Développement d’une nouvelle méthode de régionalisation basée sur le concept de « régime des débits naturels » : la méthode éco-géographique Development of a new method for flow regionalization based on the «natural flow regime» concept: The ecogeographic method

Résumé

Nous proposons une nouvelle méthode de régionalisation des débits fondée sur le concept de « régime des débits naturels » introduit en écologie aquatique : l’approche éco-géographique. Elle se distingue de deux approches de régionalisation existantes (approches hydrologique et écologique) sur les trois points suivants : le choix des variables hydrologiques, l’échelle d’analyse et la finalité de la régionalisation. En ce qui concerne le choix des variables hydrologiques, la nouvelle méthode est fondée sur le choix des caractéristiques des débits et non sur les variables hydrologiques. Ces caractéristiques des débits sont définies au moyen de l’analyse en composantes principales appliquée sur les variables hydrologiques. Contrairement aux autres approches, l’approche éco-géographique tient compte de toutes les caractéristiques des débits dans la régionalisation conformément au concept de « régime des débits naturels ». Quant à l’échelle d’analyse, à l’instar de l’approche écologique, la nouvelle méthode s’applique aussi à toutes les échelles d’analyse (annuelle, mensuelle et journalière) mais en les considérant séparément afin de tenir compte de toutes les caractéristiques de débits dans la régionalisation. Enfin, la finalité de la nouvelle méthode est de pouvoir déterminer les facteurs de variabilité spatiale des caractéristiques de débits (et non des variables hydrologiques) au moyen de l’analyse canonique des corrélations, notamment afin d’assurer une gestion durable des ressources hydriques dans un contexte de changement de l’environnement. Nous avons appliqué cette nouvelle méthode aux débits moyens annuels au Québec.

Mots clés:

• Régionalisation,
• régime des débits naturels,
• approche éco-géographique,
• débits moyens annuels,
• analyse en composantes principales,
• coefficients de corrélation,
• analyse canonique des corrélations,
• facteurs environnementaux,
• Saint-Laurent,
• Québec

Abstract

Flow regionalization has been the subject of numerous hydrologic studies. However, despite the development of regionalization methods, there are still differences in the approaches used amongst hydrologists on the one hand, and between hydrologists and experts in other fields (aquatic ecology and physical geography) on the other hand. Those differences relate to five aspects of the regionalization process: the choice of hydrologic variables, station grouping methods to produce homogeneous hydrologic regions, the choice of appropriate statistical laws to estimate quantiles for non-gauged or partially-gauged sites, the scale of flow analysis, and the ultimate purpose of the regionalization exercise. Depending on the choice of hydrologic variables, the scale of analysis and their ultimate purpose, regionalization studies may thus be divided according to two distinct approaches: the hydrologic approach and the ecologic approach.

The ultimate purpose of the hydrologic approach is to estimate flows at non-gauged or partially-gauged sites. For this reason, it has been primarily concerned with methods that allow the grouping of stations into homogeneous hydrologic regions and with the choice of statistical laws to estimate quantiles for non-gauged or partially-gauged sites. However, despite its undeniable interest from a practical point of view, this approach does not address the concerns of ecologists and geographers for three reasons: 1) the choice of hydrologic variables used for regionalization is not based on a scientific concept (this choice is arbitrary, and the variables selected do not constrain all the flow characteristics); 2) the ultimate purpose of the regionalization exercise is limited to estimating flows and is of limited interest to geographers and ecologists; 3) regionalization is performed at a daily scale, without taking into account other scales.

To make up for these limitations, ecologists have recently proposed regionalization based on the “natural flow regime” concept (the ecologic approach), which allows all fundamental flow characteristics (magnitude, frequency, duration, timing of occurrence and variability) to be taken into account. The rationale for considering all flow characteristics is that each characteristic has an effect on the behaviour of river ecosystems. Hence, regionalization based on the ecologic approach relies on a large number of hydrologic variables that define the fundamental flow characteristics. Rather than being arbitrary, the choice of variable is based on this new paradigm. Regionalization using the ecologic approach considers all time scales, and its ultimate purpose is to account for differences in the structure and biological composition of aquatic ecosystems.

However, one of the limitations of studies based on this approach is that, no matter how numerous they are, the variables used for regionalization do not constrain all flow characteristics, as required by the natural flow regime concept, so that application of this concept is incomplete. In addition, simultaneous analysis of all time scales does not allow consideration of all flow characteristics. To overcome these limitations, we propose a new regionalization approach based on the natural flow regime concept, an “ecogeographic” approach that differs from the ecologic approach in three ways. First, the proposed method is based on the use of flow characteristics rather than hydrologic variables. The reason for this is that there are an infinite number of hydrologic variables to define the five fundamental characteristics, making it impossible to account for all of them in the regionalization process. In contrast, since the number of fundamental flow characteristics is limited, they can all be taken into account, consistent with the “natural flow regime” requirements. Second, the ultimate purpose of the proposed regionalization method is to identify the physiographic and climatic factors that explain the spatial variability of these fundamental characteristics. To achieve this goal, it is necessary to analyze the different time scales (daily, monthly, annual) separately given the fact that it is impossible to constrain the effect of these various physiographic and climatic factors at all time scales. Indeed, some factors may show an effect at some time scales and not at others. This ultimate purpose addresses the concerns of geographers interested in explaining the spatial variability of such phenomena, among other things. Finally, separate analysis of the various time scales makes it possible to define all flow characteristics linked to a given time scale. As such, application of the “natural flow regime” concept to regionalization is complete.

Application of the ecogeographical method involves four separate steps: 1) the definition of the flow characteristics for the hydrologic series of interest; 2) the determination of minor and major characteristics using principal component analysis, where a “major” flow characteristic is defined as one which meets the following criterion: TVE ≥ (100% / N), where N is the total number of characteristics that define the analyzed hydrologic series and TVE is the total variance explained; 3) the grouping of stations in homogeneous hydrologic regions based on factorial scores. Homogeneous hydrologic regions are divided in two types based on the presence or absence of stations: effective homogeneous regions contain stations whereas fictive homogenous regions do not; 4) the determination of the factors that affect the spatial variability of flow characteristics. This is achieved using canonical correlation analysis, an approach that we have applied to average annual flows in Quebec watersheds.

Keys Words:

• Regionalization,
• natural flow regime,
• eco-geographic approach,
• mean annual flow,
• principal components analysis,
• coefficients of correlation,
• canonical correlation analysis,
• environmental factors,
• St. Lawrence River,
• Québec