Masterclass: Stationarity Assessment of Hydroclimatic Extremes: Methods and Applications
Convenors: Dr. Priyank J. Sharma, India, Prof. Ramesh S. V. Teegavarapu, USA, Ms. Achala Singh, India
Long-term hydroclimatic series are evaluated in research studies focused on climate change and variability assessments. In general, hydrologic design relies on the assumption of stationarity of hydroclimatic extremes and its assessment becomes an essential initial task. Stationarity, in the context of design floods, may imply their time invariance and the constant probability of failure of a given water resource structure for its entire design life. However, the assumption of stationarity may lead to over- or under design, in cases where the time series is indeed non-stationary. Stationarity, a cornerstone in hydraulic design, is now under scrutiny due to anthropogenic activities and climate change. Non-stationarity is also attributed to several factors such as human interventions (e.g., land use and cover alterations, reservoir regulations), occurrences of sporadic natural hazards (e.g., forest fires, volcanic eruptions, earthquakes), the low frequency components of oceanic-atmospheric phenomena (e.g., Pacific Decadal Oscillation, Atlantic Multidecadal Oscillation, and El Nino-Southern Oscillation), and global warming. Such non-stationarities may influence (1) changes in the distribution as a whole, wherein the average value changes without altering the variability; (2) variability changes, though the average value remains unaltered; (3) changes in both the average and the variability over time; and (4) more complex forms such as evolving asymmetry over the time or progression of a distribution from one parametric form to another. Thus, understanding of the causal mechanisms influencing changes in a time series would affirm non-stationarity, while the prevalence of uncertainty in examining the causes of such change would present a stationary narrative. Methods that rely on the presence of statistically significant trends or change points to derive inferences about stationarity may fail to check for all the time-invariant characteristics of time series. This masterclass provides an overview of conventional trend, change point and unit root methods for stationarity assessment vis-à-vis the novel non-overlapping block stratified random sampling (NBRS) approach, formulated by coupling several non-parametric tests in a multi-criteria decision-making environment, to comprehensively evaluate all the time invariant characteristics of hydroclimatic extremes.