Cet article brosse un portrait de différents types de modélisation hydrologique développés à ce jour. Nous passerons donc en revue l'hydrologie, à l'érosion hydrique des sols, au transport et aux transformations des polluants et à la qualité de l'eau en rivière. Ce bref survol, nous amène à conclure que si le développement de la modélisation hydrologique s'est fait jusqu'ici essentiellement en affinant la description des processus et en considérant des échelles spatiales et temporelles plus fines, l'étape suivante passe par l'intégration de ces divers modèles. Cette intégration permettra dès lors de considérer un ensemble de problématiques directement liées aux aspects de gestion environnementale.
This paper presents an overview of physically-based hydrological modeling approaches and a look at the future of hydrological modeling within the context of water management. It extends beyond classical hydrological modeling by surveying the modeling of water contaminants transport in porous media and surface waters, as well as soil erosion.Increasing concerns in predicting the impacts of land use management on the hydrological cycle have led researchers to construct two types of physically-based distributed models. The first type of model views the watershed as an ensemble of inter-connected reservoirs and mimics water routing with various types of discharge expressions and conceptual models (e.g., the infiltration models of Green and Ampt (1911), Holtan (1961) or Smith and Parlange (1978); the unit hydrographs of Sherman (1932) and Dooge (1973) and the geomorphological unit hydrograph of Rodriguez-Iturbe and Valdes (1979); the ground water discharge model of Beven and Kirby (1979); etc...). It is noteworthy that the pioneering Stanford Watershed Model of Crawford and Linsley (1966) led to the development of many currently used hydrological models including HBV (Bergstršm and Forsman, 1973), SLURP (Kite, 1978), TOPMODEL ( Beven and Kirby, 1979) and CEQUEAU (Morin et al., 1981), to name a few. The second type of model discretizes the watershed into an ensemble of control volumes and mimics water routing using combinations of partial differential equations for mass and momentum conservation and phenomenological models (e.g., Darcy's (1856), Dupuit's (1863), Boussinesq's (1904) and Richards (1931) equations for unsaturated and saturated flow in porous media; Saint-Venant's (1871) and Manning's (1891) equations for overland and open channel flows). Hydrological models such as SHE (Abbott et al.,1986a, b), IHDM (Calver, 1988), KINEROS (Woolhiser et al., 1990), THALES (Grayson et al.,1992) and HYDROTEL (Fortin et al., 1995), among others, represent classical examples of this type of modeling. It is noteworthy that recent advances in remote sensing and in digital elevation modeling have greatly facilitated and simplified the use of most of the hydrological models.On another front, the adverse effects of agricultural, industrial and urban runoff on surface and ground waters have motivated the development and application of different approaches to predict the fate and transport of various water contaminants in the environment (i.e., eroded soil particles, adsorbed and dissolved nutrients and pesticides as well organic matter).In soil erosion modeling, these concerns have led researchers to construct nonpoint source pollution models for evaluating the impacts of alternative land management practices on water quality. Based on the empirical Universal Soil Loss Equation (Wischmeier and Smith, 1978), the first nonpoint source models included CREAMS (Knisel et al., 1980), AGNPS (Young et al., 1987) and SWRRB (Williams et al., 1985). However, the lack of physical realism in these empirical formulations prompted the development of physically-based erosion models such as GUEST (Rose et al., 1983; Hairshine and Rose, 1992a, b), WEPP (Nearing et al., 1989), LISEM (De Roo et al., 1994) and EUROSEM (Morgan et al., 1992). The advantage of these models over the USLE resides in their ease of integration with physically-based hydrological models. Because of its close ties with the hydrological cycle and the soil erosion process (adsorbed and dissolved contaminants), the development of physically-based models for nutrient and pesticide transport benefited directly from advances in soil erosion modeling, soil chemistry and soil physics. The modeling of nitrogen transport is a representative example of this. Early modeling efforts involved the coupling of first-order kinetics models for the nitrogen cycle (Mehran and Tanji, 1974) with two types of mass conservation equation in porous media: the convection-dispersion equation and the capacity transport equation. Well known soil nitrogen dynamics models include NCSOIL (Molina et al., 1983), SOILN (Johnsson et al, 1987), EPIC (Sharpley and Williams, 1990), LEACHN and LEACHA (Hutson and Wagenet, 1991, 1992, 1993), DAISY (Hansen et al., 1991) and AgriFlux (Banton et al., 1993).The first attempt to model surface water quality goes back to the work of Streeter and Phelps (1925) who studied the impacts of a municipal waste water discharge on dissolved oxygen (DO) and biological oxygen demand (BOD) of an Ohio river. To predict DO and BOD dynamics, Streeter and Phelps assumed uniform and steady flow conditions and used first-order kinetics to model atmospheric supply of oxygen and oxygen consumption. The advances in computational power during the 70s and 80s allowed several researchers to substantially increase the complexity of the Streeter-Phelps approach. This was achieved by accounting for advection-dispersion phenomena, unsteady two and three dimensional flow conditions, as well as the effect of temperature on various chemical reactions. The QUAL2E model of Brown and Barnwell (1987) is a good example of a moderately complex water quality model where advection-dispersion and temperature effects on several water characteristics and contaminants are considered under one-dimensional steady flow conditions.At present, the state of hydrological modeling and software engineering has reached a point where it is now possible to construct spatial decision support systems (SDDS) capable of simulating the impacts of various management practices (i.e., industrial, municipal and agricultural) on the water quantity and the quality of a watershed's river network. These systems, which idealy should be user-friendly for decision makers, will be both integrated modeling systems (including a database system, hydrologic, soil erosion, agricultural-chemical transport and water quality models) and spatial data analysis systems (including a geographical information system). Currently developed SDDS include PÉGASE (Smitz et al., 1997) and GIBSI (Villeneuve et al., 1996, 1997a,b). In a sustainable water management context, the use of such systems will provide decision makers with a complete tool for exploring a variety of integrated watershed management programs.
