ABSTRACT

Flood hazard mapping is essential for reducing flood risks by identifying vulnerable areas and guiding response strategies. However, obtaining high-resolution topographic data over large areas can be costly, and reducing this expense while maintaining accuracy is crucial for effective 2D hydraulic mapping. The limited availability of topographic data in river areas makes it difficult to delineate flood inundation. This research aims to investigate the best model for extending high-resolution Digital Elevation Model (DEM) data as input for flood hazard mapping. To enhance DEM resolution, the best model was selected by comparing Root Mean Square Error (RMSE) values from various interpolation methods, namely nearest neighbour interpolation (NNI), bicubic, and bilinear in spatial analysis. The model was then validated using 40 data points. Subsequently, the best model was used to generate a DEM for a larger scale as input for 2D modeling with HEC-RAS. The flood inundation model was validated twice: first, based on water level measurements at the Automatic Water Level Recorder (AWLR) Rowotamtu, and second, based on field observations of extreme flood events. The model results show that the best interpolation method for obtaining a high-resolution DEM is bilinear. The first flood data validation produces an RMSE value of 0.113, while the second is slightly higher at 0.316. Both results indicate good model accuracy, and the flood inundation maps generated can be used to demonstrate flood hazards.

1 INTRODUCTION

Floods are among the most frequent natural hazards that cause severe damage to infrastructure and loss of life. The importance of creating reliable flood hazard maps cannot be overstated, as they are essential for effective flood risk management, planning, and mitigation strategies (Dottori et al. 2016a). Several studies have used flood models to produce 1D and 2D flood hazard maps. However, there are significant challenges in developing these maps, particularly the need for high-resolution topographic data and an accurate model. Acquiring such accurate data on a watershed scale can be time-consuming and expensive (Arseni et al. 2020).

To overcome these challenges, recent studies have turned to remote sensing technologies, which offer topographic data over large areas (Peker et al. 2024). Additionally, advances in geographic information systems (GIS) and hydrological modeling enable the integration of multi-resolution data to enhance flood hazard mapping (Hutanu et al. 2020). Flood modeling, such as hydrology and hydraulic modeling, is very important in the development and application of riverbed hydraulic analysis using HEC-RAS, which is applicable for both natural and structural hydraulic calculations (Peker et al. 2024). Tools such as HEC-RAS 6.5, which integrates RAS-Mapper, are increasingly used for two-dimensional hydraulic simulations. RAS-Mapper allows the digitization of river features and spatial visualization of inundation extents, making it an essential component in modern flood studies. Costabile et al. (2020) and Pinos and Timbe (2019) demonstrated the use of RAS-Mapper in the Purna Watershed (Navsari, India), showing that water depth, flow velocity, and surface elevation could be derived and analyzed using 2D flow modeling. Similarly, Khan et al. (2020) highlighted the potential of such tools for early warning systems, flood risk communication, and infrastructure planning.

In cases where high-resolution data are unavailable, some researchers have explored data assimilation techniques that combine lower-resolution DEMs with local measurements to enhance topographic detail (Bates et al. 2003; Pathiraja et al. 2018). One key aspect of this process is the interpolation method used to increase DEM resolution. Interpolation techniques such as bilinear, bicubic, and nearest-neighbour (NNI) have been widely applied, each offering trade-offs between computational speed and elevation accuracy (Minh et al. 2024; Śliwiński et al. 2022). Bilinear interpolation is often applied in general mapping and topographic visualization, especially when the terrain has gradual or smooth elevation changes (Kandil et al. 2024; Yu et al. 2021). Although cubic interpolation offers smoother results and is suitable for capturing terrain details, it is slower and can introduce overshooting in areas with sharp elevation changes (Ghandehari et al. 2019; Minh et al. 2024; Riazanoff et al. 2024). On the other hand, the NNI is the simplest and fastest method, but it often produces maps with a rough surface because it does not perform interpolation between points (Chen and Li 2019). However, this simple interpolation method is less suitable for areas with complex topography, such as mountainous regions, where significant elevation variations require more advanced techniques like Kriging, which takes into account the spatial correlation between data points (Cosentino et al. 2023).

Despite the frequent use of interpolation methods for DEM enhancement, very few studies have investigated how these techniques influence flood hazard mapping outputs within a full 2D hydraulic modeling framework. Most previous works focus only on elevation accuracy, without assessing how interpolation uncertainties propagate into flood extent, depth, and hazard classifications. To address this gap, this study systematically evaluates the performance of three interpolation techniques (bilinear, bicubic, and NNI) in producing high-resolution DEMs and quantifies their impact on 2D flood simulations in the Bedadung watershed, East Java, Indonesia. The findings contribute new knowledge on the suitability of interpolation methods for flood hazard mapping in data-scarce tropical watersheds, providing practical guidance for improving flood risk management in regions where high-resolution elevation data are limited but flood risk is increasing. Furthermore, by integrating multi-resolution DEMs in a practical 2D flood modeling framework, the study demonstrates how flexible data integration strategies can enhance model applicability across varying terrain complexities and data availability conditions. (Hemmati et al. 2021; Nofal and van de Lindt 2021).

2 METHODS

2.1 Description of study area

This study was conducted in the Bedadung watershed, Jember Regency, East Java, Indonesia, covering an area of 125,307 ha. Geographically, the Bedadung watershed is located between 7°57'11.96” – 8°25'3.14” S and 113°26'1.93” – 114°01'13.44” E, which administratively includes the Regency of Jember that occupies 94.89% of the watershed area. The remaining area is comprised of Bondowoso Regency, which makes up 4.22%, and Probolinggo Regency, comprising 0.89%. The northern part of the Bedadung watershed borders the Deluwang watershed and the Sampean watershed, to the east it borders the Mayang watershed, and to the west it borders the Rondodringu watershed.

Topographically, the Bedadung watershed is characterized by gentle slopes (average slope is at least less than 20%) and narrow elevation range (its most upstream point is on Mt. Argapura, +3,088 msl). The region has a tropical monsoon climate with an annual rainfall of 942–4,915 mm, a rainy season from November to May, and a dry season from June to October. Figure 1 presents the study area.

Figure 1  Study area, Bedadung watershed.

2.2 Research methodology

The research activities are divided into four main stages as presented in Figure 2. The first stage, DEM generation and interpolation, aimed to produce a high-resolution DEM as the topographic foundation for hydraulic modeling. This ensured that terrain variations along the river corridor were adequately represented by integrating total station data with DEMNAS and applying the best-performing interpolation method validated by Ground Control Points (GCPs). The second stage, hydrological modeling, was designed to simulate the watershed response and generate design flood hydrographs for different return periods. Using HEC-HMS, the model was first calibrated with observed discharge data from the Rowotamtu AWLR station and then driven by design hyetographs derived from rainfall frequency analysis and transformed through the Alternating Block Method (ABM). The third stage, hydraulic modeling, focused on converting the design hydrographs into spatial inundation patterns using HEC-RAS 2D. This included defining 2D flow areas, optimizing mesh resolution, assigning Manning’s n values, and calibrating the model against observed flood events to ensure reliability. The final stage, flood hazard mapping, aimed to classify and visualize hazard levels based on flood depth and duration, thereby producing hazard maps that support risk assessment, land-use planning, and flood mitigation strategies within the Bedadung watershed.

Figure 2  Research framework.

Figure 3 shows the location of flood depth measurements collected during the historical flood event on January 21, 2021. A total of 70 observation points were surveyed along the inundation boundary to illustrate the spatial variation in flood depth in the affected areas. The recorded inundation depth ranges from 0 to 2.0 m, with greater depths generally found in low-topographic areas. The data from this field observation is used as reference data for hydraulic model validation, as well as to evaluate the accuracy level of inundation area and flood depth simulations. 

Figure 3  Location of flood depth measurements for the historical flood on 21 Jan 2021.

2.3 Data source

Both primary and secondary data were used in the analysis. The primary data was obtained by conducting interviews with informants regarding the existing conditions of the research, location surveys, as well as secondary data, as listed in Table 1.

Table 1  Research data.

Data	Data Source	Unit Resolution
DEM	Indonesia Geospatial Portal DEMNAS (indonesia.go.id)	8.25 m × 8.25 m
Land use	Sentinel-2 Imagery https://www.arcgis.com/home/item.html?id=cfcb7609de5f478eb7666240902d4d3d	10 m × 10 m
River geometry	Public Works Agency of East Java Province
Rainfall data	Bondoyudo Baru River Basin, Water Resources	daily
Discharge data	Management Unit, East Java Public Works Agency, Indonesia	daily
HSG	FAO-UNESCO 2007 https://doi.org/10.3334/ORNLDAAC/1566	cell size: 30 m

The data related to the extent and distribution of floods, obtained from the Regional Disaster Management Agency (Badan Penanggulangan Bencana Daerah, BPBD) and interviews with local people, were used to validate the HEC-RAS modeling results.

2.4 DEM generation and interpolation

This stage involved the preparation of high-resolution topographic data. The process began with total station measurements and DEMNAS (8.25 m) data. Data preprocessing included format conversion and resolution adjustment, including the conversion of point data from a total station into raster format (TIFF) using Civil 3D. Three interpolation methods, NNI, Bilinear, and Bicubic, were tested using 40 Ground Control Points (GCPs), and the best-performing method was selected. The selected interpolation was then applied to expand coverage along the left and right riverbanks. Finally, DEMNAS data were resampled in ArcGIS to achieve a 1 m grid resolution suitable for detailed hydraulic modeling.

Since interpolation played a critical role in determining the quality of the generated DEM, the three methods tested are briefly described here. NNI is the simplest and fastest method, but it usually produces rough and less accurate results because it does not perform interpolation between points (Maleika 2024; Xu et al. 2023). The way this method works is that the new pixel takes the value from the nearest pixel on the original grid, as shown in Figure 4. No interpolation is used. The value of the new pixel is the value of the spatially closest pixel, as given in Equation 1:

(1)

Where:

V (x', y')	=	new pixel value in (x', y') coordinate, and
round (x), round (y)	=	surrounding pixel from the original grid.

Figure 4  Nearest neighbor interpolation grid.

Bilinear interpolation is often used for topography with smooth elevation changes. Its interpolation offers a balance between speed and accuracy. However, its accuracy decreases in areas with sharp elevation variations (Ghandehari et al. 2019). As a description, the bilinear interpolation grid scheme is shown in Figure 5.

Figure 5  Bilinear interpolation grid.

The way this method works is that the value of the new pixel is calculated by performing linear interpolation in two directions (horizontal and vertical) using the values from the four nearest neighbours. If the new pixel is at position (x', y'), the value of the new pixel V (x', y') is calculated using Equation 2:

(2)

Where:

x and y	=	row and column coordinates of the known points in the input image, respectively, and
u and v	=	distance difference between the required point and the known point, row and column coordinates, respectively.

The Bicubic Interpolation method is suitable for capturing finer terrain details and is usually used when high resolution is required. This method works by employing bicubic interpolation using the values of the 16 closest neighboring pixels arranged in a 4×4 pixel block, as shown in Figure 6. The interpolation is performed on both axes (x and y). Equation 3, for bicubic interpolation, is relatively more complex and is dependent upon the cubic basis functions used:

Figure 6  Bicubic interpolation grid.

(3)

Where:

ωi(x) and ωj(y)

=

cubic basis functions used for interpolation, which depend on the distance of the new pixel from the nearest neighbours.

2.5 Hydrological analysis

Hydrological modeling was performed using the HEC-HMS model. Inputs included land use maps, Hydrologic Soil Group (HSG) data, and observed rainfall records. Curve Numbers (CN) were calculated, and area rainfall was derived using Thiessen polygons. The model was first calibrated and validated using observed discharge data from the Rowotamtu AWLR station, ensuring that the simulated hydrographs accurately reproduced the measured flood events. For design simulations, rainfall frequency analysis was conducted to obtain design rainfall depths for 5-, 10-, and 25-year return periods. These rainfall depths were converted into design hyetographs using the Alternating Block Method (ABM), which were then applied in the calibrated HEC-HMS model to generate the corresponding design flood hydrographs.

2.6 HEC-RAS model

HEC-RAS 2D provides significant advantages over 1D simulations when DEM data are available, as it can represent spatial flood inundation more effectively (Hamdi et al. 2019). The output of this model can describe the depth of flood inundation, flow speed, and flood arrival time (Yalcin 2019). The 2D equations are given in Equation 4 regarding volume conservation, and Equation 5 regarding momentum conservation, governed by the Saint-Venant Equation:

(4)

(5)

Where:

A	=	flow area,
x	=	distance of flow path,
Q	=	discharge per unit length of channel,
g	=	gravitational acceleration,
t	=	time,
y	=	flow depth,
So	=	base slope, and
Sf	=	friction slope.

HEC-RAS 2D facilitates the analysis of inundation area mapping for selected flood return periods (Aryal et al. 2020; El-Naqa and Jaber 2018). Extensive high-resolution topographic data is essential for extensive flood hazard mapping using HEC-RAS to accurately represent flood inundation. However, high resolution topographic data is not easy to obtain and is expensive. To overcome the limitations of high-resolution topographic data as input for 2D models, the integration of topographic data from the watershed with DEM data is the latest advancement in this field. 

Data geometry modeling involves several components, including the creation of 2D flow areas, 2D area break lines, boundary condition lines, and the entry of Manning's coefficient values. Creating a 2D flow area begins by drawing a 2D area on the map using the polygon tool to define the boundaries of the two-dimensional flow. After drawing the 2D area, the grid mesh size was determined based on the resolution and detail requirements. In this model, multiple mesh grid sizes have been tested, including 30×30 m, 40×40 m, and 50×50 m, along with its perimeter area. The time step was determined by ensuring the Courant number remained below 1.0 to maintain numerical stability and reliability. After that, the 2D area break lines are drawn. The break lines in the 2D area are drawn by following topographic features along the river channel. The break line cell spacing values used are 25 m (minimum) and 45 m (maximum). The boundary condition lines on the map were delineated to indicate the water entry and exit points from the model area. The boundary condition types used are flow hydrograph and normal depth for upstream and downstream areas, respectively. After modeling the boundary condition lines, the relevant Manning's coefficient values are assigned for each segment or zone in the 2D area and river cross-section. The DEM results from geometric modeling from HEC-RAS are shown in Figure 7.

Figure 7  DEM results from assimilation of geometric modeling results.

2.7 Flood hazard mapping

Flood hazard maps can be used as an effective tool to minimize damage by building appropriate flood protection, warning, and evacuation systems (Kadri and Kurniyaningrum 2019). Flood hazard maps show the possibility and magnitude of flooding events in an area, which supports the creation of flood risk maps. Flood risk assessment maps show the potential consequences of a flood event in terms of affected population and assets, and the estimated economic damage. Such information is crucial for improving preparedness, land use planning, and management in flood-prone areas (Guerriero et al. 2020). One of the most widely used hydraulic/hydrodynamic models for flood hazard mapping is HEC-RAS 2D (Ongdas et al. 2020).

The flood hazard for individual flooding parameters based on the level of flood depth, hazard category, and flood duration is shown in Table 2.

Table 2  Hazard level for individual flooding parameters.

Depth of flooding (m)	Hazard category	Flood duration (hr)
0–0.2	Very Low	0–25
0.2–0.6	Low	25–50
0.6–1.5	Medium	50–100
1.5–3.5	High	100–175
> 3.5	Very High	>175

Source : Mani Murali et al. (2013).

2.8 Model performance

The model's performance was assessed by comparing the observed and simulated water level values during the calibration stage. The statistical parameters used for evaluating the model performance are Root Mean Square Error (RMSE), Coefficient of Determination (R²), and Mean Squared Error (MSE). These parameters are defined as follows:

RMSE	=	average difference between the predicted water level (or inundation height) from the model and the observed values. RMSE is particularly useful in showing the average expected error in the model predictions. The smaller the RMSE value, the better the model performs,
R²	=	quantifies the proportion of the variance in the observed data that is predictable from the model. An R² value close to 1 indicates a strong correlation between the model's predictions and the observed data, implying that the model can explain a large portion of the variability in water level or flood inundation, and
MSE	=	similar to RMSE, but without taking the square root. It represents the average of the squared differences between model predictions and actual observations. Like RMSE, lower MSE values indicate a model with better predictive performance.

3 RESULTS AND DISCUSSION

3.1 Selection of the best interpolation mode

The Wilcoxon test results indicate no statistically significant difference between the bilinear and bicubic interpolation methods (p > 0.05), suggesting comparable performance. In contrast, both the paired samples, t-test and Wilcoxon test, show that the Nearest Neighbor Interpolation (NNI) method yields significantly higher errors than bilinear (p = 0.023 and p = 0.018) and bicubic (p = 0.046 and p = 0.034). These findings confirm that bilinear interpolation performs significantly better than NNI, and is statistically comparable to bicubic.

Table 3 further supports this conclusion, with the lowest RMSE value obtained by bilinear (5.462) compared to NNI (5.533) and bicubic (5.486), consistent with field validation at 40 survey points. Consequently, bilinear interpolation can be considered the most accurate and efficient approach for this dataset, and the DEM generated from this method becomes the optimal input for HEC-RAS modeling, as shown in Figures 8 and 9. Field measurements also show variations that closely match the bilinear model, reinforcing the accuracy of the terrain data. These results are consistent with previous studies. Skala et al. (2020), emphasized that efficient interpolation methods are essential for big data processing. Shahabi (2021) also reported that bilinear and bicubic interpolation outperform NNI in flood risk estimation. Furthermore, although cubic spline interpolation can produce smooth results (Belias et al. 2022; Shafiq et al. 2024), its accuracy is not always superior (Zhang et al. 2023). Our findings underscore the significance of selecting an appropriate interpolation approach, such as bilinear, to enhance model accuracy and establish a robust foundation for more effective flood risk analysis and mitigation strategies.

Table 3  RMSE values for NNI, bilinear, and bicubic methods.

Assimilation method	RMSE value
NNI	5.533
Bilinear	5.462
Cubic	5.486

Figure 8  Measurement points for the elevation survey.

Figure 9 DEM assimilation results with (a) Nearest Neighbour Interpolation (NNI) method, (b) Bilinear method, and (c) Bicubic method.

3.2 Model calibration of 2D HEC-RAS

Calibration of flood depth modeling begins with obtaining information on historical floods. Based on the flood event on January 29, 2021, a high inundation of 4.8 m was recorded at the Bedadung AWLR post. Calibration was carried out by adjusting Manning's coefficient values in the model. In this analysis, Manning's coefficient was determined based on the land use class as presented in Table 4. The model parameter values that produced simulated flood depths that are closely matched to the observed flood depths were then used to validate the model.

Table 4  Manning's value selected.

Land cover	Allowable range of n value	Manning's value selected
Terra firma	0.10–0.16	0.13
Wetland	0.05–0.085	0.0675
Open surface water	0.025–0.05	0.0375
Cropland	0.025–0.05	0.0375
Built-up	0.12–0.2	0.16

3.3 Model validation

The simulation results from three different perimeter treatments showed that the model with a 50×50 m mesh size, and a break line spacing of a minimum of 25 and a maximum of 20, was the optimal choice, as indicated by the lowest RMSE in Table 5. This selected model was then validated against nine high waterlogging incidents that occurred on February 6, 8, 14, 15, 20, 21, 23, 25, and 27, 2023. The validation results in Table 5 show an average RMSE value of 0.113, suggesting a low prediction error. Furthermore, a Mean Error (ME) of 0.003 indicates a slight tendency for the model to overestimate the observed values, while an average R² of 0.927 indicates a strong correlation between the predicted and observed water levels. These findings demonstrate that the model performs satisfactorily in predicting flood inundation depth.

Table 5  Statistical validation results of the flood inundation model using nine observed ‎waterlogging events in February 2023.

Metric	Good Range	Score		Conclusion
RMSE	Low value		0.113		Low, indicating a good model prediction
ME	Approaching zero		0.003		Positive, indicating the tendency of the model to predict a higher value than the observed data.
R2	Close to 1		0.927		High, indicating the model fits the observed data well.

The second validation was compared the model with field conditions of historical floods gathered from the public for an extreme rain event on January 21, 2021. The distribution of historical flood points can be seen in Figure 3, and the respective flood depth and height of the puddle can be seen in Table 6. The average RMSE value is 0.316, which corresponds to a low predictive error, while the ME is -0.177, implying a slight underprediction of the flood depth. The average R² is 0.814, suggesting a strong correlation between the observed and predicted flood depths at various locations. Again, these results illustrate that the model is robust for predicting flood inundation height.

Table 6  Statistical validation results of the flood inundation model against historical flood ‎observations from the extreme rainfall event of 21 January 2021.

Metric	Good Range	Score	Conclusion
RMSE	Getting lower	0.316	Low, indicating a good model prediction capability
ME	Approaching zero	-0.177	Negative, indicating the tendency of the model to under-predict the inundation depth
R2	Close to 1	0.814	High, indicating the model fits the observation data well

3.4 Modeling extreme flooding and recurrence floods

The simulation results of the existing condition, representing the extreme flood event on 29 January 2021, indicate significant damage, with 436 houses along the riverbanks reported damaged. This condition is clearly reflected in the hydrodynamic behavior across river segments (Figure 10). In the upstream reach, with a channel width of approximately 40 m, flooding was confined to the main channel with depths exceeding 1.5 m and high flow velocities (>3.5 m/s), consistent with the steep bed slope and narrow, meandering morphology. These findings are consistent with Dottori et al. (2016b), who highlighted that upstream areas with steep slopes are more prone to high-velocity floods, increasing the risk of bank erosion and structural damage.

Figure 10  Existing condition: (a) depth of flooding (m), and (b) velocity (m/s).

In the downstream reach, the river widens to about 50 m with a braided pattern and mid-channel sediment bars, resulting in widespread shallow flooding (0.2–1.5 m) with low velocities (<1 m/s). While the direct hydraulic hazard is relatively low, exposure of settlements and infrastructure is high. This is consistent with (Salazar and Rojas González 2021), who found that tropical floodplains typically experience shallow depths (0.03–1.98 m). Lo et al. (2022) emphasized that such conditions could result in greater exposure of populations and assets.

This comparison emphasizes that the Bedadung watershed presents a distinct pattern: upstream flooding is more hazardous due to high flow energy despite its limited spatial extent, while downstream flooding poses significant risks because of its wide coverage in densely populated areas, even at relatively shallow depths. Accordingly, this study reinforces the importance of considering both depth and velocity in flood hazard mapping and underscores the need for spatially differentiated mitigation strategies for each river segment.

The verified model was utilized to simulate floods with return periods of 5-, 10-, and 25-years. The simulated floods are illustrated in Figure 11. It is evident that when discharge increases, the inundated area expands exponentially, particularly in low-lying regions with inadequate drainage systems, as noted by Yang et al. (2020). Table 6 illustrates the area impacted by flooding based on water depth and flood danger classification. Our findings align with prior research (e.g., Nisumanti et al. 2023; Nuariman and Harisuseno 2023).

Figure 11  Flood distribution with Q at (a) 5-year return period, (b) 10-year return period, and (c) 25-year return period.

The distribution of flood depths for the 5-, 10-, and 25-year return periods shows a relatively consistent pattern. The dominant categories fall within medium depth (0.6–1.5 m) at around 28% and high depth (1.5–3.5 m) at about 31%. Extreme depths of >3.5 m are also quite significant, covering 18–19% of the total affected area. Meanwhile, shallow inundation of <0.6 m accounts for approximately 22% of the affected area. This pattern indicates that most flooded areas fall within the medium–high categories, thereby representing moderate to high hazard potential. These findings are consistent with Jafar et al. (2023), who reported that flood depths in tropical floodplains generally range from 0.5–1.5 m, thus conditions in the Bedadung watershed reaching more than 3.5 m can be categorized as extreme.

In terms of flow velocity, the affected areas are dominated by very low velocity (<1 m/s) at around 48–50%, reflecting ponding in flat floodplain areas. Medium velocities of 1–3.5 m/s cover 42–43% of the area, indicating flow concentration in the main river channel. Although the percentage of areas with high velocity >3.5 m/s is relatively small (7–8%), this condition deserves serious attention as it is directly associated with bank erosion, structural damage, and threats to human safety. Slight shifts can be observed at higher return periods, where the proportion of low velocities tends to decrease, while medium–high velocities slightly increase. This is consistent with Maranzoni et al. (2023), who emphasized that higher return periods are directly proportional to the extent of affected areas.

In relation to hazard levels, the 5-year flood tends to cause greater socio-economic disruption due to the widespread exposure of shallow–medium inundation. The 10-year flood begins to show a combination of deeper inundation with medium velocity, thereby posing risks to agricultural land and infrastructure. At the 25-year flood, hazards become more intense with the expansion of high-velocity and deep inundation areas, increasing the risk of structural damage and erosion. This pattern is in line with the findings of Lyddon et al. (2025), who stressed that tropical floodplains are more vulnerable to the exposure of settlements and assets in shallow–medium inundation, while high velocities exacerbate exposure and potential disaster risk (Maranzoni et al. 2023; Zhen et al. 2022).

The sensitivity test results (Table 5) show that the 50×50 m mesh grid cell provides the most accurate and computationally efficient results, yielding the lowest RMSE values. This confirms that the selected mesh resolution adequately captures floodplain hydraulics while maintaining consistency with the roughness parameterization. Severe rainfall on January 21, 2021, has caused considerable flooding in the study area, as illustrated in Figure 10(a). The calibration of the flood model demonstrates highly accurate results. The verified model was utilized to simulate floods with return periods of 5-, 10-, and 25-years. The simulated floods are illustrated in Figures 11(b) to 11(d). It is evident that when discharge increases, the inundated area expands exponentially, particularly in low-lying regions with inadequate drainage systems, as noted by Yang et al. (2020). Table 6 illustrates the area impacted by flooding based on water depth and flood danger classification. Our findings align with prior research (e.g., Nisumanti et al. 2023; Nuariman and Harisuseno 2023).

Table 7 summarizes the inundation area in each flood hazard category for the -5, -10, and -25-year recurrence scenarios. In general, all hazard categories show an increase in inundation area as the magnitude of flooding increases, indicating an expansion of the affected area and an increase in flood severity in events with higher repetition times. The characteristics of each hazard category are described below.

Table 7  Flooded area for selected depth and return period.

Flood depth (m)	Hazard category	Flood distribution area (hectares)
		Return period of 5 years	Return period of 10 years	Return period of 25 years
0–0.2	Very low	46.59	47.33	47.55
0.2–0.6	Low	90.43	93.55	95.23
0.6–1.5	Medium	170.57	176.93	182.18
1.5–3.5	High	186.66	193.08	200.85
> 3.5	Very high	112.59	116.04	119.84
TOTAL		606.84	626.93	645.65

Very Low Category Flood Danger (0–0.2 m): The affected area for a 5-year flood reaches 46.59 hectares and increases to 47.33 hectares for a 25-year flood. This category shows a fairly moderate increase as flood discharge increases. This suggests that shallow puddles often occur in low areas, which naturally experience flooding more frequently (Yang et al. 2020).

Low Category Flood Hazard (0.2–0.6 m): Inundated areas under this category increase from 90.43 ha for a 5-year flood to 93.55 ha for a 25-year flood. This significant increase indicates that areas with medium inundation can expand more rapidly when flood discharge is higher, and this is consistent with Pertiwi et al. (2021), especially for urban areas with limited drainage (Bayas-Jiménez et al. 2022; Nadia et al. 2022).

Medium Category Flood Hazard (0.6–1.5 m): The results show that the area affected in this category increases from 170.57 ha for a 5-year flood to 176.93 ha for a 25-year flood. These results show a similar pattern, where inundation with medium depths tends to occur more frequently in areas with moderate slopes or large catchment areas (Godara et al. 2023; Varlas et al. 2019).

High Category Flood Hazard (1.5–3.5 m): The area affected by high depth floods is relatively smaller but still shows an increase from 186.66 ha for a 5-year flood to 193.08 ha for a 25-year flood. It shows that floods of greater depth occur less frequently but still have a significant impact on infrastructure and public safety (Pregnolato et al. 2022).

Very High Category Flood Danger (> 3.5 m): An increase in the inundation area in the very high category occurs from 112.59 ha for the 5-year flood to 116.04 ha for the 25-year flood. It’s noted that floods in this category were more likely to occur near major rivers or very large catchment areas (Bayas-Jiménez et al. 2022; Costache et al. 2020; Shahabi et al. 2021).

The flood modeling for the selected return periods in this study shows a clear pattern related to the increased flood risk as the return period discharge increases. These findings reinforce previous research, which indicated that higher flood discharges would result in a significant increase in inundation areas, especially in regions where the topography supports water accumulation (Shahid et al. 2022), changes in riverbed elevation and river bifurcation can alter the extent of the flood inundation (Dingle et al. 2020). Our findings on the relationship between the size of the inundated area and the flood frequency are consistent with Fleischmann et al. (2023) and Yamamoto et al. (2021). Therefore, it is justified to classify flood hazard based on water depth. The increase in inundation area for each depth category provides a deeper insight into the patterns and levels of risk for each return period discharge scenario.

These findings provide important information for future flood mitigation planning, particularly in determining areas that should be prioritized for structural mitigation efforts (such as levee construction and drainage capacity enhancement) and non-structural efforts (such as the development of early warning systems and community evacuation plans). By integrating the results of this study with earlier related findings, mitigation strategies can be more targeted toward the highest-risk areas and enhance community resilience against flood threats.

4 CONCLUSION

This research highlights the importance of flood modeling for various return periods and its impact on inundation risk. The bilinear interpolation method produced statistically significant results with the lowest RMSE, indicating its high accuracy in representing the relatively flat and low-complexity topography of the study area. Given the gentle slope and minimal terrain variation, this method is well-suited for similar floodplain environments where elevation changes are gradual. Furthermore, the integration of multi-resolution DEMs within the 2D flood modeling framework proved to be a practical strategy to enhance spatial detail while maintaining computational efficiency. This approach is particularly beneficial in data-sparse or resource-limited settings, enabling more reliable flood simulations across varying terrain and data availability scenarios. After the 2D HEC-RAS model was calibrated and validated, it was able to accurately predict the flood depth, with average RMSE values that show good model performance.

The analysis of discharge scenarios for 5-, 10-, and 25-year return periods demonstrated a significant increase in flood risk with rising discharge levels. Under the 5-year return period scenario, the initial flood inundation extent was approximately 1352.23 hectares, serving as the baseline condition. This area increased by around 20.54 hectares under the 10-year scenario, and by 42.62 hectares under the 25-year scenario, highlighting the growing impact of higher magnitude flood events on spatial flood extent. The extent of inundation across various depth categories provides crucial insight into flood patterns and their potential impacts. These findings support previous studies, reinforcing the relationship between higher flood discharges and increased inundation areas.

Future research challenges would include integrating real-time data to enhance predictive accuracy and considering other factors, such as land use changes, climate change, and river dynamics that may affect flood patterns. Further studies should also explore the use of cutting-edge technologies, such as machine learning-based modeling, to identify more complex patterns in flooding. In this way, mitigation efforts can become more effective in addressing future flood challenges.

ACKNOWLEDGMENTS

The authors would like to thank the Directorate General of Higher Education (Dirjen Dikti) for providing the grant. The research team also expresses its gratitude to the East Java Provincial Water Resources Agency (Dinas PUSDA) for supporting this work with rainfall and topographical data. Additionally, we would like to thank the Research and Community Service Institute, Universitas Jember, for their support in facilitating this research.

References

1. Arseni, M., A. Rosu, M. Calmuc, V.A. Calmuc, C. Iticescu, and L.P. Georgescu. 2020. “Development of Flood Risk and Hazard Maps for the Lower Course of the Siret River, Romania.” Sustainability 12 (16): 6588. https://doi.org/10.3390/su12166588
2. Aryal, D., L. Wang, T.R. Adhikari, J. Zhou, X. Li, M. Shrestha, Y. Wang, et al. 2020. “A model-based flood hazard mapping on the southern slope of Himalaya.” Water (Switzerland) 12 (2): 540. https://doi.org/10.3390/w12020540
3. Bates, P.D., K.J. Marks, and M.S. Horritt. 2003. “Optimal use of high-resolution topographic data in flood inundation models.” Hydrologiccal Processes 17 (3): 537–557. https://doi.org/10.1002/hyp.1113
4. Bayas-Jiménez, L., F.J. Martínez-Solano, P.L. Iglesias-Rey, and F. Boano. 2022. “Economic Analysis of Flood Risk Applied to the Rehabilitation of Drainage Networks.” Water (Switzerland) 14 (18): 2901. https://doi.org/10.3390/w14182901
5. Belias, M., M.M. Rovers, J. Hoogland, J.B. Reitsma, T.P.A. Debray, and J. IntHout. 2022. “Predicting Personalised Absolute Treatment Effects in Individual Participant Data Meta-Analysis: An Introduction to Splines.” Research Synthesis Methods 13 (2): 255–283. https://doi.org/10.1002/jrsm.1546
6. Chen, C. and Y. Li. 2019. “A Fast Global Interpolation Method for Digital Terrain Model Generation from Large LiDAR-Derived Data.” Remote Sensing 11 (11): 1–25. https://doi.org/10.3390/rs11111324
7. Cosentino, N.J., N.E. Opazo, F. Lambert, A. Osses, and E. van ’t Wout. 2023. “Global‐Krigger A Global Kriging Interpolation Toolbox With Paleoclimatology.” Geochemistry, Geophysics, Geosystems 24 (6): e2022GC010821. https://doi.org/10.1029/2022GC010821
8. Costabile, P., C. Costanzo, D. Ferraro, F. Macchione, and G. Petaccia. 2020. “Performances of the new HEC-RAS version 5 for 2-D hydrodynamic-based rainfall-runoff simulations at basin scale: Comparison with a state-of-the art model.” Water (Switzerland) 12 (9): 1–19. https://doi.org/10.3390/W12092326
9. Costache, R., A. Arabameri, T. Blaschke, Q.B. Pham, and B.T. Pham. 2020. “Novel ensembles of deep learning neural network and statistical learning for flash-flood susceptibility mapping.” Water (Switzerland) 12 (6): 1549. https://doi.org/10.3390/W12061549
10. Dingle, E.H., M.J. Creed, H.D. Sinclair, D. Gautam, N. Gourmelen, A.G.L. Borthwick, and M. Attal. 2020. “Dynamic Flood Topographies in the Terai Region of Nepal.” Earth Surface Processes and Landforms 45 (13): 3092–3102. https://doi.org/10.1002/esp.4953
11. Dottori, F., M. Lloyd, V. Martina, R. Figueiredo, M. L. V Martina, and R. Figueiredo. 2016a. “A methodology for flood susceptibility and vulnerability analysis in complex flood scenarios.” Flood Risk Management 11 (52): S632–S645. https://doi.org/10.1111/jfr3.12234
12. Dottori, F., P. Salamon, A. Bianchi, L. Alfieri, F.A. Hirpa, and L. Feyen. 2016b. “Development and Evaluation of a Framework for Global Flood Hazard Mapping.” Advances in Water Resources 94, 87–102. https://doi.org/10.1016/j.advwatres.2016.05.002
13. El-Naqa, A. and M. Jaber. 2018. “Floodplain Analysis using ArcGIS, HEC-GeoRAS and HEC-RAS in Attarat Um Al-Ghudran Oil Shale Concession Area, Jordan.” Journal of Civil & Environmental Engineering 08 (05): 1000323. https://doi.org/10.4172/2165-784x.1000323
14. Fleischmann, A.S., F. Papa, S.K. Hamilton, A. Fassoni-Andrade, S. Wongchuig, J.C. Espinoza, R.C.D. Paiva, et al. 2023. “Increased Floodplain Inundation in the Amazon since 1980.” Environmental Research Letters 18, 034024. https://doi.org/10.1088/1748-9326/acb9a7
15. Ghandehari, M., B.P. Buttenfield, and C.J.Q. Farmer. 2019. “Comparing the accuracy of estimated terrain elevations across spatial resolution.” International Journal of Remote Sensing 40 (13): 5025–5049. https://doi.org/10.1080/01431161.2019.1577581
16. Godara, N., O. Bruland, and K. Alfredsen. 2023. “Simulation of flash flood peaks in a small and steep catchment using rain‐on‐grid technique.” Journal of Flood Risk Management 16 (3): e12898. https://doi.org/10.1111/jfr3.12898
17. Guerriero, L., G. Ruzza, F. M. Guadagno, and P. Ravellino. 2020. “Flood hazard mapping incorporating multiple probability models.” Journal of Hydrology 587, 125020. https://doi.org/10.1016/j.jhydrol.2020.125020
18. Hamdi, M., D.A. El Molla, and M.A. Gad. 2019. “A Comparison Between 1D, 2D and Semi 2D Hydraulic Models.” Al-Azhar University Civil Engineering Research Magazine 41 (4): 295–305. https://www.azharcermjournal.com/CERMF1910/P19-10-26.pdf
19. Hemmati, M., H.N. Mahmoud, B.R. Ellingwood, and A.T. Crooks. 2021. “Shaping urbanization to achieve communities resilient to floods.” Environmental Research Letters 16 (9): 094033. https://doi.org/10.1088/1748-9326/ac1e3c
20. Hutanu, E., A. Mihu-Pintilie, A. Urzica, L.E. Paveluc, C.C. Stoleriu, and A. Grozavu. 2020. “Using 1D HEC-RAS Modeling and LiDAR Data to Improve Flood Hazard Maps Accuracy.” Water 12 (6): 1624. https://doi.org/10.3390/w12061624
21. Jafar, A., N. Sakke, M.T. Mapa, R. Dollah, A.K.B. Asis, F.M. Ata, M. Jaafar, and A. Abas. 2023. “Unravelling the Complexity of Flood Hazards: A Comprehensive Review of Characteristics and Risk Assessment.” ASM Science Journal 18, 1–12. https://doi.org/10.32802/ASMSCJ.2023.1404
22. Kadri, T. and E. Kurniyaningrum. 2019. “Impact of Land Use on Frequency of Floods In Upper Bekasi Watershed, Indonesia.” International Journal of Scientific Research and Technology 8 (12): 3328–3334
23. Kandil, W.M., F.H. Zarzoura, M. Salah, and M. El-Mewafi. 2024. “New approach for satellite DEM accuracy enhancement by combing machine learning, fuzzy majority voting, and weighted interpolation techniques.” Innovative Infrastructure Solutions 9, 103. https://doi.org/10.1007/s41062-024-01401-z
24. Khan, A., I. Pathan, and P.G. Agnihotri. 2020. “2-D unsteady flow modelling and inundation mapping for lower region of Purna basin using HEC-RAS.” Nature Environment and Pollution Technolology 19 (1): 277–285. https://neptjournal.com/upload-images/(28)B-3622.pdf
25. Lo, E.L., K.M. Yeager, I. Bergier, L. Domingos-Luz, A. Silva, and M.M. McGlue. 2022. “Sediment Infill of Tropical Floodplain Lakes: Rates, Controls, and Implications for Ecosystem Services.” Frontier Earth Sciences 10, 875919. https://doi.org/10.3389/feart.2022.875919
26. Lyddon, C., L. Devitt, M. Barada, G. Coxon, T. Coulthard, A. Barkwith, and P. Robins. 2025. “Climate Change Likely to Intensify Storm‐Driven Compound Flooding in an Exemplar UK Estuary.” Earth’s Future 13 (12); e2024EF005822. https://doi.org/10.1029/2024EF005822
27. Maleika, W. 2024. “Local Polynomial Interpolation Method Optimization in the Process of Digital Terrain Model Creation Based on Data Collected From a Multibeam Echosounder.” Institute of Electrical and Electronics Engineers 49 (3): 920–932. https://doi.org/10.1109/JOE.2024.3353271
28. Mani Murali, R., M. Ankita, S. Amrita, and P. Vethamony. 2013. “Coastal vulnerability assessment of Puducherry coast, India, using the analytical hierarchical process.” Natural Hazards Earth System Sciences 13 (12): 3291–3311. https://doi.org/10.5194/nhess-13-3291-2013
29. Maranzoni, A., M. D’Oria, and C. Rizzo. 2023. “Quantitative flood hazard assessment methods: A review.” Journal of Flood Risk Management 16 (1): e12855. https://doi.org/10.1111/jfr3.12855
30. Minh, N.Q., N.T.T. Huong, P.Q. Khanh, L.P. Hien, and D.T. Bui. 2024. “Impacts of Resampling and Downscaling Digital Elevation Model and Its Morphometric Factors: A Comparison of Hopfield Neural Network, Bilinear, Bicubic, and Kriging Interpolations.” Remote Sensing 16 (5): 819. https://doi.org/10.3390/rs16050819
31. Nadia, E., O. Hasan, H. Said, and A. Mohamed. 2022. “Sediment Transport Index ( STI ) modeling using the GIS at Small Agricultural Catchment.” International Journal of Innovative Research in Sciences and Engineering Studies 2 (11): 16–20.
32. Nisumanti, S., R. Baniva, and Afrizal. 2023. “Analysis of drainage network capacity in Rawa Jaya Area, Ilir Timur I, Palembang.” IOP Conference Series: Earth and Environmental Science 1173, 012039. https://doi.org/10.1088/1755-1315/1173/1/012039
33. Nofal, O.M. and J.W. van de Lindt. 2021. “High-resolution flood risk approach to quantify the impact of policy change on flood losses at community-level.” International Journal of Disaster Risk Reduction 62, 102429. https://doi.org/10.1016/j.ijdrr.2021.102429
34. Nuariman, P.A. and D. Harisuseno. 2023. “Prediction of Changes in Land Use/Land Cover and Hydrological Response in the Upper Ciliwung Watershed.” Civil and Environmental Science Journal (CIVENSE) 6 (1): 71–83. https://doi.org/10.21776/ub.civense.2023.00601.9
35. Ongdas, N., F. Akiyanova, Y. Karakulov, A. Muratbayeva, and N. Zinabdin. 2020. “Application of HEC-RAS (2D) for Flood Hazard Maps Generation for Yesil (Ishim) River in Kazakhstan.” Water (Switzerland) 12 (10): 2672. https://doi.org/10.3390/w12102672
36. Pathiraja, S., D. Anghileri, P. Burlando, A. Sharma, L. Marshall, and H. Moradkhani. 2018. “Time-varying parameter models for catchments with land use change: The importance of model structure.” Hydrology and Earth System Sciences 22 (5): 2903–2919. https://doi.org/10.5194/hess-22-2903-2018
37. Peker, İ.B., S. Gülbaz, V. Demir, O. Orhan, and N. Beden. 2024. “Integration of HEC-RAS and HEC-HMS with GIS in Flood Modeling and Flood Hazard Mapping.” Sustainability 16 (3): 1226. https://doi.org/10.3390/su16031226
38. Pertiwi, A.P., A. Roth, T. Schaffhauser, P.K. Bhola, F. Reuß, S. Stettner, C. Kuenzer, and M. Disse. 2021. “Monitoring the spring flood in lena delta with hydrodynamic modeling based on sar satellite products.” Remote Sensing 13 (22): 4695. https://doi.org/10.3390/rs13224695
39. Pinos, J. and L. Timbe. 2019. “Performance assessment of two-dimensional hydraulic models for generation of flood inundation maps in mountain river basins.” Water Science and Engineering 12 (1): 11–18. https://doi.org/10.1016/j.wse.2019.03.001
40. Pregnolato, M., A.O. Winter, D. Mascarenas, A.D. Sen, P. Bates, and M.R. Motley. 2022. “Assessing flooding impact to riverine bridges: An integrated analysis.” Natural Hazards and Earth System Sciences 22 (5): 1559–1576. https://doi.org/10.5194/nhess-22-1559-2022
41. Riazanoff, S., A. Corseaux, C. Albinet, P.A. Strobl, C. López-Vázquez, P.L. Guth, and T. Tadono. 2024. “Best BiCubic Method to Compute the Planimetric Misregistration between Images with Sub-Pixel Accuracy: Application to Digital Elevation Models.” International Journal of Geo-Information 13 (3): 96. https://doi.org/10.3390/ijgi13030096
42. Salazar, S.F. and A.M. Rojas González. 2021. “Evaluation of Debris Flows for Flood Plain Estimation in a Small Ungauged Tropical Watershed for Hurricane Otto.” Hydrology 8 (3): 122. https://doi.org/10.3390/hydrology8030122
43. Shafiq, M., M. Abbas, E.K. El-Shewy, M.A.E. Abdelrahman, N.F. Abdo, and A.A. El-Rahman. 2024. “Numerical Investigation of the Fractional Diffusion Wave Equation with the Mittag–Leffler Function.” Fractal Fractional 8 (1): 18. https://doi.org/10.3390/fractalfract8010018
44. Shahabi, H. 2021. “Flood susceptibility mapping in northern regions of Iran using advanced data mining algorithms (Case study: Haraz watershed)” Journal of Regional Planning 11 (41): 165–184.
45. Shahabi, H., A. Shirzadi, S. Ronoud, S. Asadi, B.T. Pham, F. Mansouripour, M. Geertsema, et al. 2021. “Flash flood susceptibility mapping using a novel deep learning model based on deep belief network, back propagation and genetic algorithm.” Geoscience Frontiers 12 (3): 101100. https://doi.org/10.1016/j.gsf.2020.10.007
46. Shahid, H., M. Toyoda, and S. Kato. 2022. “Impact Assessment of Changing Landcover on Flood Risk in the Indus River Basin Using the Rainfall–Runoff–Inundation (RRI).” Sustainability 14 (12) 7021. https://doi.org/10.3390/su14127021
47. Skala, V., S.A.A. Karim, and M. Cervenka. 2020. “Finding Points of Importance for Radial Basis Function Approximation of Large Scattered Data.” In: Krzhizhanovskaya, V., et al. Computational Science – ICCS 2020. Lecture Notes in Computer Science, 12142. Springer, Cham. https://doi.org/10.1007/978-3-030-50433-5_19
48. Śliwiński, D., A. Konieczna, and K. Roman. 2022. “Geostatistical Resampling of LiDAR-Derived DEM in Wide Resolution Range for Modeling in SWAT: A Case Study of Zgłowiączka River (Poland).” Remote Sensing 14 (5): 1281. https://doi.org/10.3390/rs14051281
49. Varlas, G., M.N. Anagnostou, C. Spyrou, A. Papadopoulos, J. Kalogiros, A. Mentzafou, S. Michaelides, et al. 2019. “A multi-platform hydrometeorological analysis of the flash flood event of 15 November 2017 in Attica, Greece.” Remote Sensing 11 (1): 45. https://doi.org/10.3390/rs11010045
50. Xu, T., V. Merwade, and Z. Wang. 2023. “Interpolating Hydrologic Data Using Laplace Formulation.” Remote Sensing 15 (15): 3844. https://doi.org/10.3390/rs15153844
51. Yalcin, E. 2019. “Two-dimensional hydrodynamic modelling for urban flood risk assessment using unmanned aerial vehicle imagery: A case study of Kirsehir, Turkey.” Journal of Flood Risk Management 12 (S1): 1–14. https://doi.org/10.1111/jfr3.12499
52. Yamamoto, K., T. Sayama, and Apip. 2021. “Impact of climate change on flood inundation in a tropical river basin in Indonesia.” Progress in Earth and Planetary Science 8, 1. https://doi.org/10.1186/s40645-020-00386-4
53. Yang, Y., L. Sun, R. Li, J. Yin, and D. Yu. 2020. “Linking a Storm Water Management Model to a Novel Two-Dimensional Model for Urban Pluvial Flood Modeling.” International Journal of Disaster Risk Science 11, 508–518. https://doi.org/10.1007/s13753-020-00278-7
54. Yu, W., Y. Zhang, T. Ai, and Z. Chen. 2021. “An integrated method for DEM simplification with terrain structural features and smooth morphology preserved.” International Journal Geographical Information Science 35 (2): 273–295. https://doi.org/10.1080/13658816.2020.1772479
55. Zhang, W-H., L. Dai, W. Chen, A. Sun, W-L. Zhu, and B-F. Ju. 2023. “Novel Information-Driven Smoothing Spline Linearization Novel Information-Driven Smoothing Spline Linearization.” Sensors 23 (22): 9268. https://doi.org/10.3390/s23229268
56. Zhen, Y., S. Liu, G. Zhong, Z. Zhou, and J. Liang. 2022. “Risk Assessment of Flash Flood to Buildings Using an Indicator-Based Methodology: A Case Study of Mountainous Rural Settlements in Southwest China.” Frontiers in Environmental Science 10, 1–18. https://doi.org/10.3389/fenvs.2022.931029