HESS Hydrology and Earth System Sciences HESS Hydrol. Earth Syst. Sci. 1607-7938 Copernicus Publications Göttingen, Germany 10.5194/hess-17-2669-2013 Assessing the predictive capability of randomized tree-based ensembles in streamflow modelling Galelli S. https://orcid.org/0000-0003-2316-3243 1 Castelletti A. https://orcid.org/0000-0002-7923-1498 2 3 Singapore-Delft Water Alliance, National University of Singapore 2 Engineering Drive 2, 117577, Singapore Dipartimento di Elettronica, Informazione e Bioingegneria, Politecnico di Milano Piazza L. da Vinci, 32, 20133 Milano, Italy Centre for Water Research, University of Western Australia, Crawley, Western Australia, Australia 11 07 2013 17 7 2669 2684 Copyright: © 2013 S. Galelli 2013 This work is licensed under the Creative Commons Attribution 3.0 Unported License. To view a copy of this licence, visit https://creativecommons.org/licenses/by/3.0/ This article is available from https://hess.copernicus.org/articles/17/2669/2013/hess-17-2669-2013.html The full text article is available as a PDF file from https://hess.copernicus.org/articles/17/2669/2013/hess-17-2669-2013.pdf

Combining randomization methods with ensemble prediction is emerging as an effective option to balance accuracy and computational efficiency in data-driven modelling. In this paper, we investigate the prediction capability of extremely randomized trees (Extra-Trees), in terms of accuracy, explanation ability and computational efficiency, in a streamflow modelling exercise. Extra-Trees are a totally randomized tree-based ensemble method that (i) alleviates the poor generalisation property and tendency to overfitting of traditional standalone decision trees (e.g. CART); (ii) is computationally efficient; and, (iii) allows to infer the relative importance of the input variables, which might help in the ex-post physical interpretation of the model. The Extra-Trees potential is analysed on two real-world case studies – Marina catchment (Singapore) and Canning River (Western Australia) – representing two different morphoclimatic contexts. The evaluation is performed against other tree-based methods (CART and M5) and parametric data-driven approaches (ANNs and multiple linear regression). Results show that Extra-Trees perform comparatively well to the best of the benchmarks (i.e. M5) in both the watersheds, while outperforming the other approaches in terms of computational requirement when adopted on large datasets. In addition, the ranking of the input variable provided can be given a physically meaningful interpretation.

 References Babovic, V. and Keijzer, M.: Rainfall-runoff modeling based on genetic programming, Nord. Hydrol., 5, 331–346, 2002. Bachmair, S. and Weiler, M.: Hillslope characteristics as controls of subsurface flow variability, Hydrol. Earth Syst. Sci., 16, 3699–3715, <a href="http://dx.doi.org/10.5194/hess-16-3699-2012">https://doi.org/10.5194/hess-16-3699-2012</a>, 2012. Beck, M.: Forecasting environmental change, J. Forecast., 10, 3–19, 1991. Beven, K.: How far can we go in distributed hydrological modelling?, Hydrol. Earth Syst. Sci., 5, 1–12, <a href="http://dx.doi.org/10.5194/hess-5-1-2001">https://doi.org/10.5194/hess-5-1-2001</a>, 2001. Bhattacharya, B. and Solomatine, D.: Neural networks and M5 model trees in modelling water level-discharge relationship, Neurocomputing, 63, 381–396, 2005. Box, G. E. P., Hunter, J. S., and Hunter, W. G.: Statistics for Experimenters: Design, Innovation, and Discovery, 2nd Edition, Wiley, New York, USA, 2005. Breiman, L.: Bagging predictors, Mach. Learn., 24, 123–140, 1996. Breiman, L.: Random forests, Mach. Learn., 45, 5–32, 2001. Breiman, L., Friedman, J., Olsen, R., and Stone, C.: Classification and Regression Trees, Wadsworth & Brooks, Pacific Grove, CA, 1984. Castelletti, A., Galelli, S., Restelli, M., and Soncini-Sessa, R.: Tree-based reinforcement learning for optimal water reservoir operation, Water Resour. Res., 46, W09507, <a href="http://dx.doi.org/10.1029/2009WR008898">https://doi.org/10.1029/2009WR008898</a>, 2010. Cutler, A. and Guohua, Z.: PERT – Perfect Random Trees Ensembles, Comp. Sci. Stat., 33, 490–497, 2001. Dawson, C., Brown, M., and Wilby, R.: Inductive learning approaches to rainfall-runoff modelling, Int. J. Neural Syst., 10, 43–57, 2000. Dietterich, T.: Ensemble methods in machine learning, Lect. Notes Comput Sc., 1857, 1–15, 2000. Elshorbagy, A., Corzo, G., Srinivasulu, S., and Solomatine, D. P.: Experimental investigation of the predictive capabilities of data driven modeling techniques in hydrology –Part 1: Concepts and methodology, Hydrol. Earth Syst. Sci., 14, 1931–1941, <a href="http://dx.doi.org/10.5194/hess-14-1931-2010">https://doi.org/10.5194/hess-14-1931-2010</a>, 2010a. Elshorbagy, A., Corzo, G., Srinivasulu, S., and Solomatine, D. P.: Experimental investigation of the predictive capabilities of data driven modeling techniques in hydrology –Part 2: Application, Hydrol. Earth Syst. Sci., 14, 1943-1961, <a href="http://dx.doi.org/10.5194/hess-14-1943-2010">https://doi.org/10.5194/hess-14-1943-2010</a>, 201b. Erdal, H. and Karakurt, O.: Advancing monthly streamflow prediction accuracy of CART models using ensemble learning paradigms, J. Hydrol., 477, 119–128, 2013. Fonteneau, R., Wehenkel, L., and Ernst, D.: Variable selection for dynamic treatment regimes: a reinforcement learning approach, in: Proceedings of the European Workshop on Reinforcement Learning, 30 June–4 July 2008, Villeneuve d'Ascq, France, 2008. Freund, Y. and Schapire, R.: Experiments with a new boosting algorithm, in: Proceedings of 13th International Conference on Mach. Learn., 3–6 July 1996, Bari, Italy, 148–146, 1996. Galelli, S., Goedbloed, A., Schwanenberg, D., and van Overloop, D.: Optimal real-time operation of multi-purpose urban reservoirs: a case study in Singapore, J. Water Res. Pl.-ASCE, <a href="http://dx.doi.org/10.1061/(ASCE)WR.1943-5452.0000342">https://doi.org/10.1061/(ASCE)WR.1943-5452.0000342</a>, in press, 2013. Geurts, P.: Contributions to Decision Tree Induction: Bias/Variance Tradeoff and Time Series Classification, Ph.D. thesis, University of Liège, Liège, Belgium, 2002. Geurts, P., Ernst, D., and Wehenkel, L.: Extremely randomized trees, Mach. Learn., 63, 3–42, 2006. Hejazi, M. and Cai, X.: Input variable selection for water resources systems using a modified minimum redundancy maximum relevance (mMRMR) algorithm, Adv. Water Resour., 32, 582–593, 2009. Ho, T.: Random decision forests, in: Proceedings of the Third International Conference on Document Analysis and Recognition, Vol. 1, ontreal, Quebec, 278–282, 1995. Ho, T.: The random subspace method for constructing decision forests, IEEE T. Pattern Anal., 20, 832–844, 1998. Hsu, K., Gupta, H., and Sorooshian, S.: Artificial neural-network modeling of the rainfall-runoff process, Water Resour. Res., 31, 2517–2530, 1995. Hundecha, Y., Bardossy, A., and Theisen, H.: Development of a fuzzy logic-based rainfall-runoff model, Hydrolog. Sci. J., 46, 363–376, 2001. Hwang, S., Ham, D., and Kim, J.: A new measure for assessing the efficiency of hydrological data-driven forecasting models, Hydrolog. Sci. J., 57, 1–18, 2012. Hyafil, L. and Rivest, R.: Constructing optimal binary decision trees in NP-Complete, Inform. Process. Lett., 5, 15–17, 1976. Iorgulescu, I. and Beven, K.: Nonparametric direct mapping of rainfall-runoff relationships: an alternative approach to data analysis and modeling?, Water Resour. Res., 40, W08403, <a href="http://dx.doi.org/10.1029/2004WR003094">https://doi.org/10.1029/2004WR003094</a>, 2004. Jakeman, A. and Hornberger, G.: How much complexity is warranted in a rainfall-runoff model, Water Resour. Res., 29, 2637–2649, 1993. Jekabsons, G.: M5PrimeLab – M5' Regression Tree and Model Tree Toolbox for Matlab/Octave, Technical Report ver. 1.0.1, Faculty of Computer Science and Information Technology – Riga Technical University, Riga, Latvia, 2010. Jong, K., Mary, J., Cornuéjols, A., Marchiori, E., and Sebag, M.: Ensemble feature ranking, in: Knowledge Discovery in Databases, Springer, 267–278, 2004. Jothiprakash, V. and Kote, A.: Effect of pruning and smoothing while using M5 model tree technique for reservoir inflow prediction, J. Hydrol. Eng., 16, 563–574, 2011. Kuncheva, L. and Whitaker, C.: Measures of diversity in classifier ensembles and their relationship with the ensemble accuracy, Mach. Learn., 51, 181–207, 2003. Laaha, G. and Blöschl, G.: A comparison of low flow regionalisation methods – catchment grouping, J. Hydrol., 323, 193–214, 2006. Lin, J.-Y., Cheng, C.-T., and Chau, K.-W.: Using support vector machines for long-term discharge prediction, Hydrolog. Sci. J., 51, 599–612, 2006. Maier, H. and Dandy, G.: Neural networks for the prediction and forecasting of water resources variables: a review of modelling issues and applications, Environ. Model. Softw., 15, 101–124, 2000. Quinlan, J.: Learning with continuous classes, in: Proceedings of the 5th Australian Joint Conference on Artificial Intelligence, 16–18 November, Hobart, Australia, 343–348, 1992. Rasmussen, P., Salas, J., Fagherazzi, L., Rassam, J., and Bobee, B.: Estimation and validation of contemporaneous PARMA models for streamflow simulation, Water Resour. Res., 32, 3151–3160, 1996. Romanowicz, R., Young, P., Beven, K., and Pappenberger, F.: A data based mechanistic approach to nonlinear flood routing and adaptive flood level forecasting, Adv. Water Resour., 31, 1048–1056, 2008. Sauquet, E. and Catalogne, C.: Comparison of catchment grouping methods for flow duration curve estimation at ungauged sites in France, Hydrol. Earth Syst. Sci., 15, 2421–2435, <a href="http://dx.doi.org/10.5194/hess-15-2421-2011">https://doi.org/10.5194/hess-15-2421-2011</a>, 2011. See, L., Jain, A., Dawson, C., and Abrahart, R.: Visualisation of hidden neuron behaviour in a neural network rainfall-runoff model, in: Practical Hydroinformatics, vol. 68 of Water Science and Technology Library, edited by: Abrahart, R., See, L., and Solomatine, D., Springer, Berlin, Heidelberg, 87–99, 2008. Selvalingam, S., Liong, S., and Manoharan, P.: Use of RORB and SWMM models to an urban catchment in Singapore, Adv. Water Resour., 10, 78–86, 1987. Shamseldin, A. Y., Nasr, A. E., and O'Connor, K. M.: Comparison of different forms of the Multi-layer Feed-Forward Neural Network method used for river flow forecasting, Hydrol. Earth Syst. Sci., 6, 671–684, <a href="http://dx.doi.org/10.5194/hess-6-671-2002">https://doi.org/10.5194/hess-6-671-2002</a>, 2002. Snelder, T., Lamouroux, N., Leathwick, J., Pella, H., Sauquet, E., and Shankar, U.: Predictive mapping of the natural flow regimes of France, J. Hydrol., 373, 57–67, 2009. Solomatine, D. and Dulal, K.: Model trees as an alternative to neural networks in rainfall-runoff modelling, Hydrolog. Sci. J., 48, 399–411, 2003. Solomatine, D. and Ostfeld, A.: Data-driven modelling: some past experiences and new approaches, J. Hydroinform., 10, 3–22, 2008. Solomatine, D. and Xue, Y.: M5 model trees compared to neural networks: application to flood forecasting in the upper reach of the Huai River in China, J. Hydrol. Eng., 9, 491–501, 2004. Stravs, L. and Brilly, M.: Development of a low-flow forecasting model using the M5 machine learning method, Hydrolog. Sci. J., 52, 466–477, 2007. Vezza, P., Comoglio, C., Rosso, M., and Viglione, A.: low flows regionalization in North-Western Italy, Water Resour. Manage., 24, 4049–4074, 2010. Wang, Y. and Witten, I.: Induction of model trees for predicting continuous classes, in: Proceedings of the European Conference on Mach. Learn., Prague, Czech Republic, 128–137, 1997. Wehenkel, L.: Automatic Learning Techniques in Power Systems, Kluwer Academic, Boston, USA, 1998. Wei, W. and Watkins Jr., D.: Data mining methods for hydroclimatic forecasting, Adv. Water Resour., 34, 1390–1400, 2011. Wheater, H., Jakeman, A., and Beven, K.: Progress and Directions in Rainfall-Runoff Modelling, John Wiley, Chichester, 101–132, 1993. Xie, J.: Dealing with Water Scarcity in Singapore: Institutions, Strategies, and Enforcement, China: Addressing Water Scarcity Background Paper No. 4, The World Bank – Environment and Social Development Department – East Asia and Pacific Region, Washington, DC, available at: <a href="http://www.worldbank.org/eapenvironment/ChinaWaterAAA">http://www.worldbank.org/eapenvironment/ChinaWaterAAA</a> (last access: 15 January 2013), 2006. Young, P.: Data-based mechanistic and top-down modelling, in: Proceedings of the First Biennial Meeting of the International Environmental Modelling & Software Society, Vol. I, Lugano, Suisse, 363–374, 2002. Young, P.: Top-down and data-based mechanistic modelling of rainfall–flow dynamics at the catchment scale, Hydrol. Process., 17, 2195–2217, 2003. Young, P.: Rainfall-Runoff Modeling: Transfer Function Models, John Wiley & Sons, Ltd, 2006. Young, P.: Hypothetico-inductive data-based mechanistic modeling of hydrological systems, Water Resour. Res., 49, 915–935, 2013. Young, P. and Beven, K.: Data-based mechanistic modeling and the rainfall-flow nonlinearity, Environmetrics, 3, 335–363, 1994. Young, P., Jakeman, A., and Post, D.: Recent advances in the data-based modelling and analysis of hydrological systems, Water Sci. Technol., 36, 99–116, 1997.