PROSPECTS FOR SHORT-TERM FORECASTING OF RIVER STREAMFLOW FROM SMALL WATERSHED RUNOFF USING MACHINE LEARNING METHODS

Section
Methods, models and technologies
  • Vsevolod M. Moreido Water Problems Institute of Russian Academy of Sciences, Moscow, Russia https://orcid.org/0000-0003-1763-4096
  • Boris I. Gartsman Water Problems Institute of Russian Academy of Sciences, Moscow, Russia
  • Dimitri P. Solomatine Water Problems Institute of Russian Academy of Sciences, Moscow, Russia; IHE Delft Institute for Water Education, Delft, The Netherlands
  • Zoya A. Suchilina Water Problems Institute of Russian Academy of Sciences, Moscow, Russia
Keywords: hydrological forecasting, machine learning, artificial neural networks, runoff forecasts, short-term streamflow forecasting, runoff modelling

The paper addresses prospects for short-term (from 1 to 7 days) forecasting of river streamflow runoff based on several machine learning methods: multiple linear regression (LM) model, a multilayer perceptron (MLP) artificial neural network, and a recurrent artificial neural network with long short-term memory (LSTM). Methods for expanding the set of predictors for model construction are proposed, and the possibility of random shuffling of the time-series of predictors for model calibration and verification are assessed. The object of the study is the small river of Central Russia – river Protva (Spas-Zagorie gauge). Current and lagged values of streamflow discharge at the gauge and daily precipitation at local weather stations are used as predictors for the model, as well as moisture index and evaporation rate. The obtained results show the possibility of constructing an effective operational forecasting system for short-term runoff forecasting. The study revealed the applicability of artificial neural network models, acceptable for operational practice, using all available hydrometeorological information on the catchment, as they showed the most stable results at all lead times from 1 to 7 days. In contrast to the linear model, which efficiency decreases after lead time of more than 3 days, the artificial neural networks models have higher forecast efficiency up to 7 days. The results obtained are robust for all phases of the water regime, both spring floods and summer floods. The software implementation of the models is made on the basis of open software libraries in the Python language, which makes it possible to widely use the methods for scientific research and applied problems.

Vsevolod M. Moreido,
Water Problems Institute of Russian Academy of Sciences, Moscow, Russia

E-mail: moreido@mail.ru
eLibrary (РИНЦ) SPIN-код: 5415-7452
ORCID ID: 0000-0003-1763-4096
Scopus ID: 56681224900 
Researcher ID (WoS):
D-7056-2014

Boris I. Gartsman,
Water Problems Institute of Russian Academy of Sciences, Moscow, Russia

E-mail: gartsman@inbox.ru
eLibrary SPIN-код: 7792-9120
Scopus ID: 24438012800
ORCID iD: 0000-0002-5876-7015

Dimitri P. Solomatine,
Water Problems Institute of Russian Academy of Sciences, Moscow, Russia; IHE Delft Institute for Water Education, Delft, The Netherlands

E-mail: d.solomatine@un-ihe.org
ORCID ID: 0000-0003-2031-9871
Scopus ID: 6601958160
Researcher ID (WoS):
M-9189-2013

Zoya A. Suchilina,
Water Problems Institute of Russian Academy of Sciences, Moscow, Russia

E-mail: mezozoya1@mail.ru
eLibrary (РИНЦ) SPIN-код: 8777-1177
ORCID iD: 0000-0002-2668-7408

Abadi M., Barham P., Chen J., Chen Zh., Davis A., Dean J., Devin M., Ghemawat S., Irving G., Isard M., Kudlur M., Levenberg J., Monga R., Moore Sh., Murray D.G., Steiner B., Tucker P., Vasudevan V., Warden P., Wicke M., Yu Y. Zheng X. TensorFlow: A System for Large-Scale Machine Learning. Proceedings of the 12th USENIX Symposium on Operating Systems Design and Implementation (OSDI’16) (Savannah, GA, USA, November 2-4, 2016). Publ. USENIX Association, 2016, pp. 265-283.

Abrahart R.J., Anctil F., Coulibaly P., Dawson Ch.W., Mount N.J., See L.M., Shamseldin A.Y., Solomatine D.P., Toth E., Wilby R.L. Two decades of anarchy? Emerging themes and outstanding challenges for neural network river forecasting. Progress in Physical Geography: Earth and Environment, 2012, vol. 36, iss. 4, pp. 480-513. DOI: 10.1177/0309133312444943.

Belyakova P.A., Borshch S.V., Khristoforov A.V., Yumina N.M. Prognoz maksimal'nogo stoka rek Chernomorskogo poberezh'ya Kavkaza [Forecast of the maximum runoff of the rivers of the Black Sea coast of the Caucasus]. Vodnoe khozyaistvo Rossii: problemy, tekhnologii, upravlenie [Water sector of Russia: problems, technologies, management], 2013, no. 6, pp. 4-16. (In Russian).

Beven K. Deep learning, hydrological processes and the uniqueness of place. Hydrological Processes, 2020, vol. 34, iss. 16, pp. 3608-3613. DOI: 10.1002/hyp.13805.

Bhattacharya B., Solomatine D.P. Machine learning in sedimentation modelling. Neural Networks, 2006, vol. 19, iss. 2, pp. 208-214. DOI: 10.1016/j.neunet.2006.01.007.

Borsch S.V., Khristoforov A.V. Otsenka kachestva prognozov rechnogo stoka [Hydrologic flow forecast verification]. Trudy Gidrometeorologicheskogo nauchno-issledovatel'skogo tsentra Rossiiskoi Federatsii [Proceedings of the Hydrometeorological Research Center of the Russian Federation], 2015, special issue 355. 198 p. (In Russian; abstract in English).

Emerton R.E., Stephens E.M., Pappenberger F., Pagano Th.C., Weerts A.H., Wood A.W., Salamon P., Brown J.D., Hjerdt N., Donnelly Ch., Baugh C.A., Cloke H.L. Continental and global scale flood forecasting systems. Wiley Interdisciplinary Reviews: Water, 2016, vol. 3, iss. 3, pp. 391-418. DOI: 10.1002/wat2.1137.

Fatichi S., Vivoni E.R., Ogden F.L., Ivanov V.Y., Mirus B., Gochis D., Downer Ch.W., Camporese M., Davison J.H., Ebel B., Jones N., Kim J., Mascaro G., Niswonger R., Restrepo P., Rigon R., Shen Ch., Sulis M., Tarboton D. An overview of current applications, challenges, and future trends in distributed process-based models in hydrology. Journal of Hydrology, 2016, vol. 537, pp. 45-60. DOI: 10.1016/j.jhydrol.2016.03.026.

Gelfan A.N., Motovilov Y.G. Long-term hydrological forecasting in cold regions: retrospect, current status and prospect. Geography Compass, 2009, vol. 3, iss. 5, pp. 1841-1864. DOI: 10.1111/j.1749-8198.2009.00256.x.

Halff A.H., Halff H.M., Azmoodeh M. Predicting runoff from rainfall using neural networks. Proceedings of Symposium Engineering Hydrology (San Francisco, California, July 25-30, 1993), 1993, pp. 760-765.

Haykin S. Neural Networks and Learning Machines: 3rd ed. Prentice Hall Publ., 2009. 906 p.

Haykin S. Neural networks: a comprehensive foundation. Prentice Hall PTR, Upper Saddle River, NJ, United States, 1994. 768 p.

Hochreiter S. The vanishing gradient problem during learning recurrent neural nets and problem solutions. International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems, 1998, vol. 6, no. 2, pp. 107-116. DOI: 10.1142/S0218488598000094.

Hochreiter S., Schmidhuber J. Long Short-Term Memory. Neural Computation, 1997, vol. 9, iss. 8, pp. 1735-1780. DOI: 10.1162/neco.1997.9.8.1735.

Kratzert F., Klotz D., Brenner C., Schulz K., Herrnegger M. Rainfall-runoff modelling using Long Short-Term Memory (LSTM) networks. Hydrology and Earth System Sciences, 2018, vol. 22, iss. 11, pp. 6005-6022. DOI: 10.5194/hess-22-6005-2018.

Kuchment L.S., Gel'fan A.N. Dinamiko-stokhasticheskie modeli formirovaniya rechnogo stoka [Dynamic-stochastic models of river runoff formation]. Moscow, Publ. Nauka, 1993. 101 p. (In Russian).

Motovilov Y.G. Hydrological simulation of river basins at different spatial scales: 1. Generalization and averaging algorithms. Water Resources. 2016a, vol. 43, iss. 3, pp. 429-437. DOI: 10.1134/S0097807816030118 (Russ. ed.: Motovilov Yu.G. Gidrologicheskoe modelirovanie rechnykh basseinov v razlichnykh prostranstvennykh masshtabakh. 1. Algoritmy generalizatsii i osredneniya. Vodnye Resursy, 2016a, vol. 43, no. 3, pp. 243-253. DOI: 10.7868/S0321059616030111).

Motovilov Y.G. Hydrological simulation of river basins at different spatial scales: 2. Test results. Water Resources, 2016b, vol. 43, iss. 5, pp. 743-753. DOI: 10.1134/S0097807816050092 (Russ. ed.: Motovilov Yu.G. Gidrologicheskoe modelirovanie rechnykh basseinov v razlichnykh prostranstvennykh masshtabakh. 2. Rezul'taty ispytanii. Vodnye Resursy, 2016b, vol. 43, no. 5, pp. 467-475. DOI: 10.7868/S0321059616050096).

Mozer M.C. A Focused Backpropagation Algorithm for Temporal Pattern Recognition. Complex Systems, 1989, vol. 3, iss.4, pp. 349-381.

Oudin L., Michel C., Anctil F. Which potential evapotranspiration input for a lumped rainfall-runoff model? Journal of Hydrology, 2005, vol. 303, iss. 1-4, pp. 275-289. DOI: 10.1016/j.jhydrol.2004.08.025.

Pedregosa F., Varoquaux G., Gramfort A., Michel V., Thirion B., Grisel O., Blondel M., Prettenhofer P., Weiss R., Dubourg V., Vanderplas J., Passos A., Cournapeau D., Brucher M., Perrot M., Duchesnay É. Scikit-learn: Machine Learning in Python. Journal of Machine Learning Research, 2011, vol. 12, pp. 2825-2830.

Rumelhart D.E., Hinton G.E., Williams R.J. Learning representations by back-propagating errors. Nature, 1986, vol. 323, no. 6088, pp. 533-536. DOI: 10.1038/323533a0.

Werbos P.J. Backpropagation Through Time: What It Does and How to Do It. Proceedings of the IEEE, 1990, vol. 78, no. 10, pp. 1550-1560. DOI: 10.1109/5.58337.

Zaikov B.D. Srednii stok i ego raspredelenie v godu na territorii SSSR [Average runoff and its distribution per year in the USSR]. Leningrad, Publ. Gidrometeoizdat, 1946. 148 p. (In Russian).


Abstract views: 10
PDF Downloads: 0
Published
2020-12-31
How to Cite
Vsevolod M. Moreido, Boris I. Gartsman, Dimitri P. Solomatine, & Zoya A. Suchilina. (2020). PROSPECTS FOR SHORT-TERM FORECASTING OF RIVER STREAMFLOW FROM SMALL WATERSHED RUNOFF USING MACHINE LEARNING METHODS. Hydrosphere. Hazard Processes and Phenomena, 2(4), 375-390. https://doi.org/10.34753/HS.2020.2.4.375

Downloads

Download data is not yet available.