Variability and extremes of daily rainfall during the Indian summer monsoon in the period 1901–1989

Abstract

In this study, the variability and extremes of daily rainfall events during the Indian summer monsoon are investigated on the basis of a long observational record for the period 1901–1989, originating from the India Meteorological Department. The data are daily averages for 51 blocks, extending 2.5° in the meridional and zonal directions, except for the blocks on the Indian west coast with a width of 1°. Each block contains about seven stations. In addition to some basic properties such as the seasonal mean rainfall, frequency of wet days, and rainfall intensity on wet days, the variability of daily rainfall is described via gamma distribution. Extreme daily rainfall events, on the other hand, are investigated by quantiles and suitable theoretical extreme value distributions. The spatial distribution of the rainfall intensity on wet days during the Indian summer monsoon is very similar to the spatial distribution of the overall monsoon rainfall, with maxima on the west coast of the Indian peninsula and in Assam, and a small amount of precipitation in northwest India and in the southeastern part of the Indian peninsula. The gamma distribution is suitable for characterizing the time series of daily rainfall during the Indian summer monsoon by two parameters (i.e., the scale and the shape parameter). The spatial distribution of the scale parameter has a number of similarities with the spatial distribution of the rainfall intensity but shows an additional maximum in the westernmost part of India. A substantial amount of precipitation in association with heavy rainfall events is not only found on the west coast of the Indian peninsula and in Assam, where the maxima of the rainfall intensity are located, but also in the westernmost part of India and on the north coast of the Bay of Bengal. It is found that the generalized Pareto (GPA) distribution is generally better suited for describing the behaviour of heavy rainfall events during the Indian summer monsoon than the generalized extreme value distribution. This is a consequence of the substantial temporal variability of daily rainfall during the Indian summer monsoon, both on subseasonal and interannual time scales. A method is developed to objectively determine the locally varying thresholds that the daily data have to exceed in order to enter the fits to the GPA distribution. For this purpose, the GPA distributions for a set of locally varying thresholds are compared with a suitable quantile, an alternative description of heavy rainfall events.