Author(s): Keita Shimizu; Tadashi Yamada; Tomohito Yamada
Linked Author(s):
Keywords: Probability limit method test; Confidence interval; Prediction interval; Future projection database; Climate change; Uncertainty
Abstract: The main purpose of hydrological frequency analysis is to estimate hydrological quantity with design return period. In traditional hydrological frequency analysis, we estimate T-year hydrological quantity by using observed data which were accumulated for several decades. However, observed data of extremes which were accumulated so far are so limited ones that the design hydrological quantity includes uncertainty to a large extent. Moreover, one of the difficulties caused by the shortage of extreme value data is not to predict record heavy rainfall which deviate greatly from an adopted probability distribution used for river planning. So, it is often impossible to evaluate record heavy rainfalls and these rainfalls are treated as unexpected. Therefore, we propose a new hydrological frequency analysis introducing confidence interval and prediction interval based on probability limit method test. By introducing this confidence interval into hydrological frequency analysis, uncertainty of design hydrological quantity can be quantified. Also, by introducing this prediction interval, it can be possible to predict the scale and occurrence risk of catastrophic heavy rainfalls. In addition, another difficulty of hydrological frequency analysis is how to manage non-stationarity of rainfall which is caused by climate change. To manage non-stationarity of rainfall, using Bayesian statistics is effective. In this research, Bayesian update of confidence interval and prediction interval derived from past observed data was conducted by using a large ensemble database. In this research, a theoretical framework of hydrological frequency analysis introducing confidence interval and prediction interval and update method of these intervals using Bayesian statistics are shown.
Year: 2020