Characterization of daily precipitation extremes over india

Abstract

Daily precipitation extremes play a critical role in the hydrological planning and design of major water control structures and are expected to show a changing tendency over time due to climate change. The magnitude and frequency of extreme precipitation can be assessed by studying the upper tail behavior of probability distributions of daily precipitation datasets. These unexpected low-probability events lie within the tail part and have unprecedented consequences, underscoring the importance of their accurate estimation and prediction. The primary challenge with conventional distribution fitting approaches arises from the limited availability of data pertaining to extreme events. Because of this constraint, these methods struggle to effectively model the tails of daily precipitation data, often categorizing extreme precipitation events as improbable outliers. Consequently, this leads to an underestimation of their likelihood of occurrence. An appealing approach to overcome this impediment is the assessment of the tail behavior using some modern-day techniques like advanced threshold-based approaches, Quantitative or Scalar diagnostic tools, and Graphical approaches. The thesis concerns the development of novel approaches that can assess the tail behavior of precipitation extremes, thereby overcoming the limitations associated with old approaches. Characterizing the tail behavior of the daily precipitation finds use in the design and risk assessment of water control structures, economic evaluation of flood protection projects, flood insurance assessment, land use planning and management, and operation of irrigation projects. In the context of climate change, a better understanding of the climate extremes in terms of their frequency, magnitude, and spatial and temporal variation is necessary to evaluate the implications for risk and resilience. Hence, this thesis presents the study carried out to deliver a comprehensive assessment of extreme climatic conditions in India using some novel advanced approaches. The initial part of the thesis is devoted to the application of threshold-based approaches to characterize the daily precipitation datasets over India. The investigation is carried out using the approach proposed by Papalexiou et al. (2013), where a Probability ratio mean square error (PRMSE) norm, is used to identify the best-fitting distribution to the tails of daily precipitation. Analysis related to the spatial-temporal change in the tail behavior of daily precipitation over India from pre- to post 1970 time periods as per the global climatic shift is done. The results indicate that the heavy-tailed distribution fits the tails of daily precipitation for the majority of the grids over India and an increase in the heaviness of tails of daily precipitation data over India from pre- to post-1970 time periods is observed. In the second part, an empirical index known as the “Obesity index” (OB) that can provide a quantitative comparison between two distributions by alleviating the shortcomings associated with the threshold-based approaches is developed. The OB-based approach is applied to discern the probability distribution of daily gridded precipitation data for historical (1951–2004) and future (2006–2099) periods over India into light- and heavy-tailed. Future projections of daily precipitation were obtained by downscaling simulations of the Coordinated Regional Climate Downscaling Experiment. Subsequently, a comparative analysis between the OB-based approach and threshold-based approaches by Nerantzaki and Papalexiou and Papalexiou et al. was conducted. Finally, the application of the OB based approach is extended to characterize daily precipitation in Indian Meteorological subdivisions. Furthermore, we explored the dependence of the OB on the elevation of grids. Results indicated the applicability of heavy-tailed distributions in the representation of daily precipitation over India and suggested an OB-based approach as a good alternative diagnostic tool for assessing tail behavior. The development of the Comprehensive Decision support system (DSS) was uptaken in the next part of the work, where several advanced graphical methods like Concentration profile (CP) plot, Concentration adjusted expected shortfall (CAES) plot, Zenga plot, Maximum-to-Sum plot, and Discriminant Moment ratio plot were incorporated together. Incorporation of advanced tools alleviates the limitations like lack of efficient segregation of the Lognormal distribution from the Regularly varying and Subexponential distribution families, associated with the conventional DSS. The robustness of the proposed DSS is established through a simulation experiment while the application was done to characterize the tails of daily gridded precipitation data over India. It is observed that about 98% of grids over India exhibit distributions from heavy-tailed families, which is of paramount concern as this shows higher frequency and magnitude of extreme over the Country. The final portion of the study is aimed at discussing a comprehensive framework for estimating the risk associated with the tails of the daily precipitation datasets. Inferences from the novel approach like Concentration Profile (CP) are combined with the standard results from utility theory to develop a tool known as a Concentration Map (CM), that assesses the riskiness of datasets taking into account the variability of the larger and most relevant events. Risk embedded into the tails was evaluated for gridded precipitation datasets for the historical time period (1901–2019) from Indian Meteorological Department (IMD), while the simulations from 16 General Circulation Models (GCMs) participating in the Coupled Model Intercomparison Project phase 6 (CMIP6) under four Shared Socioeconomic Pathway (SSPs), namely, SSP126, SSP245, SSP370 and SSP585 are considered for future (2020-2100). The potential spatial and temporal variation of tail risk is done by comparing tail risk estimates from CMIP6 experiments (SSP126, SSP245, SSP370, SSP585) with historical datasets. Results highlight an overall increase in tail risk, particularly in scenarios indicative of anthropogenic influences, Furthermore, the analysis is extended to assess the variation in the embedded tail risk associated with daily precipitation datasets across different meteorological subdivisions and climate zones based on a Köppen-Geiger (KG) climate classification system, during different periods. In a changing climate, understanding extreme precipitation events and their associated risks has become increasingly crucial. This study has employed advanced techniques and tools to illuminate the complexities of India's climate. The findings of this research can serve as a valuable guide for policymakers in preparing for a future marked by more frequent and severe weather events. Local decision-makers can use the information provided in this thesis to effectively address the challenges presented by shifting climate patterns and formulate appropriate adaptation strategies in their respective regions.