Application of Three Intelligent Models in Estimation of Watersheds Suspended Load (Case Study: Dareh-Roud Watershed, Ardabil Province)

آب و خاک

Article 6, Volume 34, Issue 4 - Serial Number 72, September 2020, Pages 827-845   

Document Type: Research Article

DOI: 10.22067/jsw.v34i4.86251

Authors

Y. Ataie¹; M. R. Nikpour^{* 2}; A. Kanooni¹; Y. Hoseini¹

¹University of Mohaghegh Ardabili

²University of Mohaghegh Ardabili, Ardabil, Iran

Abstract

Introduction: Suspended load estimation is utilized to study and investigate many problems of water engineering sciences such as dam reservoir design, transportation of sediments and pollution in the rivers, creation of stable channels, estimation of erosion and sedimentation around bridge piers, and watershed management. The purpose of this study was to estimate the suspended load in the Dareh-Roud watershed in Ardabil province using the rivers discharge values and the physiographic characteristics of the sub-basins. Moreover, annual suspended load and sediment specific discharge were calculated for the whole of the watershed.
Materials and Methods: In this study, the Dareh-Roud watershed in Ardebil province was considered as the study area. The flow discharge and suspended load data were collected from 16 hydrometric stations with a statistical period of 15 years from 2001-2015. The physiographic characteristics of sub-basins, including area (A), slope (S), shape factor (Sf), and curve number (CN), were achieved using ArcGIS and WMS. Five different input combinations were defined based on the effect of flow discharge variables and physiographic properties on the suspended load. Also, considering the area and slope parameters, the sub-basins were divided into two groups (i.e., the first and second groups). The performance of data-intelligent models, including Artificial Neural Networks (ANN), Adaptive Neural-Fuzzy Interference System (ANFIS), and Gene Expression Programming (GEP) models were investigated in the predict of the suspended load in the study area. Several statistical indicators, including determination coefficient (R2), root mean square error (RMSE), and Nash- Sutcliffe efficiency (NS), were utilized to evaluate the model’s efficiency.
Results and Discussion: According to the results, estimation of suspended load without using the physiographic characteristics resulted in a high error, and in contrast, the suspended load estimation was most accurate by using a combined scenario involving all physiographic aspects and flow discharge. The scatterplots indicated that in the first group, the points were concentrated around the 1:1 axis for the values of less than 20 (ton/day). However, for the greater amounts, the scattering of issues around the one-to-one line was not appropriate, which means that the models were in the condition of underestimation. Similar conditions were observed for the second group, the excellent dispersion was seen for the values of less than 1000 (ton/day), and in general, the models had underestimation conditions. However, in both groups, the dispersion of the GEP model was somewhat better than the other models. Based on the values of R2 and NS, ANN and ANFIS models had the acceptable and satisfactory accuracy for the first group. The GEP model was more reliable and efficient in estimating the suspended load of the first group. On the other hand, the efficiency of ANN and ANFIS was not acceptable for the second group. Comparison of the results of different models using the best input combination indicated that the GEP model with the highest determination coefficient (R2 = 0.68), the lowest root mean square error (RMSE = 7.69 ton/day). The NS equal to 0.55 in the validation step has shown better performance than the other models in estimating the suspended load for the first group. Similarly, for the second group, the GEP model with the highest determination coefficient (R2 = 0.72), the lowest root means square error (RMSE = 975.26 ton/day). The NS equal to 0.43 in the validation step has shown better performance than other models in estimating the suspended load.
Conclusion: In the present study, the efficiency of different intelligent models was investigated in the suspended load estimation of Dareh-roud watershed. In this regard, an extended period (i.e., during 15 years) of measured data, including flow discharge and sediment at the hydrometric stations located on the mentioned watershed, were used. In order to simulate the suspended load, five different input combinations were considered. For all models, the accuracy of suspended load estimation was improved by combining the physiographic characteristics and discharge values. Due to the higher accuracy of the GEP model, regional sediment models were achieved for the first and second groups, separately. Also, annual suspended load and sediment specific discharge were calculated for all sub-basins. According to the results, most of the suspended load of the Dareh-Roud watershed is produced and transported in its old rivers (i.e., Dareh-Roud and Qarah-Su). Based on the results of this research, in the Dareh-Roud watershed, 6.33 million tons of suspended sediments were transported during 2001-2015.

Keywords

Suspended Load; Dareh-Roud watershed; Intelligent models; Specific sediment discharge

References

Abolfathi D., Madadi A., and Asghari S. 2018. Modelling of river sediment estimation by artificial neural network method (Case study: Golroud river). Qualitative Geomorphological Research 7(2): 196-208. (In Persian with English abstract)

2- Adnan R.M., Liang Z., El-Shafie A., Zounemat-Kermani M., and Kisi O. 2019. Prediction of suspended sediment load using data-driven models. Water 11(10): 1-19.

3- Asadi M., and Fathzadeh A. 2018. The use of computational intelligence base models in suspended sediment load estimation (Case study: Gillan province). Journal of Range and Watershed Management 71(1): 45-60. (In Persian with English abstract)

4- Ayele G.T., Teshale E.Z., Yu B., Rutherfurd I.D., and Jeong J. 2017. Streamflow and sediment yield prediction for watershed prioritization in the upper Blue Nile river watershed, Ethiopia. Water 9(782): 1-29.

5- Azamathulla H.M., Caun Y.C., Aminudin A., and Chang C.K. 2013. Suspended sediment load prediction of river systems: GEP approach. Arabian Journal of Geoscience 6: 3469-3480.

6- Dehghani A.A., Zanganeh M.E., Mosaedi A., and Kouhestani N. 2015. Comparison of estimating the suspended load using rating curve and ANN (Case study: Dough river, Golestan province). Journal of Agriculture and Natural Resources Sciences 16: 1-16. (In Persian with English abstract)

7- Dolat Kordestani M., Nohegar A., and Janizadeh S. 2018. Evaluating performance of adaptive neural-fuzzy and artificial neural networks in estimating daily suspended sediment (Case study: Gero watershed). Qualitative Geomorphological Research 6(4): 120-130. (In Persian with English abstract)

8- Ferreira C. 2001. Gene expression programming: a new adaptive algorithm for solving problems. Complex Systems 13(2): 87–129.

9- Ferreira C. 2004. Genetic representation and genetic neutrality in gene expression programming. Advances in Complex Systems 5(4): 389-408.

10- Ferreira C. 2006. Gene expression programming: mathematical modeling by an artificial intelligence. 2nd edn, Springer-Verlag, Germany.

11- Ghorbani M.A., and Dehghani R. 2014. Comparison of Bayesian neural networks and artificial neural networks to estimate suspended sediments in rivers (Case study: Simineh Rood). Environment Sciences and Technology 19(2): 1-13. (In Persian with English abstract)

12- Hassanpour F., Sharifazari S., Ahmadaali K., Mohammadi S., and Sheikhalipour Z. 2019. Development of the FCM-SVR hybrid model for estimating the suspended sediment load. KSCE Journal of Civil Engineering 23: 2514-2523

13- Jang J.S.R. 1993. ANFIS: Adaptive network-based fuzzy inference system. IEEE Transaction System Management and Cybernetics 23(3): 665-686.

14- Kennedy P., Condon M., and Dowling J. 2003. Torque-ripple minimization in switched reluctant motors using a neuro-fuzzy control strategy. In: Proceedings of the IASTED International Conference on Modeling and Simulation.

15- Kheirfam H., Kheirfam B., Azhdan Y., and Hosseini S. 2018. Variability of bed and suspended load and bed to suspended load ratio in QotourChay River. Watershed Engineering and Management 10(3): 410-420. (In Persian with English abstract)

16- Kisi O., and Shiri J. 2012. River suspended sediment estimation by climatic variables implication: Comparative study among soft computing techniques. Computers and Geosciences 43: 73-82.

17- Mohammadi S. 2019. The suspended sediment load modeling by artificial neural networks, neural-fuzzy and rating curve in Halilrood watershed. Watershed Engineering and Management 11(2): 452-466. (In Persian with English abstract)

18- Montaseri M., and Ghavidel S. 2014. River flow forecasting by soft computing. Journal of Water and Soil 28(2): 394-405. (In Persian with English abstract)

19- Nikpour M.R., and Sanikhani H. 2017. Suspended load modeling of river using soft computing techniques (Case study: Dareh-rood river). Journal of Irrigation and Water Engineering 8(2): 29-44.

20- Salih S.Q., sharafati A., Khosravi K., Faris H., Kisi O., Tao H., Ali M., and Yaseen Z.M. 2019. River suspended sediment load prediction based on river discharge information: application of newly developed data mining models. Hydrological Sciences Journal 65(4): 624-637.

21- Samadzadeh R., Khayyam M., and Fazeli R.S. 2013. Modelling of estimation of the suspended Load in Ardabil Darehrud basin. Geography and Environmental Planning Journal 51(3): 153-178. (In Persian with English abstract)

22-Sanikhani H., Kissi O., Nikpour M.R., and Dinpashoh Y. 2012. Estimation of Daily Pan Evaporation using two different Adaptive Neuro-Fuzzy Computing Techniques. Water Resource Management 26(15): 4347-4365.

23- Samantaray S., and Ghose D.K. 2018. Evaluation of suspended sediment concentration using descent neural networks. Procedia Computer Science 132: 1824–1831.

24- Sheikhalipour Z., Hassanpour F., and Azimi V. 2015. Comparison of artificial intelligence methods in estimation of suspended sediment load (Case Study: Sistan River). Journal of Water and Soil Conservation 22(2): 41-60. (In Persian with English abstract)

25- Tofani P., Fakherifard A., Mosaedi A., and Dehghani A.A. 2015. Prediction of precipitation applying wavelet and ANN-wavelet. Journal of Range and Watershed Management 68(3): 553-571. (In Persian with English abstract)

26- Yilmaz B., Aras E., Nacar S., and Kankal M. 2018. Estimating suspended sediment load with multivariate adaptive regression spline, teaching-learning based optimization, and artificial bee colony models. Science of the Total Environment 639: 826-840.

27- Zhu Y.M., Lu X.X., and Zhou Y. 2007. Suspended sediment flux modeling with artificial neural network: an example of the Longchuanjiang River in the Upper Yangtze Catchment, China. Geomorphology 84: 111-125.