Hydrologic Extremes and Risk Assessment under Non-stationarity

Abstract:

In the context of hydrologic designs, robust assessment and communication of risk is crucial to ascertain a sustainable water future. Traditional methods for defining return period, risk or reliability assumes a stationary regime which may no longer be valid because of natural or man-made changes. Reformulations are suggested in recent literature to account for non-stationarity in the definition of hydrologic risk, as time evolves. This study presents a comparative analysis of design levels under non-stationarity based on time varying annual exceedance probabilities, waiting time of a hazardous event, number of hazardous events and probability of failure. A case study application is shown for peak streamflow in the flood-prone delta area of the Krishna River in India where an increasing trend in annual maximum flows are observed owing to persistent silting. Considerable disagreement is found between the design magnitudes of flood obtained by the different definitions of hydrologic risk. Such risk is also found to be highly sensitive to the assumed design life period and projections of trend in that period or beyond. Additionally, some critical points on the assumption of a deterministic non-stationary model for an observed natural process are also discussed. The findings highlight the necessity for a unifying framework for assessment and communication of hydrologic risk under transient hydro-climatic conditions. The concepts can also be extended to other applications such as regional hydrologic frequency analysis or development of precipitation intensity-duration-frequency relationships for infrastructure design.