 40% of the towns in the Gulf of Mexico. To quantify the effect of sea-level rise (excluding the effect of storm surges on flood frequency), Hague and Taylor (2021) proposed a conceptual model of tide-only inundation and a practical approach for developing tidal inundation statistics, including means, historical trends, and future predictions.

Apart from the floods, tsunamis are also a consequence of sea-level rise. Li et al. (2018) investigated tsunami risk in Macau under sea-level rise of 50 cm (by 2060) and 100 cm (by 2100) and created a probabilistic tsunami inundation map. The results demonstrated that sea-level rise could result in a significant increase in the frequency of tsunamis, doubling by 2060 and tripling by 2100. A semiempirical equation approach was used by Dawson et al. (2016) to investigate the relationship between flooding and rail transport restrictions caused by sea-level changes over the past 150 years and to estimate the extent of sea-level rise impact on the railway from London to Penzance in the United Kingdom. It was indicated that the number of rail restrictions increased significantly with rising sea levels. By the end of the century, the number of line closures will have increased to 3.6 times per year from once every 3–4 years by 2020. Although many researchers have studied flood and tsunami risks under different sea-level rise scenarios, this is a complicated multihazard problem (e.g., floods brought by hurricanes and tsunamis caused by earthquakes), which will be discussed in the following sections.

Previous structural designs did not account for the effects of climate change. In other words, climate variables were assumed to be stationary, and design loads were determined based on historical data on hazard occurrences. However, climate variables such as temperature, precipitation patterns, and sea-level rise are undergoing global changes, introducing nonstationarity in external loads, e.g., extreme wind speeds, ground snow depths, flood return periods, and precipitation. Moreover, this trend is usually biased toward hazardous conditions. For instance, traditional infrastructure designs that rely on stationary climate assumptions may substantially underestimate the frequency of extreme precipitation, resulting in increased flood risks for the infrastructure system (Cheng and AghaKouchak 2014). Consequently, it is important to integrate nonstationary climate scenarios into the new design codes to guarantee the long-term safety and performance of infrastructure.

To consider the nonstationarity climate changes and update the structural design loads, it is essential not only to model and predict future climate but also to assess the effects of climate change on the frequency and extremes of various hazards. Eventually, these effects need to be reflected in determining design loads (Orcesi et al. 2022b). In preparation for future climate change and urbanization, Gilroy and McCuen (2012) developed a methodology for evaluating nonstationary flood frequency and applied it to the Little Patuxent River in Guilford, Maryland. The results revealed that there would be a 30.2% increase in flood frequency by 2,100 for the 100-year return period. Sarhadi and Soulis (2017) developed a time-dependent risk framework that incorporated the precipitation intensity–duration–frequency to integrate the impact of climate change into infrastructure design guidelines. Hong et al. (2021) calibrated the design wind, snow, and companion load factors in the National Building Code of Canada, considering the effects of nonstationarity climate change on extreme wind and snow loads.

Climate is subject to great uncertainty. While epistemic uncertainty can be reduced through research efforts, aleatory uncertainty cannot be completely eliminated. The structural design and assessment need to include reliability, risk, and resilience (Akiyama et al. 2020; Wasko et al. 2021). To accurately quantify the predictive uncertainties, it is advisable to use more comprehensive emission scenarios and GCMs (Jeong and Sushama 2018). Lompi et al. (2021) provided a method to quantify the expected changes in future hydraulic risk in the Pamplona city catchment. The proposed method considered 12 climate models, 7 return periods, 2 emission scenarios (i.e., RCP 4.5 and RCP 8.5), and 3 time periods (i.e., 2011–2040, 2041–2070, and 2070–2100) in the EURO-CORDEX project. The study identified that the maximum flood design values decreased for the 10-year return period but increased for the 500- and 1,000-year return periods across the aforementioned three specified time periods. In the case of the RCP 8.5 emission scenario, the flood quantile is expected to increase for the return periods exceeding 50 years, and the increase in design peak flow is about 10%–30% greater than that of the RCP 4.5 emission scenario.

Regarding concrete infrastructure, reinforcement corrosion stands out as a primary factor affecting structural durability. The corrosion of reinforcement in concrete includes two forms: corrosion due to carbonation and corrosion due to chloride ion ingress. Global climate change has an impact on both reinforcement corrosion forms. For example, Köliö et al. (2014) demonstrated that CO₂ concentration and extreme precipitation have an impact on the carbonation initiation time, as well as temperature on the diffusion time, based on collected climate scenario predictions and the durability data of concrete structures in Finland. Khatami and Shafei (2021) analyzed corrosion due to chloride ion ingress in concrete bridges located in the Midwestern United States, considering climate change. They argued that the corrosion initiation time would be shortened by 13%–39% for different emission scenarios, and the crack width would expand with global warming.

Despite the differences in mechanism, carbonation corrosion and chloride-induced corrosion share great similarities in the development process. Both types can be divided into two stages: diffusion and propagation. In the diffusion stage, concrete is gradually carbonated or eroded by chloride ions, eventually leading to the breakdown of the passivation film on the reinforcement and the initiation of corrosion. During the propagation stage, reinforcement corrosion progresses over time, resulting in the deterioration of structural resistance. Therefore, to investigate the influence of climate change on the deterioration process of concrete structures, the first step is to study the impact of climate change on both the diffusion and propagation processes. The critical factors of the two stages are the corrosion initiation time and corrosion rate, respectively.

For the diffusion stage, the time-dependent effects of environmental factors, including temperature and relative humidity, are taken into account by using a modified diffusion coefficient, which is expressed as follows:where D_ref  = diffusion coefficient for reference atmospheric environment, e.g., temperature T = 20°C and relative humidity (RH) = 75%; f_T (t) = effect coefficient of temperature variation; and f_RH(t) = effect coefficient of relative humidity variation, both effect coefficients are related to time t. The Arrhenius law can be adopted to calculate the temperature effect coefficient as follows (DuraCrete 2000; Yoon et al. 2007):where T(t) = time-dependent temperature (unit: K); T_ref  = reference temperature, usually set as 20°C; E = activation energy required for diffusion; and R = gas constant.

The variation of f_T (t) with temperature is illustrated in Fig. 5(a). As the temperature increases, the f_T (t) tends to increase and both CO₂ and chloride ions have larger diffusion coefficients. This implies that both carbonation rates and chloride ions erosion rates increase under global warming, leading to a previous initiation of corrosion. The specific rate of acceleration varies depending on the magnitude of the temperature increase.

In terms of f_RH(t), the calculation models for carbonation and chloride ion erosion differ. Chloride ion erosion predominantly depends on moisture transport, meaning higher relative humidity results in a higher erosion rate, i.e., a greater f_RH(t). The f_RH(t) under chloride ion erosion can be calculated using the model proposed by Saetta et al. (1993):where RH(t) = time-dependent relative humidity; and RH_c = relative humidity at which the diffusion coefficient drops to half, usually considered as 75%. The effect of relative humidity on the diffusion rate of chloride ion erosion is depicted in Fig. 5(b), where the reference RH is equal to 75%.

In the case of carbonation, the effect of relative humidity follows a nonmonotonic pattern. When the relative humidity falls within the range of 50%–100%, an increase in relative humidity contributes to a reduction of the carbonation rate. In contrast, a reduction in relative humidity does not affect the carbonization rate when relative humidity falls between 30% and 50% (Ahmad 2003). Concrete carbonation ceases entirely at a relative humidity level below 25% (Richardson 1988). A model that considers the effect of relative humidity is provided by the International Federation for Structural Concrete as follows (fib 2006):where RH_ref = relative humidity in the reference environment; and f_e and g_e = constants of 2.5 and 5.0, respectively. It is therefore concluded that the diffusion process should be determined according to the specific condition of climate change at the infrastructure location, considering the complex influence of relative humidity. The consensus of existing research is that global warming is generally responsible for accelerated carbonation and chloride ion ingress, resulting in previous corrosion of reinforcement.

The nature of rebar corrosion is an electrochemical reaction that involves the combined effect of moisture and oxygen. The corrosion rate of rebar is influenced by not only the material properties of the rebar and concrete but also the ambient temperature, moisture, and oxygen. These environmental conditions are more difficult to determine when concrete cracks exist, which is why accurately modeling the corrosion behavior of reinforcement is a difficult task. The model provided by DuraCrete (2000) is usually adopted to simulate the corrosion rate of reinforcement under different temperature conditions:where i_corr,20 = corrosion current density at the reference condition (T = 20°C); i_corr(t) = corrosion current density at a temperature of T(t); and K = 0.025 and 0.073 at temperatures less and greater than 20°C, respectively.

Fig. 6 shows the predicted corrosion initiation time and corrosion rate of reinforced concrete (40-mm concrete cover thickness) for five emission scenarios in the AR6. Note that the influence of temperature on corrosion rate is considered in this model, whereas the effect of relative humidity is excluded. The effect of relative humidity is extremely complicated since both high and low relative humidity could accelerate the corrosion process. Even under the same relative humidity conditions, the corrosion rate of reinforcement differs significantly. Moreover, many previous studies treat carbonation and chloride ion corrosion as separate processes without accounting for their combined impact. Therefore, further investigation is needed to develop a model for reinforcement corrosion considering climate change.

The previous section provided a brief description of how climate change affects the deterioration process. This section is dedicated to reviewing the specific research progress on structural resistance and reliability assessment under the influence of climate change.

Stewart et al. (2011) evaluated the probability of corrosion initiation and damage to infrastructure under changing climate variables (i.e., CO₂ concentration and temperature), considering different emission scenarios. Taking Sydney and Darwin as examples, they investigated the corrosion risk of Australian infrastructures under climate change. Their findings indicate that the effects of climate change vary across different climate zones but generally result in accelerated corrosion propagation. If corrosion damage is defined as the presence of corrosion-induced cracks exceeding 1 mm in width, climate change could increase the likelihood of corrosion damage by up to four times in certain severe exposure scenarios. Based on this, Stewart et al. (2012) further investigated the risk of carbonation corrosion in Australian infrastructure under an emission scenario. In arid regions of Australia, the carbonation depth was observed to decrease due to moisture scarcity, with reductions of up to 15 mm in certain areas. On the contrary, in other regions, the carbonation depth could increase by 8 mm due to rising temperatures. They suggested that differences in climate conditions among regions must be considered to obtain accurate results when performing a large-scale analysis. Peng and Stewart (2016) selected three representative cities in China (i.e., Kunming, Xiamen, and Jinan) and considered three different emission scenarios to quantify the corrosion damage risk of infrastructures under climate change. They concluded that the structural carbonation depths increase rapidly due to climate change. Buildings located in temperate or cold climate areas in China may suffer an additional 7%–20% of carbonation-induced damage by the end of the 21st century due to the influence of climate change. de Larrard et al. (2014) proposed a method for quantifying the impact of climate change on the durability of concrete structures exposed to carbonation. The proposed method was applied to assess the probability of carbonation effects for several cities across France under various climate scenarios. Chen et al. (2021a) introduced a carbonation model of concrete structures that considers changes in climate variables (e.g., CO₂ concentration, temperature, and relative humidity) to investigate the extent to of climate change affects the concrete durability. The results indicate that by the end of this century, climate change could cause an additional 20%–160% concrete carbonation damage for the Chinese cities of Harbin, Qingdao, and Ningbo (representatives of severe cold, cold coastal, and temperate coastal climates).

Bastidas-Arteaga et al. (2010) investigated the influence of climate conditions on the chloride ingress into concrete structures in various chloride-contaminated environments (i.e., continental, marine, and tropical) utilizing a stochastic approach. Their study indicates that climate change has a more significant impact on structures in marine environments, leading to a 2%–18% reduction in corrosion initiation time. Subsequently, Bastidas-Arteaga et al. (2013) explored the impact of global warming on the durability of concrete structures subjected to chloride ion ingress or carbonation using time-dependent degradation models and simulations of the effects of global warming on environmental factors such as temperature, humidity, and CO₂ concentration. The conclusion is that global warming could reduce the time to failure of concrete structures by up to 31%. Xu and Yang (2023) developed a hierarchical two-tier framework to investigate how projected climate change may affect corrosion-induced damage for concrete bridges, considering the exposure and environmental conditions (i.e., temperature and relative humidity) at specific locations. Through case studies, the following conclusions can be drawn: due to climate change, the time to corrosion initiation, crack initiation, and severe cracking for Victoria and Toronto decreased by 10.9%–12.5%; under climate change impacts, the probability of corrosion initiation may increase by 0.5%–28.9%, and the likelihood of corrosion damage varies widely between different regions. In summary, although the effects of climate change can be both positive and negative for different regions or countries, it generally drives previous corrosion initiation and faster structural deterioration in most areas.

The acceleration of the deterioration process implies that the structural resistance decreases at a faster rate, further impacting the reliability of structures. Bastidas-Arteaga (2018) investigated the failure of concrete beams in the context of climate change, considering three different exposure environments and two potential future climate change scenarios. A 7% reduction in the service life of the bridge was observed due to the combined effect of vehicle fatigue loads and global warming. Mortagi and Ghosh (2020) proposed a framework for seismic vulnerability assessment of high-speed bridges considering global warming. Bridges in the central and southeastern regions of the United States were selected to study the impact of climate change on both components and the overall structure. The study concluded that the failure probability of concrete columns increased by 12.08% over a 100-year service life when only considering natural aging, while the failure probability would increase by 19.18% when further considering climate change. The impact of climate change on steel braces can be disregarded, as the difference between cases with and without considering climate change is minimal (only 3%). On the other hand, since the performance of columns is a crucial element in the seismic resistance of bridges, the heightened vulnerability of concrete columns suggests an overall increase in the seismic vulnerability of the entire bridge structural system. Mortagi and Ghosh (2022a) recently proposed an iterative approach that considers the coupled effect of carbonization and chloride ion ingress to calculate the corrosion initiation time. They performed a case study evaluating the seismic performance of offshore bridges and concluded that the corrosion initiation time obtained by considering both chloride ion ingress and carbonation is significantly earlier than that by considering chloride ion ingress only, leading to an increased seismic vulnerability of the bridge components.

Climate change impacts the intensity of wind, rain, and snow loads, as indicated in Table 3. Meanwhile, a large number of concrete infrastructures have been built in regions along rivers, coasts, or areas prone to flooding, rendering them susceptible to relevant hazards such as sea-level rise, tsunamis, and floods. Compared to historical scenarios, climate change-induced fluctuations in the intensity, frequency, and magnitude of these hazards introduce greater complexity into infrastructure loads in terms of not only the magnitude but also the spatial and temporal distributions. The characterization and description of hazards in the risk assessment of concrete infrastructure are primarily the responsibility of the domain experts rather than risk analysts. Nevertheless, the hazard models proposed by domain experts are usually complicated and computationally demanding. Despite their improved predictive accuracy, it is difficult to employ such models to analyze structural risk. The development of suitable models for climate-related hazards, applicable in the risk analysis remains to be a subject of further research.

A key challenge in hazard risk assessment is to establish connections between the impacts of climate change and hazard characteristics (e.g., the maximum annual flood flow). Based on a statistical analysis of the expected maximum annual flow of rivers, Kallias and Imam (2016) used the Monte Carlo simulation (MCS) method to estimate the failure probability of a bridge pier caused by local scour exceeding the foundation depth. In their study, climate change was represented by the gradual alteration of statistical characteristics within the expected maximum annual flow distribution. Alhamid et al. (2022b) developed a framework for probabilistic tsunami hazard assessment considering the effects and uncertainties of sea-level rise. Initially, probabilistic assessments of sea-level rise under four climate change emission scenarios (i.e., RCP 2.6, RCP 4.5, RCP 6.0, and RCP 8.5) were conducted based on available data. Subsequently, uncertainties associated with stratum movement were considered to simulate tsunami propagation scenarios during periods of sea-level rise. Finally, the agent model, relying on the radial basis function and the quasi-MCS, was used to obtain the tsunami hazard curves.

Structural vulnerability refers to the probabilistic description of structural damage under a specific hazard, such as flood, earthquake, hurricane, and so on. Vulnerability models are derived based on damage data sourced from historical records (i.e., statistical data or empirical models) or engineering models (i.e., experimental and numerical models). The empirical-based vulnerability assessments are conceptually straightforward and computationally efficient, serving as the fundamental approach to calculating the probability of structural damage. However, the empirical models demand an adequate supply of practical engineering data, and acquiring such data presents a substantial challenge in vulnerability assessment (Dawson et al. 2018). Furthermore, such empirical models are based on past hazards and will not be applicable to changing climate scenarios. The emergence of extreme climate events has promoted the development of interdisciplinary vulnerability methods (Tolo et al. 2017a). For example, Kallias and Imam (2016) developed a limit state function formula for bridges that takes climate variables into account. The failure probability of bridge piers due to local scour exceeding the foundation depth was estimated by the MCS method. The influence of climate change was evaluated quantitatively by analyzing changes in statistical data related to the expected annual maximum flow, including mean values and variance. Khandel and Soliman (2021) proposed a probabilistic vulnerability assessment framework for bridges under floods and flood-induced scour considering climate change. A neural network model was trained by historical temperature and precipitation, as well as corresponding river streamflow, to predict streamflow. Another neural network model was trained by structural finite-element results to predict structural responses (e.g., stresses and deformations). Zhu et al. (2021a, b) proposed a framework for reliability analysis of coastal bridges subjected to extreme hurricanes and waves. The artificial neural network was used to build a surrogate model for the 3D computational fluid dynamics simulations to predict the performance responses and vulnerability of bridges, considering the uncertainties in structural capacity and the effects of climate change. The results indicated that climate change could markedly elevate the failure probability of coastal bridges under hurricane waves, and bridges with longer service lives appeared to be more susceptible to the effects of climate change.

Peng and Stewart (2016) concluded that climate change intensifies the degradation of reinforced concrete structures, resulting in a significant increase in the vulnerability of concrete infrastructure. Mortagi and Ghosh (2022a) proposed a seismic vulnerability assessment framework for highway bridges that considers both aging effects and global warming. An investigation was conducted on a bridge in South Carolina, United States, with results indicating that the aging effect increased the bridge vulnerability (total loss stage) by 12.08%, while it increased by 19.18% when climate change is considered. Recently, the impact of climate change and corrosion degradation on the lifetime seismic brittleness of highway bridges has been further studied by Chirdeep et al. (2023), based on the IPCC sixth assessment report. A case study of a nonseismically designed highway bridge located close to the ocean environment shows that climate change-caused corrosion significantly increases the seismic fragility of aging bridges. Ha et al. (2017) proposed a deterioration model to reflect the structural time-dependent reliability that considers climate change under RCP 4.5, 6.0, and 8.5 emission scenarios using the gamma process. A deterioration model of concrete structures that integrates chloride corrosion, climate change, and cyclic loading was proposed by Bastidas-Arteaga (2018). The reliability of the main girders was estimated considering the uncertainties associated with climate variables. They demonstrated that the effects of climate change led to a reduction in service life ranging from 1.4% to 2.3% when fatigue damage was ignored. Xie et al. (2018) studied the impact of climate change on the durability of offshore concrete bridges. The results indicated that the rising temperatures accelerate the chloride ion penetration. According to the global warming scenario, the chloride ion concentration on the reinforcement surface was predicted to increase from 6% in 2000 to 15% in 2100, leading to a significant increase in the likelihood of erosion damage. Maniglio et al. (2021) developed a parameterized vulnerability model for port structures under hurricane-induced storm surges and waves, taking into account the potential influences of aging and degradation. Snaiki et al. (2020) proposed a hurricane risk assessment method for coastal bridges under different climate scenarios, which combines the hurricane tracking model, height-resolving analytical wind model, and machine learning-based surge model.

The damage and failure of concrete infrastructure result in significant losses, usually measured by economic costs, which is often the primary concern of decision makers. It should be noted that future risk predictions need to consider the future monetary value of the assets, and this also means that losses are time-dependent. The total loss associated with a civil infrastructure asset generally includes direct rebuilding costs C_Reb, indirect running costs C_Run, and time loss costs C_TL (Decò and Frangopol 2011), which can be expressed as follows:

The economic losses stemming directly or indirectly from future climate scenarios are projected to be substantial. For example, Jevrejeva et al. (2018) predict that in the absence of further adaptive measures to account for a 1.5°C warming and its associated sea-level rise by 2100, the worldwide annual cost of sea flooding could escalate to $10.2 trillion (1.8% of GDP). Streletskiy et al. (2019) investigated the impact of permafrost degradation on infrastructure and buildings in Russia, projecting their costs by the mid-21st century following the RCP 8.5 emission scenario. Mortagi and Ghosh (2022b) conducted a life-cycle seismic loss assessment for a continuous multispan concrete girder bridge. They found that the seismic losses of this bridge were underestimated by around 15% when the effects of aging and deterioration were neglected. Even when the effect of aging was taken into account, without considering the impact of climate change, the loss was still underestimated by approximately 13.2% under the RCP 4.5 emission scenario. On a broader scale, Alhamid et al. (2023) quantified the life-cycle economic risk associated with tsunami impact under nonstationary sea-level rise effects. Their study revealed that, with a 50% exceedance probability, the life-cycle risk could increase by 20%–30% in several cities within the Kochi prefecture. Bjarnadottir et al. (2014) investigated the impact of climate change on losses resulting from hurricanes. A framework for quantifying the potential damage risk of residential buildings exposed to hurricanes was proposed, including the hurricane wind field, hurricane-induced surge height, and hurricane vulnerability. Miami-Dade County, New Hanover County, and Galveston County in the United States underwent evaluation using the proposed framework, leading to the conclusion that climate change exerted a substantial influence on regional hurricane losses. Forcellini (2021) proposed a resilience assessment framework that considers loss and recovery models under the influence of climate change. González-Dueñas and Padgett (2022) extended the performance and recovery assessment framework for coastal communities under hurricanes, considering time-varying aspects of the hazard, depreciation, and aging or deterioration of coastal structures. Case studies have shown that the impact of climate change on hazard conditions (i.e., storm forward velocity and local sea levels) changes the probability of building collapse and recovery.

Estimating losses is complex because it requires consideration of the specific environmental and economic conditions in the region, as well as the service life of the infrastructure. For instance, increasing the thickness of concrete cover can effectively mitigate damage from chloride ion ingress. However, the cost effectiveness of this measure depends on the specific environmental exposure conditions. Bastidas-Arteaga and Stewart (2016) discovered that an adaptation strategy involving a 5 mm increase in cover thickness was advantageous for Saint-Nazaire but not for Marseille. The study by Dong and Frangopol (2016) revealed that the overall life-cycle losses were sensitive to factors such as the timing of the last earthquake, currency discount rate, and remaining service life.

Hazards may occur simultaneously, cascadingly, or cumulatively in time and space to become multihazard scenarios. This would be predictably more common in future climate scenarios, consequently increasing the potential risk of infrastructure (Li et al. 2012). Coastal regions, in particular, face a convergence of various climate-related hazards, such as tsunamis, hurricanes, sea-level rise, coastal erosion, saltwater corrosion, and so on. There are numerous multihazard events in accordance with historical records. These events can be categorized into three types: (1) specific hazards occurring in simultaneity but independent of each other (e.g., earthquakes with hurricanes or strong winds); (2) cascading hazards (e.g., earthquakes triggering tsunamis); and (3) hazards that do not occur simultaneously but threaten the same structure (e.g., floods and earthquakes) (Mahmoud and Cheng 2017). Although researchers have conducted numerous investigations concerning the concept and approach of multihazards (Bruneau et al. 2017; Gill and Malamud 2014; Zaghi et al. 2016), generally speaking, the influence of climate change has not been involved (Bruneau et al. 2017; Zaghi et al. 2016). In some literature, climate change was studied as a multihazard (Gill and Malamud 2014; Roy and Matsagar 2023).

To quantify the increasing structural vulnerability caused by multihazards, a comprehensive approach should be employed in the risk assessment of natural hazards, including those related to climate. Existing studies on multihazard risk primarily rely on static vulnerability analysis, which does not consider climate change (Gallina et al. 2016). Besides, the current vulnerability analysis is commonly conducted on models of intact structures, while, in reality, the structures that have been damaged by the primary hazard may further suffer secondary hazards. For instance, in the case of the 2011 Tohoku earthquake in Japan, multiple buildings were damaged in the mainshock‒aftershock sequences (Li et al. 2014). In most existing code provisions, the combined effects of hazards are simulated by different load combinations, but unfortunately, this frequently results in the neglect of intricate interactions between different hazards (Zaghi et al. 2016).

In recent years, there has been a notable surge in research attention toward risk assessment for climate-related multihazards. Gallina et al. (2016) conducted an insightful review of multihazard risk assessment considering the influence of climate change, covering both concepts and analysis techniques. In a subsequent study (Gallina et al. 2020), the authors proposed a multihazard risk methodology aimed at quantifying the impact of climate change on the structures exposed to various hazards on a regional scale. The method was semiquantitative on the basis of expert evaluation results. Forzieri et al. (2016) combined different damage measures induced by hazards into a multihazard indicator, enabling a quantitative assessment of structural losses due to multiple climate-related hazards. This method allows for comparability among different hazards. The findings indicated a substantial rise in climate-related hazards across Europe, particularly in coastal and floodplain regions in the southern and western parts of the continent. These areas are often densely populated and of significant economic importance. In addition, Steptoe et al. (2018) summarized the primary climate factors inducing hazards and explained the multihazard mechanism. In summary, many studies related to multiple hazards have been conducted, covering earthquakes accompanied by floods or scour (Argyroudis and Mitoulis 2021; Prasad and Banerjee 2013; Guo and Chen 2016), earthquakes accompanied by hurricanes or strong winds (Mahmoud and Cheng 2017; Roy and Matsagar 2020; Zheng et al. 2019), and earthquakes accompanied with tsunami (Akiyama et al. 2020; Carey et al. 2019; Xu et al. 2021). Nevertheless, the multihazard risk assessment of infrastructure accounting for the impacts of climate change is still scarce, which is listed in Table 4. The following sections will present the current progress in research on climate multihazard risk based on the available studies.

Scour damage to piers or abutments inevitably occurs in the life cycle of bridges situated along rivers. Bridge piers are critical components for resisting lateral forces, so any scouring of their foundation tends to reduce the seismic resistance of the bridge system. Because climate change intensifies the impact of scouring on bridges, it is crucial to consider climate change when assessing the vulnerability of bridges.

Dong and Frangopol (2016) proposed a life-cycle assessment framework for the risk and resilience of bridges under earthquakes and floods. With this framework, the impact of climate change on flood return period and intensity was considered, but the combined effect of floods and earthquakes was neglected. Damage probabilities for both hazards were individually calculated based on separate failure models and then aggregated. Yilmaz et al. (2016) performed a more elaborate study on the vulnerability of bridges, considering the coupling effect of floods and earthquakes. Similarly, Devendiran et al. (2021) analyzed the effect of climate change on the flood and earthquake vulnerability, risk, and resilience of an existing bridge. The annual maximum flow of the San Joaquin River in California was predicted until 2100 using the Coupled Model Intercomparison Project Phase 5, and the results indicated an upward trend. Moreover, the seismic vulnerability of the selected bridge spanning over the San Joaquin River was evaluated under three different flood conditions (levels in 1930–2011, in 2012–2050, and in 2051–2099). The results showed that climate change aggravated the scour damage and affected the seismic performance of the structure. Climate change had an insignificant effect on minor and moderate damages, but it heightened the seismic vulnerability of structures in cases of severe damage or collapse. The failure probability of structures for the same earthquake magnitude increases by 20%–30% under flood level of 2051–2099 compared to those from 1930 to 2011. Recently, Ge et al. (2023) integrated corrosion-induced deterioration, climate change, earthquake, and flood scour into the time-dependent seismic vulnerability of bridges. By analyzing an existing multispan concrete beam bridge, it was found that climate change and flood scour significantly increased the seismic vulnerability of deteriorating bridges.

Although the hazard scenarios that combine scour and earthquake have been studied to various extents, and the impact of scour on seismic vulnerability and risk of bridges has been evaluated (Banerjee et al. 2019), relatively few studies have considered climate change. Moreover, most previous studies are focused on the vulnerability and risk assessment of individual bridges, and there is a scarcity of research addressing the combined impacts of scour and earthquakes at the bridge network level.

Natural hazards such as tsunamis and hurricanes threaten the safety of offshore infrastructure throughout its entire life cycle, and these hazards naturally tend to become multiple hazards. For example, hurricanes are often associated with large waves and flooding. Tsunamis are typically triggered by undersea earthquakes. As the sea levels rise due to climate change, the multihazard coupling problem will become more complicated.

Mousavi et al. (2011) evaluated the impact of hurricane intensity and sea-level rise due to global warming on coastal flooding based on hydrodynamic surge models and IPCC climate scenarios. Yavuz et al. (2020) quantified the risk of tsunamis triggered by earthquakes in the eastern Mediterranean in the context of climate change. They concluded that as the sea level rose, the level of tsunami inundation in Fethiye city center and the Cairo agricultural area increased at the same earthquake magnitude. Tursina et al. (2021) examined the tsunami risk in the Banda Aceh area and discovered that a rising sea level resulted in tsunami inundation induced by earthquakes of 8.2–8.6 magnitude, being twice as extensive compared to scenarios without climate change. Alhamid et al. (2022a) developed a time-dependent tsunami hazard assessment method under the long-term progressive trend of sea level due to climate change. The impact of nonstationary sea-level rise on the tsunami hazard is predominantly determined by the geographic location of the study area; that is, the tsunami hazard tends to be lower with the increase of land elevation.

While numerous studies have evaluated the vulnerability and risk of offshore infrastructure under multihazard coupling, the coupling effects considered are usually hurricane and wave interactions or sea-level rise and tsunami interactions. To the authors’ knowledge, there is currently no framework for evaluating the combined impacts of hurricanes, tsunamis, and sea-level rise caused by climate change. This presents a significant challenge for future research.

Infrastructure risks caused by climate change can be mitigated via adaptation measures during the design and service phases, as listed in Table 5. Similarly, after reviewing the advances in infrastructure planning, design, and decision-making implementation to adapt to climate change, Buhl and Markolf (2023) emphasized that climate adaptation methods for infrastructure need to be integrated throughout the entire phases. However, climate adaptation is a complex decision-making process that needs consideration of climate variables and associated uncertainties, including but not limited to climate-related hazard analysis, structural degradation modeling, and infrastructure risk assessment.

To establish effective adaptation strategies for mitigating the effects of climate change, it has become a trend in recent years to incorporate emerging technologies such as artificial intelligence, digital twins, point clouds, and building information modeling into risk and resilience assessment and management of critical infrastructure (Argyroudis et al. 2022). These technologies help to assess the asset condition and resilience of critical infrastructure rapidly and accurately, thereby supporting decision-making and improving climate adaptation capabilities. Another effort is for government decision makers to adopt proactive infrastructure management strategies rather than emergency response after extreme hazards occur. The proactive management approach implies the prevention of potential climate change impacts on infrastructure in advance (Miao et al. 2018). Moreover, climate adaptation investments in infrastructure are huge and irreversible (Eisenack et al. 2012). Therefore, Mondoro et al. (2018) proposed a gain–loss ratio (GLR) for systematically quantifying the possible gains and losses concerning the delayed adaptation to achieve strategic flexibility. A biobjective robust model was proposed to optimize both the efficiency (i.e., benefit–cost ratio) and flexibility (i.e., GLR) for climate adaptation. The model was applied to two typical bridges spanning the Columbia and Mississippi Rivers.

To consider climate adaptation during the design phase, it is desirable to investigate the future nonstationary trends in hazard loads (e.g., sea-level rise, precipitation, hurricanes, and tsunamis) and the structural degradation mechanism. For instance, Fairburn in Wales faces a prolonged risk of flooding caused by sea-level rise, so the Shoreline Management Plan for the area has recommended coastal adaptation planning, and eventually decommissioning Fairbourne (Buser 2020). After extreme floods and hurricanes, the Federal Emergency Management Agency (FEMA) revised the flood maps and proposed new provisions for structures affected by coastal storms (AASHTO 2008; FEMA 2015). The return levels and return periods in traditional engineering designs are not suitable for the assumption of climate change. To address this issue, Rootzen and Katz (2013) defined the concept of Design Life Level (e.g., 5% of the maximum distribution of flood level over the design life period) to quantify the risks in nonstationary climates.

Another aspect is the structural capacity under design. The current degradation models for concrete infrastructure still rely extensively on historical climate data, while climate change will accelerate the corrosion of steel bars and may affect the service life. For example, the influence of climate change on the bridge deck without waterproof membrane rebar is significantly larger than on the bridge deck with waterproof membrane rebar (Guest et al. 2020a). Stewart et al. (2012) compared three adaptive strategies for corrosion control of steel rebar under climate change, namely the increase of cover thickness, the improvement of concrete strength grade, and the application of surface coatings. Therefore, greater design loads (e.g., flood level, wind load, and snow load), anticorrosion materials [e.g., stainless steel bars (Williamson et al. 2009) and glass fiber‒reinforced polymer composite bars (Manalo et al. 2021)], and urban planning adjustments can be adopted in the structure for a better climate adaptation scheme (Bastidas-Arteaga and Stewart 2019).

With the acceleration of climate change and the increasing frequency of severe climatic events, investments in maintenance for service infrastructure may not achieve the expected benefits under the changing climate variables. Climate change should be included in monitoring and maintenance, and it is desirable to use stricter design parameters to deal with various extreme weather conditions (TRB 2008). Choi et al. (2020) developed a framework for improving the adaptive capacity of existing dams under climate change, which can be used to review and rehabilitate the aging dams.

Adaptation planning should balance infrastructure investment and the planning cycle, especially for the tight funding situation. As concluded by Becker et al. (2012), most of the ports surveyed can withstand a storm with a 100-year return period without being damaged. However, if the rate of climate change accelerates, leading to changes in return periods from 100 to 30 years, such adaptations would prove inadequate (Wang et al. 2020). For irreversible investments, Koetse and Rietveld (2012) recommended delaying the decisions on infrastructure maintenance until critical parameters concerning climate change had relative deterministic characteristics since maintenance design is a long-term process and incorrect decisions can be costly. Climate adaptation can help mitigate the additional costs for future investments and maintenance, thereby lowering the overall investment and damage costs. Nasr et al. (2023) proposed a risk-based multicriteria decision analysis method, systematically addressing the prioritization of climate change impacts and bridges more severely affected by climate change.