Calcul de probabilités de dépassement d'objectifs environnementaux de rejets de sources ponctuelle et diffuse à l'aide du système de modélisation intégrée GIBSI Computation of probability of exceeding environmental load allocations from point and diffuse sources using the integrated modeling system GIBSI

Résumé

Dans un contexte de gestion intégrée de l'eau par bassin versant, on doit, avant d'implanter tout programme d'assainissement, comparer les solutions de rechange, distinguer les contributions des différentes sources de pollution aux problèmes de qualité de l'eau, et définir des objectifs environnementaux de rejet (OER) associés aux activités et aux pressions anthropiques. Dans le cas des sources de pollution d'origine ponctuelle, on détermine un OER et son risque de dépassement en fonction d'un débit d'étiage critique d'une période de retour donnée. Dans le cas de la pollution diffuse d'origine agricole, il n'existe pas de concepts similaires pour définir des OER. L'approche que nous proposons, c'est de simuler, à l'aide de chroniques météorologiques et de divers scénarios de gestion de ces rejets, les concentrations résultantes dans les cours d'eau et de calculer les probabilités de dépassement des critères de qualité de l'eau (CQE). Cette approche permet de caractériser la prise de décision pour la période de contrôle en matière du nombre moyen de jours de dépassement d'un CQE. Ceci représente une quantification plus exacte du risque car ce dernier est interprété en fonction de ce qui est mesuré en rivière, c'est-à-dire les concentrations de polluants. Dans cette étude, on illustre l'application de ce concept à partir d'un cas de figure dans le bassin versant de la rivière Chaudière (Québec, Canada) qui est simulé avec le système de modélisation intégrée GIBSI. Dans le cadre d'une analyse de scénarios de restauration des activités récréatives conduisant à des contacts directs et fréquents avec l'eau, on illustre comment on peut, en réduisant respectivement les pressions agricole et urbaine de 32 et 17%, abaisser globalement les probabilités de dépassement des critères esthétique et bactériologique de 0,32 à 0,19 et de 0,94 à 0, respectivement.

Mots-clés:

• Gestion intégrée des bassins versants,
• pollutions ponctuelle et diffuse d'origines agricole et municipale,
• objectifs environnementaux de rejet,
• qualité de l'eau,
• probabilités de dépassement de critères de qualité de l'eau

Abstract

In an integrated watershed management context, the implementation of any clean water program requires the evaluation of the contribution of pollutant loads associated with wet (nonpoint or diffuse) and dry weather (point) sources to the studied water quality problem. It is also necessary to set environmental load allocations (ELA) or total maximum daily loads (TMDL) for various anthropogenic activities and to link these loads to water pollutant concentrations. The ELA or TMDL is a numerical quantity determining the maximum load of pollutants from point and nonpoint sources as well as background sources, to receiving water bodies that will meet designated water uses (e.g., swimming or fishing) in terms of water quality standards (WQS). The American Congress included a TMDL program in the Clean Water Act of 1972 (US EPA, 1997; 1999). The outcome of the TMDL program corresponds to the drafting of a watershed management plan (NOVOTNY, 1999).

Estimation of ELA from point sources is generally based on simultaneous occurrence of severe low flows and maximum daily loads. The ensuing risk of not meeting WQS is usually linked to the inverse of the return period of the design flow. Meanwhile, estimation of ELA from diffuse sources, and associated environmental risks, is not as well established since diffuse sources of pollution generally occur during important runoff events and strongly depend on land use and management practices. These wet weather loads may be allocated using continuous hydrological modelling. Simulation results can then in return be used to link ELA from point sources and diffuse sources to pollutant concentrations and, for a given time period (e.g., summer), to evaluate the probabilities (namely the risks) of exceeding WQS designating the water use.

The objective of this study was to present, using the integrated modelling system GIBSI (VILLENEUVE et al., 1998b; MAILHOT et al.,1997), this new approach to assess the risk associated with the establisment of ELA from point and diffuse sources. A case study on the Chaudière River watershed (Quebec, Canada) was defined and simulations were performed (scenarios A-E). The case study focussed on determining whether WQS defining the designated recreational use of water requiring direct and prolonged contact were attainable. Untreated municipal waste waters from a small town (St. Martin) and nonpoint source pollution were responsible for impairment of the studied river segment. Water contaminants considered were fecal coliforms (FC) and phosphorus (P).

Two base case scenarios, A and C, were simulated using four years (1982-1985) of meteorological data to illustrate the degraded bacteriological and aesthetic conditions of the river segment due to dry and wet weather sources, respectively. Dry weather sources were assumed to solely contribute to the bacteriological impairment. Meanwhile, both wet and dry weather sources were assumed to contribute to aesthetic impairment. Scenario B was defined to examine the impact of constructing a waste water treatment plant (WWTP) for the town of St. Martin on the bacteriological conditions. Scenarios D and E were elaborated to quantify the impact of both reducing diffuse and point source loads on aesthetic conditions, respectively. For scenario C, on average 31 kg P/ha were applied on cultivated land according to local fertilization calendars. Similarly, the fertilization rate for scenarios D and E was on average 13 kg P/ha. The design stream flow for ELA from point sources was a₃₀ Q₅ (30-day low flow, 5-year return period) and estimated at 6.05 m³ /s for the studied river segment. It is noteworthy to mention that for the 1982-1985 summers (June 21 through September 20), simulated daily stream flows exceeded the design flow 31, 62, 19 and 2 times and that 150, 48, 348, and 270 Mm³ of water flowed through the studied river segment, respectively. This means the 1983 meteorological series was three times as dry as that of 1982 despite the fact it was the latter that produced the₃₀ Q₅. The ELA from both point and diffuse sources for the river segment receiving the St. Martin's effluent and the studied river segment were 0.75*10¹² UFC and 11.21 kg P and 1.05* 10¹² UFC and 15.68 kg P, respectively.

For scenario A, simulation results showed that respectively 94% and 14% of the time, the bacteriological (WQS_FC < 200 UFC/100 ml) and aesthetic (WQS_P < 0,03 P mg/l) WQS were not met over the summer season of the four-year meteorological series. Similarly, for scenario C, the simulation results indicated that 32% of the time, the aesthetic WQS was not achieved. Dry weather sources, namely WWTP and wet weather sources, namely agricultural runoff, accounted for P loads. During the driest summer, dry weather sources accounted for 63% of the P loads. For the other summers, wet weather sources accounted for 71, 88, and 78% of total P loads. For scenario B, simulation results showed that the bacteriological WQS was met 100% of the time. Meanwhile, scenario D was designed to see whether a 27% reduction of the agricultural nonpoint source load could alone restore the aesthetic properties. The simulation results showed an improvement as the probability of exceedence dropped from 0.32 to 0.27. This probability was further lowered to 0.19 when St. Martin's waste waters were treated using aerated lagoons and dephosphotation. This corresponded to an additional 17% abatement of the dry weather sources loads (scenario E).

It is noteworthy that this study did not specifically attempt to specify the reasons why WQS were not met, whether it was due to low stream flows or large point or nonpoint sources loads although the integrated modelling system allows for this. Also, this study did not attempt to determine impacts of local management scenarios instead of a systematic reduction though GIBSI also allows for this. However, this study clearly illustrated how the proposed methodology, which is rooted in a risk assessment approach based on evaluation of the probability of exceeding WQS, is well suited to characterise ELA for various anthropogenic activities. Indeed, simulation results clearly demonstrated the benefits of assessing independently the impacts of ELA from point and diffuse sources on the attainability of a designated water use. It is believed that this type of ELA assessment approach will facilitate communication with stakeholders. Ideally, these exceeding probabilities should be evaluated using long meteorological series (e.g., 30 years).

Keywords:

• Integrated watershed management,
• point and nonpoint/diffuse sources of pollution,
• probability of exceeding water quality standards,
• integrated modelling system