Framework of Spatial Flood Risk Assessment for a Case Study in Quang Binh Province, Vietnam

: Vietnam has been extensively a ﬀ ected by ﬂoods, su ﬀ ering heavy losses in human life and property. While the Vietnamese government has focused on structural measures of ﬂood defence such as levees and early warning systems, the country still lacks ﬂood risk assessment methodologies and frameworks at local and national levels. In response to this gap, this study developed a ﬂood risk assessment framework that uses historical ﬂood mark data and a high-resolution digital elevation model to create an inundation map, then combined this map with exposure and vulnerability data to develop a holistic ﬂood risk assessment map. The case study is the October 2010 ﬂood event in Quang Binh province, which caused 74 deaths, 210 injuries, 188,628 ﬂooded properties, 9019 ha of submerged and damaged agricultural land, and widespread damages to canals, levees, and roads. The ﬁnal ﬂood risk map showed a total inundation area of 64,348 ha, in which 8.3% area of low risk, 16.3% area of medium risk, 12.0% area of high risk, 37.1% area of very high risk, and 26.2% area of extremely high risk. The holistic ﬂood risk assessment map of Quang Binh province is a valuable tool and source for ﬂood preparedness activities at the local scale.


Introduction
Rivers have always been used as a water resource for food and hygiene needs as well as for agricultural and industrial harvesting. Land use planning around the world is often linked to hydrology and water supply [1]. Alluvial plains are ideal places for the development of urban activities as they are Floods often occur from September to November every year. When storms and tropical low pressures occur together with heavy rains, high tides cause floods in the plains and flash floods in mountainous and hilly areas. The flood damage categories throughout the period of 1989-2015 of Quang Binh province are summarised in Figure 2. The October 2010 flood event was the largest and most damaging recorded event during the period, which caused 74 flood fatalities, 188,628 damaged houses, and many other damages. This flood event was chosen as the case study for this research.

Flood Risk Assessment Framework
In response to the impact of floods around the world (including Vietnam [25]), this study proposed a unique approach to flood risk assessment. The concept of flood risk is often considered to involve the three elements of hazard, exposure, and vulnerability [26][27][28][29]. The combination is illustrated as in Equation (1). Flood exposure and flood vulnerability should be combined with flood hazard in assessing flood risk to provide a comprehensive resource reference for decision-makers in flood risk management [30]. Flood hazard can be determined as the potential for harm, loss, or damage of an event occurring at one location [28,31]. Exposure to flood hazard is defined by the potential for personal danger or property damage occurring during flood events [32,33]. Vulnerability is specified by the characteristics of a community that make it vulnerable to the damage of a flood hazard [26,34].
This study aimed to assess the nature and extent of flood risk by analysing potential flood hazards and evaluating existing conditions of flood exposure and vulnerability that potentially harm people, property, and livelihoods as shown in Equation (1). The case study is the October 2010 flood event of Quang Binh province, a hazardous flood event. The indicators or criteria of flood hazard, exposure, and vulnerability are adapted a variety of studies and based on an analysis of the existing

Flood Risk Assessment Framework
In response to the impact of floods around the world (including Vietnam [25]), this study proposed a unique approach to flood risk assessment. The concept of flood risk is often considered to involve the three elements of hazard, exposure, and vulnerability [26][27][28][29]. The combination is illustrated as in Equation (1). Flood exposure and flood vulnerability should be combined with flood hazard in assessing flood risk to provide a comprehensive resource reference for decision-makers in flood risk management [30]. Flood hazard can be determined as the potential for harm, loss, or damage of an event occurring at one location [28,31]. Exposure to flood hazard is defined by the potential for personal danger or property damage occurring during flood events [32,33]. Vulnerability is specified by the characteristics of a community that make it vulnerable to the damage of a flood hazard [26,34].
This study aimed to assess the nature and extent of flood risk by analysing potential flood hazards and evaluating existing conditions of flood exposure and vulnerability that potentially harm people, property, and livelihoods as shown in Equation (1). The case study is the October 2010 flood event of Quang Binh province, a hazardous flood event. The indicators or criteria of flood hazard, exposure, and vulnerability are adapted a variety of studies and based on an analysis of the existing data in Quang Binh province. The risk assessment result is integrated into a GIS framework to provide a flood risk map. The incorporation of flood risk assessment into a GIS framework has been applied at global, regional, and local scales in many recent studies [29,[35][36][37][38]. Meanwhile, there have been a few applications of geospatial assessment tools, including GIS, to assess the flood risk in Vietnam [18][19][20]39]. The proposed flood risk assessment framework of this study is described in Figure 3. The framework presents the relevant relationships between the flood risk components, flood risk assessment, and flood risk management that is proposed for Quang Binh province. This is the first time that a holistic flood risk assessment combining hazard, exposure, and vulnerability indicators has been conducted in the study area.
Sustainability 2020, 11, x FOR PEER REVIEW 5 of 17 data in Quang Binh province. The risk assessment result is integrated into a GIS framework to provide a flood risk map. The incorporation of flood risk assessment into a GIS framework has been applied at global, regional, and local scales in many recent studies [29,[35][36][37][38]. Meanwhile, there have been a few applications of geospatial assessment tools, including GIS, to assess the flood risk in Vietnam [18][19][20]39]. The proposed flood risk assessment framework of this study is described in Figure 3. The framework presents the relevant relationships between the flood risk components, flood risk assessment, and flood risk management that is proposed for Quang Binh province. This is the first time that a holistic flood risk assessment combining hazard, exposure, and vulnerability indicators has been conducted in the study area. Flood risk assessment and management framework for Quang Binh province adapted from several studies [26,31].
In this study, based on a critical analysis of the available data in Quang Binh province, we used indicators applied to flood hazard, exposure, and vulnerability as shown in Figure 3. This framework might also assess the risk of flooding in other provinces in Vietnam. Historical flood marks, land use map, Digital Elevation Model (DEM), river network map, transportation network map, and socialeconomic data are used in this study for mapping flood risk assessment. Various GIS techniques are applied for analysing and overlaying data and establishing spatial relationships using distributed information. In addition, the Analytical Hierarchy Process (AHP) is applied to give spatial decisions and provides the weights to analyse the input indicators [40][41][42][43]. Flood risk assessment and flood risk mapping that include the analysis of flood hazard, exposure, and vulnerability can provide a useful resource for flood risk management, mitigation actions, and governance ( Figure 3). The systematic approach and management aim to minimise potential harm and loss of flood risk.
In this study, based on a critical analysis of the available data in Quang Binh province, we used indicators applied to flood hazard, exposure, and vulnerability as shown in Figure 3. This framework might also assess the risk of flooding in other provinces in Vietnam. Historical flood marks, land use map, Digital Elevation Model (DEM), river network map, transportation network map, and social-economic data are used in this study for mapping flood risk assessment. Various GIS techniques are applied for analysing and overlaying data and establishing spatial relationships using distributed information. In addition, the Analytical Hierarchy Process (AHP) is applied to give spatial decisions and provides the weights to analyse the input indicators [40][41][42][43]. Flood risk assessment and flood risk mapping that include the analysis of flood hazard, exposure, and vulnerability can provide a useful resource for flood risk management, mitigation actions, and governance ( Figure 3). The systematic approach and management aim to minimise potential harm and loss of flood risk.

Multi-Criteria Decision-Making Analysis Model
Multi-criteria decision analysis methods (MCDA) allow working with quantitative variables and are applied to decision-making processes. These methods have potential applications in solving flood risk management issues (e.g., formulating their preferences and measuring these priorities [44]).
In this study, AHP is selected to weight criteria and subcriteria for flood risk assessment model. Several advantages of AHP include the direct opinion involvement, simple GIS integration [52], criteria and sub-criteria systematisation [53], and consistency in judgement [54]. Besides these advantages, this approach has three main limitations of subjective preference in the evaluation [55], a large number of pairwise comparisons [56], and vague criteria [57]. However, these shortcomings remain in almost MCDA methods [53,57].
AHP is a theory of tangible criteria measurement proposed by T. Saaty [58]. The weight of the criteria is evaluated through algorithms and pairwise comparison matrices. AHP basically supports the decision-making process by quantifying alternative priorities for decision-makers [56]. This is a powerful and flexible technique to support setting priorities and improving decision-making processes. This method has been widely used in many areas such as economics, planning, education, environment, transportation, resource allocation, and management [59,60]. More recently, it has been applied to flood risk assessment studies [42,45,61,62]. Figure 4 illustrates the AHP model used to assess the flood risk for Quang Binh province. The flood risk is assessed by the combination of flood hazard (flood depth indicator), flood exposure indicators of population density, land-use classification, and distance to river, and flood vulnerability indicators of road density and poverty rate. The AHP is performed by the following three main steps: Step 1: Construct a hierarchical decision model as in Figure 4.
Step 2: Develop a paired comparison matrix for criteria or sub-criteria of the decision model as in Equation (2) based on subjective judgment and reciprocal judgement axiom.
Step 3: Obtain the relative importance or weights of criteria and sub-criteria.
In this study, AHP is selected to weight criteria and subcriteria for flood risk assessment model. Several advantages of AHP include the direct opinion involvement, simple GIS integration [52], criteria and sub-criteria systematisation [53], and consistency in judgement [54]. Besides these advantages, this approach has three main limitations of subjective preference in the evaluation [55], a large number of pairwise comparisons [56], and vague criteria [57]. However, these shortcomings remain in almost MCDA methods [53,57].
AHP is a theory of tangible criteria measurement proposed by T. Saaty [58]. The weight of the criteria is evaluated through algorithms and pairwise comparison matrices. AHP basically supports the decision-making process by quantifying alternative priorities for decision-makers [56]. This is a powerful and flexible technique to support setting priorities and improving decision-making processes. This method has been widely used in many areas such as economics, planning, education, environment, transportation, resource allocation, and management [59,60]. More recently, it has been applied to flood risk assessment studies [42,45,61,62].
Figures 4 illustrates the AHP model used to assess the flood risk for Quang Binh province. The flood risk is assessed by the combination of flood hazard (flood depth indicator), flood exposure indicators of population density, land-use classification, and distance to river, and flood vulnerability indicators of road density and poverty rate. The AHP is performed by the following three main steps: Step 1: Construct a hierarchical decision model as in Figure 4.
Step 2: Develop a paired comparison matrix for criteria or sub-criteria of the decision model as in Equation (2) based on subjective judgment and reciprocal judgement axiom.
Step 3: Obtain the relative importance or weights of criteria and sub-criteria.
AHP operates by setting priorities for multi-criteria, which are judged by experts to derive the best decision [63]. The weights of criteria in AHP method rely upon the subjective judgment of several experts [64][65][66][67], or the author experience-based assessment [42,61,62]. In this study, we applied the author experience-based assessment. We also referenced the related studies which used the AHP to assess flood risk criteria [42,55,57,[64][65][66][67].  The paired comparison matrix in Step 2 is A = a ij , i j = 1, 2, . . . , n. The entries a ij is defined by reciprocal judgement rule, if a ij = a, then a ji = 1/a, a > 0 in Equation (2).

Data Used
The historical flood in October 2010 in Quang Binh resulted in 74 deaths, 210 injuries, 188,628 flooded properties, 9019 ha of submerged and damaged agricultural land, and widespread damage to canals, levees, embankments, and road (see Figure 3). The flood frequency of 2010 flood event was estimated at 5% or 20 years of return period at Gianh river basin, and at 2% [68] or 50 years of return period at Nhat Le river basin [69]. Flood marks of 2010 flood event were collected by Quang Binh Centre for Hydrometeorological Forecasting. Flood marks were mainly distributed along Gianh and Nhat Le river basins, which are frequently flooded areas of Quang Binh province. The flood mark data then were integrated with a high-resolution DEM to establish flood inundation map. The DEM was collected from the Department of Surveying and Mapping, Vietnam.
Three criteria were selected to estimate flood exposure: land use categories, population density, and distance to the river. The land use map was provided by Quang Binh Environment and Natural resource Departement. The land-use categories include residential areas, construction areas, transportation areas, mining areas, agricultural areas, grassland, water bodies, forest, plants in residential areas, woodland, and bare soil. In this study, the land use criterion was used to estimate flood exposure. Land use categories were reclassified into four classes: homestead and built up, agriculture land, forest and vegetation, and water bodies. The classification is often used for flood risk assessment models [42,43]. Population density is the most important criterion in flood exposure analysis since it is determined by human settlements. This data was collected in the 2019 statistical yearbooks of eight districts of Quang Binh province. More densely populated areas are at higher risk when floods occur [10]. The distance to rivers is taken from a river network map using Euclidian distance tool in ArcGIS software. People living near river systems are also at higher risk than others [42].
Two criteria were used in the present study to analyse flood vulnerability: poverty rate and road density. The poor are more vulnerable to natural hazards [14] and the poverty rate is often considered a structural cause of flood vulnerability. The poverty rate data was collected in the 2019 statistical yearbooks. Roads or transportation system play an important role in disaster response activities (e.g., evacuation) and recovery activities [70]. Many other criteria could be added to assess flood vulnerability, such as healthcare facilities, disabilities, income, gender, age, and adaptive indicators. However, such data is not currently available in the research area.

Flood Inundation Map
This study used flood mark data from the October 2010 flood event and a 5 m resolution DEM to create a flood inundation map using spatial analyst techniques. The flood inundation map in Figure 5 is the result of the geospatial modelling techniques which was established in the study of Luu et al. [20].

Flood Exposure Analysis
Flood exposure criteria, including land use categories, population density, and distance to rivers criteria are shown in Figure 6. The weights of criteria and sub-criteria are derived from the AHP model ( Figure 4). We used Super Decision software [71] to calculate AHP algorithms based on subjective judgements. The distance to rivers is acquired from river and stream network data, It is classified into four categories: less than 1 km from river systems, 1-2 km, 2-3 km, and greater than 3 km. The closer it is to rivers, the more dangerous it is when exposed to floods. The higher score is given to closer distance to the river system. The land-use categories are reclassified into agricultural land, homestead and built-up, water bodies, and forest and vegetation. The potential impact of flooding is very high on residences and infrastructure involving people, so the relative weight is set to be the highest for the homestead and built-up category. The second highest weight is achieved for agriculture land because it is linked to the livelihood of the communities. The forest and vegetation and water bodies have the lowest weights since they do not pose a threat to people. The population density is the most crucial criterion since it is directly linked to people. The higher the population density is, the higher the weight is distributed (Table 1).

Flood Exposure Analysis
Flood exposure criteria, including land use categories, population density, and distance to rivers criteria are shown in Figure 6. The weights of criteria and sub-criteria are derived from the AHP model ( Figure 4). We used Super Decision software [71] to calculate AHP algorithms based on subjective judgements. The distance to rivers is acquired from river and stream network data, It is classified into four categories: less than 1 km from river systems, 1-2 km, 2-3 km, and greater than 3 km. The closer it is to rivers, the more dangerous it is when exposed to floods. The higher score is given to closer distance to the river system. The land-use categories are reclassified into agricultural land, homestead and built-up, water bodies, and forest and vegetation. The potential impact of flooding is very high on residences and infrastructure involving people, so the relative weight is set to be the highest for the homestead and built-up category. The second highest weight is achieved for agriculture land because it is linked to the livelihood of the communities. The forest and vegetation and water bodies have the lowest weights since they do not pose a threat to people. The population density is the most crucial criterion since it is directly linked to people. The higher the population density is, the higher the weight is distributed (Table 1).

Flood Vulnerability Analysis
Flood vulnerability criteria, including poverty rate and road density, are shown in Figure 7. The weights of criteria and sub-criteria are derived from the AHP model ( Figure 4). We used Super Decision software [71] to calculate AHP algorithms based on subjective judgements. The poverty rate and flood vulnerability are interrelated in Vietnam [18]. A higher poverty rate receives a higher weight for flood vulnerability. The road density is derived from intersecting transportation network and commune boundary. The lower road density area has a higher vulnerability score ( Table 2).

Flood Vulnerability Analysis
Flood vulnerability criteria, including poverty rate and road density, are shown in Figure 7. The weights of criteria and sub-criteria are derived from the AHP model ( Figure 4). We used Super Decision software [71] to calculate AHP algorithms based on subjective judgements. The poverty rate and flood vulnerability are interrelated in Vietnam [18]. A higher poverty rate receives a higher weight for flood vulnerability. The road density is derived from intersecting transportation network and commune boundary. The lower road density area has a higher vulnerability score (Table 2).

Flood Risk Assessment
Flood risk is assessed based on the framework in Figure 1 and Equation (1). In this study, a flood inundation map is considered a flood hazard map. Flood exposure and vulnerability maps are generated using Weighted Sum technique in ArcGIS software. The results of flood inundation, vulnerability, and exposure are displayed in Figure 8. The final flood risk assessment map is generated based on a combination of flood inundation (hazard), exposure, and vulnerability maps.
The flood risk assessment result is displayed in Figure 9. In the map, the flood risk score is normalised within the range of 0-1. Areas located near and along Nhat Le and Gianh river basins are at higher risk of flooding. The total inundation area is 64,348 ha, in which 5361 ha of low risk (0.14-0.312), 10,518 ha of medium risk (0.312-0.484), 7702 ha of high risk (0.484-0.656), 23,901 ha of very high risk (0.656-0.828), and 16,868 ha of extremely high risk (0.828-1.000). Alternatively, the distribution of risk levels in the research area is 8.3% of low risk, 16.3% of medium risk, 12.0% of high risk, 37.1% of very high risk, and 26.2% of extremely high risk.

Flood Risk Assessment
Flood risk is assessed based on the framework in Figure 1 and Equation (1). In this study, a flood inundation map is considered a flood hazard map. Flood exposure and vulnerability maps are generated using Weighted Sum technique in ArcGIS software. The results of flood inundation, vulnerability, and exposure are displayed in Figure 8. The final flood risk assessment map is generated based on a combination of flood inundation (hazard), exposure, and vulnerability maps.
The flood risk assessment result is displayed in Figure 9. In the map, the flood risk score is normalised within the range of 0-1. Areas located near and along Nhat Le and Gianh river basins are at higher risk of flooding. The total inundation area is 64,348 ha, in which 5361 ha of low risk (0.14-0.312), 10,518 ha of medium risk (0.312-0.484), 7702 ha of high risk (0.484-0.656), 23,901 ha of very high risk (0.656-0.828), and 16,868 ha of extremely high risk (0.828-1.000). Alternatively, the distribution of risk levels in the research area is 8.3% of low risk, 16.3% of medium risk, 12.0% of high risk, 37.1% of very high risk, and 26.2% of extremely high risk. Sustainability 2020, 11, x FOR PEER REVIEW 11 of 17

Discussion
Although studies on flood risk analysis have been increasingly conducted in Vietnam, there is a lack of research on detailed flood risk assessment maps, particularly at local levels. In this study, we developed a detailed flood risk assessment map for Quang Binh province of Vietnam. The flood risk map was generated by combing flood hazard, exposure and vulnerability maps. Some studies have focused on flood risk assessment in terms of combining flood vulnerability and flood hazard [70,72,73]. However, flood risk is often considered by the combination of hazard, exposure, and vulnerability, which fully reflects aspects of flood risk. This approach has been applied in many studies at both global [30,36,74] and local scales [29,37,75].
Successive flood events have significantly impacted on the residents' livelihood and socio-economic development in Quang Binh province (see Figure 2). A practical risk management approach can help to reduce the adverse impacts of flood risk in the area. Before a disaster ever materialises, we can work to reduce risks and discuss the avoidance of activities that actually create risk [76]. The holistic flood risk assessment map would provide a useful tool and source for flood preparedness activities at the local scale.
The study has been based on the regulations for the establishment of component maps for constructing flood risk assessment. These maps are gathered from reliable and high-resolution data sources. Maps are constructed in raster format using geographic information technology application for the study area. The research results include a new modelling methodology for developing flood risk assessment mapping tool using historical flood marks, high-resolution DEM, and MCDA approach. MCDA methods are often used to incorporate the three components of hazard, exposure, and vulnerability due to several advantages of the methods such as criteria and sub-criteria systematization [53], suitable for GIS integration [77], and consistency in judgement [54]. The methodology would contribute not only to the development of theoretical and methodological modelling of flood risk assessments, but also as a tool to assist managers, decision-makers, and policy-makers in developing flood risk management action plans.
In this study, flood risk is assessed with the integration of various indicators of flood depth, population density, land use category, distance to rivers, poverty rate, and road density using AHP and spatial analysis techniques. Some previous studies in the field of flood risk analysis in Vietnam focused on the flood hazard assessments (for example, [17,39,78,79]). The present study used AHP method and spatial techniques to combine flood inundation map with flood exposure and vulnerability data to provide an integrated flood risk assessment map.
The flooded areas calculated from the model of this study are compared with the flood areas of an equivalent flood in Quang Binh, which were calculated in another study to ensure the reliability of the model. The flood frequency of November 1999 and October 2010 flood events are approximately the same [80]. The inundation map of October 2010 flood event was generated in this study using historical flood mark data and high-resolution DEM. The inundation map of November 1999 flood event was generated in the study [81]. We compared the flooded areas in communes along the Nhat Le river basin of October 2010 flood event and November 1999 flood event. The result in Figure 10 shows that the inundation areas of October 2010 flood event, which was created in this study by using historical flood marks and a high-resolution DEM, are quite compatible with the inundation areas of November 1999 flood event, which was created in the study of Nguyen and Phan [81].
Floods have severely affected communities' livelihoods and socioeconomic development in Quang Binh over the years (Figure 10). Low-land areas, including agricultural areas along Nhat Le and Gianh river basins, are often subjected to flooding in annual rainy seasons ( Figure 5). This study developed a holistic flood risk assessment map incorporating hazard, exposure, and vulnerability for Quang Binh province that could provide helpful information for decision-makers and policy-makers to implement and improve flood mitigation and response measures for the area. The map is also essential for accurate communication about the local flood risk situation in the floodplain for affected communities. The study developed a geospatial database and a theoretical framework for developing flood risk assessment maps using historical flood marks, DEM, other geospatial data and social-economic data. The framework does not require time series meteorological and streamflow data or updated river cross-section data, which are not available in many data-scarce areas. Therefore, the framework could potentially be applied to other provinces, especially those in central Vietnam with similar topographic and climatic conditions to create flood risk assessment maps.
Sustainability 2020, 11, x FOR PEER REVIEW 13 of 17 updated river cross-section data, which are not available in many data-scarce areas. Therefore, the framework could potentially be applied to other provinces, especially those in central Vietnam with similar topographic and climatic conditions to create flood risk assessment maps. Besides these strengths, the present study has some limitations of flood vulnerability data and MCDA approach. More data could be added to analyse flood vulnerability such as healthcare facilities, gender, and persons with disabilities; however, such data is not available in the research area. The MCDA approach, in general, and AHP, in particular, require subjective judgments in weighting indicators [55] and subjective model validation [45].

Conclusions
The present study provides a new approach to assess the flood risk for the local area of Quang Binh province. We used historical flood marks and a high-resolution DEM to create an inundation map, and then combined the inundation map with exposure and vulnerability data to create a flood risk assessment map using spatial multicriteria decision analysis techniques. The detailed flood risk assessment map could support the implementation of strategies and specific actions of government authorities to control, reduce, and transfer the flood risks. The framework developed in this study could provide a methodology to rapidly simulate flood risk assessment maps using available data in local areas, especially in areas where there is insufficient data for hydraulic models. In addition, it is a potential way to engage local decision-makers in the approach and framework of this study for further investigation and validation of flood risk in the study area.  Besides these strengths, the present study has some limitations of flood vulnerability data and MCDA approach. More data could be added to analyse flood vulnerability such as healthcare facilities, gender, and persons with disabilities; however, such data is not available in the research area. The MCDA approach, in general, and AHP, in particular, require subjective judgments in weighting indicators [55] and subjective model validation [45].

Conclusions
The present study provides a new approach to assess the flood risk for the local area of Quang Binh province. We used historical flood marks and a high-resolution DEM to create an inundation map, and then combined the inundation map with exposure and vulnerability data to create a flood risk assessment map using spatial multicriteria decision analysis techniques. The detailed flood risk assessment map could support the implementation of strategies and specific actions of government authorities to control, reduce, and transfer the flood risks. The framework developed in this study could provide a methodology to rapidly simulate flood risk assessment maps using available data in local areas, especially in areas where there is insufficient data for hydraulic models. In addition, it is a potential way to engage local decision-makers in the approach and framework of this study for further investigation and validation of flood risk in the study area.