Prediction of flow duration curves in ungauged basins

Abstract

The magnitude of streamflow in a river varies across sites depending on the geomorphological and climatic factors influencing hydrological processes in the contributing catchments. Knowledge of streamflow availability at target locations (hot spots) over different time frames is necessary for assessing the capacity to meet water demands for various purposes (e.g., designing hydropower schemes, devising strategies for flood control, and framing environmental flow standards to maintain ecosystems). Towards this, hydrologists construct a Flow Duration Curve (FDC) using historical flows at each hotspot to assess dependable streamflows visually. The curve provides information on the percentage of time that any specified magnitude of streamflow is equalled or exceeded. Hydrologists use regionalization procedures to predict FDC when the hotspots of interest are ungauged/sparsely gauged. The prediction involves the transfer of catchment and FDC-related information from an identified group of resembling gauged sites to the hotspot. For this purpose, various regionalization procedures are available, but none is established to be universally effective/efficient. The past studies have examined the efficacy of various regionalization procedures in predicting FDCs at ungauged sites in certain parts of the globe. However, such attempts on the Indian river basins are scanty, covering catchments that are widely scattered. Hence, there is a need to identify effective river basin-specific strategies for FDC prediction. Towards this, investigations are undertaken in this thesis on the catchments of the flood-prone Mahanadi River basin. The efficacy of various parametric and nonparametric FDC regionalization procedures is explored in predicting FDCs at ungauged and sparsely gauged target locations. Catchment-related attributes significantly influencing the flows are identified for establishing catchment similarity and estimating flows (quantiles) corresponding to different exceedance probabilities for predicting FDC at an ungauged site. In addition, the effect of various criteria for determining catchment similarity is explored, and the optimal number of similar sites/catchments required for effective transfer of regional information is determined. In the case of sparsely gauged sites, data from similar neighboring sites is used to arrive at FDC corresponding to a long-term, contemporaneous/concurrent period for use in regional hydrologic studies. Another novel contribution of this thesis is developing a Multiplicative Cascade model for predicting FDCs that harnesses the multifractal nature of streamflows. Knowledge of temporal variability in streamflow at ungauged target locations might be necessary for various practical applications, and FDC alone may not serve this purpose. To address this, one could explore generating the entire streamflow time series from the predicted FDC. An attempt is made in this direction by adapting an existing method and exploring multiple options for selecting neighbouring sites for drawing regional information. Their effectiveness is illustrated through application to the Mahanadi River basin. Future projections of FDCs in gauged and ungauged catchments of the Mahanadi River basin are derived by forcing a hydrological model (Soil and Water Assessment Tool) developed for the basin with hydrometeorological variables corresponding to two CMIP6 climate change scenarios: SSP 245 (intermediate scenario) and SSP 585 (worst case scenario). Future projections of water balance components in the basin are also determined. The research contributions from the work (including predicted streamflow and FDCs at ungauged catchments) are expected to aid in studies related to the long-term water resources planning and management in the Mahanadi River basin.