Integrated assessment of flood risk in Arial Khan floodplain of Bangladesh under changing climate and socioeconomic conditions

In the assessment of flood risk, the future flood hazard due to climate change is often tied to the present socioeconomic conditions. This makes an implicit assumption that the drivers of risk, other than the hazard, remain constant with time. Therefore, such risk assessment does not provide a realistic outlook for devising plausible mitigation strategies and plans. In this study, flood risk was assessed from an integrated perspective by considering both physical hazard, and socioeconomic exposure and vulnerability—all changing with time. The flood hazard in the Arial Khan River floodplain in the southcentral Bangladesh was simulated with a two‐dimensional hydrodynamic model, and the exposure and vulnerability were projected using different statistical techniques. Principal component analysis was conducted to assign weights to the indicators of hazard, exposure, sensitivity, and adaptive capacity. The results show that the flood depth, duration, and extent would increase from the baseline to 2080s under regional concentration pathway (RCP) 2.6 and RCP 8.5 scenarios. The sensitivity and vulnerability would decrease, reflecting an improved adaptive capacity. The low‐risk areas could increase from 62% in the baseline to 85%–91% in 2080s depending on the RCPs. The approach followed can be applied elsewhere in developing countries, particularly in riverine floodplain settings.

It is anticipated that the greatest chance would be in increasing rainfall during the monsoon season due to CC (Karim et al., 2020). This increase in rainfall is more likely to cause frequent floods with a high probability of severity (Tabari, 2020). River flooding, triggered by heavy rain and excess runoff from upstream, is identified as one of the leading causes of economic losses and human deaths (Bronstert, 2003;Dottori et al., 2016;Malik & Ahmad, 2014). It is anticipated that the changes in timing of rainfall events would modify flooding experiences across the world, creating a great uncertainty in future flood risk (Ashley et al., 2005;Wheater & Evans, 2009). Also, flooding on a regular basis would exacerbate the existing drainage problems. This worsening would also lead to a reduction in river conveyance capacity, which in turn would lengthen the flood duration.
Bangladesh, a South Asian tropical monsoon country, is highly vulnerable to the impacts of CC-induced hazards like flood owing to its unique geographic location, high population density, widespread poverty, and overwhelming dependency on natural resources (Mondal et al., 2018;Zaman & Mondal, 2020). According to the Germanwatch's Global Climate Risk Index, Bangladesh is one of the 10 most-affected countries by climate-related risks in recent decades (Eckstein et al., 2021). As the country is situated downstream of the Ganges-Brahmaputra-Meghna (GBM) basins, it is highly susceptible to flood risk due to the basins' hydro-meteorological and topographical characteristics (Shaw et al., 2013). The country is crisscrossed by more than 400 rivers that convey water from the Himalayas to the Bay of Bengal (Basak et al., 2015;Rouf, 2015). Normal floods inundate 20%-25% of the country, and severe floods inundate up to 67% (Kundzewicz et al., 2014). Due to CC, temperature, and precipitation would increase, which would have a higher impact on flood risk, changing the GBM rivers' hydrological cycle (Masood et al., 2015;Mirza et al., 2003;Mohammed et al., 2018;Whitehead et al., 2015).
Many studies on possible flood risk due to hydroclimatic hazards were conducted both in Bangladesh and elsewhere around the world. Some of these considered the effects of CC, whereas others did not. In many studies, the role of socioeconomic factors in risk portfolio was not also considered. For example, the potential effects of CC were not considered in the assessment of storm surge inundation risk in the Ganges tidal floodplain in Bangladesh (Chowdhury & Karim, 1996), creation of hazard maps due to the flood in Bangladesh (Islam & Sado, 2000), investigation of socioeconomic vulnerability and adaptation aspects of flooding in Bangladesh (Brouwer et al., 2007), analysis of flood risk in the Hoang Long River basin in Vietnam (Tu & Tingsanchali, 2010), mapping flood hazard in the Swannanoa River watershed in North Carolina, USA (Ahmadisharaf et al., 2017), assessment of flood hazard and risk for the mid-eastern part of Dhaka, Bangladesh (Masood & Takeuchi, 2012), and assessment of flood damage and risk of the boro rice due to the pre-monsoon flash flood in the haor basin of Bangladesh (Hossain, 2013). Again, socioeconomic vulnerability was not considered in Islam and Sado (2000), Hossain (2013), De Silva et al. (2016) for analysis of flood inundation in the Lower Kelani River basin in Sri Lanka, Ahmadisharaf et al. (2017), Shrestha and Lohpaisankrit (2017) for assessment of flood hazard in the Yang River basin of Thailand, and Gusain et al. (2020) for evaluation of CC impacts on flood hazards in the Mahanadi River basin of India. Although socioeconomic change was considered in Hall et al. (2006) for coastal flood risk analysis in England and Wales, socioeconomic vulnerability was not considered. The identification of the elements at risk, and hence the evaluation of socioeconomic exposures to a hazard, are important in risk analysis. But the exposures were not considered in Chowdhury and Karim (1996), Brouwer et al. (2007), and Tu and Tingsanchali (2010). Moreover, Tingsanchali and Karim (2005) reported the flood hazard and risk analysis for the southwest region, including the Arial Khan floodplain, in Bangladesh, but did not consider either the CC effects or the changes in socioeconomic conditions. Rahman (2019) studied the flood hazard and vulnerability in the Old Brahmaputra River for present and future CC scenarios. The future flow was generated for the RCP 8.5 scenario using the HEC-HMS model for the Brahmaputra River basin which was then used in a one-dimensional-twodimensional (1D-2D) coupled HEC-RAS model for hazard analysis. The flood risk was calculated for baseline and future by multiplying the flood hazard with the exposure and vulnerability of the particular area. Another recent study (Roy, 2019) is available on flood hazard and risk for future CC scenarios, however, the changes in socioeconomic conditions were not considered.
Flood repercussions may become more severe in the future because of the ever-growing socioeconomic changes, posing a threat to the populations living in the floodplains (Nur & Shrestha, 2017;Toda et al., 2017). The government of Bangladesh has undertaken a long-term flood management strategy to deal with the impending flood situation and mitigate the future flood losses and damages (General Economics Division, 2020). Socioeconomic vulnerability is at the very heart of the management strategy strongly influencing the mitigation measures (Dewan, 2015;Rufat et al., 2015). The adverse effects of disasters are mainly determined by the vulnerability and exposure of societies and social-ecological systems (Cardona et al., 2012). People and other assets must be exposed to hazards for these events to become disasters, otherwise, the risk will be nil (Vatsa, 2004). Nevertheless, while assessing the future flood risks, most of the above studies focused on future climate projections with current socioeconomic conditions (Ebi et al., 2016;Preston et al., 2011;Rohat et al., 2018) by making an implicit assumption that the drivers of risk, other than CC, would remain constant (Jurgilevich et al., 2017). Thus, though the dynamism of vulnerability was long recognized, future socioeconomic conditions were seldom accounted for. In this study, flood risk was assessed using an integrated approach, which focuses on both the physical and social aspects of the flood event as they are equally important (Adger, 2006;De Sherbinin, 2014;Gain et al., 2015;Qi et al., 2022;Rahman, 2006). The physical aspects include the flood attributes such as affected area, depth and duration, and the social aspects include the features of human life and property that face the harmful consequences due to the occurrence of the flood (de Moel et al., 2015;Huda et al., 2022). Thus, along with a hardengineered approach to flood problem, the soft floodproofing and nonstructural measures to flood risk as well as socioeconomic vulnerability reduction and capacity enhancement are highly emphasized in the approach (Gain et al., 2017;Rahman, 2006). The IPCC AR6 conceptualization of risk (IPCC, 2021), elaborated later, is in this line and hence this study followed it. The risk assessment using the AR6 concept as well as considering the future CC and socioeconomic circumstances is important for flood management at the international, national, regional, and local levels.

| Study area
This study assesses the flood hazard, vulnerability, and risk in the Arial Khan River floodplain in southcentral Bangladesh. The study was conducted at different unions (the bottom tier of local administration) of the Madaripur Sadar upazila (middle tier) of the Madaripur district (top tier). The upazila is considered flood vulnerable to existing and future CC scenarios as the Arial Khan River traverses through the upazila from the northwestern to the southeastern direction. Also, the upazila is an agricultural and economic hub of the southcentral zone of Bangladesh. Besides, the floodplain of the Arial Khan River has become overly populated with continuing socioeconomic development and agricultural activities. The area is frequently flooded during the monsoon (June-October) by the riverine flood, which is the main cause of the suffering of the local people. As the area is in close proximity to two mighty rivers, the Padma in the north and the Meghna in the east, it faces devastating floods when the peaks of these rivers synchronize. A rise in sea level due to CC could directly affect the area through the Meghna River, further worsening the flooding condition. Hence, the flood modeling and risk assessment for the Arial Khan River basin is important for better management of the floodplain. The Madaripur Sadar upazila comprises an area of 314 km 2 , and has 15 unions and one municipality as shown in Figure 1.
The study area is part of the floodplain of the Arial Khan River, which is one of the important rivers in the country's southcentral region. The river branches off the Padma River at 51.5 km downstream of the Ganges-Brahmaputra confluence. It flows through the Faridpur and Madaripur districts before discharging into the Tentulia Channel in the south. The hydrology of the study area is linked to the hydrology of the Arial Khan River. The floodplain is frequently flooded as there is no flood control embankment along the riverbank. This causes severe hardship for the people living in the floodplain (Tingsanchali & Karim, 2005). The morphological processes and instabilities of the Arial Khan River have been the subject of a few studies (Akter et al., 2013;Ferdoush et al., 2022;Mamun, 2008;Winkley et al., 1994). The flood danger level of the river at the Madaripur gage station (SW5) is 4.20 m above the public works datum (PWD), which is crossed frequently during the monsoon. During the historical flood of 1998, the water level was recorded to be 5.80 m PWD, and the area remained flooded for about a month (FFWC, 2019).

| Assessment of flood hazard
Flood hazard assessment involved the estimation of adverse effects of flooding in the study area. For that, the most important parameters of flood, such as flood depth, flood duration, and inundation extent, were necessary. Such parameters were simulated by using a coupled 1Dand 2D hydrodynamic model (HEC-RAS) for both baseline and future CC scenarios. Observed discharge in an upstream gage of the river and observed water level in a downstream gage was used to calibrate the model for 1 year and validate for another year. Manning's roughness coefficient was the main calibration parameter. Floodplain inundation was verified with the Moderate Resolution Imaging Spectroradiometer (MODIS) images. The details on the modeling and flow generation for the RCP 2.6 and RCP 8.5 CC scenarios are reported in Roy (2019). By importing the output raster files from the RAS-Mapper, the mean values of these parameters for different unions were obtained. The imported files consisted of polygon-type vector features on flood depth, flood duration and inundated area for both twodimensional floodplains and one-dimensional river channels.
The parameters were then subjected to a pretreatment to eliminate the differences in units and dimensions (Wang et al., 2015). There are four types of data pretreatment: mean centering, differentiation, normalization, and auto-scaling (Amrhein et al., 1996). The pretreatment was done by normalization of the data to a common, comparable unit-less scale (1-100). Normalization of individual variables provides a linear transformation preserving the rank and correlation structure of the original data and allows for the variables with differences to be used together (Tran et al., 2010). The following normalization equation (Rahman, 2019) was used: The indicators need to be assigned different weights to avoid the uncertainty of equal weighting given the diversity of the indicators used. There are different approaches to assign weights. Ibrahim et al. (2017) used a weight of 0.7 for flood depth and 0.3 for inundation area while studying the flood vulnerability in the Kelantan River basin in Malaysia. Al Amin et al. (2017) provided a greater weightage to flood depth while simulating the flood risk from the Musi River in Indonesia. We assigned the weights to the hazard indicators using the principal component analysis (PCA) (Dormann et al., 2013;Nandi et al., 2016).
The technique of PCA has already been extensively applied in flood hazard assessment (Akukwe & Ogbodo, 2015;Carreau & Guinot, 2021;Sung & Liaw, 2021) and in socioeconomic vulnerability assessments at regional, national, and global levels (Abson et al., 2012;Sarmah et al., 2020). The objective of PCA is to explain potential relations between a set of F I G U R E 1 Location and union boundaries of the study area in the Arial Khan River floodplain in southcentral Bangladesh independent variables, such as socioeconomic indicators, and a latent dependent variable, which is in this case the vulnerability level of each union. The indicators were tested for potential correlations between the independent variables, known as factor loadings, which are equivalent to the standardized regression coefficients (β weights) in multiple regressions (Beaumont, 2012). The higher values of the factor loadings (correlation) mean a closer relationship with the principal components. The components presenting eigenvalue higher than one were used for explaining the independent indicators (Abson et al., 2012;Everitt & Hothorn, 2011). Weights were chosen to maximize the explained proportion of the variance in the original set of indicators. The weights of the indicators of flood hazard were found to be 0.41 for depth, 0.37 for duration and 0.22 for extent from PCA.
Hazard score (H) for each union was obtained by multiplying the normalized values (H i ) with their respective fractional weights (W i ) as below: The hazard scores were distributed into 5 categories: 0-20 for very low, 21-40 for low, 41-60 for medium, 61-80 for high, and above 80 for very high hazards. From these, the percentage areas under different hazard categories were calculated.

| Assessment of exposure
Hydro-climatic risk due to CC depends on the exposure of the elements at risk. The higher the exposure, the higher the risk, and vice versa. The exposure of the study area for present and future socioeconomic conditions was assessed at the union level. The number of people, number of households, and total cropped area were selected as the elements at risk for assessing the exposure. Past data for the years 1981, 1991, 2001, and 2011 were gathered from various census and survey reports as well as from local government offices, and future data were generated based on projections. Population can be projected using various methods (Gawatre et al., 2016). In this study, the logistic curve method was used because of its wide acceptability (Akhter et al., 2017;Zabadi et al., 2017). According to this method, there would be a positive population growth if there are enough environmental resources to support the increased population. The growth would slow down with the decrease in resources. Logistic model illustrates how a population may increase until it reaches the carrying capacity of its environment (Edwards & Edwards, 2011). The method is useful in the case of limited space and economic opportunity under the assumption that population growth occurs under normal situations and is not affected by extraordinary changes, such as epidemic, war or natural disaster. If P 0 , P 1, and P 2 are the population of an area at time t 0 , t 1, and t 2 , respectively, the saturation population P S ð Þ, equivalent to the carrying capacity of the area, is calculated as: and the projected population P t ð Þ at time t as: where, The population for each union was projected separately for different decades and then averaged for 2011-2040, 2041-2070, and 2071-2100 to obtain the values for 2020s, 2050s, and 2080s, respectively. The other two exposure indicators were projected using trend analysis. The indicators were then normalized as described earlier, and the weights were assigned to the normalized indicators by applying PCA. The weights were found to be 0.39 for the population, 0.27 for the number of households, and 0.34 for the total cropped area. Finally, unionwise exposure (E) was calculated by a weighted sum technique: where, E i is the normalized score of an indicator and W i is the corresponding weight.

| Vulnerability concepts and factors
Vulnerability has various connotations in different disciplines. It conveys a negative connotation and indicates the susceptibility to be harmed (Adger, 2006). The susceptibility can be due to the internal characteristics of a system, or both internal characteristics and external shocks and stresses. The internal characteristics include sensitivity and adaptive capacity, and the external drivers are the hazard phenomena. While the third and fourth assessment reports of IPCC as well as Adger (2006), UNEP (2007), Varis et al. (2012) and Khan et al. (2019) advocate for the latter, the fifth and sixth assessment reports as well as  and Abson et al. (2012) suggest the former. This study conceptualizes vulnerability as the pre-event, inherent characteristics or qualities of a system and is akin to contextual or starting point vulnerability (Kelly & Adger, 2000;O'Brien et al., 2007). It is conceptualized differently from the damage, impact, or vulnerability function used in hydrology, water resources development, and sometimes, in hazard and disaster literature (Mondal et al., 2010;Mondal & Wasimi, 2007;United Nations, 1991;Zaman & Mondal, 2020). The vulnerability of an area is influenced by several aspects including human condition, infrastructure, land use, social imbalances, and economic pattern (Nasiri & Shahmohammadi-Kalalagh, 2013). Vulnerability can be represented by objective material measures, such as mortality, income, wealth, or access to education, which depend on the nature of the vulnerability being represented (Adger, 2006). The vulnerability of a place combines three distinct issues, namely elements of potential exposure or risk, societal coping response, and more inherently the geographical location (Cutter, 1996). Vulnerability indicators also vary for different calamities, for example, housing materials, building standards, building safety, income level, type of tenure, and location of dwelling relative to seismic zone can be the indicators of vulnerability for earthquakes, whereas crop profiles, physical access to markets, and market behavior can be the indicators for drought (Wisner et al., 2004). Consideration of different aspects of vulnerability ensures the right selection of its indicators. Women, children, elderly people, and the people with partial ability are more vulnerable in times of natural hazards (Cutter et al., 2003;Flanagan et al., 2011). Education and human resources are important for developing a community's adaptive capacity to face the adverse impacts of disasters (Roy & Blaschke, 2015). Significant interdependence between education and food security makes the illiterate group less secure due to lack of technological knowledge in farming and livelihood adaptations, whereas educated farmers are more adaptive to CC and embrace numerous employment options (Aryal et al., 2020).
The ability of groups or communities to cope with external stresses minimizes their vulnerability to disasters (Adger, 2006). The condition of housing, presence of flood shelters, length of unpaved road, and the existence of electricity supply are directly related to vulnerability. Katcha and jhupri types of household structures provide minimum protection during natural hazards like flood and render increased vulnerability (Laila, 2013). Flood shelters provide accommodation during a disaster and are used as the school buildings during the normal time. Thus, they contribute to increasing the adaptive capacity of the locality, not only by providing accommodation but also by supporting education.
The economic condition of a community is an important determinant of how quickly it can adapt to the effects of natural hazards (Roy & Blaschke, 2015). Higher cropping intensity reduces vulnerability by giving alternate cropping availability to the people after any natural disaster. Communities with diverse economic activities are more able to adjust to the effects of natural disasters because of their potentiality of switching to alternate income-generating activities in disastrous situations (Roy & Blaschke, 2015). The livelihood status is represented by incorporating people engaged in household works, agriculture, and industries and services. Health is an important factor contributing to socioeconomic vulnerability and is directly related to the percentage of households using tap water and sanitation facilities, and the number of hospitals and clinics providing healthcare facilities to a locality.

| Vulnerability indicators, projection, and assessment
In this study, vulnerability was considered as a function of sensitivity and adaptive capacity (IPCC, 2021). The selection of the indicators was guided by their suitability to represent the vulnerability due to the riverine flood. The availability of data in the previous censuses and statistical yearbooks also guided the selection. Eight socioeconomic domains were considered, namely population, gender, health, education, housing and infrastructure, land use, economic, and livelihood, where each domain included select indicators based on McLaughlin et al. (2002), Roy and Blaschke (2015) and Rahman (2019). A total of 10 sensitivity indicators and 8 adaptive capacity indicators were selected. The categorization of the indicators into sensitivity and adaptive capacity was done following Cutter and Finch (2008), Khan et al. (2019), and Roy and Blaschke (2015). Table 1 shows the sensitivity and adaptive capacity indicators used, and their weights calculated with the PCA technique.
This study required the projection of the socioeconomic indicators of vulnerability for the future to address the influence of socioeconomic conditions on flood risk. To project the value of each indicator related to population, firstly relevant data of the indicator for the years 1981, 1991, 2001, and 2011 were collected for each union. The ratio of the value of an indicator to the total population for each year was obtained and the ratio was then projected for the future with trend analysis. Using the predicted ratio and total population, the future value of the indicator for each union was obtained. For other indicators, not related to population, union-wise data for the years 1981, 1991, 2001, and 2011 were collected and the best-fitted lines were drawn.
To assess the sensitivity and adaptive capacity and thus the vulnerability of the 15 unions, the values of each indicator were normalized to a common scale of 1-100 to make them compatible. Then the weights were assigned to the normalized indicators by applying PCA separately to the sensitivity and adaptive capacity indicators. Finally, union-wise sensitivity (S), adaptive capacity (AC) and vulnerability (V ) (IPCC, 2021; Sharma & Ravindranath, 2019) were estimated by: where, S i and AC i are the normalized scores of a sensitivity and an adaptive capacity indicator, respectively, and W i is the corresponding weight.

| Assessment of risk
The flood risk for the study area was assessed following the IPCC concept of climate risk given in its 6th assessment report (IPCC, 2021), where flood risk (R) is comprised of hazard (H), exposure (E) and vulnerability (V ), and is calculated as: Although there are other approaches for risk assessment, such as the probability of failure in hydrology, and the expected loss and damage in natural disaster (United Nations, 1991;USAID, 2011), the IPCC approach integrates physical hazard with socioeconomic exposure and vulnerability. However, the approach is semi-quantitative in nature and does not explicitly incorporate the probability of hazard. Although both multiplicative and additive formulations are used in vulnerability assessment, the use of multiplicative formulation is almost universal in risk assessment (Rahman, 2006;Smith, 2004;United Nations, 1991). This multiplicative approach of integrated flood risk assessment was also adopted in many studies (e.g., Allen et al., 2016;Rakib et al., 2017) as it aggregates items of different scales and units. It was also suggested to use the multiplicative equation for the flood risk assessment due to CC, as it allows lesser compensation for low value indicators, whereas the additive approach provides a constant trade-off among the components (Choi, 2019). Finally, hazard, exposure, vulnerability, and risk were classified into five categories: 0-20, 21-40, 41-60, 61-80, and 81-100 for very low, low, medium, high and very high, respectively. Based on this categorization, union-wise hazard, exposure, vulnerability, and risk maps were prepared.

| Flood hazard
The flood hazard results from the combined effects of flood depth, duration, and extent. The flood depths for the baseline, 2020s, 2050s, and 2080s are given in Table 2. The flood depth is found to be reflective of not only the land elevation, but also the proximity of a union to the river. The maximum flood depth is found in the Paurashava, which might experience a depth of up to 4.57 m in 2080s. The minimum flood depth is found in the Kendua union, where the depth might be around 1.40 m in 2080s. Also, there is an increasing trend in flood depth from baseline to 2080s. The increase is more prominent for the RCP 8.5 scenario compared to the RCP 2.6 scenario due to the higher flow expected under RCP 8.5. The duration of the flood is found to be higher for the coming decades under both scenarios. Unlike the flood depth, the duration does not depend on the proximity to the river. This could be due to the difficulty in draining out the floodwater from a place farther from the river. Table 3 shows how the duration of the flood varies across the unions for different years. At the highest, Dhurail might experience a flood duration of about 32 days in 2080s for the RCP 8.5 scenario. The duration of flood is found not to vary much over time as flood duration is a function of the topography of the area and surface dynamics.
The flood extent exhibits the same pattern as the flood depth. A comparison of variation of inundation area over the years in each union is presented in Table 4. The Pachkhola and Kalikapur unions are found to be inundated more than the other areas. From baseline to 2080s, the inundation extent varies from 14.71 to 20.43 km 2 in the Pachkhola union and from 10.34 to 15.93 km 2 in the Kalikapur union. In Mustafapur, the inundation extent is the lowest as it is farthest from the river. Table 5 shows the areas under different hazard categories. The flood hazard increases from baseline to 2080s and the area with high-intensity hazard also increases. In the RCP 2.6 scenario, 13.5% of the total area might be under high-hazard class and the rest might be under very low-to medium-hazard classes in 2020s. There might be no administrative unit under very high-hazard class in 2020s, whereas in 2080s, the high-hazard zone could comprise 47.3% of the total area. In the RCP 8.5 scenario, 20.9% of the total area might be under very high-hazard class in 2080s. Also, the rate of increase in hazard is more prominent in the RCP 8.5 scenario compared to the RCP 2.6 scenario. Most of the unions might degrade to very high-or high-hazard class in 2050s and 2080s for the RCP 8.5 scenario.

| Flood exposure
The past and projected population and number of households in different unions are shown in Figures 2 and 3, T A B L E 2 Mean flood depths for the baseline, 2020s, 2050s and 2080s under regional concentration pathway (RCP) 2.6 and RCP 8. respectively. The figures indicate that the population and number of households would increase in the study area in the future. But the total cropped area would decrease ( Figure 4). Figure 5 shows the progression of exposure from baseline to 2080s. The Rasti union has a very low exposure and the Pachkhola, Khoajpur, Jhaudi, and Kendua unions have a very high exposure to flood in the baseline. The Pachkhola, Khoajpur, and Kendua unions maintain a very high exposure for the future as well. The Pachkhola and Khaojpur unions are in close proximity to the river which makes these two unions more exposed to floods. The Jhaudi union could shift to the high-exposure class from the very high-exposure class in 2020s and remain there for the rest of the century as the total cropped area is projected to decrease at a rate of about 20 ha per year. In all the periods, more than 70% of the study area lies under the high and very high exposures to flooding compared to the low and medium exposures.

| Flood vulnerability
The socioeconomic conditions of the study area would change with time. The projection of the socioeconomic indicators indicates that the disabled population, unemployed people, people engaged in agricultural and household activities, unmetalled road, number of families with katcha and jhupri housing, floating population, dependency ratio, and poverty rate would decrease in the future. The female-to-male ratio could either increase or decrease depending on the union. These indicate a remarkable decrease in the sensitivity of the study area to flood hazard in the future. As a result, parts of the study area could shift from a very high-sensitive zone to a highor medium-sensitive zone (Figure 6). Along with this, all the indicators of adaptive capacity show improvement over time, which is a reflection of the ongoing socioeconomic development in the country. As the sensitivity of the areas to flood would gradually decrease and the adaptive capacity would increase, the socioeconomic vulnerability would decrease with time.
The Pearpur, Jhaudi, and Khoajpur unions show very high vulnerability in the baseline. However, a sharp drop in the vulnerability to a very low or low category is expected for these unions in the future due to the projected decreases in sensitivity (e.g., poverty rate would decrease to 6% from 66% in the baseline, and unemployment would be almost nil from 30%), and increases in adaptive capacity (e.g., literacy rate would be almost 100% from 30%), which would play a positive role in decreasing the vulnerability in these unions. Around 20% of the study area was under very high vulnerability, which would drop to almost nil in 2050s. A significant part of the study area would shift from medium to low or very low, and from high to medium vulnerability. A significant increase could occur in the very lowvulnerability category from baseline to 2080s. About 26% of the total area was under the very low-vulnerability category in the baseline which could rise to 85% in 2080s depicting a gradual decrease in the vulnerability in the future decades.

| Flood risk
The risk maps for the baseline, 2020s, 2050s, and 2080s are shown in Figures 7 and 8 for the RCP 2.6 and RCP 8.5 scenarios, respectively. Six unions and the Paurashava area are found to be under very low-risk zone, and two unions (Kalikapur and Khoajpur) under very high-risk zone in the baseline. Under the RCP 2.6 scenario F I G U R E 4 Past and projected total cropped areas in the study floodplain (Figure 7), by 2080s, the risk of Kalikapur would reduce significantly and it could fall under a very low-risk category. For Khoajpur, the risk would remain the same for 2020s and 2050s, but in 2080s it could be under the medium-risk category. The number of unions under very low-risk zone could increase with time from 7 unions in 2020s to 12 in 2050s and 13 in 2080s.
Under the RCP 8.5 scenario (Figure 8), the risk of Kalikapur could reduce significantly and it could fall under very low-risk category by 2080s. For Khoajpur, the risk could remain the same in 2020s and 2050s, but could reduce to high-risk category in 2080s. The number of unions under very low-risk zone could increase with time from 6 unions in 2020s to 10 in 2050s and 12 in 2080s.
The areal extent of the very low-risk zone is around 33% in the baseline which could turn out to be 85% under the RCP 2.6 scenario and 80% under the RCP 8.5 in 2080s ( Table 6). The areal extent of the very high-risk zone is around 16% in the baseline which could turn out to be almost nil in 2080s under both scenarios. Also, there could be no area under high risk in 2080s for RCP 2.6, but there could be 9% area under such risk for RCP 8.5.

| DISCUSSION
Hydrodynamic modeling has become a popular tool in flood simulation Iqbal et al., 2022) in Bangladesh. Simulated flood hazard is used as a surrogate for flood risk ignoring the dynamic characteristics of socioeconomic exposure and vulnerability. As highlighted in the IPCC 6th assessment report, flood risk is also a function of exposure and vulnerability. However, there are less works on projection of socioeconomic conditions than on projection of hydro-climatic hazards. As a result, the future risk outlook due to CC often remains highly uncertain and practically less useful. For example, a normal monsoon flood can cause high risk if the socioeconomic condition, such as katcha housing, of the elements at risk is very poor. On the other hand, an improved housing condition can render low risk even in a high-flood situation. This study focuses on this dynamic and integrated manifestation of risk.
The socioeconomic indicators of the study area show that the sensitivity would decrease, and the adaptive capacity would increase in the future. As a result, more areas would fall in very low-to low-risk classes, even though the hazard intensity would be high or very high. Figure 9 compares the hazard and risk maps for RCP 2.6, and Figure 10 compares the same for RCP 8.5. As seen from the figures, the unions adjacent to the Arial Khan River would be more hazard-prone in the future. For example, the inundation area would increase by 52% in the Dhurail union, 54% in the Kalikapur union, and 39% in the Panchkhola union. These unions would fall under the very high-hazard category. Yet, the unions would remain under a very low-risk zone due to the decreases in disabled population, poverty rate, unemployment, and other sensitivity indicators, and the increases in crop productivity, literacy, income diversification, and other adaptive capacity indicators. As an example, the literacy rate would increase from 30% in the baseline to almost 100% in 2080s, and the unemployment rate would reduce from 25% in the baseline to almost nil. Due to such improvements, though seven unions would fall under the highhazard category, there would be almost no area in the high-risk category in 2080s under RCP 2.6. Three unions would fall under the very high-hazard and six unions under the high-hazard category, but only one union would fall under the high-risk category in 2080s under RCP 8.5.
Poor people including farmers and fishers who live close to the river suffer the most due to flooding as they experience higher exposure and risk (Brouwer et al., 2007). Predictions of the socioeconomic indicators based on the past data indicate that the unemployment and poverty situations could improve significantly, and the crop productivity and diversity, literacy rate, employment opportunities in industries and services, etc., could increase in the future. For example, the number of people employed in industries and services in a union was 80 in 1981, which increased to 106 in 1991, 218 in 2001 and 527 in 2011. Such increases could improve the adaptive capacity of local people and reduce the flood risk. As the income level would rise, the households would be able to flood-proof their houses. Moreover, the locals will be able to follow more adaptive coping strategies during floods, which would further reduce the flood damage (Koks et al., 2015).
To contain the future flood risk within a low-risk class in the study area, the current impetus of socioeconomic development should be enhanced or at least be maintained to ensure low sensitivity and high adaptive capacity. Along with this, both structural and nonstructural flood mitigation measures should be undertaken. Physical structures, such as flood control embankments, drainage channels and sluices, are needed along both the left and right banks of the river to reduce flood hazards in the very high-and highhazard unions (Dhurail, Kalikapur, Panchkhola, etc.). Such structures would help control and regulate the flood flow and thus modify the flood hazard. The crest level of the embankment should be raised so that the embankment can withstand the projected flood depth under the RCP 8.5 scenario. The drainage system should be improved, and the hydraulic performance of the system should be enhanced to facilitate quick drainage of flood water.
Among the nonstructural measures, bringing changes in the functional design of a building, for example raising the plinth of the building above the flood level in the very high-and high-hazard areas can be an effective flood-F I G U R E 7 Flood risks for different areas of the study floodplain under regional concentration pathway (RCP) 2.6 scenario for (a) baseline (b) 2020s (c) 2050s, and (d) 2080s.
T A B L E 6 Areas under different risk zones for regional concentration pathway (RCP) 2.6 and RCP 8. proofing approach to increase the adaptive capacity. Also, the habitable rooms and the most valuable assets can be kept at a higher level. Floodplain zoning with option for flood water storage in low-lands and river-corridors, regulation of development activities (e.g., filling of and encroachment on rivers, khals, canals, etc.) which hamper the river-floodplain connectivity, strengthening the disaster preparedness and response programs, and enhancing the flood forecasting capability with longer lead-time and 2D flood modeling for the floodplain, and improving the early warning and dissemination system involving locals can help reduce the flood damages and risk.
This study incorporates hydrologic and hydrodynamic aspects in flood simulation. Hazard parameters, such as flood flow velocity, recession period of floodwater, combined impact of flood depth and velocity, and shear stress on the floodplain, could also be included for a more comprehensive assessment of flood hazard. The incorporation of local rainfall might improve the accuracy of the simulated flood inundation. To identify vulnerability to flooding, only overall socioeconomic aspects of vulnerability were considered. However, to obtain a comprehensive view, vulnerability of specific crop, infrastructure, or economic sector to flood could be studied separately and then integrated.
F I G U R E 9 Comparison of flood (a) hazard and (b) risk for different areas of the study floodplain under regional concentration pathway (RCP) 2.6 scenario during 2080s.
F I G U R E 1 0 Comparison of flood (a) hazard and (b) risk for different areas of the study floodplain under regional concentration pathway (RCP) 8.5 scenario during 2080s.

| CONCLUSIONS
Natural and human-induced global warming and CC might intensify the magnitude and frequency of extreme events, such as monsoon precipitation in South Asia. This would potentially increase the fluvial flow and render the riverine floodplain to more frequent and intense flooding. Changes in socioeconomic exposure and vulnerability would interplay with dynamic physical hazard to shape the future flood risk. Such dynamic characteristics of flood hazard, socioeconomic exposure, vulnerability, and risk under existing and future CC scenarios were studied for the Arial Khan River floodplain in southcentral Bangladesh following the recent IPCC concepts.
Increasing trends are found in the flood depth, duration and flooded areas from baseline to 2080s. The flood depth, duration and extent could increase by 0.46 m, 5 days and 62% for the RCP 2.6 scenario and 0.75 m, 8 days and 80% for the RCP 8.5 scenario. Due to such increases in the flood parameters, the combined very low-and low-hazard zones would decrease from 71% to 21% and the combined high-and very high-hazard zones would increase from 14% to 47% for RCP 2.6. For RCP 8.5, the very low-and low-hazard zones would decrease from 56% to 22% and the high-and very high-hazard zones would increase from 14% to 58%. Thus, the hydroclimatic analysis suggests that the flood hazard would increase with time due to CC.
The socioeconomic exposure of the study area to flood could increase in the future due to the projected increases in population and households, though the net cropped area could decrease. In contrast to hazard and exposure, almost all the sensitivity parameters indicate a decrease in sensitivity to flood in the future. As a result, parts of the study area could shift from the very high-sensitive zone to the high-and medium-sensitive zones. Commensurate to sensitivity, all the adaptive capacity indicators would improve with time and hence the vulnerability would decrease. Due to the combined influences of flood, exposure and vulnerability, the low-risk flood zones would increase from 62% to 85%-91% from baseline to 2080s and the high-risk zones would shrink from 31% to 0%-9%.
The findings of this study provide useful information on flood hazard and risk in the Arial Khan floodplain. The union-wise flood risk maps prepared for the present and future CC scenarios could help the government line agencies to devise effective, economical, and sustainable flood management strategies. The integrated approach followed in flood risk assessment could emerge as a methodological guideline in risk management for the development planners, policymakers, implementers and researchers, not only in Bangladesh but also in other developing countries.