Development of a 2D hydrodynamic model for inundation assessment and flood early warning system: a case of depressed Ghed region, India

Natural disasters occur more frequently and with greater intensity (Singh et al. 2021). Natural disasters threaten human happiness and economic development, claim lives, and wreak havoc on societal and physical infrastructure (Sahu 2016; Inman & Lyons 2020). People are more vulnerable to frequent and severe disasters, particularly in the poorest communities, which commonly have the lowest infrastructure, mainly rely on agriculture for their livelihoods, and typically lack basic resources. (Wang et al. 2019; Pangali Sharma et al. 2022). In addition, it may cause alterations in personal preferences, such as increased risk aversion, prosocial conduct, and impatience, which may have a favourable or unfavourable impact on how vulnerable a person experiences (Nandalal 2009; Wu et al. 2021; Garrote 2022). Over one-third of the world's territory and nearly 82% of its population are affected by flooding (Li et al. 2015; Rangari et al. 2019). The Gujarat government passed the Gujarat State Disaster Management Act 2003 to reduce the number of fatalities and the adverse effects of disasters on the state's overall socioeconomic growth (Patel & Srivastava 2013; Thakkar et al. 2017; Pathan et al. 2022).

The development of hydrodynamic modelling is required for accurate mapping and flood assessment. Numerous programs, including TUFLOW, Mike, HEC-RAS (hydrological engineering centre-river analysis system), RASPLOT, RUNUP, and Coastal Hazard Analysis Modelling Program (CHAMP), are available for flood modelling (Johnston & Smakhtin 2014; Teng et al. 2017). Especially for semi-arid regions, HEC-RAS is a widely used software because of its interface and optimized input data (Mangukiya et al. 2022; Arash & Yasi 2023). An Early flood warning is essential for evacuation and damage control in vulnerable areas (DeVries et al. 2020; Asitatikie et al. 2022). In addition, flood early warning systems are being used more frequently as a precautionary measure globally since timely flood information can assist people living in downstream areas, reduce human casualties, and save movable goods (Kumar et al. 2020; Rai et al. 2020; Bettin 2023).

In this research, the Ghed region is assessed for flood risk analysis. The study region's topography with its cup-shaped geometry and area that experiences more inundation than flood makes the research distinctive (Trambadia et al. 2022). The water remains here for more months, even after the monsoon season. In addition, this research emphasizes dealing with open-source data for two-dimensional hydrodynamic analysis. The 2D flood model was developed to assess the downstream Ozat River basin flood. The river basin has fan-shaped catchment that exhibits the flash flood from this characteristic (Thameemul Hajaj et al. 2019). The Ghed region is located near the river mouth of the Ozat River, having more than 200 km² area. The HEC-RAS was employed for 2D hydrodynamic modelling with 14 days of simulations from 14 August 2017 to 27 August 2017. For model validation, limited observed data are available on the water depth at key places in an exposed flood zone. In this regard, various statistics are performed for a comparative study like, R², root mean square error (RMSE), Nash–Sutcliffe efficiency (NSE), confidence interval, and coefficient of variations. End of the simulations, different hazard maps are generated for the analysis. The arrival time map is a substantial map for the flood early warning system (FEWS), especially for framing the decision-making system.

The Ozat River basin is in the deep western part of Gujarat, India. The area frequently floods as a result of its unique cup-shaped terrain. The term ‘Ghado’ in Gujarati, which denotes a pot of water, is where Ghed originates. This area is located near the river mouth of Ozat River (69°50′N and 70°50′N longitude, 21°10′E and 21°40′E latitudes), which is a coastal zone of Saurashtra. The total length of the river is 145 km from the origin to the river mouth near Navi Bandar. The river's main stream splits into two portions: one flows straight towards the sea near Pata village and another skewed towards the Ghed and meets the Arabian Sea near Navi Bandar. Near the origin place of the river, the mainstream has sufficient width (230 m), but it narrows noticeably 82 km away and even more so close to the Ghed. The straight river path is 82.4 km long, while its tributaries towards the Pata and Navi Bandar are 36.6 and 62.6 km long, respectively.

The data used in this study are presented in Table 1 with the listed data's source, purpose, and features. The digital elevation model (DEM), acquired from the United States Geological Survey (USGS) official website with the 30-meter spatial resolution, is the primary step for flood modelling. The 13 topographical maps are utilized from the SOI to delineate minor reaches. The hydrograph data as an input parameter of the inflow boundary are acquired from the Junagadh irrigation department. At this location, a broad crested vertical gated weir with 17 gates and a length of 235 m is built. Its height is 16.94 m. For validating purposes, the water depth study was executed at key regional places, and water surface elevation (WSE) data were obtained from the Junagadh irrigation department. In the context of consistency of a 2D model, the land use land cover (LULC) map was prepared from the latest cloud-free Landsat image collected from the cloud source of the Google Earth Engine.

The flood inundation mapping needs hydrodynamic modelling, and various data related to terrain, runoff, and observed flow at multiple locations. In this regard, the hydraulic model HEC-RAS was used for flood assessment for this case study. The numerous case studies and convenient interface of the HEC-RAS lead this research to effective outcomes.

The LULC map is essential for flood effect analysis on different land classes. The LULC map was generated with the Landsat 8 top of the atmosphere imagery using the Google Earth Engine. The Google Earth Engine is the most effective spatial analysis platform because it can operate with raw satellite data attributable to the cloud computing facility. The total study area is 870.11 km^2, and the current research area covers the classes of water body, cropland, built-up area, forest, and barren land of 50.03, 502.02, 103.89, 75.01, and 139.16 km², respectively.

The most critical parameter for hydrodynamic modelling is manning roughness, particularly when the terrain has various land cover classifications. In this situation, the Indian territory uses the Dam Rehabilitation and Improvement Project (DRIP) criteria as a reference system for employing roughness values. According to this DRIP recommendation, the appropriate manning numbers for water bodies, cropland, built-up areas, forests, and barren land are 0.04, 0.035, 0.1, 0.160, and 0.1, respectively. These values were used for 2D modelling to allocate manning regions in HEC-RAS for validation with the limited observed dataset by trial and error.

The hydrodynamic modelling using HEC-RAS was simulated for 14 days in a semi-arid region in the western part of Gujarat, India. The average rainfall is 708 mm, and it has recently become in the increasing mode as per the Indian Meteorological Department of India. The Ghed region (study area) has an unique geometry of depressed topography and is located near the coastal zone of Saurashtra.

The various statistics measured for water depth comparison are presented in Table 2; R², RMSE, and NSE were employed for this analysis. Due to their distance from the river valley area, the values close to the villages of Sarama and Kadachh diverge from the modelled dataset and have a marginally lower level of agreement. As shown in Table 2, the Koylana and Kadachh villages closely matched the observed dataset. Data measured closer to the valley area show a consensus agreement than those measured farther away.

The water arrival times at significant locations along the study region are shown in Table 3. As previously mentioned, the simulation window's flood starts at 36 h after which the water disperses over the area and takes a variable amount of time, as shown in the table, to reach key sites.

This flood inundation assessment case study was conducted using just a limited number of observed datasets and comparisons. However, advanced databases, high-resolution satellite imagery, and Unmanned Aerial Vehicle (UAV)-based surveys are required for flood assessment research. The validation of the hydrodynamic model is vital for future event predictions and decision-making systems. The specific limitations are described as follows:

• The more precise DEM is needed for studies on flood assessment, while the current study examined only a 30-m spatial resolution DEM. In this context, the authors advised that future flood events in the Ghed should be evaluated with UAV-based DEM for terrain file generation.

• The hydrodynamic model needs the boundary condition, especially in the HEC-RAS, for data inputs in these locations. Although the normal depth was chosen as the downstream boundary condition in the present study, the software interface offers other options for accurate modelling, including WSE and tidal data. For the current situation, the tidal wave data are not accessible near the river mouth. For accurate validation of results, the authors advise monitoring the tidal data or WSE observation close to the river mouth.

• As discussed in the Results section, the water scattering in the large cup-shaped Ghed region and gets steady over days. In this instance, the water diverts from the Ozat River's main stream and travels through the low-lying terrain. In this regard, the water depth observation at four places is available to validate the modelled data. The authors suggest mandating more observation sites together with frequent data collection systems for studies on water depth.

• The inflow point of water is near the Shapur weir, where the river gauge is available for discharge data recording. This place is located nearly 82 km away from the river mouth, and after 19 km from this site, the river's main stream splits into two portions; the water depth indicator is present, but a discharge recording facility is not available. Establishing the new gauge discharge stations on bifurcated reaches is highly suggested.

• The inflow hydrograph is directly employed in the modelling of upstream boundary conditions for the input datasets. However, modelling of rainfall–runoff is hampered by the absence of hourly rainfall data. For precise modelling and data validation before use in a hydrodynamic study, the establishment of new rain gauge stations for frequent data recording areas is advised.

• The small cross drainage structures like stop logs, small weirs, and check dams need to consider for modelling; however, in the present case, the geometry data are not available for these structures. After the inflow point, a few structures are constructed in the downstream region and the periphery of the Ghed area. Although these structures have a relatively small impact on hydraulic modelling, they must be taken into account for accurate modelling. It is recommended that the design data be quantified and made available to researchers.

The present case study effectively employed the hydrodynamic model for the semi-arid region and non-perennial Ozat River. A simulation time of 14 days is adequate for hydrodynamic modelling with an area of 850–1,000 km². The observed dataset for water depth has an exceptional degree of agreement with the data from models. The average values for R², RMSE, and NSE, respectively, are 0.80, 0.26, and 0.99, according to the statistics shown in Results section. The water depth map reveals the key findings for the flood depth analysis during simulations.

The village built-up areas are safe from peak flow conditions, but following the bifurcation, the situation worsens since the reaches shrink as they approach the river mouth. The water depth at Ghed is roughly between 0.5 and 1 m; it is not more extensive, but the water stays longer and becomes more consistent here. The water arrival scenario across the simulation period is shown on the arrival time map for FEWS. The water begins spreading out from the split reaches and scattering in Ghed 36 h after the simulations begin. The depressed area of Ghed contains significant farmland and farm dwellings. The 19 h of leg time is available for the intimation to this cup-shaped area for leaving this place and arriving at the village area. The two villages (Kadachh and Mander) are close to the coastline zone, which causes the surrounding area to flood during the peak flow conditions. The periphery of these villages is lower elevated, and the water level settles after the flood. These areas need to be connected to a higher-elevation road network, and the residents of these areas require early warning notifications via a digital network. The hydrodynamic model results reveal that more than 200 km² of areas frequently flooded in the typical flow condition. Discussions with villagers during study area visits in the winter (4 months following the monsoon), when most of the land was covered with chickpeas, revealed that a similar crop pattern was caused by prolonged flooding. This type of crop design degrades the quality of the soil and leaves the subsurface layer without nourishment.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.