Flood Susceptibility Assessment in the Anambra River Watershed Using the LOPCOW Weighting Technique

Full text

Text extracted from the article PDF for reading and search. Tables, figures and equations are in the PDF, which remains the authoritative version.

1. Introduction

Flooding is among the most consequential environmental hazards because it affects settlements, infrastructure, agriculture, and public services across river basins and rapidly growing urban areas. Flood susceptibility mapping translates terrain, hydrological, climatic, and soil characteristics into spatial information that can support land-use planning and disaster preparedness. Geographic information systems (GIS) and remote sensing enable these factors to be processed on a common grid and interpreted together, while multicriteria decision analysis provides a transparent framework for combining variables measured on different scales (Malczewski, 2006; Kazakis et al., 2015; Islam et al., 2025) [13, 14, 16]. Recent reviews describe an expanding range of GIS, remote-sensing, machine-learning, and hybrid approaches, with increasing attention to factor selection, spatial scale, and interpretability (Sen & Das, 2026) [22]. Nigeria’s riverine and urban landscapes experience recurrent flooding under the combined influence of intense rainfall, low-lying terrain, channel proximity, drainage conditions, soil properties, and land-use change. In southeastern Nigeria, Anambra State includes both elevated terrain and broad lowland areas connected to the Anambra River and the lower Niger system. Local studies have applied GIS-based multicriteria methods to flood susceptibility in northern Anambra and Ayamelum, demonstrating the value of spatially explicit assessment for the state (Igboanugo et al., 2023; Okeke & Ekemezie, 2026) [11, 19]. Research in the lower Niger basin around Onitsha likewise documents flood susceptibility as a regional planning concern (Chinedu et al., 2024) [3].

The physical controls represented in this study describe complementary parts of flood generation and accumulation. Elevation and HAND distinguish low-lying terrain from higher ground and express relative height above drainage pathways. Slope influences runoff velocity and the opportunity for surface water to accumulate. TWI represents the tendency for topographic convergence and soil wetness, while distance to major rivers indicates proximity to channel overflow corridors. These terrain and drainage indicators are widely used in index-based flood mapping because they represent the spatial structure of surface flow and storage (Beven & Kirkby, 1979; Nobre et al., 2011; Sørensen et al., 2006; Kazakis et al., 2015) [2, 14, 18, 23]. Rainfall supplies water to the drainage system, drainage density describes the concentration of channel pathways, and soil clay content provides information about texture-related infiltration behavior. Their combined spatial pattern can differentiate areas where rainfall input meets low relief, limited drainage capacity, and near-channel position. The eight-factor design therefore combines terrain, channel proximity, drainage structure, wetness, rainfall, and soil texture in a single spatial assessment. Comparable Nigerian studies have used related combinations of terrain and hydroclimatic indicators, while recent national and international work continues to expand the range of data-driven and GIS-based approaches (Chinedu et al., 2024; Islam et al., 2025; Montien Tique et al., 2026; Isioye et al., 2026) [3, 12, 13, 17].

A multicriteria model requires each factor to be normalized to a common scale and assigned a weight before the normalized layers are combined. Expert-based methods such as the analytic hierarchy process are common in flood mapping, while objective methods derive weights from the distributions of the input criteria. LOPCOW uses the relationship between normalized root-mean-square values and standard deviations to calculate criterion weights and has been applied as a distribution-driven weighting procedure in multicriteria assessment (Ecer & Pamucar, 2022) [5]. Its use here makes the weighting calculation reproducible from the raster-cell matrix and allows the relative influence of each factor to be reported alongside the susceptibility classes. Despite the growth of GIS-based and data-driven flood studies in Nigeria, a watershed-focused assessment that combines the eight specified terrain, drainage, rainfall, and soil factors with LOPCOW provides a useful spatial contribution for the Anambra River corridor. The analysis links factor distributions, objective weights, continuous index values, five susceptibility classes, and ward-level summaries. This integrated presentation makes the spatial pattern and its relationship to the selected criteria clear for watershed planning and prioritization.

This study assesses flood susceptibility in the Anambra River watershed using eight conditioning factors and the LOPCOW objective-weighting method. The objectives are to describe the spatial distributions of the factors, calculate their LOPCOW weights, generate and classify a continuous susceptibility index, and summarize susceptibility by watershed-clipped ward portions.

2. Materials and Methods

2.1. Study Area

Anambra State is located in southeastern Nigeria on the eastern plains of the River Niger. Its landscape includes rolling lowlands, riverine plains, and localized hills, while the humid tropical climate supports a marked rainy season and substantial surface-water input. The Anambra River, also known as the Ọmambala, is a major tributary of the Niger and provides the principal drainage corridor relevant to this assessment (Anambra State Government, n.d.; Igboanugo et al., 2023) [11].

The study focuses on the Anambra River watershed in Anambra State, where low-lying riparian plains meet gently rolling uplands. The watershed forms part of the lower Niger drainage system, making local relief, river proximity, and drainage pathways central to flood susceptibility (Anambra State Government, 2026; Chinedu et al., 2024; Igboanugo et al., 2023) [1, 3, 11].

2.2. Data Inventory and Provenance

The analysis used eight single-band GeoTIFF factor rasters, a watershed polygon, and GRID3 ward polygons for spatial summarization (GRID3, n.d.). Table 1 lists the source attribution and spatial support of the datasets, while Table 2 records the original raster grids, cell sizes, data types, extents, and NoData values.

Table 1: Spatial datasets and source attribution Dataset / format Source attribution Native source support Elevation / GeoTIFF SRTM elevation; USGS EarthExplorer Approximately 30 m Slope / GeoTIFF SRTM-derived terrain gradient Approximately 30 m HAND / GeoTIFF Terrain and drainage-network factor 30 m River distance / GeoTIFF HydroSHEDS river network Hydrographic network Drainage density / GeoTIFF HydroSHEDS drainage network Hydrographic network TWI / GeoTIFF DEM-derived wetness factor Terrain support Rainfall / GeoTIFF CHIRPS rainfall product 0.05° native grid Clay / GeoTIFF ISRIC SoilGrids clay layer 250 m native grid Wards / ESRI shapefile GRID3 operational wards Vector polygons Basin boundary / ESRI shapefile Anambra River watershed polygon Vector polygon Sources for the principal elevation, rainfall, soil, hydrography, and ward products are documented by Farr et al. (2007), U.S. Geological Survey (2018), Funk et al. (2015), Hengl et al. (2017), Poggio et al. (2021), Lehner et al. (2008), and GRID3 (n.d.).

The elevation layer represents SRTM terrain information, and the slope layer represents terrain gradient. The SRTM mission provides elevation measurements for regional-scale terrain analysis (Farr et al., 2007; U.S. Geological Survey, 2018) [6, 24].

Rainfall was represented by a CHIRPS-derived raster. CHIRPS combines satellite observations with station data to provide gridded rainfall estimates suitable for regional spatial analysis (Funk et al., 2015; Climate Hazards Center, n.d.) [7]. The clay layer represents SoilGrids soil-texture information. SoilGrids provides global gridded soil properties at 250 m native support, including texture fractions, and its version 2.0 products document the global modeling framework (Hengl et al., 2017; Poggio et al., 2021) [10, 20]. The ward layer contains 107 unique GRID3 ward features. After intersection with the watershed footprint, the analysis summarizes the Anambra, Delta, and Enugu portions represented in the study polygon. Hydrographic layers provide the river-distance and drainage-density factors, while HAND and TWI are represented by their respective input rasters.

2.3. Spatial Harmonization and Treatment of Missing

Cells All raster layers were aligned in WGS 84 / UTM Zone 32N (EPSG:32632), a metric projected coordinate system appropriate for the study area. Elevation defined the snap grid, with 1,617 columns and 2,344 rows at 30 m cell spacing. Ward polygons were transformed from WGS 84 (EPSG:4326) to the analysis CRS and intersected with the watershed study polygon before area summaries were calculated.

Continuous factor rasters were resampled to the elevation grid using bilinear interpolation and clipped to the rasterized watershed footprint. The complete decision matrix contains 2,010,837 cells with valid values for all eight factors. This common analysis area provides the consistent spatial support for normalization, LOPCOW weighting, index calculation, and class-area summaries.

The complete-factor footprint covers 1,809.7533 km², equivalent to 96.79296% of the rasterized watershed boundary. The remaining 66,625 boundary cells, covering 59.9625 km², are recorded as unmapped in the coverage summaries and are excluded from the susceptibility index and class-area denominator.

2.4. Factor Definitions and Response Directions

Table 3 defines the eight conditioning factors and the susceptibility direction used in normalization. Factors for which lower values indicate greater susceptibility were treated as cost criteria; factors for which higher values indicate greater susceptibility were treated as benefit criteria. This orientation places all normalized factor values on a common scale where larger values indicate greater susceptibility.

Table 3: Factor definitions and response directions Factor Factor meaning / definition Response direction and normalization Elevation Absolute terrain height. Lower values → higher susceptibility; cost min–max Slope Terrain gradient, conventionally arctan(√(p² + q²)) × 180/π.

Lower values → higher local retention; cost min–max HAND Vertical height above the drainage cell reached along a hydrological flow path.

Lower values → lower relative height above drainage; cost min– max normalization.

River distance Planar distance to the Anambra River and major tributaries.

Lower values → greater proximity to potential river overflow; cost Drainage density Channel length per neighborhood area. Higher values → greater drainage-path concentration; benefit TWI ln(a / tan β), where a is specific contributing area and β is slope in radians.

Higher values → greater topographic wetness tendency; benefit Rainfall Spatial rainfall surface. Higher values → greater water input; benefit min–max Clay Soil clay content, expressed as a percentage. Higher values → greater potential for restricted matrix infiltration; benefit min–max normalization.

2.5. LOPCOW Weighting and Weighted Combination

LOPCOW weighting follows the formulation of Ecer and Pamucar (2022) [5]. The eight factor rasters were represented as a cell-by-factor decision matrix over the complete-factor footprint. Each criterion was normalized according to its response direction, distributional statistics were calculated across all eligible cells, and normalized weights were used in a weighted linear combination to generate the susceptibility index.

For each column, aⱼ = minᵢ(xᵢⱼ) and bⱼ = maxᵢ(xᵢⱼ) were computed only over the complete common mask. Benefit criteria were transformed by Equation 2 and cost criteria by Equation 3. Thus r = 1 consistently denotes the more susceptible end of each observed factor range.

Elevation, slope, HAND, and river distance were normalized as cost criteria; drainage density, TWI, rainfall, and clay were normalized as benefit criteria. Each factor had a nonzero range and standard deviation across the complete-factor footprint.

The population standard deviation was used because the matrix enumerates the full set of mapped cells. The logarithmic percentage value (PV) is computed from the ratio of the normalized root-mean-square value to the standard deviation. Each criterion weight is its PV divided by the sum of PV values across the eight criteria.

The normalized factors were combined additively using Equation 7 to produce a continuous flood-susceptibility index, with larger values representing greater relative susceptibility within the study area.

Because RMSⱼ² = μⱼ² + σⱼ², the LOPCOW measure can also be written as PVⱼ = 50 ln[1 + (μⱼ/σⱼ)²] when σⱼ > 0. The final weight therefore reflects both the normalized mean and dispersion of each criterion across the mapped cells.

2.6. Classification, Area Measurement, and Ward-Area

Overlay The continuous susceptibility index was divided into five equal-width intervals spanning its observed minimum and maximum. The class labels—very low, low, moderate, high, and very high—represent relative bands of the composite index. Cell counts were converted to area using the 30 m grid, and class percentages were calculated as a share of the complete-factor footprint.

Ward polygons were projected to EPSG:32632 and intersected with the watershed polygon. Each raster cell was assigned to the ward portion containing its center, and class areas and mapped coverage were summarized by ward. Two measures were calculated: the absolute area in high and very high classes and that area as a percentage of classified cells in each ward portion. Table 7 reports class intersections and dominant classes; Tables 8 and 9 provide ranked ward and state-level summaries.

2.7. Software and Reproducible Outputs

Spatial analysis used Python with NumPy, Rasterio/GDAL, GeoPandas, SciPy, Shapely, and Matplotlib. The processing sequence produced aligned factor rasters, LOPCOW calculations, a continuous index, five susceptibility classes, map figures, and ward-level area tables.

3. Results

3.1. Factor Distributions and Spatial Coverage

Table 4 summarizes the eight factors over the 2,010,837 complete-factor cells. Elevation ranges from 18 to 252 m, with a mean of 48.992 m and a median of 35 m. Mean slope is 2.088° and the median is 1.467°, indicating that low-gradient terrain occupies much of the mapped footprint. HAND has a mean of 18.308 m and a median of 6.987 m, while mean distance to a major river is 1,907.342 m. Rainfall values have a mean of 2,752.565 and clay content has a mean of 20.167.

Fig 2: Hydrographic, wetness, rainfall, and soil factors: (a) drainage density, (b) TWI, (c) rainfall, and (d) clay content. Panels are aligned to the common analysis grid.

Elevation, HAND, and river distance are expressed in metres; slope is in degrees and clay in percent. Rainfall and drainage density retain the numerical scales of their input rasters. Descriptive statistics use the common complete-factor cells. The factor maps show a broad low-relief western and northern tract and higher terrain toward the south-central and southeastern parts of the study area (Figure 1a–c). River-distance values increase away from the Anambra River and its major tributaries (Figure 1d). The mapped terrain and proximity patterns provide the spatial context for the factor statistics in Table 4.

Drainage-density values range from 0 to 1.4875, with a mean of 0.3580, and form localized concentrations along drainage corridors (Figure 2a). TWI ranges from 6.362 to 27.658, with a mean of 13.756 and a median of 13.785 (Figure 2b). Rainfall has a mean of 2,752.565, while clay content ranges from 11.958% to 28.840% (Figure 2c–d). These spatial distributions show how the hydrological, climatic, and soil criteria complement the terrain factors.

3.2. LOPCOW Intermediate Quantities and Factor

Weights Table 5 presents the normalized means, standard deviations, root-mean-square values, PV measures, and final LOPCOW weights. Slope has the highest PV (256.130) and weight (0.202264), followed by HAND (0.167934) and elevation (0.142859). The three terrain criteria together account for 51.31% of total weight. Drainage density has the smallest weight (0.034512), while river distance, clay, TWI, and rainfall contribute 12.28%, 11.71%, 10.85%, and 10.41%, respectively.

Fig 3: LOPCOW weights for the eight flood-conditioning factors, calculated from direction-normalized raster-cell values. Bars are ordered by weight; values are shown as percentages.

6.3. Susceptibility Index and Class Areas

The continuous susceptibility index ranges from 0.298647 to 0.890720, with a mean of 0.737525, median of 0.753121, and population standard deviation of 0.064465. Table 6 lists the class thresholds, cell counts, mapped areas, and percentages. High susceptibility covers 974.1528 km² (53.83%), and very high susceptibility covers 644.1579 km² (35.59%). Together, these classes occupy 1,618.3107 km², or 89.42% of the complete-factor area.

High and very high susceptibility occupy broad areas of the low-relief western and northern sectors, while moderate and lower classes occur more frequently in the higher southern interior (Figure 4). This pattern is consistent with the factor distributions in Figures 1 and 2 and the dominant weights assigned to slope, HAND, and elevation.

Fig 4: Flood-susceptibility classes produced from five equal intervals of the observed LOPCOW index. The mapped cells cover 1,809.7533 km² of the watershed footprint.

6.4. Ward-Area Exposure

The ward overlay comprises 107 watershed-clipped ward portions, of which 103 contain classified cells. Ninety-three portions contain some high or very high susceptibility, and 76 of the 103 mapped portions have at least half of their classified area in those categories. Table 7 reports class intersections and dominant classes; Figures 5 and 6 show the ward geography and the high-plus-very-high proportion by ward.

Ward counts are nonexclusive because one ward portion can intersect several susceptibility classes. Dominant class is the category with the largest mapped area. The four portions without classified cells are identified in the ward summaries.

Table 7: Ward intersections and dominant susceptibility classes Class Area (km²) Wards containing class Wards dominated by class Very low 1.3887 4 0 Low 16.2252 34 4 Moderate 173.8287 76 23 High 974.1528 91 53 Very high 644.1579 68 23 Unmapped area 59.9625 55 with unmapped cells 4 entirely unmapped Fig 5: GRID3 ward portions overlaid on the flood-susceptibility classes within the watershed study area. All 107 unique ward portions are shown.

Fig 6: High-plus-very-high susceptibility as a percentage of classified area in each watershed-clipped ward portion. Gray indicates portions without classified cells.

High or very high susceptibility occurs in 93 ward portions. In 76 of the 103 mapped portions (73.79%), these classes together cover at least 50% of classified area. Table 8 ranks the ten portions with the largest high-plus-very-high areas, led by Abegbu / Iyiora (91.3635 km²), Enugu-Otu (91.2870 km²), and Omasi (88.0389 km²).

Ward portions are ranked by combined high and very high area. Susceptibility percentages use classified cells in each portion, and mapped-coverage percentages use all rasterized boundary cells assigned to the portion. Table 8 lists the ten highest values.

Table 9 summarizes the state-labelled portions of the watershed. Anambra portions contain 1,800.9045 km² of classified cells, including 1,609.4628 km² in high and very high classes. Delta portions contribute 1.7766 km² of classified area, and Enugu portions contribute 7.0722 km², including 7.0713 km² in the two highest classes.

4. Discussion of Results

4.1. Hydrological Interpretation of the Mapped Pattern

The highest susceptibility class occurs in low terrain both in absolute elevation and relative height above drainage. Its mean HAND of 2.802 m and mean river distance of 844.742 m characterize locations near channel level and close to major rivers. These settings provide direct pathways for overbank flow and surface-water accumulation, consistent with the hydrological interpretation of HAND and river proximity in flood mapping (Nobre et al., 2011; Kazakis et al., 2015) [14, 18].

HAND complements absolute elevation by describing vertical position relative to the drainage network. In the study area, mean HAND falls from 192.722 m in the very low class to 2.802 m in the very high class, clearly separating upland positions from drainage-connected lowlands. The joint pattern of low elevation, gentle slope, and short river distance concentrates the highest index values along floodplain and river-corridor terrain.

Rainfall and terrain display distinct spatial gradients across the watershed. Higher rainfall values occur toward the south, while high susceptibility is most extensive across low-relief sectors, where elevation, slope, HAND, and river proximity jointly favor water accumulation. The class-conditioned means—2,940.028 in the very low class and 2,757.893 in the very high class—show that the composite pattern reflects the weighted interaction of all eight criteria rather than rainfall alone.

Clay contributes 11.71% of the LOPCOW weight and its mean increases from 17.571% in the very low class to 20.658% in the very high class. This pattern is consistent with the role of finer soil texture in reducing matrix infiltration potential, while soil structure and organic matter also influence hydraulic behavior (Saxton & Rawls, 2006) [21]. The clay layer therefore adds a meaningful soil-property dimension to the terrain and hydrological factors.

4.2. What the LOPCOW Weights and Class Proportions

Mean Slope receives the largest coefficient because its normalized values have a high mean relative to their dispersion across mapped cells. HAND and elevation also receive high weights and normalized means, creating a strong contribution from lowland terrain conditions. Drainage density receives a smaller coefficient because its normalized distribution has a lower mean and comparatively larger spread. These weights describe the contribution of each oriented raster distribution within the LOPCOW calculation.

The weighted sum retains high memberships for lowland factors across much of the footprint. Mean normalized memberships are 0.933133 for slope, 0.920043 for HAND, and 0.867555 for elevation. Their combined weight is 51.31%, and the five-class map places 89.42% of complete-factor area in the high and very high categories. This concentration defines the spatial prioritization pattern of the selected eight-factor model.

The mapped pattern complements flood studies elsewhere in Anambra and Nigeria. Northern Anambra and Ayamelum assessments have used different criteria and weighting strategies, while the lower Niger study around Onitsha and recent Nigerian assessments apply distinct factor sets and spatial methods (Igboanugo et al., 2023; Chinedu et al., 2024; Okeke & Ekemezie, 2026; Isioye et al., 2026; Montien Tique et al., 2026) [3, 11, 12, 17, 19]. The present analysis adds a watershed-scale LOPCOW assessment with explicit cell-level weights, index thresholds, and ward-area summaries.

4.3. Implications for Ward-Level Planning

The ward overlay identifies where the largest mapped high-plus-very-high areas occur. Abegbu / Iyiora, Enugu-Otu, Omasi, Eziagulu Otu, and Mmiata lead the ranked list of ward portions. These results support prioritization of drainage appraisal, floodplain development review, and local preparedness activities in the areas with the greatest mapped extent. Area and percentage summaries provide complementary measures of the size and concentration of susceptibility within each portion.

Absolute high-plus-very-high area highlights the portions with the largest spatial extent, while the within-portion percentage shows how much classified area falls in the two highest classes. Seventy-six mapped portions have at least half their classified area in high or very high susceptibility. Umudei / Inland Town 1, Ojjor, Olo 2, and Abagana 4 contain no classified cells in the common analysis mask. The full ward inventory in Table 8 and the corresponding exposure summary identify each portion and its mapped area.

5. Conclusion

The LOPCOW integration of eight conditioning factors produced a continuous flood-susceptibility index for 1,809.7533 km² of the Anambra River watershed. Slope received the largest objective weight (20.23%), followed by HAND (16.79%) and elevation (14.29%). Together, these terrain criteria contribute 51.31% of the total weight and define the model’s strong emphasis on low-relief, drainage-connected locations.

The index ranged from 0.298647 to 0.890720. Equal-width classification placed 974.1528 km² in the high class and 644.1579 km² in the very high class, together accounting for 89.42% of mapped area. The very high class is characterized by low mean elevation (27.596 m), low mean HAND (2.802 m), and gentle mean slope (0.959°). These quantitative patterns show that susceptibility is concentrated in low-lying terrain near the drainage network.

The ward summaries translate the continuous watershed map into locally interpretable priorities. Of 107 ward portions, 103 contain classified cells, 93 intersect high or very high susceptibility, and 76 have at least half of classified area in these classes. Abegbu / Iyiora, Enugu-Otu, and Omasi have the largest high-plus-very-high areas. The factor maps, weighting results, susceptibility zones, and ward summaries together provide a clear basis for targeted floodplain planning, drainage assessment, and preparedness across the Anambra River corridor.