Spatial Variability of Long-Term Dependencies in the Precipitation for a
Basin-Scale under the Detection of Climate Change
Abstract
Long-term dependencies may be one of the reasons for the spatial
variability in precipitation frequencies. This study assesses the
long-term dependencies in precipitation time series at a basin-scale
using the wavelet-based fractal decomposition technique. The gridded
precipitation datasets (0.25deg x 0.25deg) from the India Meteorological
Department (IMD) for the year, 1901 to 2018 have been used. In order to
find the climate change point (i.e., the year in terms of annual series)
from each grid point, the mean-based change point detection is
performed. Based on the change points, the input for the wavelet
analysis is generated into two series, the series -1 (before change
point) and the series – 2 (after change point). The results of the
climate change points are different for every location, and the
corresponding length of the series also gets changed. In order to handle
the non-stationarity associated with the time series datasets, the
method of wavelet decomposition is used. The Discrete Wavelet Transform
(DWT) based fractal decomposition of time series is performed by taking
the Daubechies (db1 to db10) mother wavelet along with the varying scale
and translation parameters. Both the results of the series wavelet
coefficients are compared using the scaled ratio method and the relative
shift in the cumulative distribution functions (CDF). Comparing the time
series datasets before and after the change point reveals the
significance of long-term dependencies at each location. The results of
the spatial variability and its patterns explain the long-term
dependencies and their significance at a basin-scale, which may support
various scientific studies and development.