Developing the Regional Nonstationary IDF Curves Using NGN-ProNEVA Framework

Intensity-Duration-Frequency (IDF) curves are known as practical tools in the construction of infrastructures. However, developing such curves is restricted in many regions due to sparse and insufficient record length of in-site rainfall observations. It is hence recommended to generate the regional curves. According to the hydro-climate variability and change during the recent decades, it is essential to consider the non-stationarity of hydro-climate variables. In this research, the Neural Gas network (NGN) coupled with ProNEVA have been applied to develop non-stationary regional IDFs. The l-moments approach was used to plot regional stationary IDF curves to compare the results of non-stationary IDFs for homogenous regions. The results showed that the regional nonstationary curves had overestimated rainfall intensity compared with the regional stationary curves which can attribute to the decreasing trend of rainfall over the study area. The average value of overestimation in the return period of 2 years was equal to 50 percent. This overestimation was more significant for lower return periods, which indicates that the nonstationary approach is more important for short-duration events. The return period of 100 years is equal to 25 percent in region two, and in region one, it is equal to 20 and 43 percent, respectively.

Automated summary

The Neural Gas network coupled with ProNEVA was used to develop non-stationary regional IDFs. The results showed that the regional nonstationary curves overestimated rainfall intensity, especially for short-duration events.