Asymmetric Relationship of Nino Indices with Rainfall Extremes over Western Ghats and Coastal Region of Karnataka

Abstract

Climate variability and change has increased extreme rainfall events. There is an underreporting and limited analysis, which often have significant impact with extreme rainfall events at regional scale. The magnitude of variability of the rainfall extremes varies according to locations. Among subdivisions of Western Ghats of India maximum rainfall occurs over Coastal Karnataka. Examining the extreme events of rainfall provide an idea of the probable occurrence of severity conditions in future in the context of changing climate. Extreme rainfall indices to identify the variation of rainfall patterns such as the number of rainy days, total rainfall, daily intensity index, one and five-day maximum rainfall, dry spells and threshold intensity rainfall frequency indices were considered as per the norms suggested by Expert Team on Climate Change Detection (ETCCDI) of Intergovernmental Panel on Climate Change (IPCC). These rainfall extremes indices are analyzed using IMD gridded high resolution daily rainfall data for the period 1901-2013. Statistical trend analysis techniques namely Mann–Kendall test applied for extreme rainfall indices and Theil-Sen estimator perceive nature and magnitude of slope in rainfall indices. The trends show contrasting spatial variations of extreme rainfall indices in Coastal region (low land) and Western Ghats (high land) regions of Karnataka. The changes in daily rainfall events in the lowland region primarily indicate statistically significant (varies from 95% to 99.9% confidence level) positive trends in the annual total rainfall, 1-day, and 5-day maximum rainfall, frequency of very heavy rainfall, and heavy rainfall as well as medium rainfall events. The seasonal variation of rainfall exhibits mixed trend, however significantly rising trend is witnessed in the southern coastal plains and the adjacent Western Ghats region during the pre-monsoon. But, southern coastal plains show a decreasing trend in the monsoon period (JJAS). Furthermore, the overall annual rainfall strongly correlated with all the rainfall indices in both regions, especially with indices that represent heavy rainfall events which are responsible for the total increase of rainfall. The interannual variability of rainfall and its extreme events over study region is observed to be associated with ENSO cycle, whereas Nino indices are asymmetric over the study region. The trends in ETCCDI extreme rainfall indices analyzed as an issue of climate change and the possible teleconnection with the ENSO mode as a concern of natural climatic variability have been analyzed over the study region. Nevertheless, differences are foundii between the spatial extent of correlation coefficients and their magnitudes. Using most significant time lag between the extreme rainfall indices (dependent variable) and the November-January (ONDJ) seasonal average of Niño indices (independent variable). The best model with the highest coefficient of determination was identified by Step wise regression analysis. The teleconnection between the Niño indices (Niño 1+2, Niño 3, Niño 3.4 and Niño 4) and the rainfall extremes with 0-year and 1-year ahead are at different phases, regional response of rainfall extremes to these indices are dissimilar. This analysis provides insights into regional response of rainfall extremes to global climate indices over the study region. The large-scale phenomenon over the pacific ocean with rainfall over the study region provide a scientific basis for understanding and developing credibility in future regional climate. A significant lag correlation between the summer monsoon rainfall and Niño indices was revealed by the seasonal lead-lag correlation analysis, Niño 3(t-4) at 90% confidence level, remaining Niño 3.4(t-2), Niño 4(t-2), and Niño 4(t-3) at 95% confidence level shows a significant relationship at respective lag period from onset of summer monsoon rainfall. In order to investigate the combined lagged effects of the potential climate predictors for monsoon rainfall using multiple linear regression as a linear method compared to neural network as a nonlinear method have been employed to examine the predictability of the summer monsoon rainfall. The principal component analysis of predictors aids to represent in one-dimensional space using the eigen vector that corresponds to the covariance matrix’s largest eigen value. Whereas first principal component explains about 72% of the variance of the predictors. Thus, PC1 considered as predictor in regression equation and input layer in neural network models to avoid over fitting. The attained prediction on the basis of the overall performance of the prediction models, feed forward neural network model shows a better prediction compared to other models with a good correlation coefficient and RMSE of 0.53 and 1.6 for training case, and 0.72 and 1.63 for testing case, respectively. From the time series analysis for period 1951-2013 of standardized monsoon rainfall Index selected the positive episodes values having standardized value greater than +1 (excess) and similarly with negative episodes values with standardized values less than -1 (deficit). The mean anomalous SST values for the region Nino 3.4 for the season DJF (- 2) for positive episode is 0.1719oC and the negative episode is -0.5133oC. The two SST means are significantly different at confidence level of 87.15% through the Student’s t-test.iii In awaken of climate change, this study is a contribution in the on-going research of extreme events over mountainous terrain including disaster management study. The sequential daily rainfall extremes and other atmospheric parameters may be utilized for the now-casting of extreme rainfall events. Further the relationship between topography and other atmospheric parameters influence for rainfall extremes should be studied separately to get better insight. This research may also be useful for the modifications in rainfall extremes retrieval methods over the Western Ghats mountainous terrain.