Projections of Extreme Dry and Wet Spells in the 21st Century India Using Stationary and Non-stationary Standardized Precipitation Indices

Abstract

The conventional approach to projecting meteorological extremes involves the application of indices such as the Standardized Precipitation Index (SPI) to rainfall simulated by General Circulation Models (GCMs). However, the sensitivity of SPI to the length of the records and the poor skills of GCMs in simulating rainfall are major drawbacks, leading to implausible projections. It is imperative to quantify and address these limitations before implementation of the approach. Here, we project the frequency of extreme dry and wet spells during the 21st century over India, incorporating special measures to alleviate the aforementioned limitations of the approach. We deploy kernel regression-based statistical downscaling to obtain improved 0.25-degree resolution monthly rainfall projections for India based on five GCMs and three emission scenarios (RCP2.6, RCP4.5, and RCP8.5) belonging to phase five of the Coupled Model Intercomparison Project. We also establish that the Standardized non-stationarity Precipitation Index (SnsPI), which incorporates changing climatic conditions considering linearly varying non-stationary scale parameter of the gamma distribution, is less sensitive to the length of the records as compared to SPI and we use both indices to obtain the frequency of future meteorological extremes. The results show an increase in the occurrences of extreme dry spells (EDS) over central, southeast coast, eastern region and some parts of northeast India. Differences between SPI and SnsPI based on the sensitivity are observed over, central India, where SPI overestimates EDS. Also, both the indices show diametrically opposite trends for areas under the influence of extreme wet spells in the future (2070-2099).