Introduction
In recent decades, as climate change has intensified, flood risks pose a greater threat to countries than ever before. These events have resulted in significant losses, including the tragic loss of thousands of lives and substantial economic damages due to severe and flash floods. Rapid estimation of flood extents is one of the most important and crucial things in emergency management and reducing damages caused by floods. To mitigate risks, damages, and challenges, leveraging available data and meteorological forecasts is crucial for better identifying flood extents and pinpointing the exact locations of impacts. This information empowers us to take prompt and necessary actions to reduce risks and minimize damages effectively. With the advancement of computers, achieving a method for precise and rapid flood mapping is now possible. For this purpose, extensive studies have been conducted in the past, and various hydraulic, hydrological, and empirical models have been proposed and examined. Existing models and methods for delineating flood zones are often hydraulic, involving heavy computational load and consuming significant time for calculations. Alternatively, they may be hydrological or empirical, lacking desirable accuracy and sometimes speed, making it challenging to provide precise and timely information on flood inundation areas.
Methodology
In this study, the 2D-shallow water equations were solved by developing a new user-friendly module of STE software. The impact of parallel processing and various computational grid spacing on the speed and accuracy of the obtained flood-prone areas has been investigated. The model results were compared with Sentinel-2 satellite and HEC-RAS results for the 2019 flood in the Golestan province, Gorganrood River. In order to model 2D surface water flow, the equations under consideration, based on assumptions derived from the Navier-Stokes equations, are known as the shallow water equations. The key to simplifying these equations is to locally approximate the inertial term based on the assumption of the negligible contribution of the convective acceleration term compared to the other terms of the equations. Therefore, it is possible to neglect the mentioned term, and the simplified equations, named local inertial approximation, can be written as the relations (4) to (6). These equations have been solved in the developed software in this study using the finite difference method in both implicit and semi-implicit schemes.
In this study, new user-friendly module of STE software have been undertaken with an extension named STEGIS. This extension is tasked with conducting geographical analyses and performing all GIS operations and two-dimensional calculations, presenting the results, and analyzing them. In the development of this extension, efforts have been made to incorporate all the necessary tools for users in a way that eliminates the need for separate software such as QGIS, Arcmap, HEC-RAS, HEC-HMS, etc., during hydraulic, hydrological, and other modeling and analyses. Therefore, the aim is to make all the necessary tools for working with raster, vector, and geographical files, as well as performing hydraulic, hydrological, and topographic calculations and analyses, readily accessible to the user in this extension and making it easier to use. For more information, please visit www.en.ste.hwstr.ir.
Given the binary nature of the observed data (flood zone extracted from satellite imagery) being either zero or one, it is necessary to assess the accuracy of the models using specific parameters designed for evaluating binary data. For this purpose, parameters such as accuracy percentage and point percentage have been employed.
Results and Discussion
In this study, the 2D module of the STE software was developed to investigate the impact of parallel computing and various computational grid spacing on the speed and accuracy of flood simulations. Flood zones and their extents were estimated using the two-dimensional shallow water equations. The obtained results were compared, discussed, and evaluated against the outcomes of HEC-RAS software and the flood zone recorded by Sentinel-2 satellite for the flood event in the Gorganrood River, Golestan province, in March-April 2019. In summary, the results of this research demonstrated that the developed STE software provides more accurate results compared to HEC-RAS software across all computational grid spacings. Meanwhile, the time required to complete flood simulation in the STE software is nearly 95% less than that in HEC-RAS software. In both software applications, precision and the time required to complete the simulation increased with a reduction in the computational grid spacings. Values ranging from 10 to 15 times the size of reference topographic map cells have proven to be the optimal range for computational grid spacings in both software applications. Utilizing these values, computation can provide sufficient accuracy along with an appropriate time frame for completing the simulations. The time required to complete the simulation using various discretization methods in the STE software is nearly the same, but these methods significantly impact the accuracy of the simulation. The Hybrid method provides higher accuracy at different computational grid spacings. Parallel computing and simultaneous computation of different sections of the two-dimensional solution domain only result in increased computational speed and reduced time needed for completing simulation, especially in shorter computational grid spacings. This approach has no impact on the accuracy of the simulation and other considerations. The algorithm implemented in the STE software for adjusting the time step has successfully, maximized the time step in order to minimize the overall simulation time and prevented computational errors in water volume estimation which makes the computations more stable.
Conclusion
Results indicate that the developed software in this study, apart from better accuracy, demonstrates significantly higher speed in flood simulations and delineation of flood zones. It has the capability to reduce the time required to complete simulation and provide precise flood extents by up to 95% compared to the HEC-RAS software. Parallel computing proves highly effective in reducing the necessary time for modeling completion. Moreover, increasing the computational grid spacings up to 15 times the cell sizes of the reference topographic map results in both satisfactory simulation speed and good accuracy.