Distributed Hydrological Model for Assessing Flood Hazards in Laos

Many natural disasters have recently occurred in Laos. Among them, flooding has been the greatest problem. Land use change (deforestation and urbanization) and climate change have played significant roles, and it is important to understand the impacts of these changes on flooding. We have developed an integrated hazard map based on a combination of four hazard maps of flooding, land use change and climate change to assess hazard areas at the national scale. The hazard map was developed using the analytical hierarchy process (AHP) and a hazard index. Finally, we divided the map into four hazard area categories, which include low, medium, intermediate and high. Based on this analysis, the integrated hazard map of Laos indicates that low hazard areas cover 87.44% of the total area, medium hazard areas cover 8.12%, and intermediate and high hazard areas respectively cover 2.42% and 2% of the land area. We compared the results with historical events to confirm that the proposed methodology is valid.

Over the last few decades, floods have become one of the main problems causing damage to human and animal lives, infrastructure, and agricultural and economic areas [3]. Furthermore, in the near future, socioeconomic development, climate change and the decrease in forest density could significantly increase the impact of floods [2] [4]. However, the impact of floods can be reduced by developing a flood risk map. Generally, flood risk maps are generated based on a combination of factors related to hazards (physical) and vulnerability (socioeconomic). Current global flood models can be unsuccessful in precisely predicting the dynamics of climate change and socioeconomic development [4].
Nevertheless, knowledge of flood hazards plays an important role in the strategy for minimizing flood risks such as reducing damage in agricultural areas and losses of animal and human lives. In addition, decision-makers can create future development plans. Flood hazards can be quantified by considering the occurrence probability of potentially damaging events. The occurrence probability can serve as an indicator for verifying the sensitivity of such areas. Many researchers have developed reliable flood hazard maps as a basis for generating flood risk maps to support flood mitigation plans [5] [6]. However, many countries in the Southeast Asia area still lack flood hazard maps.
To date, there has only been limited assessment, analysis or projections regarding potential climate change impacts on the physical and social environment in Lao PDR. The Intergovernmental Panel on Climate Change (IPCC) has found that floods will tend to occur more frequently in Southeast Asia in the future [7]. The occurrence of climate change can lead to increases in heavy rainfall and flooding [8], and floods are one of the most dangerous hazards to the economy and human lives, especially for developing countries like Laos.
Hazards can be single or combined in their origins and effects on an area. In recent decades, many researchers have paid attention to the used of multi-hazard assessment, which focuses on all scales. For example, Marzocchiet [9] used the multi-hazard assessment at the medium scale in their studies. Large scale studies involving the multi-hazard assessment include those of [10] [11]. In addition, several models have been developed for estimating the risk of multi-hazards, but most of these models are very data intensive.
which are a powerful tool for managing large volumes of spatial data, integrating data from different sources and performing analyses [12] [13]. Nevertheless, for decision-making purposes, GIS applications are inefficient for use in multiple criteria analyses. To use GIS for multi-criteria decision-making, it is necessary to combine GIS with multi-criteria decision analysis (MCDA). Several studies have shown that coupling GIS with MCDA can be used to generate a flood hazard map. Among MCDA methods, the analytic hierarchy process is the most popular because it is a user-friendly, convenient method that provides accurate results and is suitable for other hazard studies [13] [14]. Until recently, flood hazard mapping using AHP and GIS has been applied using GIS-based map information such as slope, flow accumulation, elevation, land use, and rainfall intensity [15].
Regarding the region, there is a difference between the impact of land use change and climate change to flood [16] [17]. Characteristic of hazards in each region can lead to the transformation of future land use to lower the impact of the flood. Therefore, it is important to understand the characteristic of hazards in each region. Understanding the factor that driven regional sensitivity is essential for future mitigation and adaptation strategies for instance based on Laos national report [18]. The magnitude of hazard is increasing in the northern and southern part of Laos because of the climate change.
The main objective of this study is to propose a reliable and capable of AHP method to integrate hazard maps to gather and the integrated hazard map can be used to identify sensitive areas over the region with limited available data. The modeling method combines different hazard maps including flood, land use change, and climate change. AHP is used as a tool to weigh the priority of the maps. The proposed methodology provides an integrated hazard map that can be used as a guide map, which provides important information for developing countermeasures for floods, as well as other natural hazards. The map index is shown on a scale from zero, which means low hazard to one, which means high hazard probability. This is also the first time a hazard map for the entire country of Laos has been developed. Another advantage of this proposed study is that the AHP weights used to develop unified hazard maps are based on the decision maker's design criteria and the priorities. It is helpful to identify hazard areas and focus on potential impact areas.

Study Area and Dataset
Located in the center of the Indochinese Peninsula, Laos is situated between longitudes 100˚ and 108˚E and latitudes 14˚ and 23˚N. The country has a total area of 236,800 km 2 , with the Mekong River flowing through approximately 1900 km of its territory from north to south and with over 800 km comprising a natural border with Thailand.
Land use in Laos is classified as forests (65.2%), vegetation or agricultural Journal of Water Resource and Protection areas (33.8%), and water bodies and bare land (1%). Almost all agricultural areas are paddies. Land use type is one of the factors used to determine water infiltration. Soil data are based on the Harmonized World Soil Database, which is a 30 arc-second raster database with over 16,000 different soil mapping units; it combines existing regional and national soil information updates worldwide with the information contained within the 1:5,000,000 scale FAO-UNESCO Soil Map of the World. The original soil type data based on the global soil unit (SU) was converted to the soil texture class. The soil type data plays an important role in the infiltration factor of the hydrological distribution. Based on the government the country can separately into 3 regional areas namely, northern, central and southern region (Figure 1).
To simulate hydrological distributed models, a hydrological and meteorological dataset from the Mekong River Commission is used. The daily meteorological data at 40 stations for the period of 1970 to 2000 is used for the rainfall-runoff simulation and for calibration and validation of the distributed hydrological model ( Figure 1). The parameters include precipitation, soil types, and elevation. In addition to the rainfall data, daily maximum data were selected to analyze the rainfall intensity for 50 and 100-year return periods ( Figure 2).
The rainfall intensity for many return periods was calculated using the Log-Pearson Type III.

Methodology
We have developed a model for the multi-criteria analysis of hazard maps. The AHP hazard index aims to support the identification of areas that are sensitive to flood hazards. Figure 3 shows the methodology used. The first step is the data collection stage, in which spatial data in the target area and the judgment of several experts on hazard assessment criteria are collected. Phase I includes modeling and the analysis of hazard maps such as flooding, land use changes, and climate changes. In Phase II, we calculate the weights of the hazard maps criteria. During this phase, we used AHP as a tool to measure the weights of the criteria, and the consistency ratio is computed to verify the expert judgment consistency. Finally, Phase III covers the integration of the hazard maps and the criteria weights, which generates an integrated flood hazard map. Using this integrated map, sensitive areas may be easily determined, which provides support for decision makers to design plans. The hydrological model was used to estimate 3 hazard maps, which include floods, climate changes, and land use changes.

Floods
The method is based on the distributed hydrological model proposed by Kazama [19] and developed by Kashiwa [20]. The hydrological processes considered in this model include precipitation, infiltration, surface runoff, base water flow and water balance in each layer. The model technically consists of a set of hydrological parameters describing the catchment properties and algorithms describing the physical processes. For each grid cell, two layers are considered in the vertical direction: the base water layer and the surface layer. The model incorporates a direct flow and base flow model used to estimate the river flow. Direct flow and base flow are calculated using kinematic wave concepts, which process meteoric water runoff using momentum and a continuity equation. Calculations of base flow use precipitation data as input and Manning's equation to determine the base flow rate, as well as infiltration from the upper layer as input data and the storage function method to determine the sub-base outflow. The infiltration rate was determinedly based on soil type. For distributed system modeling, information on river catchment geological and topographical characteristics is required to derive or measure the necessary parameters. River basin characteristics are described by a dataset (e.g., elevation, flow direction, catchment area and stream network, and land use type), which are derived from the DEM. Manning's roughness coefficients were calibrated by trial and error; three basins, which include the northern Ou river (Figure 4(a)), the central Sane river (Figure 4(b)) and the southern Sedone river (Figure 4(c)) in Laos of 2000 were chosen as example for comparison between observed and simulated discharge. The comparison is shown in Figure 4. The performance of this model was determined using two commonly used statistical performance measures. The first is the coefficient of determination R 2 , and the second is the Nash-Sutcliffe efficiency E, all the stations' coefficients of 2000 are shown in Table 1 as an example. The overall results of the validation are satisfactory since all the values are Journal of Water Resource and Protection above 0.6, which indicates a satisfactory fit between observed and simulated hydrographs [21].  We propose a hazard index, which is adapted from the relationship between velocity and flood depth [22], which is shown in Table 2. By considering the water depth of every grid in the flood map, we converted the map to a hazard index.
The scenario is that the water velocity from flooding areas is low, and the depth can be transformed into a hazard index. The index is scaled from zero to one, with zero describing the lowest risk and one describing the highest risk.

Land Use Change
Soil, topography and land cover are the most important factors that determine rainfall runoff, which leads to the scale of flood events in the catchment area. Land use changes may result in flood-drought disasters or ecological problems [23]. A land use change scenario has been generated. First, the forested area becomes an agricultural area with the assumption that the ground slope is less than 12˚ [24]. Second, there is an expansion of urban areas based on the increased population probability, and then, a shift from rural areas to urban areas occurs. For this scenario, data from the NASA Global Grid of Probabilities of Urban Expansion were used.  1  6  11  16  21  26  31  36  41  46  51  56  61  66  71  76  81  86  91  96  101  106  111  116  121  126  131  136  141  146  151  156  161  166  171  176  181  186  191  196  201  206  211  216  221  226  231  236  241  246  251  256  261  266  271  276  281  286  291  296  301  306  311  316  321  326  331  336  341  346  351  356  361

Climate Change
Climate change is expected to increase both the magnitude and frequency of extreme precipitation events, which may lead to more intense and frequent river flooding. Several studies have shown that climate change has been a contributing factor to flooding by raising the precipitation level relative to the average annual rainfall [25] [26]. We used the inverse distance weight (IDW) spatial analysis function in ArcGIS to distribute the rainfall, which is shown in Figure 2.

Analytical Hierarchy Process (AHP)
AHP is a powerful tool for multi-criteria decision-making [27]. To provide the relative weights of the criteria, it is necessary to define each criterion's relative importance, and thus, a pair-wise comparison matrix for each criterion is created to enable significance comparisons. We have 3 criteria, which include Flood, Land use change and Climate change, and thus, the matrix is 3 by 3, and the diagonal elements are equal to 1. The value of each row of pair-wise comparisons is determined based on expert judgments.

Relative Weight of Criteria
To obtain the criteria relative importance value, expert judgments are required. We where RI is relative important of pairwise i A and i B are the responses to the questionnaire and m is the number of samples.
Based on the data gathered from the questionnaire, a pair-wise comparison matrix was constructed with normalized values of each parameter from Table 3.
When we compare the inverse of the pair-wise values, the scale value is the reciprocal value. For example, the value for flooding vs. land use change is 4.90, and thus, the value for land use change compared to flooding is 1/4.90 = 0.20; the parameter averages for each row in the normalized matrix are computed to obtain the corresponding weight i w of each criterion, which is shown in Table 3.

Consistency Check
In practice, it is impossible to expect the decision maker to provide a pair-wise comparison matrix that is completely consistent. Therefore, after obtaining i w , the consistency needs to be evaluated.
The consistency ratio is evaluated as follows: where CR is the consistency ratio, CI is the consistency index and RI is a random index that is dependent on the sample size, which is shown in where max λ is the maximum eigenvalue of the comparison matrix and n is number of criteria. max λ is calculated according to: where A is a pair-wise comparison matrix from Table 3.
From Equation (5), we find i v ; from Equation (4), we can find max λ = 3.10 and CI = 0.05. Finally, the consistency ratio was calculated to be CR = 0.086.
Since, the CR value is lower than the threshold (0.1), this indicates that the expert judgments are reasonably consistent.

AHP Based Hazard Index
The AHP method was used to weight the priority of each hazard on the national scale but on the regional scale, the priority of each hazard is different. Therefore, on the regional scale, the weight of each hazard is defined by:

Hazard's Weight in Regional Scale
where HW is the weight of the hazards in regional scale, avgHW is the average weight of the hazards index, hazard is either fl , lu or cc ( fl is flood hazard map, lu is land use hazard map and cc is climate change hazard map).

Flood Hazard Map
A distributed hydrological model was used to simulate a flood hazard map for all of Laos. We considered the greatest water depth in every grid cell, which was determined by contributing factors during the simulation, and these included the 100-year return periods of rainfall, land types, soil hydrologic characteristics, and elevation. The results are shown in Figure 5

Land Use Change Hazard Map
The results in Figure 5(b) show the overall impact of the hazard areas, which are growing significantly; this is mostly because of the loss of forest area that slows the rainfall runoff. Without the forest area, all rainfall runoff runs directly downstream without storage or other factors to slow it down. Therefore, the hazard areas downstream are expanding. The total area where land use change occurs can be divided into 71.88%, 12.68%, 7.94% and 7.5% of low, medium, intermediate and high hazard areas, respectively. Intermediate and high hazard areas can be further divided: 89.32% of intermediate areas are in the forest, 10.55% are in agricultural areas, and 0.12% are in urban areas. We found that 90.52% of high hazard areas are in forests, 9.32% are in agricultural areas and 0.12% are in an urban area. In addition, we analyzed the increase of total hazard index between flood and land use change hazard map to identify the sensitivity of the area to land use change in 3 different regional areas namely, northern area, central area, and southern area. The average hazard indexes in the northern, central and southern region are 0.12, 0.16, and 0.13 respectively.

Climate Change Hazard Map
Developing countries in tropical regions are highly susceptible to floods. These regions already have high levels of precipitation, and the hydrologic cycle is significantly interlinked and sensitive to the weather. The objective here is to find areas that are sensitive to intense rainfall. Therefore, 50 and 100-year return period rainfall events were used to determine the sensitivity of these areas. From the results, it appears that low hazard areas cover 69.4% of the total area, medium hazard areas cover 12.57%, and intermediate and high hazard areas respectively cover 10.18% and 7.85%. As shown in Figure 5(c), for the percentage increase in water depth from 50 and 100-year return period rainfall events, many areas were susceptible to the change in rainfall intensity. Those areas show a percentage increase in water depth, which leads to an increase in flood hazards that need to be considered in a flood management plan or met with countermeasures. The average hazard indexes in the northern, central and southern region are 0.18, 0.11, and 0.21 respectively.

Integrated Hazard Map
The main objective of this paper is to integrate four existing hazard maps. Many studies have estimated the risk of different hazards in various zones using GIS as a tool to integrate and analyze data from different sources. The weight of each hazard is defined using the analytical hierarchy process (AHP), and a flood hazard map is created, which is shown in Figure 6.

Land Use Change Hazard Map
The results in Figure 5(b) show the overall impact of the hazard areas, which are growing significantly; this is mostly because of the loss of forest area that slows the rainfall runoff. Without the forest area, all rainfall runoff runs directly downstream without storage or other factors to slow it down. Therefore, the hazard areas downstream are expanding. The total area where land use change occurs can be divided into 71.88%, 12.68%, 7.94% and 7.5% of low, medium, intermediate and high hazard areas, respectively. Intermediate and high hazard areas can be further divided: 89.32% of intermediate areas are in the forest, 10.55% are in agricultural areas, and 0.12% are in urban areas. We found that 90.52% of high hazard areas are in forests, 9.32% are in agricultural areas and 0.12% are in an urban area. In addition, we analyzed the increase of total hazard index between flood and land use change hazard map to identify the sensitivity of the area to land use change in 3 different regional areas namely, northern area, central area, and southern area. The average hazard indexes in the northern, central and southern region are 0.12, 0.16, and 0.13 respectively.

Discussion
According to Table 5, the high hazard areas have shallow water depth, and most of them include forest areas. Similarly, forest areas comprise the largest percentage of the area, which is followed by agricultural and urban areas. Comparing the hazard areas before integration (flood, land use change and climate change) and after integration (integrated hazard map by AHP), we found that most of the high hazard areas (90.73%) were located in forest areas, which increased compared to the hazard maps before integration (flood (90.47%), land use change (90.54%) and climate change (89.80%); 8.93% of the integrated hazard map was located in urban areas, which increased from the climate change hazard map (8.16%) but decreases when compared to the flood hazard map (9.38%) and land use change hazard map (9.21%); and 0.34% of the integrated hazard map was located in agricultural areas, which increased from the flood (0.15%) and land use change hazard map (0.25%) but decreased from the climate hazard map (2.04%). Although, each region has its own sensitivity to hazard, the priority of each hazard from the AHP method can reflect the overall situation of hazard in national scale. Based on these results, it is important to focus priorities on flood protection and management plans in high hazard areas that are in urban areas to minimize the consequences of flood damage. Journal of Water Resource and Protection To validate the hazard weights of AHP method, we compared them with the hazard weight from the average hazard index in Equation (7) for intermediate and high hazard indices. The weights from the AHP method were acquired from experts' judgments. The experts' judgments rely on their involvement in hazard events that occurred in accessible areas such as urban, agricultural and paddy field areas; the experts did not take into consideration the hazard events that occurred in areas that are hard to access such as mountainous and forest areas.
The areas were divided into 3 regions (southern, central and northern) to monitor the hazard events because the same hazard can have different magnitudes depending on the region. Therefore, we used Equation (7)  were 13 km 2 and 12 km 2 , respectively, and after land use changes, these areas increased to 32 km 2 and 34 km 2 , respectively (Figure 8(b)). The results show that in terms of the impact area, the agricultural areas have a more significant impact than urban areas if the trend in changing land use is according to our assumptions. To minimize the impact from the expansion of agricultural and urban areas, those areas located in hazard areas must be avoided.
Flooding can be alleviated using agricultural land to store water. For example, an agricultural area can act as a sub-catchment to reduce the impact of flooding on urban and rural areas. In addition, it is necessary to include strategies for flood mitigation downstream from urban areas. With planning, areas prone to flooding will be able to store floodwaters in rural areas. Moreover, water from flooded areas can be used for irrigation systems to develop a tolerance to climate Journal of Water Resource and Protection change. The advantage of the proposed integrated hazard map is its ability to provide an overall assessment of flood hazard areas at the national scale. The application can also be extended to assess hazard zones in urban areas.

Conclusions
The main purpose of this study is to develop an integrated hazard map that is reliable on a national scale. The result is an integrated map of flooding, land use change and climate change hazards developed using AHP to perform the integration. The flood hazard map was generated using a hydrological model, which was scaled with a hazard index. The results show that urban areas have few high hazard flooding indexes when compared with other land use types. The land use change hazard map was generated based on the scenario that all forested areas with less than 12 degree slopes are deforested for agricultural use. The high hazard areas increase from approximately 5.34% in the flood hazard map to 7.5% of the total area. This indicates that forest density is a significant factor in preventing flooding in our study area. The climate change hazard map was generated by examining the differences in rainfall between the 50 and 100-year return period maps to show the areas susceptible to changes in rainfall intensity. The results show the hazard magnitude increased in northern and southern regions.
Additionally, weights from the AHP meted provide a capability to show the susceptibility to hazards in each area similar to that of the Laos national report.
Therefore, the AHP method is reliable and can be adopted for integrated hazard maps.
This study provides an important and reliable methodology for the development of integrated hazard maps using multi-criteria decision analysis, such as AHP. Furthermore, exposure data on population and economic losses in a hazard area can provide more detail and improve the results.