Abstracts
Résumé
L'analyse de sensibilité constitue une étape importante dans le développement et la mise en place d'un modèle de simulation. Cette analyse de sensibilité du modèle CEQÉROSS effectuée sur un petit bassin versant agricole du Québec (Canada) montre que la dimension du maillage utilisé pour découper le bassin versant de Lennoxville en éléments de calcul exerce une influence directe, mais modérée, sur la charge sédimentaire transportée à l'exutoire. Les paramètres de calage associés à l'érosivité des précipitations et les variables d'entrée décrivant la topographie influencent considérablement les résultats de la simulation pour la période considérée et, de ce fait, nécessitent une plus grande attention lors des étapes de préparation des données et de calage du modèle. La forme linéaire de l'équation de production de sédiments en rivière exerce un rôle majeur sur les charges totales annuelles simulées. L'importance relative des paramètres associés à la granulométrie des sédiments transportés en rivière indique que la capacité du cours d'eau à transporter les particules grossières limite la charge totale annuelle évacuée à l'exutoire du bassin versant de Lennoxville pendant la période d'étude. Ces résultats doivent cependant être confrontés à d'autres analyses de sensibilité afin de décider des modifications éventuelles à apporter aux équations du modèle.
Mots-clés:
- Érosion,
- sédiments,
- bassin versant,
- analyse de sensibilité,
- CEQÉROSS
Abstract
Sensitivity analysis is a significant step in the development of computer models used to simulate soil erosion and nonpoint source pollution because it allows to evaluate objectively the importance of changes which occur in the response of a model (output) following changes brought in the parameters and variables of the input data. The objective of this study consists of using sensitivity analysis in order to verify the influence of calibration parameters, input variables and the size of the simulation units (i.e. grid squares) on the annual suspended sediment load (net erosion) of a small agricultural watershed (78 ha) located in Lennoxville, province of Quebec (Canada) for the period October 1991 to September 1992. This sensitivity analysis uses the results of an hydro-sedimentological simulation carried out with the CEQÉROSS computer model (DUCHEMIN et al. 2001). The sensitivity analysis takes into account calibration parameters and input variables selected for their ability to influence soil and river erosion and sediment transport (table 2). The analytical procedure consists in using a sensitivity index to quantify the response of the model to extreme (minimum and maximum) values of parameters and variables. The principal results of the sensitivity analysis are presented in figure 3.
The sensitivity analysis of the annual load of suspended sediments indicate that the size of the grid squares used to segment the watershed (i.e. RESOLUTION) had a direct but rather weak influence on net erosion. The results indicate that the erosivity calibration parameters EXPRP and COEFAM, the transport capacity parameters FCTS and PSDS and the topographic input variable PENTD associated to soil erosion, have a considerable influence on the annual load of suspended sediments and, therefore require a greater attention to be brought during the steps of data preparation and model calibration. The results indicate also that calibration parameters FPUSS1, FPUSS2, FSSRIV1 and FSSRIV2 associated to river erosion are the most influential. The linear form of the river sediment production equation contribute to increase the influence of these parameters on the output of the simulation. The fraction of fine sediments (POSS1) transported in the river plays an insignificant role on the annual load of suspended sediments simulated by the CEQÉROSS model. The annual suspended sediment load varies conversely to the increase of coarse sediments (POSS4). The river capacity to transport the coarse grained particles hinders the annual suspended load evacuated from the Lennoxville watershed. The results of the sensitivity analysis must be confronted with additional sensitivity analysis (i.e. larger watershed, longer period of simulation) using the CEQÉROSS model in order to decide if modifications should be undertaken to the model equations.
Keywords:
- Erosion,
- sediments,
- watershed,
- sensitivity analysis,
- CEQÉROSS