Abstracts
Résumé
L’Algérie, caractérisée par un climat semi-aride, est menacée par l’érosion des terres agricoles qui provoque l’augmentation du transport solide et l’envasement croissant des barrages. Cet article décrit une nouvelle méthode d’estimation des flux de matières en suspension (MES) au niveau d’un barrage algérien (barrage de Beni Amrane) basée sur la logique floue. Cette dernière utilise des termes flous tels que « faible », « moyen » et « élevé », pour décomposer le processus débit-MES en plusieurs sous-ensembles flous et d’en déduire les quantités de matières solides en fonction du débit observé de la rivière. Les performances de cette méthode ont été évaluées en période de calage, mais aussi en période de validation, pour mieux juger de la capacité prédictive du modèle à ces deux échelles. En comparant la logique floue avec un modèle empirique régressif utilisant une relation de puissance, nous avons démontré la robustesse du modèle flou en tant qu’outil de quantification du transport solide.
Mots clés:
- érosion,
- transport solide,
- modèle empirique,
- logique floue,
- barrage de Beni Amrane,
- Algérie
Summary
Sediment transport and erosion is a complex natural process that is strongly influenced by human activities such as deforestation, agriculture and urbanization. In particular, suspended sediments play a key role in controlling water quality and they can cause a major reduction in the capacity of a stream for handling floods. In Algeria, increasing erosion and suspended loads are responsible for serious problems in agricultural land and hydraulic reservoirs, since the suspended load and its sedimentation lead to flooding and dam silting. Water and soil conservation practices, such as contour ridges and areas of reforestation, were introduced in many regions of Algeria in order to decrease erosion and to collect runoff in hill-slope catchments.
Relationships for water discharge and suspended sediment load can be divided into three types: empirical models that allow quantification of erosion on annual time scales, such as the Wischmeier and Smith soil loss equation; conceptual models, which include several reservoirs estimating sediment load on different time scales; and finally, physically-based models, which introduce physical laws such as the Saint-Venant equation. These models represent another category, and allow the estimation of sediment load in different areas of the watershed and supply spatial results. These models also take into account numerous variables that are difficult to obtain on a regional scale.
The objective of this research was to develop runoff-suspended sediment models for the Beni Amrane reservoir. This reservoir is located in the Isser watershed, situated in northern Algeria. This basin covers an area of 4,000 km2 and is characterized by a semi-arid climate and a very high soil erosion rate, exceeding 2,000 tons/km2/year. The Beni Amrane reservoir represents an important dam as it supplies the Keddara dam, which in turn supplies the town of Algiers with drinking water.
In the present study, two approaches to suspended sediment simulation were applied on hourly time scales for suspended sediment concentrations, and on daily time scales for water discharge and solid discharge analysis. The first approach is an empirical regression model based on a rating curve and uses a relationship between the observed runoff and the sediment concentration values. The model uses only two parameters, with the second being based on fuzzy logic. Fuzzy logic is already used in many scientific domains, and represents a new simulation technique based on artificial intelligence. Fuzzy variables were used to organize knowledge that is expressed ‘linguistically’ into a formal analysis (for example ‘high suspended sediment’, ‘average suspended sediment’ and ‘low suspended sediment’). The simulation results confirm the performances and robustness of the fuzzy logic model for the two time scales. In fact, the Nash criterion, which is the principal validation criterion for the models, displayed high performances in calibration and validation periods. The neurofuzzy model (fuzzy logic with neural supervised learning) offers a simulation advantage. On an hourly time scale, while increasing the number of fuzzy rules, the model results in good precision with the observed suspended sediments.
The fuzzy logic model results showed that the Nash criterion for two periods (calibration and validation) was greater than 88%, and the peaks of suspended sediments were generally correctly reproduced for the four episodes studied. This is in contrast to the empirical model, where the Nash efficiency was generally weak and decreased during the validation period. In this latter period, the Nash criterion was often negative, the global error was high and the maximum concentration peak was underestimated.
On a daily time scale, knowing the complexity of the runoff-suspended sediment process, we have analyzed these two models for solid discharge simulation. The study was carried out on daily solid discharge data collected from the gauging station on the Isser River (1986 to 1989). While based on the same validation criteria, i.e. the Nash efficiency and the global error, the fuzzy logic model appeared more robust than the empirical model. The fuzzy logic model produced better estimates of the daily sediment yield than the empirical model during calibration and validation periods, and it represents a high prediction power. Thus, we have validated the fuzzy logic model as a tool for simulation of runoff of suspended sediments, and it can be explored to predict sediment loading and silting in Algerian reservoirs.
Key-words:
- erosion,
- suspended sediment,
- empirical model,
- fuzzy logic,
- Beni Amrane dam,
- Algeria
Appendices
Références bibliographiques
- ANRH, 2000. Actualisation des volumes régularisables. Bulletin Agence Nationale des Ressources Hydrauliques (Algérie), N° 12/2000.
- ANRH, 1990. Étude hydrologique du barrage de Beni Amrane. Étude COBA, 40 p.
- ARABI M., 1991. Influence de quatre systèmes de production sur le ruissellement et l’érosion en milieu montagnard méditerranéen algérien. Thèse de doctorat, Université de Grenoble, 276 p.
- BENKHALED A. et B. REMINI, 2003. Analyse de la relation de puissance : débit solide - débit liquide à l’échelle du bassin versant de l’Oued Wahrane (Algérie). Rev. Sci. Eau, 16, 333-356.
- BENOIT E., 1993. Capteurs symboliques et capteurs flous : un nouveau pas vers l’intelligence artificielle. Thèse de Doctorat, Université Joseph Fourier, Grenoble, 164 p.
- BERENJI R., MALKANI A. et C. COPELAND, 1985. Tether control using fuzzy reinforcement learning, IEEE/IFES, Yokohama, Japan, 1315-1319.
- BERSINI H. et V. GORRINI, 1993. FUNNY, fuzzy or neural net methods for adaptive process control. Proceedings of EUFIT, Allemagne, 55-61, 1993.
- BOUCHON-MEUNIER B., 1995. La logique floue et ses applications, Addison-Wesley France, Paris, 257 p.
- DECHEMI N., BENKACI T. et A. ISSOLAH, 2003. Modélisation des débits mensuels par les modèles conceptuels et les systèmes neuro-flous. Rev. Sci. Eau, 16, 407-424.
- DEMMAK A., 1982. Contribution à l’étude de l’érosion et des transports solides en Algérie septentrionale. Thèse de Doctorat, Université Pierre et Marie Curie – Paris, 323 p.
- DRIANKOV D., HELLENDOOM H. et M. REINFRANK, 1993. An introduction to fuzzy control, Springer-Verlag, New York, États-Unis, 1993, 323 p.
- GAFREJ R., 1993. Modélisation conceptuelle du transfert des matières en suspension, effet d’échelles spatio-temporelles. These de Doctorat, Université Paris 6, CEMAGREF, 180 p.
- GLORENNEC P.Y, BARRET C. et M. BRUNNET, 1992. Application of neuro-fuzzy network to identification and control of nonlinear dynamical systems, Proceedings of IPMU, 1992, Mallorca, Spain, 507-510.
- JANG J., 1993. ANFIS: Adaptive-network-based fuzzy Inference Systems. IEEE SMC, 23, 665-685.
- MAMDANI E.H., 1974. Application of fuzzy algorithms for control of a simple dynamic Plant. Proceedings of the IEE Control and Science, 121, 12, 1585-1588, 1974.
- NAKOULA Y., 1997. Apprentissage des modèles linguistiques flous par jeu de règles pondérées. Thèse de Doctorat, Université de Savoie, 158 p.
- NASH J.E. et J.V. SUTCLIFFE, 1970. River flow forecasting through conceptual model. Part 1 - a discussion of principles. J. Hydrol., 10, 282-290.
- PICOUET C., HINGRAY B. et J.C., OLIVRY, 2001. Empirical and conceptual modelling of the suspended sediment dynamics in a large tropical African river: the Upper Niger river basin. J. Hydrol., 250, 19-39.
- ROOSE E., 1994. Introduction à la GCES. Bull. Pédol. FAO, Rome, 70, 420 p.
- ROOSE E., 2004. Évolution historique des stratégies de lutte antiérosive. Vers la gestion conservatoire de l’eau, de la biomasse et de la fertilité des sols, Sécheresse, 1, 9-18.
- SUGENO M., 1985. Industrial applications of fuzzy control, Elsevier sciences publications, New York, États-Unis, 269 p.
- VAN SICKLE J. et R. BESCHTA, 1983. Supply based models of suspended sediment transport in Streams. Water Resour. Res., 19, 768-778.
- WALLING D.E. et B.W. WEBB, 1981. The reliability of suspended sediment load data. Erosion and sediment transport measurement, Proceedings of the Florence Syrnposium, IAHS Publ., 133, 177-194, 1981.
- WISCHMEIER W.H. et D.D. SMITH, 1960. A universal soil loss estimating equation to guide conservation farm planning. Proceedings 7th Int. Congress Soil Sci. Soc., 1, 418-425, 1960.
- ZADEH L., 1965. Fuzzy sets. Inf. Contr., 8, 338-353.
- ZADEH L., 1971. Quantitative fuzzy semantics, Inf. Sci., 3, 159-176.