Lack of uniform trends but increasing spatial variability in observed Indian rainfall extremes

Abstract

Recent studies disagree on how rainfall extremes over India have changed in space and time over the past half century 1-4 , as well as on whether the changes observed are due to global warming 5,6 or regional urbanization 7 . Although a uniform and consistent decrease in moderate rainfall has been reported 1,3 , a lack of agreement about trends in heavy rainfall may be due in part to differences in the characterization and spatial averaging of extremes. Here we use extreme value theory [8][9][10][11][12][13][14][15] to examine trends in Indian rainfall over the past half century in the context of long-term, low-frequency variability. We show that when generalized extreme value theory 8,16-18 is applied to annual maximum rainfall over India, no statistically significant spatially uniform trends are observed, in agreement with previous studies using different approaches 2-4 . Furthermore, our space-time regression analysis of the return levels points to increasing spatial variability of rainfall extremes over India. Our findings highlight the need for systematic examination of global versus regional drivers of trends in Indian rainfall extremes, and may help to inform flood hazard preparedness and water resource management in the region.There is considerable debate in the recent literature about the nature of space-time trends in extreme rainfall over India 1-3 and their attribution to aspects of global change, specifically, global climate change 5,6 versus regional urbanization patterns 7 . Previous researchers have drawn a variety of conclusions 1-4 about trends in rainfall extremes during the Indian monsoon from a regular 1 • ×1 • (or similar) gridded daily rainfall dataset over India for 1951-2003. The differences can probably be attributed to the corresponding definitions of extremes, levels of spatial aggregation and areas of coverage. The use of fixed thresholds over a 12 • × 10 • box labelled as Central India suggested an increasing trend in rainfall extremes concurrent with decreasing moderate rainfall, resulting in no discernible net trends 1 . However, the use of variable percentile-based thresholds over each individual 1 • × 1 • grid 2 , analysis based on homogeneous regions 3 and analysis with a percentile-based definition of the frequency and intensity of rainfall extremes 4 showed no discernible spatially uniform trends in rainfall extremes over India. Field significance tests do not statistically support the hypothesis of increasing trends in heavy rain events 2,3 , and the region labelled as Central India in ref. 1 may not be meteorologically homogeneous 19 . However, the use of additional data (1901-2004; ref. 5) confirmed the findings of ref. 1 when identical definitions of extremes, aggregations and coverage were used. The often conflicting insights about Indian rainfall extremes in the recent literature point to the importance of effective characterization of extremes, especially for understanding and communicating their relevance to impacts and policy.Here we show that rainfall extremes over I...

Trends in Rainfall and Peak Flows for some River Basins in India

“…They have reported that in spite of considerable year-to-year variability, there are significant increases in the frequency and the intensity of extreme monsoon rain events in central India over the past 50 years and the observed trends suggest enhanced risks associated with extreme rainfall over India in the coming decades. The findings of Ghosh et al 23,24 were similar to those of Goswami et al 19 . Lacombe and McCartney 25 found changes in rainfall pattern aligning with the geography of anthropogenic atmospheric disturbances and confirmed the paramount role of global warming in recent changes for daily gridded rainfall data .…”

Logit-normal mixed model for Indian monsoon precipitation

“…While validating their 2006 study, Ghosh et al (2012) indicated increasing spatial variability in observed Indian rainfall extremes. They also found that moderate rainfall increased in central India despite a decreasing trend in occurrence of moderate rainfall.…”

Investigating the variation and non-stationarity in precipitation extremes based on the concept of event-based extreme precipitation

“…In practice, the percentile-based method were widely used in many regions over the world because of its efficiency and simplicity (Bell et al, 2004;Wang and Zhou, 2005;Du et al, 2013). In most cases, the EP was based on the 90th (Gemmer et al, 2011), 95th (Bell et al, 2004;Zhai et al, 2005), 97.5th (Wang and Zhou, 2005) or 99th (Gemmer et al, 2011) percentiles of daily precipitation data, which represent moderate to extremely unusual events and were demonstrated to be appropriate in various regions (Alexander et al, 2006;Ghosh et al, 2012;Vittal et al, 2013). Three indicators, i.e., frequency (F), intensity (I) and extreme volume (EV), are often used to describe the characteristics and detect the trends of EP.…”

“…However, the functional relationships between parameters and the time need to be specified as a priori, and such information might not be available or persistent over time due to the complexity of climate system (Kao and Ganguly, 2011;Ghosh et al, 2012). Alternatively, the distributional parameters were estimated separately for a series of Y-years moving window, rather than for the entire extreme series (Kao and Ganguly, 2011;Ghosh et al, 2012;Vittal et al, 2013).…”

“…However, the functional relationships between parameters and the time need to be specified as a priori, and such information might not be available or persistent over time due to the complexity of climate system (Kao and Ganguly, 2011;Ghosh et al, 2012). Alternatively, the distributional parameters were estimated separately for a series of Y-years moving window, rather than for the entire extreme series (Kao and Ganguly, 2011;Ghosh et al, 2012;Vittal et al, 2013). A choice of 30 years of window size was recommended by Kao and Ganguly (2011) considering that it was expected to smooth out the effects of most multi-decadal climate oscillators and providing more confidence for low frequency extremes (Kharin et al, 2007).…”