Floods are catastrophic natural disasters that cause a substantial toll on human lives, infrastructure, and the economy. Structural and non-structural measures are developed for planning flood mitigation strategies. Flood inundation mapping is valuable information for decision-makers and authorities to develop flood mitigation strategies and resource allocation. This study uses the HEC-RAS 2D model for flood inundation mapping in the Krishna River Basin. Digital elevation models (DEMs) of 12.5 and 30 m resolutions were used to model the inundation map. The study also investigated the effect of change in upstream boundary data on the inundated area. The simulated results with 12.5 m resolution DEM are found in good agreement with the validation data and conform to the inundated areas with the available reports. This study proves the 2D capabilities of HEC-RAS and helps the experts with better management practices.
Flood inundation mapping was performed on the Krishna River Basin using the HEC-RAS 2D model.
Digital elevation models (DEMs) of 12.5 m resolution and 30 m resolution were incorporated to perform the inundation mapping.
The study investigated the influence of varying the upstream boundary data on the inundated area.
The simulated results with 12.5 m resolution DEM give a good agreement with the data available for the validation.
Floods are prevalent and recurring natural hazards that disrupt social and economic activities. This hazard leads to many fatalities and extensive damage to livelihood systems, property, infrastructure, and utility services (Amarasinghe et al. 2020; Band et al. 2020). This can be attributed to various factors, including rising urbanization, increased developmental and economic activities in flood plains, and global warming. Recent research has also shown that large-scale human interference in nature's order (deforestation, increased sedimentation rate in river channels, intrusion of human settlement in riverbank areas, etc.) increases flood occurrence (Patel & Srivastava 2013; Kvočka et al. 2015; Malik & Abdalla 2016; Rafiq et al. 2016; Kumar & Singh 2021). Flood mitigation requires the implementation of structural and non-structural measures to attenuate and avoid flood risk. To minimize the economic loss and adverse effects of floods, measures like flood forecasting (Madsen 2003; Wang et al. 2017; Tamiru & Wagari 2021), inundation modeling (Kadam & Sen 2012; Timbadiya et al. 2015; Vojtek et al. 2019; Komolafe 2021), flood hazard and risk mapping should be employed to identify flood-susceptible zones (Sahoo & Sreeja 2017; Farooq et al. 2019; Pal et al. 2022; Puno et al. 2022). Managers and policymakers experts must have information on flood depth and danger to public infrastructure in flood-prone areas to plan flood risk management. Flood inundation modeling is one of the most effective ways to plan flood mitigation strategies and identify flood-prone areas for flood risk management (Kadam & Sen 2012; Timbadiya et al. 2015; Teng et al. 2017; Jacob et al. 2020; Nkwunonwo et al. 2020; Tamiru & Wagari 2021). Flood inundation models are developed to help people better understand, analyze, and anticipate floods and their effects on socio-economic infrastructure.
Numerous mathematical models have been developed for flood inundation, depending upon the spatial extent, dimensionality, and mathematical complexity (Nkwunonwo et al. 2020). Numerical models, HEC-RAS, HEC-HMS, ISIS 1D, MIKE Hydro, SOBEK 1D, and TUFLOW 1D, have been used extensively for one-dimensional (1D) flood simulation, but failed to provide thorough information regarding the flow field and have the flaw of dropping out important features during simulation (Bates 2004; Chatterjee et al. 2008; Seyoum et al. 2012; Villazón et al. 2013; Gharbi et al. 2016). In recent years, two-dimensional (2D) models such as DIVAST, HEC-RAS 2D, ISIS 2D, MIKE 21, and TELEMAC 2D have received the most research attention for 2D flow simulations (Lin et al. 2006; Achilleas 2013; Ahn et al. 2019; Rangari et al. 2019). Jahandideh-Tehrani et al. (2020) give a comparison table of 1D, 2D, and 3D models, with their advantages and disadvantages.
The uncertainty involved in flood inundation modeling, and the complex and unpredictable character of floods mark a challenging task for hydrologic and hydraulic studies at high spatiotemporal resolutions (Merz & Thieken 2005). Merwade et al. (2008) and Bales & Wagner (2009) highlight the sources of uncertainty associated with flood inundation mapping. Uncertainty in model input (digital elevation model (DEM), channel bathymetry, roughness parameters, cross-section spacing, and hydraulic structures), estimation of design discharge, and choice of model selection and flow condition, are the major uncertainties involved. Sensitivity to topographic data and DEMs is discussed by various researchers (Singh 2005; Yan et al. 2013; Farooq et al. 2019; Prakash Mohanty et al. 2020). Prakash Mohanty et al. (2020) recommend use of the resampled and freely available Carto DEMs, where precise DEMs, like LiDAR DEMs, are not available. Vojtek et al. (2019) concluded that the sensitivity of flood mapping to hydrologic modeling is greater than that to hydraulic modeling.
In this study, HEC-RAS 2D (Hydrologic Engineering Center's River Analysis System) model was employed for inundation mapping. Past studies (Patel et al. 2017; Rangari et al. 2019; Kumar et al. 2020; Tamiru & Wagari 2021; Adane & Abate 2022) have successfully applied the latest versions of the HEC-RAS 2D model for flood inundation mapping and proved its ability to predict accurate results. However, these studies lack variation in inundation area with changes in areal DEM resolution and/or upstream boundary conditions (BCs). Thus, this study's main objectives were to study the effect of DEM resolution on inundation mapping, along with flood zone depths, and change in the inundation area with change in upstream boundary data.
The present work is carried out using version 6.0.0 of the HEC-RAS 2D model developed by the U.S. Army Corps of Engineers Center. This study identifies the flooded areas in the Karnataka, Telangana, and Andhra Pradesh regions of the Krishna River Basin, India, by developing an inundation map for the study area.
MATERIALS AND METHODOLOGY
The Krishna River rises in the Western Ghats north of Mahabaleshwar, Maharashtra, India at an elevation of 1,337 m. It flows through four states – Maharashtra, Karnataka, Telangana, and Andhra Pradesh – and discharges into the Bay of Bengal, covering approximately 8% of India's topographic area. The basin lies between longitudes 73°17′00″ E and 81°09′00″ E and latitudes 13°10′00″ N and 19°22′00″ N on the Deccan Plateau. Ghataprabha, Malaprabha, Tungabhadra, Bhima, Musi, and Munneru are Krishna River's primary tributaries. In 2009, for the first time in 60 years, heavy rain caused flash floods in north Karnataka and Andhra Pradesh, affecting approximately two million people and claiming about 210 lives.
Data collection and preparation
Hydraulic modeling with HEC-RAS involves geometric data to incorporate the physical characteristics of flood plain into the model. To achieve this objective, a DEM of the study area and discharge file for the boundary condition information are needed. The DEM should be of good quality to accurately utilize the topographic data and transform the flood plain accordingly. DEMs of 12.5 m and 30 m resolutions were incorporated for the inundation mapping. The freely available 12.5 m resolution DEM consisting of a Synthetic Aperture Radar (SAR) data set from ALSO PALSAR satellite was acquired from the Alaska Satellite Facility (ASF) website. The 30 m resolution DEM was downloaded from the U.S. Geological Survey Earth Explorer website with Shuttle Radar Topographic Mission (SRTM) data. The discharge files and water level data for the gauging stations were acquired from the Central Water Commission (CWC), Krishna Godavari Basin Organization (KGBO), Hyderabad, Telangana, India. The results from HEC-RAS 2D simulations are validated with data available from reports and past studies (Andhra Pradesh Water Resources Development Corporation 2009; Padmanabhan 2009a; Yarrakula et al. 2016).
Numerical model: HEC-RAS
HEC-RAS is one of the commonest hydraulic models used for inundation modeling. The model allows 1D steady flow simulations, and 1D and 2D unsteady flow calculations, sediment transport/mobile bed computations, and water temperature/water quality modeling (Brunner & CEIWR-HEC 2021). HEC-RAS 2D employs shallow water equations to describe the motion of water in terms of depth-averaged 2D velocity and water depth. The diffusion wave approximation approach is used for computing the flow field in the 2D mesh, as it leads to shorter computation time and may reduce model instability (Brunner 2016). The model area is discretized into grid cells, and HEC-RAS generates a detailed hydraulic property table for each cell and cell face. The water surface profiles provided by the model using several hydraulic design features can help decision-makers to invest resources effectively to prepare for catastrophes and improve the quality of life, by analyzing the extent of flooding and flood inundation zones. Arc-GIS v 10.8 was used to prepare flood depth maps.
Hydraulic modeling and inundation mapping are performed using RAS Mapper, a spatial data integration and mapping tool in HEC-RAS. In HEC-RAS Mapper, the development of an RAS terrain, laying out the geometric data, extracting the terrain data, and visualization of results in the form of maps and tables can be inaugurated. For this study, a new terrain model was implemented with DEMs of 12.5 m and 30 m resolution.
2D flow area and mesh generation
When simulating floods, hydraulic models are calibrated by varying Manning's roughness coefficient for channels and floodplains. For this study, the model was implemented using the values considered in the previous studies of the basin (Pallavi et al. 2022; Vashist & Singh 2022). Simulation run time and output accuracy are determined by mesh cell size and model simulation time step. Studies suggest limiting the mesh to one million cells because exceeding this might cause severe performance faults owing to memory allocation issues in small units (Goodell & Warren 2006; Goodell 2015). For the 12.5 m resolution DEM, 284,096 cells were generated for cell size dx and dy = 100 m and Manning's roughness coefficient (n) as 0.03. For the 30 m DEM, 211,983 cells were generated with similar specifications. There should only be one computation point in each mesh; otherwise, the mesh will show errors and have to be generated again for a smooth run.
Unsteady flow analysis
After mesh generation, BCs are established for the 2D flow area to perform the unsteady flow analysis. Three BCs, two upstream (Huvinhedgi and Yadgir) and one downstream (K. Agraharam), are added near the 2D flow area as shown in Figures 2(b) and 3(b) (blue lines). Flow data are added for the unsteady flow analysis, providing all the necessary details. For the upstream boundaries (Huvinhedgi and Yadgir), a time series file of discharge in the form of a hydrograph is supplied to the model. For the downstream boundary at Agraharam, a normal depth is provided.
Programs such as geometry preprocessor, unsteady flow simulation, postprocessor, flood plain mapping, and simulation period must be specified for the unsteady flow analysis. The computational settings containing computation interval, hydrograph output interval, mapping output interval, and detailed output interval must be furnished for the simulations. In the unsteady flow calculations, computation interval is one of the imperative parameters used (Patel et al. 2017). The computation interval/time step must be chosen in such a way that it is sufficient to maintain the accuracy and stability criteria as per the Courant condition (Brunner & CEIWR-HEC 2021) and produces acceptable results. The computation interval was varied from 30 minutes to 1 second to determine the optimum computation interval, and maintain the accuracy and stability criteria.
RESULTS AND DISCUSSION
Inundation area with DEM 12.5 m
Inundation area with DEM 30 m
The model was run for various computation intervals, and 1 second was found to be the optimum and thus adopted for the simulation. The 12.5 m DEM model took longer to generate its simulation than the 30 m DEM model. A 40–50% increase in computing time was observed for both DEMs as the computation interval was changed from 1 minute to 1 second.
Flood depth map
Effect of upstream boundary data on the inundated area
|Discharge decrease (%) .||Inundated area decrease (%) .|
|Discharge decrease (%) .||Inundated area decrease (%) .|
Validation of results
In developing countries, result validation is complex because data are scarce but some situation reports were available (Situation Report South India Floods 2009; Padmanabhan 2009a, 2009b), that were used for validation in this study. The reports enabled the flooded areas in Telangana, Karnataka, and Andhra Pradesh states to be identified. The areas worst affected were Bijapur, Gulbarga, and Raichur in Karnataka, and Kurnool and Mahbubnagar in Andhra Pradesh, of which Gulbarga, Raichur, and Mahbubnagar are discussed. As per the simulations done on 12.5 m DEM, the Mahbubnagar district area was worst flooded. The observations of worst-flooded district (from 12.5 m DEM) conform to the actual reports. Simultaneously, the simulation with 30 m DEM yielded the Gulbarga district as most flooded, while it was reported as the second-most affected district by the actual reports. The confluence of the Krishna and Tungabhadra rivers is the primary cause of flooding in Mahbubnagar, and nearby districts of Andhra Pradesh and Telangana. The Jurala dam also causes flooding in these areas but, due to lack of data, the dam's effect was not considered in this study.
This study demonstrates the use of HEC-RAS 2D to prepare a flood inundation map for the Krishna River Basin. Simulations were run with DEM resolutions of 12.5 m and 30 m to develop the inundation map for the floods in 2009. The main conclusion drawn from the study are:
The simulated results based on the 12.5 m resolution DEM showed that 322.49 km2 was inundated by the flood. The flood affected about 155 km2 in the Mahbubnagar district, the worst affected by the flood, which is in line with available reports.
The simulated results based on the 30 m resolution DEM showed that the flood inundated 281.03 km2, approximately 116 km2 in the Mahbubnagar district – i.e., this DEM under-reported the inundated area.
A 40–50% increase in simulation time was observed as the computation interval was reduced from 1 minute to 1 second for both DEMs. In addition, the computation time with the 12.5 m DEM was higher than for 30 m DEM.
The effect of upstream BCs shows that decreases in discharge reduce the inundated area in linear proportion.
Alaska's 12.5 m resolution DEM modeled the flood inundation area better than the USGS 30 m DEM. The areas identified from the inundation map with 12.5 m resolution DEM are validated by the available reports. If reasonable results are needed with less computational time, the use of a 30 m resolution DEM is suggested.
The authors are extremely thankful to the office of the Chief Engineer, Krishna and Godavari Basin Organization, Central Water Commission, Krishna Godavari Bhavan, Hyderabad, Telangana for providing data.
Conceptualization: K.V. and K.K.S.; Modeling simulation and investigation: K.V. and K.K.S.; Writing – original draft preparation: K.V.; Writing – review and editing: K.K.S.; and Supervision: K.K.S.
DATA AVAILABILITY STATEMENT
Data cannot be made publicly available; readers should contact the corresponding author for details.
CONFLICT OF INTEREST
The authors declare there is no conflict.