6 th International Conference on Hydroinformatics - Liong, Phoon & Babovic (eds) © 2004 World Scientific Publishing Company, ISBN 981-238-787-0 1 A DATA MINING APPROACH TO MODELLING SEDIMENT TRANSPORT B. BHATTACHARYA 1 R.K. PRICE 2 D.P. SOLOMATINE 3 1,2,3 Department of Hydroinformatics and Knowledge Management, UNESCO-IHE Institute for Water Education, P.O. Box 3015, 2601 DA Delft, The Netherlands. Email: {bha, rkp, sol} @ihe.nl. Even though numerous models for predicting sediment transport rates are available their dependability is often questionable. Data mining (DM), which is particularly useful in modelling processes about which adequate knowledge of the physics is limited, is presented as a tool complimentary to modelling sediment transport. This paper reports on the use of DM methods such as artificial neural networks and model trees in modelling bed-load and total-load transport using measured data. The predictive accuracy of these models is compared with that of some well-known existing models. A conclusion is reached that the DM models are able to learn the complex transport process from the available data. INTRODUCTION A reasonable estimate of sediment transport rates in alluvial streams is important in the context of a number of water management issues. Even though extensive research over the last fifty years has produced a plethora of bed-load, suspended-load and total-load transport models the predictive accuracy of these models has barely increased. The adequacy of these models has been reviewed by ASCE [1], Gomez and Church [2], Yalin [3], Van Rijn [4]-[6], etc. Sediment transport is an immensely complex process and the expression of the transport process through a deterministic mathematical framework may not be possible in the foreseeable future. In parallel with research into sediment transport has been the emergence of new modelling paradigms such as data mining (DM). This has opened up new opportunities for modelling processes about which either the level of available knowledge is too limited to put the relevant information in a mathematical framework or too little data is available for calibrating an appropriate model. DM is presently being utilised in almost all branches of science as an alternative and complementary to the more traditional physically-based modelling system. Use of artificial neural networks (ANN) remains in the forefront of this complementary modelling practice. The recent successful applications of DM methods to modelling water engineering problems (e.g. ASCE, [7]) present DM as a suitable potential candidate to modelling sediment transport. This paper