Simplification rationnelle des outils hydrologiques de gestion : recommandations méthodologiques pour la construction de modèles semi-empiriques à origine mécaniste Rational Simplification of Hydrological Mnagement Tools: Methodological Recommandations for Building Semi-Empirical Models of Mechanistic Origin

Résumé

Les modèles de simulation hydrologiques sont reconnus comme des outils mathématiques très performants mais généralement d'application difficile, principalement à cause du nombre élevé des paramètres requis. À l'inverse, les outils actuels de gestion sont généralement développés à l'aide d'approches empiriques limitant leur application. De plus, leurs paramètres ne sont pas des grandeurs mesurables et doivent être ajustés pour chaque situation. Une approche est proposée pour le développement d'outils semi empiriques de gestion. Elle consiste à simuler un grand nombre de scénarios en utilisant un modèle complexe de simulation puis à rationaliser l'information obtenue pour développer un nouveau modèle semi empirique. L'exemple illustrant cette approche concerne l'évaluation des flux d'eau ruisselée à la surface des champs, lessivée vers la nappe souterraine et drainée par les drains agricoles spécifiquement pour le contexte du Québec. À partir des résultats de simulation de 4500 scénarios, une simplification.rationalisation a permis de réduire à 120 le nombre de scénarios de référence à l'aide desquels peuvent être évalués tous les scénarios possibles par de simples interpolations linéaires. Une application de l'algorithme résultant sur un site du Québec a montré la bonne concordance entre les résultats calculés et mesurés. À la fois l'ordre de grandeur du ruissellement et du drainage et leur grandeur relative sont bien évalués.

Mots-clés:

• Flux,
• bilan en eau,
• gestion,
• modélisation,
• logiciel

Abstract

Considering the complexity of the water cycle in soil systems, models are used more than ever in parallel with field investigations to assist in the decision making process (KHAKURAL et ROBERT, 1993). Most available models are either too complicated (many non-measurable parameters) or too simple (empirical or site-specific) to be used as management tools. Such tools should conform to known theory and should be structured to enable efficient analysis of field situations with minimal requirements for parameters (CARSEL et al 1984). However, if the mechanistic models are very performing tools with regards to their representation of the processes and for the accuracy and reliability of their results, they are criticized for their complexity and for the large number of parameters they require. For this reason, their potential application as management tools cannot be recommended especially in preliminary investigations when the methodology has to be straight forward and rapidly implemented. On the other hand, existing management tools are often developed using an empirical approach for a specific context which considerably limits their transferability to different situations. Moreover, their empirical parameters often cannot be measured for the new situations, and must be adjusted for each new application.

A new approach conciliating the qualities of both kinds of tools was elaborated for the development of management tools. This approach consists in using mechanistic models for simulating a set of possible situations and in rationalizing the information obtained by simulation through regression analyses or other methods. An example of this methodology is presented in this paper with the development of the hydrological part (runoff, leaching and drainage) of a management tool dedicated to the evaluation of nutrient losses related to manure applications. Developed for the Quebec conditions, 4500 theoretical situations were considered corresponding to ten climates, nine soil textures, 25 crops and two slope values. Independently, agricultural management practices and drainage were taken into account.

For the mechanistic simulation of the water budget in the 4500 theoretical situations, the hydrologic module of the mechanistic-stochastic model AgriFlux was used (BANTON et al. 1993b). Because of the important field variability of most parameters, the stochastic AgriFlux model incorporates the variability resulting from field heterogeneity, measurement errors and intrinsic uncertainty related to parameter definition. The soil profile is divided in plot scale homogeneous horizons (or compartments) and a daily time step is used in the calculations. The water budget module in AgriFlux is named HydriFlux and simulates all the water-related processes (precipitations, snowmelt, infiltration, runoff, water uptake by plants, evaporation, percolation and drainage) using characteristic water contents and unsaturated hydraulic conductivity.

In the example presented, the simulation results obtained by running HydriFlux have shown that the soil water fluxes (runoff and percolation) vary as linear functions of both the annual rain volume (the most important characteristic of the climate) and the logarithm of the saturated hydraulic conductivity (the most important characteristic of the soil type). A reduction of the number of crops could also be achieved by taking into account the water needs and the water uptake curves of the crops. This rationalization-simplification reduced the number of theoretical simulations to be stored in the management tool to 120 (2 climates x 3 textures x 10 crops x 2 slopes). These represent only 2.7% of the initial situations simulated by the mechanistic HydriFlux model. The different water fluxes are stored in the management tool as tables in which direct interpolations are performed to calculate the fluxes corresponding to all the potential intermediary situations. Such developed management tool presents good qualities at the same time for its calculation speed, for its easy parameterization, for the reliability of its evaluation (through the evaluation of the mechanistic model) and for its high transferability and applicability to various situations. The calculations are rapidly done and their programming can be very easily made by using a spreadsheet software.

An application of this evaluation method has been done on an experimental site located in Quebec (ENRIGHT et MADRAMOOTOO, 1994), the only one for which both the runoff and the drainage have been measured during many years (1989 to 1991, April to December). The application on two fields (1.84 et 4.63 ha) has shown a good concordance between the calculated and measured results, as well for the magnitude of the fluxes than for the relative importance of these fluxes. Moreover, this application has shown that the variability of the measured values is higher than the calculated ones, attesting of the great influence of the variations in climatic, soil, crop and management conditions on the water budget. However, the good evaluation of the fluxes (for relative and absolute values) confirms the reliability of the proposed approach and of the simplification.

Keywords:

• Flux,
• water budget,
• management,
• modeling,
• software,
• AgriFlux