A Framework for Evaluating Hydrologic Risk of Dams and Dam-Break Floods Under Data-Sparse and Climate Change Scenarios

Abstract

Design Flood Hydrograph (DFH) desirable for studies related to the planning, design, operation, and risk assessment of a dam is often derived using the Unit Hydrograph (UH) approach. It involves the convolution of UH of the dam’s catchment with its Design Storm Hyetograph (DSH). The UH approach assumes that (i) the intensity of rainfall is constant within the effective duration of the storm, and (ii) the rainfall is uniformly distributed over the catchment. To avert the violation of the first assumption, the rainfall observations should be obtained at fine temporal resolution, as it facilitates considering shorter duration uniform intensity pulses in DSH. Likewise, to avoid undermining the second assumption, the rainfall observations should be obtained at fine-spatial resolution that facilitates considering smaller drainage areas for developing UH. Against this backdrop, a new methodology is proposed for deriving hourly rainfall at fine spatial resolution in data-sparse regions from remote sensing precipitation products. Its utility is demonstrated through a case study on the Karnataka state (India) through split-sample validation. The DSH required for convolution with UH could be derived corresponding to a specified return period from Intensity-Duration-Frequency (IDF) relationships for dams having low to significant hazard potential. Those relationships can be represented in the form of empirical and semi-empirical formulations and there is an ambiguity in their choice, as none of them is established to be universally superior. Against this backdrop, a novel methodology is proposed that represents the IDF relationship as a non-separable function of return period and duration. Its effectiveness is elucidated through Monte Carlo simulation (MCS) experiments and application to precipitation records from the United States and India. In the case of an ungauged catchment, UH required for convolution could be obtained synthetically using GcIUH derived based on geomorphometric characteristics of the catchment. A new variant of GcIUH referred to as EGcIUH is contributed to obtain UH (and consequently DFH) under climate change scenarios. The proposed variant considers the self-similarity hypothesis to alleviate the uncertainty in parameter estimates of GcIUH arising from stream network demarcation from a DEM. Furthermore, it overcomes the ambiguity associated with the choice of equations for estimating the dynamic parameter (characteristic flow velocity) of GcIUH by specifying an appropriate equation to be considered for different precipitation intensity ranges. The potential of EGcIUH in simulating runoff hydrograph resulting from an observed isolated storm event is demonstrated for typical catchments in the United States and India by considering finer (sub-daily) as well as coarser (daily) resolution pulses in the storm’s hyetograph. Investigations with several (eighteen) distributions revealed that the shape of EGcIUH can be represented by various probability density functions other than the conventional gamma distribution. Conventional methodologies used for risk analysis of dams do not account for uncertain, ambiguous, and vague information on three risk indicators (likelihood of dam-break flood, severity of consequent hazard, and exposure to flood risk). Further, in quantifying exposure to flood risk there aren’t any attempts to consider the socio-environmental and ecological consequences/losses, whose effects are irreversible. It is because of the lack of methodologies that have provision to consider such qualitative losses in conjunction with the conventionally considered quantitative (e.g., economic) losses. A fuzzy framework is developed for risk analysis of dam-break floods that addresses these issues. It involves estimating a newly proposed Modified Aggregate Risk Index (MARI) that harnesses the advantages of both static and variable fuzzy set theories to integrate uncertain/ambiguous/vague information on the aforementioned three risk indicators. The MARI is developed by modifying a conventional index, which solely relies on the static fuzzy set theory. MARI makes use of the exposure index (Ie) which is also novel, it assimilates information on both qualitative and quantitative exposure indicators utilizing weights that could be determined using different options (e.g., conventional and fuzzy analytical hierarchy process, simplex binary comparison method). The effectiveness of the proposed framework is illustrated by utilizing it to determine future projected changes in Dam Break Flood (DBF) risk for a typical large (Hemavathi) dam in India, considering climate simulations from eleven GCMs, corresponding to the four climate change scenarios. Village-wise risk maps prepared using the proposed MARI indicate vulnerable villages and towns that are susceptible to potential hazard due to DBF. The maps would be useful to dam owners and policymakers in prioritizing villages at high risk for implementing flood risk mitigation/management strategies.