Determination of Intensity-Duration-Frequency Relation by Particle Swarm Optimization and Genetic Programming


Başakın E. E. , Ekmekcioğlu Ö. , Özger M. , Çıtakoğlu H.

II. International Applied Statistics Conference (UYIK-2021), Tokat, Turkey, 29 June - 02 July 2021, pp.1-8

  • Publication Type: Conference Paper / Full Text
  • City: Tokat
  • Country: Turkey
  • Page Numbers: pp.1-8

Abstract

Rainfall intensity values at different frequencies are required to design the water structures. Thus, obtaining the intensity-duration-frequency relationship has vital importance. In this study, an equation was obtained in which the rainfall intensity can be calculated with the help of the intensity-durationfrequency values of long-term standard periods of Kayseri. Rainfall intensity is designed as a function of both rainfall duration and frequency. Genetic programming (GP) was used to generate the equation and obtained equation validated with the data taken from other stations in the region, such as Nevşehir, Niğde and Yozgat. The observed and predicted rainfall intensity values were evaluated with two different performance indicator, namely mean square error (MSE) and coefficient of efficiency (NSE) criteria. For Kayseri station, the MSE and NSE are calculated as 0.004 and 0.99, respectively. Also, the MSE and NSE values obtained for Nevşehir and Niğde stations are 0.028 and 0.91, and 0.067 and 0.76, respectively, while for Yozgat station, MSE was determined as 0.027 and NSE was found as 0.95. Furthermore, one of the empirical equations in the literature was used to perform a comparison. The parameter optimization for the empirical equation was carried out using particle swarm optimization (PSO). Finally, it was concluded that the equation obtained with GP has higher accuracy than the obtained equation with the help of PSO.