资源环境科技发展态势分析平台

Cluster analysis based regionalization procedure is proposed to construct a regional envelope curve for use with Hershfield probable maximum precipitation (PMP) estimation method. The procedure suggests (1) delineation of the study area into homogeneous PMP zones (regions) by performing cluster analysis on attributes affecting PMP, and (2) construction of regional envelope curve for each of the zones, and use of the curve(s) as the basis to estimate regional frequency factor (RFF) and PMP for site(s) in the study area. Conventional practice suggests construction of a single envelope curve for entire study area, which could be a political/physiographic area, or a river basin. This practice could yield unreliable RFF and PMP estimates if there is considerable variation in topography, landscape and/or climate in the study area. Effectiveness of the proposed regionalization procedure is demonstrated through a case study on 0.25 degrees resolution rainfall data covering India. Global K-means (GKM) and global fuzzy c-means (GFCM) methods are considered for cluster analysis. The number of delineated homogeneous PMP zones is 20 with GKM method and 17 with GFCM method. The PMP estimates obtained for 0.25 degrees resolution grids using the proposed procedure are generally found to be lower than those obtained by considering (1) one/two envelope curves for the entire India determined based on zones considered in previous studies, and (2) envelope curves prepared for four major river basins (Mahanadi, Godavari, Krishna and Cauvery) in peninsular India. The PMP estimates are often considered as the basis to arrive at probable maximum flood (PMF) for design and risk assessment of large-scale hydrological structures in river basins. Hence, PMF estimates could be expected to be lower when based on PMP estimates obtained using the proposed procedure, which is likely to have economic implications in design of the structures.