| dc.description.abstract | Probable maximum precipitation (PMP) estimation has significance, as floods triggered by PMP are widely used for risk analysis of critical infrastructures (e.g., large dams and nuclear power plants), whose failure could have catastrophic consequences for the environment, ecology, and economy. Conventionally, PMP estimates are obtained using data-intensive stationary hydrometeorological methods, which require information on several climate variables. Less data-intensive statistical methods are preferred when reasonably long precipitation records are available. However, if even the precipitation records are sparse, practitioners encounter impediments in arriving at effective PMP estimates with the use of conventional methods. To address this, a novel variant of the Bethlahmy (non-parametric) method is contributed in this thesis. Its potential is illustrated over existing statistical methods (original Bethlahmy method, Hershfield method, and conventional probabilistic approach) and their variants through Monte Carlo Simulation experiments and application to 37,872 stations across the globe. Another issue that deserves attention is the development of a methodology for the quantification of uncertainty in PMP estimates, which is at a nascent stage. Very recently, some methods attempted to quantify uncertainty in analysis with the Hershfield method (HM), which is opted when sites in a region have reasonably long precipitation records. Their focus has been only on sampling uncertainty, though there are additional sources of uncertainty due to new variants that have emerged for analysis at various stages of the method. To address this research gap, a novel imprecise probability framework is contributed, which facilitates (i) quantifying the overall uncertainty in HM-based PMP estimates arising from multiple sources and (ii) disintegrating the uncertainty into contributions from individual sources and their combinations. Four uncertainty sources were considered, which include (i) the already identified sampling uncertainty and additional uncertainty arising from the options for (ii) defining meaningful zones of sites in the study area, (iii) preparation of envelope curve space, and (iv) construction of the curve. The efficacy of the proposed framework is demonstrated through a case study over two major flood-prone river basins (Mahanadi and Godavari) in India. Results reveal that the major contribution to the overall uncertainty is not from any single source but from combinations of multiple sources. Results further revealed that HM-based PMP estimates improved with (i) delineation of the study area into zones, (ii) omission of outlier sites, and (iii) use of peaks over threshold series of precipitation instead of annual maximum series. In recent decades, there has been an impetus for customization of the conventional stationary PMP estimation methods to account for non-stationarities arising in hydrometeorological variables due to climate change. Limited attempts made in this direction have several shortcomings. To alleviate those, a novel climate change factor-based moisture maximization method (CCF-MMM) is proposed. It customizes the conventional moisture maximization method (MMM) by introducing a climate change factor (CCF), which is determined using five proposed options. The first three options are developed by leveraging ideas underlying existing strategies, whereas the last two options are built on ideas underlying the recent concepts of the non-stationary return period of floods. The efficacy of the CCF-MMM over conventional MMM and its existing non-stationary variants is illustrated through a case study on 24 hydrometeorological subzones covering India. | en_US |
