Performance evaluation of various data driven techniques for infilling missing streamflow data across Turkey’s rivers

Öz

Missing data with gaps is always an obstacle to effective planning and management of water resources. Complete and reliable hydrological time series are necessary for the optimal design of water resources. A study was conducted to fill in missing streamflow data of 54 observation stations across Turkey. This process was done with the aid of various statistical estimation methods. Estimations were performed by using Linear regression (LR), Artificial neural network (ANN), adaptive neuro-fuzzy inference system (ANFIS), Support vector machine (SVM), Multivariate Adaptive regression splines (MARS), and K-nearest neighbor (KNN) methods. Performances of infilling methods were evaluated based on four performance criteria; namely, root mean squared error (RMSE), coefficient of determination (R2), mean absolute error (MAE), and the Kling–Gupta efficiency (KGE) during training and test periods. Reliable and long streamflow data from surrounding stations were selected as input to fill in missing streamflow data for an output station. The results revealed that a single method cannot be specified as the best-fit method for the study area. During the test phase, the R2 ranged from 0.54 to 0.99, and the KGE range was between 0.62 and 0.98. This study showed that especially SVM and MARS methods are suitable for estimating missing streamflow data in Turkey’s rivers. These findings will provide reliable streamflow data that can be used in hydrological modeling and water resources planning and management.

Anahtar Kelimeler

• Missing values; streamflow; Support vector machine; Multivariate adaptive regression splines; Turkey

Türkiye'nin nehirlerinde eksik akım verilerinin tamamlanması için çeşitli veri odaklı tekniklerin performans değerlendirmesi

Öz

Eksik veriler, su kaynaklarının etkin bir şekilde planlanması ve yönetilmesinin önünde her zaman bir engel teşkil etmektedir. Su kaynaklarının optimal tasarımı için eksiksiz ve güvenilir hidrolojik zaman serileri gereklidir. Türkiye genelinde 54 gözlem istasyonunun eksik akış verilerinin doldurulması için bir çalışma yapılmıştır. Doğrusal regresyon (LR), yapay sinir ağı (ANN), uyarlanabilir nöro-bulanık çıkarım sistemi (ANFIS), Destek vektör makinesi (SVM), Çok değişkenli uyarlanabilir regresyon eğrileri (MARS) ve K-en yakın komşu (KNN) kullanılarak tahminler gerçekleştirilmiştir. Yöntemlerinin performansları dört performans kriterine göre değerlendirilmiştir; bunlar, ortalama kare hata (RMSE), belirleme katsayısı (R2), ortalama mutlak hata (MAE) ve Kling-Gupta verimliliği (KGE) dir. Bir istasyonda eksik akış verilerinin doldurulması için, çevredeki istasyonlardan alınan güvenilir ve uzun akış verileri girdi olarak seçilmiştir. Sonuçlar, tek bir yöntemin çalışma alanı için en uygun yöntem olarak belirlenemeyeceğini ortaya koymuştur. Test aşamasında, R2 0,54 ile 0,99 arasında ve KGE aralığı 0,62 ile 0,98 arasındadır. Bu çalışma, özellikle SVM ve MARS yöntemlerinin Türkiye'deki nehirlerdeki eksik akış verilerinin tahmin edilmesi için uygun olduğunu göstermiştir. Bu bulgular, hidrolojik modelleme ve su kaynakları planlaması ve yönetiminde kullanılabilecek güvenilir akış verileri sağlayacaktır.

Anahtar Kelimeler

• akım
• destek vektör makineleri
• Çok değişkenli uyarlanabilir regresyon eğrileri
• Turkiye
• Eksik değerler

Kaynakça

1. [1] Kuriqi, A., Ali, R., Pham, QB., et al 2020. Seasonality shift and streamflow flow variability trends in central India. Acta Geophys 68:1461–1475. https://doi.org/10.1007/s11600-020-00475-4.
2. [2] Dikbas, F., Yasar, M. 2020. Data-Driven Modeling of Flows of Antalya Basin and Reconstruction of Missing Data. Iran J Sci Technol - Trans Civ Eng 44:1335–1344. https://doi.org/10.1007/s40996-019-00331-6
3. [3] Ergen, K., Kentel, E. 2016. An integrated map correlation method and multiple-source sites drainage-area ratio method for estimating streamflows at ungauged catchments: A case study of the Western Black Sea Region, Turkey. J Environ Manage 166:309–320. https://doi.org/10.1016/j.jenvman.2015.10.036
4. [4] Dembélé, M., Oriani, F., Tumbulto, J., et al 2019. Gap-filling of daily streamflow time series using Direct Sampling in various hydroclimatic settings. J Hydrol 569:573–586.
5. [5] Kim. J.W., Pachepsky, Y.A. 2010. Reconstructing missing daily precipitation data using regression trees and artificial neural networks for SWAT streamflow simulation. J Hydrol 394:305–314. https://doi.org/10.1016/j.jhydrol.2010.09.005
6. [6] Xia, Y., Fabian, P., Stohl, A., Winterhalter, M. 1999 Forest climatology: Estimation of missing values for Bavaria, Germany. Agric For Meteorol 96:131–144. https://doi.org/10.1016/S0168-1923(99)00056-8
7. [7] Ng, W.W., Panu, U.S., Lennox, W.C. 2009. Comparative Studies in Problems of Missing Extreme Daily Streamflow Records. J Hydrol Eng 14:91–100. https://doi.org/10.1061/(asce)1084-0699(2009)14:1(91)
8. [8] Nayak, P.C., Sudheer, K.P., Rangan, D.M., Ramasastri, K.S. 2004 A neuro-fuzzy computing technique for modeling hydrological time series. J Hydrol 291:52–66. https://doi.org/10.1016/j.jhydrol.2003.12.010

9. [9] Yilmaz, A.G., Muttil, N. 2014. Runoff Estimation by Machine Learning Methods and Application to the Euphrates Basin in Turkey. J Hydrol Eng 19:1015–1025. https://doi.org/10.1061/(asce)he.1943-5584.0000869
10. [10] Dastorani, M.T., Moghadamnia, A., Piri, J., Rico-Ramirez, M. 2010. Application of ANN and ANFIS models for reconstructing missing flow data. Environ Monit Assess 166:421–434. https://doi.org/10.1007/s10661-009-1012-8
11. [11] Kim, M., Baek, S., Ligaray, M., et al 2015. Comparative studies of different imputation methods for recovering streamflow observation. Water (Switzerland) 7:6847–6860. https://doi.org/10.3390/w7126663
12. [12] de Souza, G.R., Bello, I.P., Corrêa, F.V., de Oliveira, L.F.C. 2020. Artificial Neural Networks for Filling Missing Streamflow Data in Rio do Carmo Basin, Minas Gerais, Brazil. Brazilian Arch Biol Technol 63:1–8. https://doi.org/10.1590/1678-4324-2020180522
13. [13] Tabari, H., Sabziparvar, A.A., Ahmadi, M. 2011 Comparison of artificial neural network and multivariate linear regression methods for estimation of daily soil temperature in an arid region. Meteorol Atmos Phys 110:135–142. https://doi.org/10.1007/s00703-010-0110-z
14. [14] Uysal, G., Şorman, A.A., Şensoy, A. 2016 Streamflow Forecasting Using Different Neural Network Models With Satellite Data for a Snow Dominated Region in Turkey. Procedia Engineering 154 1185 – 1192.
15. [15] Sun, Y., Niu, J., Sivakumar, B. 2019. A comparative study of models for short-term streamflow forecasting with emphasis on wavelet-based approach. Stoch Environ Res Risk Assess 33:1875–1891. https://doi.org/10.1007/s00477-019-01734-7
16. [16] Jang, J.R. 1993 ANFIS : Adap tive-Ne twork-Based Fuzzy Inference System. 23
17. [17] Karaboga, D., Kaya, E. 2019. Adaptive network based fuzzy inference system (ANFIS) training approaches: a comprehensive survey. Artif Intell Rev 52:2263–2293. https://doi.org/10.1007/s10462-017-9610-2
18. [18] Kisi, O., Nia, A.M., Gosheh, M.G., et al 2012. Intermittent Streamflow Forecasting by Using Several Data Driven Techniques. Water Resour Manag 26:457–474. https://doi.org/10.1007/s11269-011-9926-7
19. [19] Cortes, C., Vapnik, V. Support-vector networks. Mach Learn 20, 273–297 (1995). https://doi.org/10.1007/BF00994018
20. [20] Parisouj, P., Mohebzadeh, H., Lee, T. 2020. Employing Machine Learning Algorithms for Streamflow Prediction: A Case Study of Four River Basins with Different Climatic Zones in the United States. Water Resour Manag 34:4113–4131. https://doi.org/10.1007/s11269-020-02659-5
21. [21] Dibike, Y.B., Velickov, S., Solomatine, D., Abbott, M.B. 2001. Model Induction with Support Vector Machines:Introduction and Applications. J Comput Civ Eng 15:208–216. https://doi.org/10.1061/(asce)0887-3801(2001)15:3(208)
22. [22] Lin, J.Y., Cheng, C.T., Chau, K.W. 2006. Using support vector machines for long-term discharge prediction. Hydrol Sci J 51:599–612. https://doi.org/10.1623/hysj.51.4.599
23. [23] Friedman, J.H. 1991 Multivariate Adaptive Regression Splines. Annals of Statistics, 19, 1-67. https://doi.org/10.1214/aos/1176347963
24. [24] Mehdizadeh, S., Fathian, F., Safari, M.J.S., Adamowski, J.F. 2019. Comparative assessment of time series and artificial intelligence models to estimate monthly streamflow: A local and external data analysis approach. J Hydrol 579:. https://doi.org/10.1016/j.jhydrol.2019.124225
25. [25] Zhang, W., Goh, A.T.C. 2016. Multivariate adaptive regression splines and neural network models for prediction of pile drivability. Geosci Front 7:45–52. https://doi.org/10.1016/j.gsf.2014.10.003
26. [26] Alizamir, M., Heddam, S., Kim, S., et al 2021. Prediction of daily chlorophyll-a concentration in rivers by water quality parameters using an efficient data-driven model: online sequential extreme learning machine.ActaGeophys. https://doi.org/10.1007/s11600-021-00678-3
27. [27] Khazaee, Poul, A., Shourian, M., Ebrahimi, H.2019. A Comparative Study of MLR, KNN, ANN and ANFIS Models with Wavelet Transform in Monthly Stream Flow Prediction. Water Resour Manag 33:2907–2923. https://doi.org/10.1007/s11269-019-02273-0
28. [28] Rainer, S., Kenneth, P. 1997. Differential Evolution: A Simple and Efficient Heuristic for Global Optimization over Continuous Spaces. J Glob Optim 11:341
29. [29] Jiang ,Z., Ma, W. 2018. Integrating differential evolution optimization to cognitive diagnostic model estimation. Front Psychol 9:1–9. https://doi.org/10.3389/fpsyg.2018.02142
30. [30] Mullen, K.M., Ardia, D., Gil, D.L., et al 2011. DEoptim: An R package for global optimization by differential evolution. J Stat Softw 40:1–26. https://doi.org/10.18637/jss.v040.i06
31. [31] Sleziak, P., Holko, L., Danko, M., Parajka, J. 2020. Uncertainty in the number of calibration repetitions of a hydrologic model in varying climatic conditions. Water (Switzerland) 12:. https://doi.org/10.3390/W12092362
32. [32] Tang, S., Jiang, J., Zheng, Y., et al 2021. Robustness analysis of storm water quality modelling with LID infrastructures from natural event-based field monitoring. Sci Total Environ 753:142007. https://doi.org/10.1016/j.scitotenv.2020.142007
33. [33] Yilmaz, M., Tosunoglu, F., Demirel, M.C. 2021. Comparison of conventional and differential evolution-based parameter estimation methods on the flood frequency analysis. Acta Geophys 69:1887–1900. https://doi.org/10.1007/s11600-021-00645-y
34. [34] Kadiolu, M. 2000 Regional variability of seasonal precipitation over Turkey. Int J Climatol 20:1743–1760. https://doi.org/10.1002/1097-0088(20001130)20:14<1743::AID-JOC584>3.0.CO;2-G
35. [35] Güçlü, Y.S. 2018. Multiple Şen-innovative trend analyses and partial Mann-Kendall test. J Hydrol 566:685–704. https://doi.org/10.1016/j.jhydrol.2018.09.034
36. [36] Wambua, R.M., Mutua, B.M., Raude, J.M. 2016. Prediction of Missing Hydro-Meteorological Data Series Using Artificial Neural Networks ( ANN ) for Upper Tana River Basin , Kenya. Am J Water Resour 4:35–43. https://doi.org/10.12691/ajwr-4-2-2