Affiliations: [a] Centre for Environmental Science and Engineering, Indian Institute of Technology Bombay (IIT-B), Mumbai 400 076, India | [b] Applied Statistics and Informatics, IIT-B, Mumbai 400 076, India | [c] Interdisciplinary Programme in Climate Studies, IIT-B, Mumbai 400 076, India | [d] Centre for Urban Science and Engineering, IIT-B, Mumbai 400 076, India | [e] Department of Civil Engineering, IIT-B, Mumbai 400 076, India
Abstract: The Nonstationary analysis drew formidable attention to the flood frequency analysis (FFA) research community due to analytically perceivable impacts of climate change, urbanisation and concomitant land use pattern on the flood event series. Albeit, the inclusion of nonstationarity in FFA significantly enhanced the accurate estimation of the return period, however, its application is questionable when the flood variables (FV) are not having persisting significant nonstationarity. In such cases, the assumption of stationarity is still valid and will direct to accurate estimation of the flood quantiles. Hence, prior to conducting the comprehensive FFA, it is vital to inspect the existence of stationarity/nonstationarity in the FV. This can be accomplished by a comprehensive trend analysis. The aim of present study is to emphasize the importance of a comprehensive trend analysis during FFA by proposing a framework to conduct the same. Further, the proposed framework has been demonstrated on unregulated daily streamflow series of two gauging stations, at the Kanawha Fall of Kanawha River, West Virginia, USA, and at the Baltara gauging station of Kosi River, Bihar, India. The results show that the annual maxima (AM) delineated flood peak series has a significant trend in both the gauging stations, providing sufficient evidence of nonstationarity, which is modelled by first- and second-order nonstationary analyses. A comparison between first-order and second-order nonstationarity analyses has also been performed, which suggests higher order nonstationary analysis might give more accurate information on the occurrence of flood extremes. Overall, our study highlights that the proposed framework is an important initial step before initiating FFA to avoid the ambiguity between the selection of stationary and nonstationary analysis.
