THE DETECTION AND MODELING OF MULTINONSTATIONARITY FOR ACCURATE ASSESSMENT OF LONG-TERM HYDROLOGIC RISK
Gilroy, Kristin Leigh
McCuen, Richard H.
MetadataShow full item record
Climate change and urbanization are nonstationary factors that influence hydrologic data, which results in the concept of multinonstationarity in hydrologic data. Methods to deal with important aspects of multinonstationarity do not exist. Currently, a statistical method to detect multinonstationarity in a hydrologic time series is needed. Likewise, flood mitigation methods, such as infrastructure designs and the national flood insurance policy, are based on the assumption of stationarity and, therefore, may not provide expected levels of protection in a nonstationary environment. The goal of this study was to provide a method to detect and model multinonstationarity in hydrologic data, as well as to assess the change in risk associated with multinonstationarity. A statistical test was developed to identify multiple change points within a time series, which is necessary to achieve optimum modeling accuracy for hydrologic data in a nonstationary environment. A procedure was developed to incorporate multinonstationarity into the existing flood frequency analysis method based on two nonstationary factors: urbanization and climate change. Finally, a flood risk assessment was conducted in which the risks as well as the performance of a flood mitigation system were compared for stationary and multinonstationary environments. The results showed that the incorporation of multinonstationarity into the current flood frequency analysis creates a noticeable difference in the magnitude of floods for the same return period as well as the associated risk. Based on the developed method, engineers and policy makers can begin to analyze the hydrologic and risk sensitivity of communities to nonstationarity. If the sensitivities of the system are understood, the factors, such as urbanization and emissions rates that influence climate change, can potentially be controlled to mitigate the consequences. Therefore, while many uncertainties exist in regards to the future conditions of these nonstationary factors, through methods such as those proposed in this study, the range of possibilities will be better understood and lead to more informed decisions to mitigate future risks.