当前位置：
X-MOL 学术
›
J. Hydrol.
›
论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Connecting flow duration curve and precipitation duration curve based on the relationship deduced from machine learning in the watersheds of northern China
Journal of Hydrology ( IF 6.4 ) Pub Date : 2024-04-20 , DOI: 10.1016/j.jhydrol.2024.131235 Lan Ma , Dengfeng Liu , Jinkai Luan , Guanghui Ming , Xianmeng Meng , Qiang Huang
Journal of Hydrology ( IF 6.4 ) Pub Date : 2024-04-20 , DOI: 10.1016/j.jhydrol.2024.131235 Lan Ma , Dengfeng Liu , Jinkai Luan , Guanghui Ming , Xianmeng Meng , Qiang Huang
|
Obtaining reliable hydrological data for ungauged watersheds has always been a significant challenge in the water resources management. The flow duration curve (FDC) and precipitation duration curve (PDC) are classical methods to address this challenge. However, traditional FDC and PDC methods mostly rely on pre-assumed probability distribution function form, which may result in sub-optimal model parameters and consequently sub-optimal simulation results of hydrological data, and the common function forms are not applicable to deal with the data including zero-values, which often occurs in the discharge data of arid region and daily rainfall data. To address the issues in FDC and PDC , this study firstly determines the most suitable function form to describe FDC and PDC from various distribution functions, i.e., Log Normal, Generalized Pareto Distribution and H2018 function, from 1960 to 2010 in 7 watersheds in the north of China. The multiple methods including Random Forest, polynomial, Multilayer Perceptron and K-Nearest Neighbor are employed to simulate the parameters of the functions, aiming to deduce the optimal parameters of FDC functions. The results show that, H2018 function form has the best performance in simulating the PDC and FDC and in general, parameter a and b of H2018 function indicate an opposite trend, and parameter k declines in mostly condition at the annual scale from 1960 to 2010. The results of different methods indicate that the Random Forest method is the most efficient and accurate approach to deduce FDC parameters from measured precipitation and climate factors, compared with polynomial, Multilayer Perceptron and K-Nearest Neighbor. Based on the relationship between PDC parameters and FDC parameters, the FDC of ungauged basins can be deduced from the observed precipitation data. The research findings provide a new method to estimate FDC for the water resources management in the ungauged basins.
更新日期：2024-04-20
京公网安备 11010802027423号