The study focuses on the Nong Peung wetland, located in the Pakxan district of Bolikhamxay province, central Laos (Figure 1). This area extends from 18˚20'30.155''N to 18˚25'21.162''N latitude and 103˚38'9.033''E to 103˚44'18.773''E longitude. The wetland covers approximately 63.85 Km², at an elevation of 140 - 184 meters above sea level and encompasses 14 villages.

The region’s hydro-climatology is dominated by a strong monsoon system that arrives in late July, leading to prolonged periods of intense rainfall. This precipitation frequently causes local rivers to exceed their bankfull capacity, resulting in widespread inundation of adjacent low-lying areas [7]. The specific topography of Pakxan town, combined with the hydrology of the Nong Peung wetland, makes the urban center particularly vulnerable to flooding.

The Nong Peung river system, which includes a network of lakes, ponds, canals, and marshes, supports a diverse ecosystem of aquatic and terrestrial species. It also serves as a critical water resource for surrounding communities, ultimately flowing into the Nam San and Mekong Rivers. Given the history of significant flood events in the area, a comprehensive flood risk assessment is essential for developing strategies to mitigate potential damage to local communities and infrastructure.

Figure 1. Study area Nong Peung Wetland in Pakxan District, Bolikhamxay Province, Laos.

The methodology for this study was contingent upon the acquisition and processing of multiple geospatial datasets to model urban flood risk. Nine parameters, identified as critical indicators of flood susceptibility through a comprehensive literature review, were derived from various sources (summarized in Table 1). The core datasets included:

Topography: A Digital Elevation Model (DEM) with a 30-meter spatial resolution was obtained from the USGS Earth Explorer. This was instrumental in generating essential topographic parameters, including elevation (El), slope (Sl), and the Topographic Wetness Index (TWI).

Soil: Soil type (ST) information was acquired from the Laos Digital Soil Map, a high-resolution (10 m) ESRI Shapefile provided by the Lao Knowledge for Development (https://en.data.k4d.la). Soil characteristics are a key determinant of surface runoff and infiltration rates.

Land Use/Land Cover (LULC): A high-resolution LULC map for 2023 was sourced from the Esri Land Cover collection within the ArcGIS Living Atlas. Derived from Sentinel-2A imagery, this dataset enabled the analysis of land use patterns, which significantly influence flood dynamics.

Vegetation: The Normalized Difference Vegetation Index (NDVI), a metric for vegetation health, was calculated from Sentinel-2A satellite imagery provided by the EU’s Copernicus program (https://www.copernicus.eu/en), acquired 24 April 2023.

Hydrology: Key hydrological factors, specifically distance to river (DR) and drainage density (DD), were derived from the DEM and existing hydrographic data.

Precipitation: Rainfall (Rf) data, a primary driver of flooding, was sourced from the high-resolution monthly and annual datasets provided by the Climate Research Unit, accessed via NASA POWER (https://www.earthdata.nasa.gov).

Table 1. Data types and brief description.

The nine parameters selected for this study—encompassing critical geomorphological, hydrological, and land cover attributes—were chosen based on their established, direct influence on inundation dynamics. These factors (e.g., elevation, slope, rainfall, LULC) are widely validated in flood susceptibility literature as primary drivers of surface runoff and water accumulation, particularly within tropical, monsoon-driven wetland systems similar to the Nong Peung.

All geospatial data were processed and analyzed using ArcGIS 10.8 to generate the nine thematic factor maps required for the Multi-Criteria Decision Analysis (MCDA). Topographic and Hydrological Derivatives: The DEM served as the primary source for several factors. It was first hydrologically corrected by filling sinks, from which flow direction and flow accumulation rasters were derived using the Hydrology toolset. The slope and curvature maps were calculated directly from the DEM. Rainfall: To create a continuous rainfall map, mean annual precipitation data from 14 local and regional meteorology stations (https://www.earthdata.nasa.gov) were interpolated using the Inverse Distance Weighted (IDW) method. The resulting raster was clipped to the study area boundary. Drainage and Proximity: The drainage network was delineated from the flow accumulation raster, and this was used to compute the drainage density map via the Line Density tool. The distance to river map was generated by applying the Euclidean Distance tool to a river network layer sourced from the Lao Knowledge for Development (https://en.data.k4d.la). Topographic Wetness Index (TWI): The TWI map was generated using the Raster Calculator tool, applying the standard formula proposed by Moore et al. (1991) (Equation (1)).

where W represents the topographic wetness index, A_s represents the cumulative upslope area draining through a point (per unit contour length), and B represents the local slope angle in degrees.

To create the Land Use and Land Cover (LULC) factor map, we first extracted the LULC map of the district from the land use and land cover data available on the ESRI website. Next, we used ArcGIS software to assign appropriate names to each LULC class, as the downloaded map represents these classes by numbers.

To create the soil type factor map, we first identified the soil types of the watershed. This was done by extracting data using ArcGIS 10.8 software from the Digital Soil Map of Laos, which we obtained from the Lao Knowledge for Development. Once the soil types were extracted as a vector soil map, we converted it into raster format.

The NDVI map for the district was created using a Sentinel 2A satellite image, which was downloaded from the EU’s Copernicus website. The map was generated by applying the equation provided below (Equation (2)) with ArcGIS 10.8 software.

The NDVI, or Normalized Difference Vegetation Index, is calculated using surface spectral reflectance values. In this context, NIR refers to the reflectance in the near-infrared band, which corresponds to band 8 in the Sentinel 2A satellite imagery. RED refers to the reflectance in the red band, which is band 4 in the same satellite imagery.

After preparing all flood-controlling factors in raster format, we reclassified these factors into five common measurement scales ranging from 1 (very low susceptibility to flooding) to 5 (very high susceptibility to flooding). The data was rescaled to a consistent spatial resolution of 10 meters. A higher classification value (5) indicates areas that are more susceptible to flooding, while a lower value (1) corresponds to areas that are less susceptible. Since there is no standard reclassification scale for flood-controlling factors, the classification of all factors was determined based on previous studies and the local context of the study area.

Map Standardization and Flood Risk Modeling: Following their generation, all nine factor maps (El, Sl, Rf, DR, DD, TWI, LULC, ST, and NDVI) were standardized. Each map was reclassified into five flood susceptibility categories: Very Low (1), Low (2), Moderate (3), High (4), and Very High (5). The area corresponding to each susceptibility level for every criterion is detailed in Table 2. All raster layers were subsequently resampled to a uniform spatial resolution of 10 meters. Finally, a weighted overlay analysis, integrating the factor weights derived from the Analytical Hierarchy Process (AHP), was performed to produce the final flood risk map for the Nong Peung wetland.

The relative weights of the eleven factors influencing flood susceptibility in the Nong Peung wetland were derived using the Analytical Hierarchy Process (AHP). AHP is a multi-criterion decision-making (MCDM) method that systematically quantifies the relative importance of multiple criteria [1]-[4][11][12][16]. The process involved three stages. First, a pairwise comparison matrix was constructed using Saaty’s 1-to-9 scale to score the relative importance of one factor over another. Second, this matrix was normalized by dividing each element by its column total. Third, the final criteria weights were determined by calculating the row-wise average of the normalized matrix. To validate the judgments, the consistency of the matrix was evaluated. This required the calculation of the Consistency Index (CI) (Equation (3)) and the Consistency Ratio (CR) (Equation (4)).

where n is the number of factors (11), λ max is the principal eigenvalue, and RI is the standard Random Index for n = 11 . A CR value below the 0.10 threshold signifies that the matrix judgments are consistent and acceptable (Saaty 1987).

To ensure a transparent weighting process, the pairwise comparison judgments were derived from a hybrid approach. This involved: 1) a comprehensive review of established flood susceptibility literature to determine the general importance of factors, and 2) a consensus of the authors’ expert judgment to adapt these principles to the specific hydrological context of the Nong Peung wetland.

After preparing and reclassifying each flood-controlling factor to a common measurement scale ranging from 1 (very low) to 5 (very high) using ArcGIS software, we then applied the Analytic Hierarchy Process (AHP) to weight these factors. Subsequently, the spatial layers were integrated and overlaid together in the Spatial Analyst Extension of the ArcGIS environment using the weighted overlay technique. We employed the equation (Equation (5)) presented below to derive the flood susceptibility map of the study area. Several previous studies [1]-[5][11][12][15]-[17] have utilized this equation (Equation (5)) to generate flood susceptibility maps.

In this context, (FS) represents flood susceptibility, (n) denotes the number of decision criteria, (x_i) refers to a specific normalized criterion, and (w_i) indicates the corresponding weight assigned to that criterion. The values of each cell or pixel in the raster layers are multiplied by their respective weights, which are derived from AHP (Analytic Hierarchy Process) analysis. These weighted values are then summed together to produce the flood susceptibility output raster map.

To assess the predictive accuracy of the generated Flood Susceptibility Map (FSM), a validation process was conducted using an independent dataset of known flood locations. A flood inventory map was prepared comprising 45 “Flood plain samples”. These locations were identified based on historical inundation records obtained from the Pakxan District Agricultural and Forestry Office (DAFO) and supplemented by field-based GPS surveys conducted in 2023.

The validation was performed by spatially correlating this inventory of known flood points with the five susceptibility classes (Very Low to Very High) of the final FSM. The model’s reliability is confirmed by the high concentration of these validation points falling within the areas classified as ‘High’ and ‘Very High’ susceptibility. Furthermore, to quantitatively measure the model’s performance, the Receiver Operating Characteristic (ROC) curve was employed [18]. The Area Under the Curve (AUC) was calculated, which provides a robust measure of the model’s ability to correctly distinguish between flood-prone and non-flood-prone areas.

The spatial analysis identified the distribution and characteristics of the nine factors conditioning flood susceptibility across the Nong Pueng wetland. The reclassified thematic maps revealed significant spatial heterogeneity, with each factor contributing differently to the overall flood risk. A detailed breakdown of each factor’s contribution is provided below, with quantitative data on area and percentage coverage presented in Table 2.

Elevation: The topography of the study area ranges from 140 m to 184 m. The reclassified Digital Elevation Model (DEM) shows a clear distinction between higher-elevation upstream areas and low-lying downstream plains (Figure 2(a)). Areas classified with ‘Very High’ susceptibility (140 - 151 m) are concentrated in these downstream regions, which are prone to water accumulation. Conversely, the ‘Very Low’ susceptibility class (169 - 184 m) corresponds to the highest ground, primarily in the upper catchment.

Slope: The slope analysis indicates that a substantial portion of the watershed is characterized by gentle terrain, which facilitates water stagnation. A total of 30.56% of the area (19.22 ha) has a slope of 0˚ - 2.5˚, falling into the ‘Very High’ susceptibility class (Figure 2(b)). These flat areas are predominantly located in the northeastern, central, and southwestern parts of the study area. In contrast, areas with low to very low susceptibility (slope > 8.7˚) constitute only a small fraction of the landscape (8.54%).

Rainfall: Mean annual rainfall varies considerably, from 955 mm to 2380 mm. The ‘High’ (1830 - 2060 mm) and ‘Very High’ (2070 - 2380 mm) susceptibility classes collectively cover nearly half of the study area (47.33%). Spatially, the western parts of the region receive the highest precipitation volumes, rendering them more vulnerable to rainfall-induced flooding compared to the eastern parts (Figure 2(c)).

Distance from river: Proximity to the river network is a critical determinant of flood risk. The analysis shows that areas within 130 m of a river channel are classified as ‘Very Highly’ susceptible to inundation. These high-risk zones are concentrated along the primary river corridors and their immediate floodplains (Figure 2(d)). As distance from the river increases, the susceptibility level decreases, with areas over 620 m away classified as having ‘Very Low’ susceptibility.

Table 2. Flood conditioning factors, their classification, rating values, area coverage, and percentage.

Drainage Density: The highest drainage densities (18 - 30 km/km²) correlated with ‘Very High’ susceptibility, impacting 2.58% of the area (Figure 3(a)). The lowest densities (0 - 2.24 km/km²) had ‘Very Low’ susceptibility and covered the majority of the study area (59.74%).

Topographic Wetness Index (TWI): Areas with the highest TWI values (3.11 - 9.23), indicating a high potential to accumulate water, corresponded to ‘Very High’ susceptibility (3.25% of area). The lowest TWI values (−7.07 - −4.06) were rated ‘Very Low’ (42.26%) (Figure 3(b)).

Land Use and Land Cover (LULC): ‘Waterbody’ was identified as the ‘Very High’ susceptibility class, though it comprised only 0.01% of the area. ‘Bare ground’ was classified as ‘Very Low’ susceptibility (9.16%) (Figure 3(c)).

(a) (b)

(c) (d)

Figure 2. (a) Elevation; (b) Slope; (c) Rainfall; (d) Distance from river maps.

NDVI: The Normalized Difference Vegetation Index (NDVI) results showed that the lowest values (−0.6 - 0.068), indicative of sparse vegetation or water, had ‘Very High’ susceptibility (19.53% of area) (Figure 4(a)). Conversely, the highest NDVI values (0.41 - 0.58), representing dense vegetation, had ‘Very Low’ susceptibility (6.6%).

(a) (b)

(c) (d)

Figure 3. (a) Drainage density; (b) Topography witness index; (c) Land use land cover; (d) Soil type.

Soil Type: The (CMe)/(CMg) soil class showed ‘Very High’ susceptibility and was the most dominant high-susceptibility class, covering 35.81% of the study area. The (ACf)/(LVg) soil type was rated ‘Very Low’ (1.15%) (Figure 3(d)). It seems the highest flood susceptibility was consistently associated with low elevations, flat slopes, proximity to rivers, high rainfall, high drainage density, high TWI values, water bodies, (CMe)/(CMg) soil types, and low NDVI values.

(a) (b)

Figure 4. (a) Normalize different vegetation index (NDVI), (b) Villages around Nong-Peung wetland.

The final weights for the nine flood-controlling factors, derived from the AHP pairwise comparison (Tables 3-5), quantify their relative influence on flood susceptibility. The results indicate that geomorphological characteristics are the primary drivers of flooding in the study area. Slope (24.35%) and elevation (21.92%) received the highest weights, collectively accounting for over 46% of the model’s influence. This suggests that low-lying, flat terrain is the most critical condition for inundation. Hydrological factors, including annual rainfall (15.56%) and distance from rivers (13.83%), also showed significant influence. In contrast, surface and soil characteristics such as LULC (4.36%), soil type (2.67%), and NDVI (2.49%) had the lowest relative importance. A consistency check (Table 5) was performed to validate the AHP matrix. The resulting Consistency Ratio (CR) was 0.074 (7.4%), derived from a Consistency Index (CI) of 0.11 and a Random Index (RI) of 1.46 for n = 9. As this CR value is well below the 0.10 threshold stipulated by Saaty (1987), the pairwise judgments were deemed consistent and reliable for use in the weighted overlay.4.2. Flood Susceptibility Map (FSM) Analysis. The final Flood Susceptibility Map (FSM) was generated by integrating the nine thematic factor maps using the AHP-derived weights. The FSM categorizes the Nong Peung wetland into five susceptibility classes, with their areal distribution detailed in Table 5. A key finding is that a majority of the study area, 61.4% (totaling 39.22 Km²), is classified as having ‘Moderate’ (27.56%), ‘High’ (20.06%), or ‘Very High’ (13.80%) susceptibility to flooding. The remaining 38.58% of the area exhibits ‘Low’ (24.84%) or ‘Very Low’ (13.74%) susceptibility. Spatially, the FSM (Figure 5) reveals that the most vulnerable areas are concentrated in the northern, western, central, and southern regions of the district. This spatial pattern is logical, as these areas correspond to the low-lying (low elevation) and flat (low slope) regions, which were the two most heavily weighted factors in the model. This confluence of high-risk geomorphology and proximity to drainage networks creates significant susceptibility to inundation.

Table 3. Estimated areas of flood susceptibility within the study area.

The results indicate that several areas are particularly vulnerable to flooding (Table 3). These include: B. Thasykhai, B. Nachick, B. Sysa ath, and B. Phonehsa ath in the northern part; B. Syvilai, B. Myxai, B. Symoungkhoun, B. Phosy, B. Phoxai, B. Pakxanhneua, and B. Pakxantai in the western part; B. Phonhsa ath also in the central part; B. Nasommor in the eastern part and B. Pakpeung in the southern part. These areas were identified as the most vulnerable to flooding within the study area. The analysis of flood susceptibility, as shown in (Figure 5), reveals that the areas of B. Sysaath, B. Phonhsath, B. Pakxanhneua, B. Pakxanhtai, and B. Pakpeung, located in the western, central, and southern parts of the district, are the most vulnerable to flooding.