Computational & Technology Resources
an online resource for computational,
engineering & technology publications
Civil-Comp Proceedings
ISSN 1759-3433
CCP: 92
Edited by: B.H.V. Topping and Y. Tsompanakis
Paper 14

Calibrating Conceptual Rainfall-Runoff Models using a Real Genetic Algorithm

A.R. Awad1 and I. Von Poser2

1Department of Environmental Engineering, Faculty of Civil Engineering, Tishreen University, Lattakia, Syria
2Ingenieurtechnik, Merck K Ga A, Darmstadt, Germany

Full Bibliographic Reference for this paper
A.R. Awad, I. Von Poser, "Calibrating Conceptual Rainfall-Runoff Models using a Real Genetic Algorithm", in B.H.V. Topping, Y. Tsompanakis, (Editors), "Proceedings of the First International Conference on Soft Computing Technology in Civil, Structural and Environmental Engineering", Civil-Comp Press, Stirlingshire, UK, Paper 14, 2009. doi:10.4203/ccp.92.14
Keywords: real genetic algorithm, rainfall-runoff modeling, Alkabir Alshimali catchment data, model calibration, sequential simplex method.

The paper provides an accurate estimate of the factors affecting the urban rainfall-runoff quantity and quality using real genetic algorithms. In this paper a new genetic algorithm (GA) technique for optimal solving the calibrating problem of rainfall-runoff models has been applied. This methodology is considered original because the real representation in it takes all kind of codes (binary, integer, or real) without any need to transfer from one to another, i.e. our GA, using a chromosome as a unit, works equally well with integer and non-integer decision variables. This is a specific characteristic that distinguishes our real-coding GA from other (binary, Gray, or integer) coding GAs. Thus the encoding scheme used in this paper has proved to have the property of matching the effective problem search space with the GA search space.

Our rainfall-runoff model is very similar to Xinanjiang Rainfall-runoff model developed by Zhao et al.. It consists of two components dealing with water balance and routing, respectively [1]. A total of seven parameters are thus contained in the model. All of them are to be calibrated by minimizing the residual variance defined as the sum of squares of differences between computed and observed discharges.

The model developed by the authors was applied to calibrate the data of the developed rainfall-runoff model for the Al-Kabeer Al-Shimali river catchment (1200km2) where the terrain rolls gently and the climate is moderate, east of Lattakia city, using the real genetic algorithm for optimization. As for the case study, there are seven decision variables to be made concerning the hydrological model. A real string made of substrings is used for representing the problem in a suitable form for use within a GA. This string (chromosome) of seven genes represents a rainfall-runoff model consisting of seven variables, i.e. all seven parameters of the model are optimized.

The parameters used for implementing our GA technique in the real example are the population size of 80 with probabilities of 0.8 for crossover and 0.04 for mutation. The optimization was carried out over 50 generations.

All of the seven parameters of the model have been optimized in ten runs. Each run started from a different set of randomly selected initial points in the search space and with 4000 objective function evaluation. All ten runs proved to be able to locate the global optima.

The paper ends with the already known fact that the real GA can provide an efficient and robust means for calibrating the newly developed rainfall-runoff model. The real GA can further contribute in calibrating other hydrological models and can also be very useful in solving many different inverse problems and operation research problems in environmental modeling.

R.J. Zhao, Y.L. Zhuang, L.R. Fang, X.R. Liu, Q.S. Zhang, "The Xinanjiang model, Hydrological Forecasting", Proceedings of the Oxford Symposium, IAHS AISH Publ., 129, 351-356, 1980.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £78 +P&P)