Evidences of Significant Nonstationarity in Precipitation Extremes over Urbanizing Areas in India

Abstract:

The statistical assumption of stationarity in hydrologic extreme time/event series has been relied heavily in frequency analysis. However, due to the analytically perceivable impacts of climate change, urbanization and concomitant land use pattern, assumption of stationarity in hydrologic time series will draw erroneous results, which inturn effects the policy and decision-making. Past studies provided sufficient evidences on changes in the characteristics of Indian monsoon rainfall extremes and further it has been attributed to climate change and urbanization, which indicates the presence of significant nonstationary in the Indian monsoon extremes. Therefore, a comprehensive nonstationary frequency analysis must be conducted all over India to obtain realistic return periods. The present study aims to conduct a nonstationary frequency analysis of the precipitation extremes over India at 1^o resolution for a period of 1901-2004, with the implementation of the Generalized Additive Model for Location, Scale and Shape (GAMLSS) parameters. A cluster of 74 GAMLSS models has been developed by considering nonstationary in different combinations of distribution parameters and regression techniques (families of parametric polynomials and nonparametric/smoothing cubic spline), which overcomes the limitations of the previous studies. Further, for identification of urban, urbanizing and rural grids, an population density data has been utilized. The results showed the significant differences in the stationary and nonstationary return periods for the urbanizing grids, when compared to urbanized and rural grids. The results give implications of presence of nonstationary in the precipitation extremes more prominently in urbanizing areas compare to urbanized and rural areas.