DİNAMİK SU BÜTÇESİ MODELİNE DESTEK VEKTÖR REGRESYONU ENTEGRASYONU

Abstract

Yağış-akış modelleri kapsamında ele alınan modeller içerisinden kavramsal modeller havza dinamiğini atanan parametreler yardımıyla benzeştirmeye çalışırken, kapalı kutu modelleri ise fiziksel süreci dikkate almadan veri işleme esaslı uygulanmaktadır. Her iki yöntemin de birbirine göre avantajlı ve dezavantajlı yönleri bulunmaktadır. Örneğin kavramsal modellerin bazı parametreleri doğrusal tanımlandıklarında simülasyonlarda yanlılıklar gözlenebilmektedir. Diğer yandan, kapalı kutu modelleri tutarlı bir simülasyon için gecikmeli yağış değerlerine ihtiyaç duymaktadır. Bu nedenle çalışmada her iki yaklaşımın iyi yönlerini birleştiren hibrit bir model yapısının ortaya konması amaçlanmıştır. Bu kapsamda, dinamik su bütçesi adı verilen kavramsal bir yağış-akış modelinin doğrusal davranış gösteren yeraltısuyu depolama elemanı yerine destek vektör makinesi eklenerek beş parametreli hibrit bir model oluşturulmuştur. Destek vektör makinesi ilavesi ile doğrusal olmayan haritalama yetisi kazanan model Balıkesir’in İkizcetepeler Baraj Havzası’nda uygulanmıştır. Hibrit modelin kavramsal modele kıyasla kalibrasyon ve validasyon dönemlerinde sırasıyla %21 ve %14 daha düşük hata performansı vermesi istatistiksel açıdan anlamlı bulunmuştur.

Keywords

• Yağış-Akış Modellemesi
• Kavramsal Model
• Makine Öğrenmesi
• Hibrit Modelleme
• İkizcetepeler Barajı

Integrating Support Vector Regression into Dynamic Water Budget Model

Abstract

Among the various rainfall-runoff models, conceptual ones can simulate the basin dynamics by means of assigned parameters, while black-box models are applied as data-driven techniques which take no account of the physical process. Both types involve some advantages and shortcomings relative to each other. For instance, as some parameters assigned in conceptual ones are linear, the runoff simulations can be biased. Besides, black-box models generally require antecedent precipitation data to get a robust simulation. Therefore, in the study, it is intended to propose a hybrid model structure integrating the prominent aspects of both approaches. In this concept, the linear groundwater storage of the dynamic water budget model, one of the conceptual types, was eliminated and a support vector regression was included instead, and thus, a hybrid model with five parameters was built. The model, which achieved nonlinear mapping capability with the inclusion of support vector regression, was implemented for Ikizcetepeler Dam located at Balikesir. It was found statistically significant that hybrid model provided relatively lower error performance as 21% and 14% in calibration and validation periods, respectively, when it was compared to that of the conceptual one.

Keywords

• Rainfall-Runoff Modelling
• Conceptual Model
• Machine Learning
• Hybrid Modeling
• Ikizcetepeler Dam

References

1. 1. Anctil, F., Michel, C., Perrin, C. ve Andréassian, V., (2004) A soil moisture index as an auxiliary ANN input for stream flow forecasting, Journal of Hydrology, 286 (1-4), 155–167. doi: 10.1016/j.jhydrol.2003.09.006
2. 2. Ersoy, Z.B. (2021) Dinamik Su Bütçesi Modeline Makine Öğrenmesi Entegrasyonu ile Aylık Akış Tahminlerinin İyileştirilmesi. Balıkesir Üniversitesi Fen Bilimleri Enstitüsü, Yüksek Lisans Tezi, Balıkesir.
3. 3. Ersoy, Z.B., Okkan, U., ve Fistikoglu, O. (2021) Hybridizing a Conceptual Hydrological Model with Neural Networks to Enhance Runoff Prediction, Manchester Journal of Artificial Intelligence and Applied Sciences, 02, 176-178.
4. 4. Humphrey, G.B., Gibbs, M.S., Dandy, G.C., ve Maier, H.R., (2016) A hybrid approach to monthly streamflow forecasting: Integrating hydrological model outputs into a Bayesian artificial neural network, Journal of Hydrology, 540, 623–640. doi: 10.1016/j.jhydrol.2016.06.026
5. 5. Kumanlioglu, A.A., ve Fistikoglu, O., (2019) Performance enhancement of a conceptual hydrological model by integrating artificial intelligence, Journal of Hydrologic Engineering, 24 (11), 04019047. doi: 10.1061/(ASCE)HE.1943-5584.0001850
6. 6. Moriasi, D.N., Arnold, J.G., Van Liew, M.W., Bingner, R.L., Harmel, R.D., ve Veith, T.L., (2007) Model evaluation guidelines for systematic quantification of accuracy in watershed simulations, Transactions of the ASABE, 50(3), 885-900. doi: 10.13031/2013.23153
7. 7. Nash, J.E., ve Sutcliffe, J.V., (1970) River flow forecasting through conceptual models, Part I- A discussion of principles, Journal of Hydrology, 10(3), 282-290. doi: 10.1016/0022-1694(70)90255-6
8. 8. Noori, N., ve Kalin, L., (2016) Coupling SWAT and ANN models for enhanced daily streamflow prediction, Journal of Hydrology, 533, 141–151. doi: 10.1016/j.jhydrol.2015.11.050

9. 9. Okkan, U., Ersoy, Z.B., Kumanlioglu A.A., ve Fistikoglu, O. (2021) Embedding machine learning techniques into a conceptual model to improve monthly runoff simulation: a nested hybrid rainfall-runoff modeling, Journal of Hydrology, 598, 126433. doi: 10.1016/j.jhydrol.2021.126433
10. 10. Okkan, U., ve Kirdemir, U. (2020) Towards a hybrid algorithm for the robust calibration of rainfall–runoff models, Journal of Hydroinformatics, 22(4), 876-899. doi: 10.2166/hydro.2020.016
11. 11. Okkan, U., ve Serbes, Z.A., (2012) Rainfall-runoff modeling using least squares support vector machines, Environmetrics, 23(6), 549–564. doi: 10.1002/env.2154
12. 12. Senbeta, D.A., Shamseldin, A.Y., ve O’Connor, K.M., (1999) Modification of the probability-distributed interacting storage capacity model, Journal of Hydrology, 224(3-4), 149–168. doi: 10.1016/S0022-1694(99)00127-4
13. 13. Tongal, H., ve Booij, M.J., (2018) Simulation and forecasting of streamflows using machine learning models coupled with base flow separation, Journal of Hydrology, 564(2018), 266-282. doi: 10.1016/j.jhydrol.2018.07.004
14. 14. Zhang, L., Potter, N., Hickel, K., Zhang, Y., ve Shao, Q., (2008) Water balance modeling over variable time scales based on the Budyko framework - Model development and testing, Journal of Hydrology, 360(1-4), 117–131. doi: 10.1016/j.jhydrol.2008.07.021