Document Type : Research Paper
Authors
1 Department of Water Engineering, Agricultural Sciences and Natural Resources University of Khuzestan, Iran
2 Department of Nature Engineering, Agricultural Sciences and Natural Resources University of Khuzestan, Iran
3 Head of Innovation and Technology Development - Khuzestan Water and Power Organization, Khuzestan, Iran
Abstract
Keywords
Main Subjects
Effective flood management requires flood mapping, estimating potential damages and risks in flood-prone areas, and designing a comprehensive plan to mitigate flood risks. Understanding the phenomena and the impacts of changes on flow conditions, as well as predicting hydraulic events in rivers, plays a significant role in minimizing damages and losses. Modern methodologies leverage techniques such as remote sensing, geographic information systems (GIS), and hydraulic and hydrological models to simulate river flows. Given the lack of high-accuracy topographic maps in many areas of the country, this study aims to investigate the impact of the quality of topographic maps for the Karun River basin, including the riverbed and floodplains areas from Molasani to Farsiat, on flood inundation mapping.
The study area covers approximately 110 km of the Karun River, including three hydrometric stations: Molasani, Ahvaz, and Farsiat. For two-dimensional modeling in the HEC-RAS environment, a digital elevation map (DEM) of the study area is essential. Therefore, DEMs of the Karun River with varying resolutions were prepared using existing survey data and aerial imagery. In aerial maps, riverbed elevation is represented as the water surface elevation. Due to the unavailability of a detailed elevation map for the Karun Riverbed, the riverbed was constructed in the GIS environment based on existing cross sections. For evaluating the impact of topographic map quality on floodplain mapping, maps with resolutions of 30 m, 50 m, 100 m, and 150 m were used. Subsequently, flood inundation map was generated using HEC-RAS model based on the different DEMs. To investigate the efficiency of the different DEMs with varying resolutions, Sentinel-2 satellite imagery and 12 quantitative metrics were employed. These metrics include Proportion Correct (PC), Threat Score (TS), Odds Ratio (θ), Bias, False Alarm Ratio (FAR), Hit Rate (H), False Alarm Rate (F), Extremal Dependence Index (EDI), Heidke Skill Score (HSS), Pierce Skill Score (PSS), Success Ratio, and Odds Ratio Skill Score (ORSS).
Analysis of the performance of 16 scenarios modeled in HEC-RAS with varying pixel resolutions for the river and floodplain, focusing on the PC metric, showed that models with smaller river pixel sizes (30 m and 50 m) consistently achieved the highest PC values. For 30 m pixels, the PC was approximately 0.799, while for 50 m pixels, it was slightly lower (ranging between 0.785 and 0.791). Examination of the TS metric, which is suitable for rare event prediction, revealed that models with the highest river pixel resolution (30 m) consistently achieved the highest TS values (approximately 0.67) across all floodplain pixel sizes, indicating strong performance. Additionally, higher-resolution river pixels consistently yielded the highest Odds Ratios (θ), reflecting high prediction reliability. For 30 m river pixels, θ started at 17.69 for 30 m floodplain pixels and slightly decreased to 17.51 for 150 m floodplain pixels. Increasing river pixel size from 30 m to 150 m led to a consistent rise in Bias, indicating over-prediction tendencies in larger pixel sizes. FAR also increased significantly with larger river pixels, signifying more false alarms. For smaller river and floodplain pixels, FAR remained relatively low, ranging from 0.25 to 0.29, indicating fewer false alarms at higher resolutions. The ORSS analysis showed that smaller pixel sizes for both river and floodplain consistently yielded higher ORSS values, demonstrating superior skill.
Based on the main effect analysis of river pixel size, PC and TS scores decreased as river pixel size increased, particularly for floodplain pixel sizes of 100 m and 150 m. The Heidke, Pierce, and Gilbert skill scores also decreased with larger river pixel sizes, with Gilbert’s score showing a steep decline for larger river pixels, reflecting weak flood prediction performance at lower resolutions. Bias increased with larger river pixel sizes, indicating a tendency for over-prediction. FAR followed a similar rising trend. Regarding the main effect of floodplain pixel size, PC and TS scores declined as floodplain pixel size increased, particularly for river pixel sizes of 100 m and 150 m. The drop in TS suggests that lower floodplain resolution reduces the model’s ability to accurately predict floods. All skill scores decreased with larger floodplain pixel sizes, especially in scenarios with 100 m and 150 m river pixels. Increased floodplain pixel sizes also resulted in higher Bias, indicating a greater tendency for over-prediction in lower floodplain resolutions.
Conceptualization, Methodology, Formal analysis, Writing Original Draft, J.Z.; Methodology, Writing - Review & Editing, A.J.; Software, Writing - Review & Editing, M.Ch. and MJ.N. All authors have read and agreed to the published version of the manuscript.
Data available on request from the authors.
The authors would like to thank the reviewers and editor for their critical comments that helped to improve the paper. The authors gratefully acknowledge the support and facilities provided by the Agricultural Sciences and Natural Resources University of Khuzestan [Grant number:1/411/1078].
The authors avoided data fabrication, falsification, plagiarism, and misconduct.
The author declares no conflict of interest.
)