Uncertainty Analysis of Artificial Intelligence Models in Forecasting River Flow (Case Study: Karun River)

Document Type : Research Article

Authors

1 Graduate Student, Water Engineering Department, Faculty of Civil and Surveying Engineering, Graduate University of Advanced Technology, Kerman, Iran

2 Department of Water Engineering, Faculty of Civil and Survey Engineering, Graduate University of Advanced Technology, Kerman

3 Assistant Professor, Department of Ecology, Institute of Science and High Technology and Environmental Science, Graduate University of Advanced Technology, Kerman, Iran

Abstract

An accurate estimation of the discharge flow of natural streams plays a key role in irrigation planning, the design of bridges embedded in waterways, the management of reservoirs and dams, and the design of flood-warning systems. In recent decades, several studies have been conducted using artificial intelligence (AI) to accurately estimate the flow. Despite the proven accuracy level of AI methods, in many cases, there are uncertainties that occur for a variety of reasons. Insufficient knowledge of these uncertainties in the flow modeling process can have irreversible effects. In this research, monthly flow data, measured from Armand hydrometric station, located in Karun River basin, during a-28 year period (from 1980 to 2008) were used. First, the Thomas and Fiering method was used to generate flow series data and consider them as input variables. Then, flow forecast modeling was performed by three AI methods, namely Model Tree (MT), Gene Expression Programming (GEP), and Multivariate Adaptive Regression Spline (MARS). Statistical indicators such as correlation coefficient (R) and Root Mean Square Error (RMSE) were used to evaluate the accuracy of the models. In terms of the training stage, the MARS model with R=0.839 and RMSE=28.624 m3/s performed better than the other two models. Additionally, in the testing stage, the MT model with R=0.784 and RMSE=34.441 m3/s showed a more appropriate performance than other models. The Monte Carlo simulation method was used to calculate the uncertainty of the models, so the results showed that the r-factor parameter, which was the average width of the confidence band in the MT model, was equal to 1.67, indicating a lower and more optimal number compared to MARS (1.92) and GEP (2.025) models. Moreover, the usability of the statistical criterion for quantifying uncertainty, (known as 95PPU) indicated that the GEP model with 95PPU of 64% was selected as a more appropriate percentage than MARS (61%) and MT-M5 (55%). 

Keywords

Main Subjects


[1] R.J. Abrahart, L. See, Neural network vs. ARMA modelling: constructing benchmark case studies of river flow prediction. In GeoComputation’98. Proceedings of the Third International Conference on GeoComputation, University of Bristol, United Kingdom (1998) 17-19.
[2] H. K. Cigizoglu, Estimation, forecasting and extrapolation of river flows by artificial neural networks. Hydrological Sciences Journal, 48(3) (2003) 349-361.
[3] C.L. Wu, K.W. Chau, C.Fan, Prediction of rainfall time series using modular artificial neural networks coupled with data-preprocessing techniques. Journal of Hydrology, 389(1-2) (2010) 146-167.
[4] W. Huang, B. Xu, A. Chan‐Hilton, Forecasting flows in Apalachicola River using neural networks. Hydrological processes, 18(13) (2004) 2545-2564.
[5] R. Noori, A.R. Karbassi, A. Moghaddamnia, D. Han, M.H. Zokaei-Ashtiani, A.Farokhnia, M. Ghafari Gousheh, Assessment of input variables determination on the SVM model performance using PCA, Gamma test, and forward selection techniques for monthly stream flow prediction. Journal of Hydrology. 401(3-4) (2011) 177-189.
[6] Z.M. Yaseen, A. El-Shafie, O. Jaafar, H.A. Afan, K.N. Sayl, Artificial intelligence based models for stream-flow forecasting: 2000–2015. Journal of Hydrology, 530 (2015) 829-844.
[7] F. Azarpira, S. SHahabi, Evaluating the capability of hybrid data-driven approaches to forecast monthly streamflow using hydrometric and meteorological variables. Journal of Hydroinformatics. 23(6) (2021) 1165-1181.
[8] M. Montaseri, S. Zamanzad Ghavidel, Streamflow Forecasting using soft computing techniques. Water and soil Journal. 28(2) (2014) 394-405. (In Persian)
[9] B. Saghafian, S. Anvari, S. Morid, Effect of Southern Oscillation Index and spatially distributed climate data on improving the accuracy of Artificial Neural Network, Adaptive Neuro-Fuzzy Inference System and K-Nearest Neighbour streamflow forecasting models. Expert System. 30(4) (2013). 367-380.
[10] A. Danandeh Mehr, An improved gene expression programming model for streamflow forecasting in intermittent streams. Journal of Hydrology. 563 (2018) 669-678.
[11] R.M. Adnan, Z .Liang, K.S. Parmar, K. Soni, O. Kisi, Modeling monthly streamflow in mountainous basin by MARS, GMDH-NN and DENFIS using hydro climatic data. Neural Computing and Applications. 33(7) (2021). 2853-2871. 
[12] A. Bensoussan, N. Farhi, Uncertainties and risks in water resources management. The economics of sustainable development, (Chapter: Uncertainties and risks in water resources management), Publisher: Econom (2010) 163-79.
[13] M. Dehghani, B. Saghafian, F. Nasiri Saleh, A. Farokhnia, R. Noori, Uncertainty analysis of streamflow drought forecast using artificial neural networks and Monte‐Carlo simulation, International Journal of Climatology, 34(4), (2014) 1169-80.
[14] S. Anvari, J. Mousavi, S. Morid, A Multilevel Uncertainty-Based Approach for Optimal Irrigation Scheduling. Advances in Hydroinformatic. (2018) 359-372.  
[15] S. Anvari, J.H. Kim, M. Moghaddasi, The role of meteorological and hydrological uncertainties in the performance of optimal water allocation approaches, Irrigation and Drainage, 68(2) (2019) 342-53.
[16] M. Moghaddasi, S. Anvari, N. Akhondi, A trade-off analysis of adaptive and non-adaptive future optimized rule curves based on simulation algorithm and hedging rules. Theoretical and Applied Climatology, 148(1) (2022) 65-78.
[17] M. Abbasi, A. Farokhnia, M. Bahreinimotlagh, R. Roozbahani, A Hybrid of Random Forest and Deep Auto-Encoder with Support Vector Re‐gression Methods for Accuracy improvement and uncertainty reduction of long-term streamflow prediction. Journal of Hydrology, 597 (2021) 125717.
[18] S. Anvari, E. Rashedi, S. Lotfi, A Coupled Metaheuristic Algorithm and Artificial Intelligence for Long-Lead Stream Flow Forecasting. International Journal of Optimization in Civil Engineering, 12(1) (2022) 91-104.‎
[19] R. Zamani, F. Ahmadi, F., Radmanesh Comparison of the Gene Expression Programming, Nonlinear Time Series and Artificial Neural Network in Estimating the River Daily Flow (Case Study: The Karun River), Journal of Water and Soil; 28 (6) (2015) 1172-1182 (In Persian).
[20] Friedman, J.H. (1991). Multivariate Adaptive Regression Splines. The Annals of Statistics. 19: 1-67.
[21] West, A.M., Evangelista, P.H., Jarnevich, C.S & Schulte, D. (2018). A tale of two wildfires: testing detection and prediction of invasive species distributions using models fit with topographic and spectral indices. Landscape Ecology. 33: 969-984.
[22] Ferreira, C. (2001). Gene Expression Programming:  a new adaptive algorithm for Solving Problems. Complex Systems 13(2): 87–129.
[23] Mollahasani, A., Alavi, A.H & Gandomi, A.H. (2011). Empirical modeling of plate load test moduli of soil via gene expression programming. Computers and Geotechnics. 38(2): 281-286.
[24] Quinlan, J.R. (1992). Learning with continuous classes. 5th Australian joint conference on artificial intelligence. Vol 92.
[25] Wang, Y. and Witten, I. H. (1996).  Induction of model trees for predicting continuous classes.
[26] Mirhashemi, S.H., Panahi, M. and Zareei L (2020). Evaluation of M5P Algorithm for Estimation of Potential Evapotranspiration, Minimum and Maximum Temperature (Case study: Sari Weather Station). Journal of Meteorology and Atmospheric Sciences 2(4):  287-295 (In Persian).
[26] Yi, H. S., Lee, B., Park, S., Kwak, K. C., & An, K. G. (2019). Prediction of short-term algal bloom using the M5P model-tree and extreme learning machine. Environmental Engineering Research, 24(3), 404-411.
[27] Eckhardt K., Breuer L., Frede H.G. (2003). Parameter uncertainty and the significance of simulated land use change effects, Journal of Hydrology, 273: 164 -176.
[28] Talebizadeh M, Morid S, Ayyoubzadeh SA, Ghasemzadeh M. Uncertainty analysis in sediment load modeling using ANN and SWAT model, Water Resour Manag 2010; 24(9): 1747-61.
[29] Thomas, H.A & Fiering, M.B. (1962). Mathematical synthesis of streamflow sequence for the analysis of river basins simulations. In: Maass, A. et al. (Eds.), Design of Water Resources Systems. Harvard University.
[30] Khoi, D. N., Thom, V. T., Quang, C. N. X., & Phi, H. L. (2017). Parameter uncertainty analysis for simulating streamflow in the upper Dong Nai river basin. La Houille Blanche, (1), 14-23.
[31] Maheepala, S., & Perera, B. J. C. (1996). Monthly hydrologic data generation by disaggregation. Journal of hydrology, 178(1-4), 277-291.
[32] McMahon TA, Adeloye A.J. (2005) Water resources yield. Water Resources Publications, Littleton.