M. Nazeri Tahroudi; Y. Ramezani; C. De Michele; R. Mirabbasi Najafabadi
Abstract
Introduction: The erosion, sediment transport, and estimation problems in the streamflow are the most complicated and essential subjects in the river engineering studies. It is important to model and predict these parameters correctly to determine the effective life of the hydraulic structures ...
Read More
Introduction: The erosion, sediment transport, and estimation problems in the streamflow are the most complicated and essential subjects in the river engineering studies. It is important to model and predict these parameters correctly to determine the effective life of the hydraulic structures and drainage networks. On the other hand, river flow discharge is considered as one of the main components of water resources, which affects sediments. The increasing need of urban and rural communities for limited resources on the one hand, and issues related to climate change and atmospheric precipitation over the past few years, more and more attention is paid to the attitude of surface flows. The phenomena of erosion, sediment transport, and estimation of sediment load in the rivers due to its damages are one of the most critical and complex issues of river engineering. The primary goal of the frequency analysis is to relate extreme events to their frequency using probability distributions. In the frequency analysis of meteorological and hydrological events, the observed data would be analyzed for a long time at a basin. In these analysis, the assumption of independence and stationarity is considered. In fact, the basic assumption is that the studied data are spatially and temporally independent. The main issue is identifying actual distribution across different exiting distributions when using the frequency distribution to estimate the magnitude of the event. There is no appropriate general distribution for all types of rainfall regimes, river flows, etc. On the other hand, in order to analyze the frequency of a similar case, there is no agreement on the use of a particular distribution function. The experience gained so far in the field of statistical analysis of hydrological data shows that some data are more consistent with some specific statistical distributions.
Materials and Methods: In this study, the frequency analysis of total sediment load of the Zarinehrood basin was investigated in the south-east of Lake Urmia with consideration of the peak flow discharge at the Chalekhmaz hydrometric station during the statistical period of 1992-2016 using copula functions. At first, the correlation of these data was investigated using Kendall Tau correlation statistics, and the correlation coefficient was calculated as 0.75. In this study, Ali-Mikhail-Haq, Clayton, Frank, Galambos, Gumbel-Hougaard, Plackett, and Farlie-Gumbel-Morgenstern copula functions were used. In the conventional method of estimating the return period of extreme values, different statistical distributions are fitted on the studied data. After fitting the statistical distributions on the data series, the accuracy of each distribution is evaluated by one of the goodness of fit tests, such as the Kolmogorov-Smirnov test. After statistically controlling the goodness of fit test and determining the acceptable distributions, the root means square error (RMSE) and the Nash-Sutcliffe criterion are calculated to select the best fit model. Each of the fitting distributions that have the highest Nash-Sutcliffe (NS) criteria, and the lowest RMSE is chosen as an appropriate distribution.
Results and Discussion: With the fitting of 65 different distribution functions into the series, the Weibull distribution for total sediment load values and generalized Pareto distribution for peak flow discharge values were selected based on the evaluation criteria as appropriate marginal distributions. The results of the evaluation of the accuracy and efficiency of copula functions were studied by using root mean square error, Nash-Sutcliffe, BIAS, and AIC statistics. In this regard, the results were compared with the experimental copula functions. Finally, the Galambos copula was selected from the candidate copulas as superior copula function. The conditional and joint return period of the total sediment load based copula was proposed with a probability of 10 to 90 percent.
In univariate mode, the lowest probability of exceedance is 50%. In a bivariate mode, this possibility is presented more accurately. With a possibility of exceedance of 50%, it can be observed that the total sediment load at Chalekhmaz station during the studied period is about 400 tons per day. This is about 56% higher than its univariate mode, which is the average long-term of the total sediment load is closer to the Chalekhmaz station.
Conclusion: By comparing the bivariate analysis and its return period with univariate mode, the results indicated that more accurate calculation. Also the results showed the estimation of total sediment load is closer to the total sediment load of the Chalekhmaz station in bivariate analysis mode. Also, the results showed that in univariate mode, estimation of total sediment load at Chalekhmaz station was less than its actual value during the two-year return period. Regarding the results, the generated return curves can be used as the type curves for the management of water resources in the basin.
