Floods and water quality in Canada: A review of the interactions with urbanization, agriculture and forestry

Abstract

Water quality remains a major issue in Canada. This paper reviews recent research on the impacts of urbanization, agriculture and forestry on water quality in Canada. Specific water quality issues such as mining, sewage treatment and waste treatment are not included in this paper. For each land use, a brief summary of the dominant processes linking runoff to water quality is provided and recent findings are summarized. With respect to urbanized watersheds, the relatively large proportion of impervious areas, lower vegetation cover and the presence of high-density drainage systems alter surface water routing and timing of peak flows. High concentrations of heavy metals are considered to be the most important water quality problem in urban runoff, but nutrients, pathogens, concentration of pharmaceuticals and water temperature also often contribute. In watersheds dominated by agricultural activities, overland flow is an important vector of pollutants, but subsurface flow such as macropore and tile-drain flows also play a role in the alteration of water quality during or after high runoff events. Nutrients, pesticides, pathogens and sediments remain important topics of research in agricultural watersheds, and the modelling effort has significantly increased in the last few decades. Beneficial management practices (BMPs) are being tested and applied at a local scale, mostly on experimental watersheds. Forestry-related activities also affect water quality. In forested watersheds, studies have been ongoing for many decades, but have decreased in intensity in the last 15 years. Sediment delivery and water temperature can be strongly affected in watersheds with significant clear-cut logging and riparian buffer strips and sylviculture remain the main mitigation BMPs. There is a need for an increase in the monitoring effort for most water quality variables in Canada. The authors recommend that flow-dependent monitoring frameworks should be further developed and implemented in the future.

La qualité de l’eau demeure une préoccupation majeure au Canada. Cet article présente une synthèse des travaux de recherche récents sur les impacts de l’urbanisation, de l’agriculture et de la foresterie sur la qualité de l’eau au Canada. Les problématiques de la qualité de l’eau liées d’autres activités spécifiques comme les exploitations minières, le traitement des eaux usées et le traitement des matières résiduelles ne sont pas incluses dans cette synthèse. Pour chacune des trois activités, un bref résumé des processus prédominants qui décrivent les liens entre les écoulements et la qualité de l’eau est fourni et les conclusions des travaux récents sont décrites. En ce qui a trait aux bassins versants urbanisés, la grande proportion de surface imperméable, la faible densité de végétation et la présence d’importants réseaux de drainage denses modulent le ruissellement de surface et modifient certaines caractéristiques de la crue telle que l’occurrence du débit maximum. Des concentrations élevées de métaux lourds sont considérés comme étant le problème le plus criant sur les bassins versants urbains, bien que les nutriments, les pathogènes et les concentrations élevées de produits pharmaceutiques, de même que la thermie des cours d’eau urbains, soient aussi des causes de la détérioration du milieu aquatique. Dans les bassins versants dominés par l’agriculture, le ruissellement de surface demeure un important vecteur de pollution, mais le ruissellement hypodermique, qui peut être influencé par les macropores du sol et les drains agricoles, joue aussi un rôle dans la dégradation de la qualité de l’eau pendant et suite à des crues. Les nutriments, pesticides, pathogènes et sédiments demeurent des sujets importants de recherche dans les bassins versants agricoles, pour lesquels le développement de modèles s’est significativement accéléré au cours des dernières décennies. Les pratiques de gestions bénéfiques (PGB) sont présentement testées et appliquées à l’échelle locale, principalement sur des bassins versants expérimentaux. Les activités forestières peuvent aussi avoir des impacts sur la qualité de l’eau. Dans les bassins versants à dominance forestière, la recherche se poursuit depuis plusieurs décennies, mais l’effort de recherche a augmenté au cours des 15 dernières années. Les dynamiques sédimentaire et thermique de l’eau en rivière peuvent être fortement influencées négativement par les coupes à blanc et positivement par la présence d’une zone tampon végétalisée le long des cours d’eau. La sylviculture est la principale pratique de gestion bénéfique utilisée. Les suivis de qualité de l’eau au Canada devraient augmenter au cours des prochaines années. Les auteurs recommandent la mise en œuvre de protocoles de suivi des variables de qualité en fonction du débit, pour une meilleure gestion de la qualité de l’eau.

Introduction

Water quality remains at the forefront of societal priorities, not only in the context of drinking water, but also for irrigation and agricultural use, as well as a key component of aquatic ecosystem health. In Canada, extreme weather-related water quality events have generated significant scientific and media interest. Indeed, the joint occurrence of drought-induced severe low flows and poor water quality in urban settings has drawn interest, given human and ecosystem vulnerability to the adverse effects of poor water quality (Beck 2005). Similarly, significant runoff events and water quality issues – although the processes that determine the links between runoff and water quality are complex – have also been reported in the scientific literature and mass media (e.g. McCullough et al. 2012; Michalak et al. 2013).

Water quality can decrease when mobilized sources of contaminants are transported away from their original locations by overland flow and subsurface flow, and discharged in surface and ground waters. The transfer can be characterized by the way chemical, microbiological and sediment outputs entered the aquatic environment; that is, through a point (e.g. effluent release through drains, ditches or pipes) or diffusively (e.g. non-point entries or pathways such as overland flow and subsurface flow or sediment resuspension). Mobilization occurs when fine particles (i.e. sediments, soil aggregates) are either detached from the soil or washed off an impervious surface, or when chemicals or pathogens are dissolved through water interactions with soil and impervious areas. Transfer can be altered by modifying the hydrological connectivity and sediment connectivity – that is, the transport and discharge of dissolved pollutants and sediment-associated pollutants, respectively. In urban watersheds, water quality tends to be governed by variability in both dry and wet conditions, while in agricultural and forested watersheds it is mostly governed by wet conditions producing significant overland flow (e.g. Rousseau et al. 2002).

This paper presents a review of the links between precipitation-runoff, land use and water quality, with an emphasis on Canadian watersheds with the following dominant land uses: urban, agricultural and forestry. For each dominant land use type, a brief summary of recent research on the interaction between high flows and water quality completed in Canada or applicable to the Canadian context is provided. The emphasis is on key water quality variable categories: water temperature, nutrients, major ions and sediment. When available, information on metals and pathogens is included.

Urbanization

In urbanized systems, the hydrological cycle is highly conditioned by the significant proportion of impervious areas, and by the installation of drainage systems. Networks of pipes for potable water supply and wastewater collection co-exist in parallel systems. These changes in water routing, combined with the generation of various pollutants from urban activities and land surfaces, govern the water quality of receiving watercourses, but also alter their physical characteristics (e.g. amplified erosion, increase in flood amplitude and decrease in base flow, etc.; Stephens et al. 2002). Two types of sewer networks can be implemented in urban areas to drain stormwater runoff, namely separate and combined sewers. During rainfall events, when the capacity of the conveying pipes and/or the capacity of the treatment plant are exceeded in combined sewers, overflows to watercourses occur. The impacts of separate and combined sewers on receiving watercourses are very different. Even if separate sewers are now the accepted practice in most Canadian provinces (e.g. Alberta Environment 2001; Ontario Ministry of the Environment 1997), their superiority over combined sewers in terms of water quality impacts is still controversial and depends mainly on the considered pollutant (De Toffol et al. 2007).

Flooding in urbanized areas can occur either in the pipe drainage network (e.g. sewer backups), or in the stream/river flood plain, or both, and can be caused by a variety of rainfall events. Indeed, short and intense rainfall events can lead to high runoff on small and highly impervious watersheds, while long rainfall events can affect larger and less impervious watersheds. High water levels in rivers due to upstream conditions (e.g. snowmelt) can also alter the conveyance capacity of sewer pipes and produce flooding in some urban areas even during minor rainfall events.

Characterization of urban runoff water quality

The water quality of urban runoff has been characterized in many studies since the pioneering Nationwide Urban Runoff Program (NURP) research project conducted by the United States Environmental Protection Agency (US EPA 1983). Heavy metals (especially copper, lead and zinc) have been recognized as the predominant pollutants. It is also now well known that high levels of coliform bacteria are present in urban runoff, indicating the likely presence of pathogenic micro-organisms and viruses. Makepeace et al. (1995) produced an extensive review of 140 papers published from 1968 to 1995 containing water quality analyses of stormwater. Like many others, they showed that the concentration ranges can vary by many orders of magnitude for some contaminants. From their review, the list of contaminants for which the upper concentration limit is 10 times the water quality guidelines (WQGs) for drinking water or protection of aquatic life includes: total solids, total suspended solids, some inorganic chemicals (aluminum, beryllium, cadmium, chloride, chromium, copper, iron, lead, mercury, silver and zinc), low dissolved oxygen, some organic chemicals, fecal coliforms, fecal Streptococci and Enterococci. Emerging organic compounds (EOCs) have been identified as important contaminants in urban rivers over the last 15 years. Of these contaminants, pharmaceuticals and personal care products (PPCPs), including endocrine-disrupting chemicals (EDCs), have received much attention. These compounds are found mainly in wastewater. They may enter urban stormwater runoff directly from rainwater (Pal et al. 2014), from re-use of reclaimed domestic wastewaters for irrigation (Ellis 2006) or, most often, from combined sewer overflows (CSOs; Marsalek 2008). Knowledge of organic contaminants in stormwater is rather limited compared to wastewater (Jenberg et al. 2013), as only a few Canadian researchers have studied this issue. Marsalek (2008) provided an excellent summary of sources and pathways of PPCPs in urban waters and an exhaustive overview of Canadian research in this area until 2008. Since then, Sauvé et al. (2012) took water samples from streams, brooks and storm sewer outfall pipes across the Island of Montréal and analyzed them for carbamazepine, a drug that is used for various psychiatric treatments. All samples contained various concentrations of this tracer, indicating a contamination of stormwater by wastewater.

Concentrations of contaminants in runoff depend mainly on the watershed land use and on sewer type (separate or combined). In watersheds with dense road networks, chlorides from applied road salts can alter river water quality all year long, as they accumulate in groundwater (Eyles and Meriano 2010; Perera et al. 2010). For 28 subwatersheds located in the Greater Vancouver area, Brydon et al. (2009) found significant correlations between catchment imperviousness/road density, and metals in river bed sediments.

Stormwater runoff was characterized by different field studies in Canada. Marsalek et al. (1999) studied the toxicity of stormwater runoff from 14 urban sites in Ontario. They found that only 40% of the samples did not cause any toxic response, and that samples from highway runoff were the most toxic. By combining stormwater quality characterization with rainfall measurements, McLeod et al. (2006) realized that urban runoff for the entire City of Saskatoon contributed more total suspended solids and total Kjeldahl nitrogen load to the South Saskatchewan River than the effluents from the wastewater treatment plant and chemical industry. Sauvé et al. (2012) sampled 85 storm sewer outfalls on the Island of Montréal; all samples contained various concentrations of caffeine and fecal coliforms, indicating sanitary contamination of all 85 storm sewers. Combined sewer overflows (CSOs) were also characterized by different authors in Canada. Ridal et al. (2010) demonstrated that total mercury and methylmercury concentrations were significantly higher in CSO discharges than in the St. Lawrence River, upstream of these discharges (near Cornwall). In two CSO outfalls located in the Montreal area, Madoux-Humery et al. (2013) measured E. coli, total suspended solids, caffeine and carbamazepine, an antiepileptic medication. They concluded that the concentrations of these variables in CSO vary temporally and spatially, and that snowmelt leads to high concentrations and long-duration events.

Impact of urban runoff on natural watercourses

The impact of urban runoff and/or urbanization on natural watercourses has been the subject of several studies over the last decade. Finkebine et al. (2000) conducted field surveys in streams near Vancouver, located in watersheds which were urbanized approximately 20 years earlier. They observed that construction activities may lead to substantial increases in fine sediments discharged by stormwater runoff into receiving watercourses. Channel erosion is also enhanced in urbanizing streams, due to the increased frequency of bankfull discharge. These modifications can negatively affect aquatic habitats, as well as modifications in food sources, water quality, flow regime and biotic interactions (Marsalek 2006). Lafont et al. (2007) reported that stormwater outfalls over 30 sites in Ontario impacted sediment physico-chemical conditions of watercourses, but not to levels that could be harmful to the majority of benthic invertebrates.

Urbanization also affects water temperature. First, the surface runoff component of an urban flood hydrograph is often more important than that of an equivalent forested watershed. In the latter, the cooler interflow and baseflow components will decrease the thermal load (in the summer), compared to equivalent surface runoff in cities. Impervious areas often have low albedo, which may lead to a relative increase in heat transfer from the soil to the water. The information on such impacts is scant in Canada, but some studies have been conducted in the USA. For instance, Wardynski et al. (2013) have shown that the thermal load associated with urban runoff can be attenuated using several stormwater control measures such as permeable pavement and internal storage. In Canada, van Buren et al. (2000) developed a model to predict thermal enhancement of stormwater runoff by paved surfaces. The authors observed that paved surfaces increased runoff temperature, while runoff in sewer pipes contributed to loss of heat. Olding et al. (2004) monitored streams in three subwatersheds undergoing urbanization over a 5-year period in Ontario; they did not observe any increasing trend in water temperature during this period despite the increase in imperviousness due to watershed urbanization.

Beneficial management practices

Recognition of the poor quality of stormwater runoff and its impacts on urban streams has led to the implementation of mitigation measures and the development of a new paradigm for urban stormwater management. They are described in many documents, ranging from the pioneer Stormwater planning guidebook for British Columbia (Stephens et al. 2002) to the more recent Guide de gestion des eaux pluviales in Quebec (Ministère du Développement durable, de l’Environnement et des Parcs [MDDEP] and Ministère des Affaires municipales, des régions et de l’Occupation du territoire [MAMROT] 2011). Stated briefly, urban stormwater management should tend to reproduce, as much as possible, the natural water cycle, while minimizing the impact of urbanization on urban stream water quality. These objectives can be attained by the construction of infrastructures that enable retention, evaporation, infiltration and improvement of the water quality of stormwater runoff. The efficiency of these infrastructures, often referred to as green infrastructures or best management practices (BMPs), has been studied by many researchers in Canada, including Farrell and Scheckenberger (2003) and Brydon et al. (2006), for wetlands; Brydon et al. (2009), Mungasavalli and Viraraghavan (2006) and Olding et al. (2004) for detention ponds; Drake et al. (2013) for permeable pavement; and Abida and Sabourin (2006) for perforated pipes. Results vary, but generally conclude in improvements in water quality, and reductions in peak flow.

Water quality modeling

The conception of stormwater facilities and the evaluation of their long-term impacts on water quality require the use of stormwater quality models. Models for runoff quality can be classified into four main types, namely: (1) deterministic or probabilistic models reproducing processes of contaminant build-up and washoff; (2) regression models; (3) export-coefficient models; and (4) event mean-concentration models. Yuan et al. (2001) developed a model that simulates build-up and washoff of suspended solids (SS) in urban areas. Assuming that transport of metal species is linked to SS transport, they used the model to predict lead and zinc loads from a highway catchment. Behera et al. (2006), and Chen and Adams (2006a, 2007) developed a novel approach to assess stormwater pollution from urban catchments within a probabilistic framework. Their methodology combines probability density functions of rainfall event characteristics, a runoff coefficient based rainfall-runoff transformation, and pollutant build-up and washoff representations to derive analytical models that can be used to compute event or annual washoff loads, along with their probability distributions. They applied their model to compute total suspended solids (TSS), total solids (TS), total Kjeldahl nitrogen, total phosphorus, chemical oxygen demand (COD) and aluminum, copper, iron and zinc loads from three urban catchments located in the Toronto area (Behera et al. 2006; Chen and Adams 2006a, 2007), and extended the approach to assess the fraction of pollutants removed from storage facilities (Chen and Adams 2006b). A regression model, the second type of model, was applied by Pfeifer and Bennett (2011) to compute phosphorus (P) concentrations and fluxes in seven urban streams, on the island of Montreal, as a function of watershed land use and imperviousness fraction, which were retained as the most effective variables to asses river concentrations. A model of the third type, namely an export coefficient model, was applied by Winter and Duthie (2000) to assess the influence of land use on P loads to a stream in Southern Ontario. The model attributes an export coefficient (in kg/ha/year) to each land use type and computes the loads from each land use. Using this approach, Winter and Duthie (2000) found that runoff from urban areas provided the greatest contribution of P loads to the studied stream. The last type of model, event mean concentration models, computes pollutant loads by multiplying simulated water runoff by mean concentrations. An example of such a model is given in Thériault and Duchesne (2012), who assessed the annual loads of fecal coliforms from areas drained by different sewer types to an urban river located in Quebec City. Water quality models can also be extended to stormwater detention facilities, such as in Kuzin and Adams (2010), Vallet et al. (2014) and Zhang and Gou (2014).

Climate change

In Canada, the increase in intensity (and/or frequency) of heavy rainfall events due to climate change (CC) has been demonstrated by Kharin et al. (2007), Mailhot et al. (2007), Mladjic et al. (2011), Cheng et al. (2012), Mailhot et al. (2012) and Burn and Taleghani (2013). Modifications in climate will likely lead to more frequent hydraulic dysfunctions of urban drainage systems (backups, flooding, etc.). Examples of sources discussing this in Canada include He et al. (2006), Mailhot et al. (2008b) and Vidil (2012). In light of these changes, design criteria for urban drainage systems should be revised (Mailhot and Duchesne 2010). Although the impacts of CC on the urban water cycle have been the subject of many recent studies, its impacts on urban runoff water quality are still poorly understood. This could be due to the fact that these kinds of studies usually rely on simulation models, and that these models rarely take into account all of the processes that impact urban water quality (Langeveld et al. 2013). Moreover, most of the build-up/washoff models use the inter-event time as an explanatory variable, while little is known about the future evolution of this variable in the context of CC. Some researchers, however, have demonstrated that the volume (Kuchenbecker et al. 2010, in Germany; cited in Bendel et al. 2013), frequency (Bendel et al. 2013, in Germany; Fortier and Mailhot 2014, May and October in Canada) or mean annual duration (Fortier and Mailhot 2014, in Canada) of CSOs should increase in the future climate. Logically, these increases will cause water quality to deteriorate in urban rivers – impacts that could be more severe as a result of increased water temperature. To overcome the current limitations of water quality simulation models and climate projections, impacts of CC on urban surface water quality in Canada could be assessed based on the method proposed by Langeveld et al. (2013). These authors analyzed intensive monitoring data for extreme meteorological conditions that resemble expected climate changes (high quantities of rainfall after long dry periods combined with high temperatures).

Agriculture

During runoff events, agricultural pollutants can be transferred to surface waters by overland and subsurface flows (Haygarth and Jarvis 2002) as well as drainage ditches, and resulting concentrations and exported loads ultimately depend on: rainfall intensity, antecedent soil conditions (moisture, nutrient levels), topographic factors (hillslope plan and curvature profiles [Noël et al. 2014], and slope length), geology, soil erodibility, land cover, tillage practices and water management. These loads can be mitigated through BMPs such as nutrient management (MacKay and Hewitt 2010), integrated pest management (Maredia et al. 2003), tillage practices (Hilliard et al. 2002), grassed waterways, and edge-of-field and riparian vegetated filters (e.g. Gumiere et al. 2013; Nigel et al. 2013).

Despite regulatory and voluntary implementations of BMPs, impaired surface and ground water quality conditions persist in many watersheds throughout Canada; consequently, agriculture has been singled out as one of the major contributors to their degradation (Eilers et al. 2010; Council of Canadian Academies 2013).

Overland flow and subsurface flow can cause significant adverse off-farm socio-economic and environmental impacts through the local depletion of oxygen due to organic waste and the release of adsorbed and dissolved nutrients, pesticides, pathogens and toxins in surface and ground waters (Eilers et al. 2010). These may include loss of recreational potential due to algae blooms, increased cost of treating water supplies for municipalities and industries, loss of biological diversity induced by eutrophication, or destruction of fish spawning habitat caused by sedimentation, to name a few. Modelling work and recent analyses of observed data have shown that for large agricultural watersheds, the bulk of sediment-associated pollutants enters the stream network through the spring period and a few runoff events during the rest of the year (e.g. Lake Winnipeg Stewardship Board 2006; Quilbé et al. 2006; Mailhot et al. 2008a; Salvano et al. 2009; McCullough et al. 2012). This behaviour has also been observed for micro-watersheds where most of the sediment and nutrient discharge occur during a few events over the course of the snow-free period (Ratté-Fortin et al. 2015). These results highlight the need to implement flow-dependent monitoring systems throughout the agricultural landscape (e.g. Madrid and Zayas 2007; Birgand et al. 2011; Xing et al. 2012). This framework has long been recognized as a way forward, but because of maintenance and labour costs, governments have not been able to deploy the required resources throughout their territories. Research teams have been able to implement such systems on study watersheds, and subsequent analysis and statistical and deterministic modelling have been used to assess the impact of various land covers and BMPs on water quality (Agriculture and Agri-Food Canada 2007, 2011).

Certainly, the bulk of the literature has focused on nutrients (namely P and nitrogen), pesticides, pathogens and sediments, but has focused to a lesser extent on temperature. The literature consulted does not establish a direct link between pollutant discharges and runoff, but rather reports on loads and concentrations over the course of seasons and years.

Nutrients (phosphorus and nitrogen)

As illustrated by Leinweber et al. (2002), the soil chemistry of P is complex and not fully understood, and the current worldwide problems associated with this key nutrient (especially build up and saturation of agricultural soils) are paramount, leading to the eutrophication of surface waters and the ensuing impairment of recreational activities. Once dissolved P and colloidal particle-associated P are mobilized through solubilization (mostly desorption) and detachment processes (mostly erosion), respectively, they are transferred via surface and subsurface pathways. Earlier studies focussed on transfer through overland flow because it was thought that P was mostly adsorbed to eroded soil particles and to a much lesser extent in a dissolved form. However, recent studies have shown that all forms of P could be transferred via subsurface flow (preferential flows, tile drainage flow), making it even more challenging to design adequate BMPs (Leinweber et al. 2002). Eilers et al. (2010) reported that overland and subsurface flows are responsible for the risk of P losses to surface water in eastern Canada, while overland flow represents the dominant risk of water contamination in the Prairies (see also van Bochove et al. 2012). It is now widely acknowledged that surface waters in Canada – that is, water courses of agricultural watersheds as well as large lakes (Lake Winnipeg, MA; Lake Simcoe, ON; Lake Diefenbaker, SK) and small lakes in Québec – have high P concentrations, and, in most instances, exceed the WQG to prevent eutrophication (0.03 mg P/L in MDDEP 2008; (Council of Canadian Academies 2013). For example, the modelling work of Rousseau et al. (2013) on a 718-km² Quebec watershed (36% of the surface area being farmed intensively) corroborates these findings. The mean annual simulated total in-stream P loss near the watershed outlet was 2.7 kg/ha/year (per ha of agricultural land), representing 80% of the simulated overland total P yield (i.e. delivery ratio of 80%). Meanwhile, for the simulated period of 1970–2006 (1 May to 31 October of each year), the probability of exceeding the WQG was 0.91.

Hatch et al. (2002) report that mobile nitrogen (N;(i.e. nitrates and organic N adsorbed to eroded soil particles) from farmland can reach the aquatic environment through surface pathways and subsurface pathways. Nitrate–nitrite (NO₃^––NO₂^–) transfer in subsurface pathways can occur throughout the year even under cold weather and snow cover conditions. However, wet conditions predominantly govern nitrate–nitrite transfer throughout the spring and following N application. In general, concentrations of nitrate in surface water do not exceed drinking water WQG (10 mg NO₃^–-N/L in MDDEP 2008), which is not the case for groundwater where it can build up and become harmful to human health. In Canada, shallow and deep wells in British Columbia, Ontario, Quebec and the Maritime provinces have been contaminated (Council of Canadian Academies 2013). According to Eilers et al. (2010), the amount of N likely to be transferred is on the order of the residual N – that is, the amount of N remaining in the top 60 cm of soil at the end of the cropping season. The aforementioned watershed modelling study of Rousseau et al. (2013) corroborates these findings during the period between 1 May and 31 October of the years 1970 to 2006. The simulated mean annual overland loss of nitrate was 16 kg/ha/year (agricultural ha), which represented over 60% of total N loss and 60% of the annual N application rates in excess of crop demand. The simulation results showed that the nitrate WQG was not exceeded near the watershed outlet, which was not the case for nitrite, where the WQG for the protection of aquatic life was exceeded by 47% over the course of the study period (0.02 mg NO₂^–-N /L).

Pesticides

Pesticides come in a wide range of active ingredients with distinct physico-chemical properties, and accordingly are applied at various rates. They usually find their way out of the root zone or crop foliage (i.e. the intended treated zone and surface) through various pathways, namely spray drift, volatilization, chemical and biological degradation, subsurface flow, overland flow, and uptake by plants and animals (Gevao and Jones 2002). Gevao and Jones (2002) mentioned that the overall loss of pesticides through each of the aforementioned pathways is essentially governed by the physico-chemical properties of the soil and the chemical properties of the active ingredients, as well as environmental and management factors such as timing and intensity of rainfall following spraying, and application rates and formulation type. Dissipation patterns in the environment can be anticipated from an envelope of decay curves. At one end, there are pesticides that have short half-lives and decay rapidly – that is, highly volatile, water-soluble or easily degradable compounds (e.g. soil fumigants). They are usually transferred via overland and subsurface flows. At the other end, there are non-volatile, water-insoluble and recalcitrant pesticides – extremely sorbed to soil particles and organic matter. They are likely to be mobilized by erosion and transferred in overland flow (e.g. chlorinated pesticides). Most pesticides used in agriculture fall within these two behaviors. Once in the aquatic environment, pesticides can have a range of harmful effects on living organisms and human health, from lethal to chronic effects. In Canada, WQGs exist, but not for the majority of pesticides. Furthermore, due to advances in pesticide science and regulations, more effective active ingredients requiring low application rates are now being used (e.g. glyphosate), which make them hard to detect. The Council of Canadian Academies (2013) highlighted that herbicides (94%) constitute the major group of pesticide use, exceeding by far fungicides (4%) and insecticides (2%). For example, 80% of pesticides used in British Columbia, the Prairies, Ontario and Quebec are herbicides, while 50% are fungicides in the Maritimes. Pesticides have been found in the surface waters of most Canadian agricultural watersheds (e.g. Xing et al. 2012) and, in some instances, associated to fish kills in the Maritimes (e.g. Gormley et al. 2005), but according to Eilers et al. (2010), the risk of groundwater contamination is much less, with 99% of farmland classified as very low risk in 2006. In a watershed modelling study, Rousseau et al. (2012) showed that in-stream loads of pesticides through river segments of the Yamaska River watershed divided by the mass of pesticide applied on upstream land were mostly below 1%, with only two river segments ranging between 1 and 3%. The study also illustrated that concentration values at the outlet of six Canadian watersheds for eight pesticides following implementation of BMPs were usually less than ecological thresholds of good condition, when available. Upstream river segments were at greater risk of having concentrations above a given ecological threshold, because of limited stream flows and overland loads of pesticides.

Pathogens

As reported by the Council of Canadian Academies (2013) and Eilers et al. (2010), agricultural watersheds across the country are very likely to have pathogens in surface water and shallow groundwater. Transfer of pathogens into surface or groundwater can be diverse and summarized as follows (Jones et al. 2002): overland and subsurface flows from agricultural fields fertilized with animal manure, contamination of recreational water by wildlife, and point sources such as wastewater from dairy parlours or slaughter houses. The common microorganisms are Campylobacter, Listeria monocytogeneses, E. coli 0157:H7, Giardia and Cryptosporidium. The number of pathogens in the environment usually decline after release from the host, but under certain conditions (temperature, oxygen concentration), they can survive for a long time. The most well-known and documented case of contaminated agricultural overland flow in Canada – Walkerton, Ontario, in 2000 – was responsible for the deaths of seven people and over 2000 serious cases of illness. E. coli 057:H7 and Campylobacter jenuni transferred from a livestock farm by overland flow and deficient training of the drinking water treatment plant operators were found to be the main causes of this tragedy (O'Connor 2002; Holme 2003). Considering the risk to human health and the poor understanding of the fate and occurrence of pathogenic species in surface and ground waters, the Council of Canadian Academies (2013) recommends that more active research should be dedicated to this problem.

Sediments

Intensive agriculture has been recognized as one of the major causes of accelerated cropland erosion and, in instances where livestock have access to watercourses, of damaged stream banks leading to bank erosion (Harrod and Theurer 2002). Benoy et al. (2012) stated that agriculture can be considered to affect 40% of watercourses in temperate regions of Canada. Overland flow, stream flow and sometimes subsurface flow (transport of colloidal soil particles through tile drains) represent the key pathways. The transfer of sediments in surface water is driven by rainfall intensity, soil moisture condition, soil texture, soil structure (e.g. soil compaction reducing infiltration), applications of animal slurry and manure, slope length and spring melt (e.g. Rousseau et al. 2013; Mailhot et al. 2008a). Hould-Gosselin et al. (2014) introduced an application of MHYDAS-Erosion (Gumiere et al. 2010), an event-based model, to a small agricultural watershed under temperate climate conditions (Quebec, Canada). The results revealed a bimodal behavior of the watershed. During high-intensity rainfall events, most of the sediments that reached the outlet originated from cropland, whereas during low-intensity events, most of the sediments originated from the drainage network. The sedimentological connectivity can be broken through the implementation of BMPs such as reduced tillage, contour cropping, cover crops and settling ponds, to name a few. Arable soils are particularly vulnerable to accelerated erosion before seeding and after harvesting. Since erosion selectively removes fine soil particles, discharge of sediment-associated nutrients, pesticides and microbes represents a threat for the aquatic environment (see previous sub-sections). In Canada, WQGs exist for suspended sediments. For example, in Quebec, guideline for the protection of aquatic life allows an average of 25 mg/L above the natural background concentration in a given stretch of a river (MDDEP 2008). In the aforementioned modelling study of Rousseau et al. (2013) of the Beaurivage watershed, the probability of exceeding the WQG for suspended sediments (estimated at 70 mg/L) was 31% during the period between 1 May and 31 October of the years 1970 to 2006. Recent research has focused on establishing physical and ecological guidelines for bedded sediments (i.e. deposited sediments) in agricultural watersheds in New Brunswick and Prince Edward Island in the Atlantic Maritimes of Canada (Benoy et al. 2012). However, a lot of research will need to be carried out to develop and validate deposited sediment thresholds for streams in landscapes with watersheds dominated by agriculture.

Temperature

Streams and rivers flowing through agricultural land are generally more exposed to incoming solar radiation and, accordingly, runoff will also be characterized by higher temperature than in forested watersheds. Riparian buffer strips are the main mitigation measure. There is a dearth of documented studies relating agriculture, hydrology and water temperature. Few studies have focused on water temperature modelling in this context. For instance, as part of the National Agri-Environmental Health Analysis and Reporting Program (NAHARP), Daigle et al. (2010) developed a water temperature model for monthly maxima in the Okanagan drainage basin. The approach consisted of interpolating monthly maxima in a multivariate physiographic space defined in part by land use information. This information included the percentage of the drainage area used for agriculture. The model was able to estimate monthly maximum temperatures for the May to July season with a root mean square error (RMSE) between 0.8 and 2.1°C.

Climate change

While there is no doubt that significant land cover changes affect runoff and low flows (e.g. Quilbé et al. 2008; Savary et al. 2009), there is high level of uncertainties on how water quality could be affected by changing climate conditions in Canada. Yet there are few studies that have primarily focused on this issue. One such study is that of Gagnon et al. (2014). In their investigation, they assessed the impact of climate change on pesticide transport by surface runoff in southern Québec for the 1981–2040 period. The crop enemies investigated were: for corn, weeds; and for apple orchard, three insect pests and two diseases. A total of 23 climate simulations, 19 sites, and 11 active ingredients were considered. The relationship between climate and phenology was accounted for by bioclimatic models within the Computer Centre for Agricultural Pest Forecasting (CIPRA) software. Exported loads of pesticides were evaluated at the field scale using the Pesticide Root Zone Model (PRZM), simulating both hydrology and chemical transport. A stochastic model was developed to account for PRZM parameter uncertainty. For the 2011–2040 period, application dates would be advanced from 3 to 7 days on average with respect to the 1981–2010 period. However, there is no significant climate change impact on maximum daily rainfall during the application window, mainly due to the high variability of extreme rainfall events. These findings suggest that for the studied sites and the crop enemies considered, climate change should not significantly affect pesticides losses throughout the 1981–2040 period. Meanwhile, in the USA, Bosch et al. (2014a) used the SWAT model along with global climate model outputs to assess the combined impact of agricultural BMPs and climate change on runoff, nutrients (N, P) and sediment loads from four agricultural Lake Erie watersheds. Results showed that contaminant loads are expected to increase, but the study does not explicitly report on the links between loads and ensuing water quality conditions with respect to WQGs for N and P. The study also illustrated that BMPs will play key roles in the future, but they are unlikely to be as effective under the studied climate projections. Nevertheless, their adoption rates are expected to potentially mitigate anticipated increases in sediment and nutrient loads, but maintaining loads at current levels will not reduce algal blooms in Lake Erie. That being said, little is known about the impacts of changing climate conditions on the quality of water in agricultural watersheds. Thus, it is expected that significant research activities in the years to come will provide insights on the impacts to water quality and the adaptation of agricultural BMPs with respect to new technologies and emerging markets (e.g. Mehdi et al. 2013).

Forestry

The present section focuses on deforestation as the major driver of alteration of water quality. Impacts of industries such as pulp and paper plants and sawmills are excluded from this discussion.

A number of water quality variables are influenced by the changes in hydrology that can occur following timber harvesting. They include changes in the temperature and sediment regimes of the river, as well as the magnitude and timing of nutrient and major ions export and loads.

Temperature

In contrast with the agricultural context, there have been a significant number of studies related to water temperature and forestry in Canada. Such studies were more abundant in the 1980s and 1990s than they are today, as a number of experimental forested watersheds were established prior to or during those decades, including some on the Canadian West Coast (Carnation Creek: Hartman et al. 1996) and in the Atlantic Region (Catamaran Brook: Cunjak et al. 1990). In Carnation Creek, the thermal impact of streamside logging was studied. Water temperature rises were important during the winter, prior to the flood season (Scrivener and Andersen 1984). In Catamaran Brook, logging was less intensive than in Carnation Creek and buffer strips mitigated the thermal impact of deforestation. In a sub-basin of Catamaran Brook where 23.4% of the area was logged, there were significant changes in the relation between precipitation and peak flow after logging. The water temperature modelling performed on Catamaran Brook showed that during the flood period, interflow water temperature variability can lead to increased water temperatures on the order of 1°C for scenarios in which a relatively high percentage (>50%) of watershed was logged (St-Hilaire et al. 2000). In Québec, Prévost et al. (1999) reported important increases in weekly maximum temperatures (7°C) in drained forested peatlands. More recently, Guenther et al. (2014) investigated stream and bed temperature variability in a watershed located 60 km east of Vancouver. The paired-catchment study revealed that post-harvest temperature increases were on the order of 1.6 to 3°C during the low-flow summer period. Gomi et al. (2006) investigated daily maximum temperature increases in four partially afforested sub-watersheds in the same region. They measured water temperature increases between 2 and 8°C. They also found that the recovery rates toward pre-harvest temperatures were more rapid in late summer than during the flood period. However, according to these authors, this difference in recovery rates was not associated with the different hydrological conditions, but rather with variation in leaf area in the riparian vegetation, modulating the incoming radiation.

Sediments

Removal of vegetation and soil disturbance associated with deforestation are known to influence the sediment outputs of watersheds. One of the main factors that determine the amplitude of the change in sediment delivery is the ratio of deforested to watershed area. For instance, Jolicoeur et al. (2007) found that while the suspended sediment delivery downstream of cut blocks located near a first-order tributary of Catamaran Brook (NB) was very high (23% of area logged, maximum recorded suspended sediment concentration of 404 mg/L), such was not the case for the third-order Catamaran brook (less than 10% logged area), where no significant change was found in SS concentrations before and after logging. Terrain slope is one of the major dampening factors in sediment delivery, as noted by Swanson et al. (1986). They did not find any significant changes in sediment loads in a Marmott Creek sub-watershed, after 23% of its area was logged. They suggested that given its relative flat topography, much of the Boreal plain would not be prone to increases in sediment loads after logging. Such is not the case in the BC interior, where MacDonald et al. (2003) reported an important increase in suspended sediment yield (as high as 74%, compared to pre-harvesting) during the second postharvest year (1998) in a sub-boreal watershed. While low slope can mitigate potential increases in sediment loads after logging, high road density and a large number of stream crossings typically exacerbate the impact of logging on sediment loads. High sediment loads can affect salmonid habitats by smoltering redds with fine sediments. This can lead to habitat loss and/or costly stream restoration programs (Ogston et al. 2014). A study by Spillios (1999) monitored fine sediments deposited downstream of crossings in 15 first- to third-order streams in central Alberta. His results indicated more fine sediments downstream of narrow stream crossings (<2.5m wide) than upstream. In a study conducted on paired watersheds in southeastern BC, Jordan (2006) concluded that erosion from forest roads was important, while the same process was negligible on cut blocks.

Nutrients

In deforested areas, the disturbed soil matrix is often more conducive to quickflow, which reduces the residence time of water in the soil. The pathways and chemical interactions that determine their concentrations will also be influenced directly or indirectly by changes in soil humidity acidity and temperature, as well as the soil structure, all of which can be altered by logging activity (NCASI 2009).

The underlying processes differ for each type of nutrients. For instance, P will be greatly affected by soil properties and erosion processes, while N mobility depends in part on the microbial activity that is associated with nitrification (Chanasyk et al. 2003). Logging practices can also influence ion concentrations and delivery to the river system. For instance, Duchesne and Houle (2006) studied a 56-ha boreal shield watershed in Québec for 7 years. They noted that whole-tree harvesting can deplete the soil base cation pool by as much as 66% over time. This percentage decreases to 47–57% when stem-only harvesting is practiced. Changes in soil quality (and consequently in water quality) often occur over a relatively long period after logging, as observed in a study by Pennock and Van Kessel (1997) in central Saskatchewan. They found that over a period of 6 to 20 years after a clear-cut, soil organic carbon had decreased by 24%, soil N by 27% and soluble organic P by 15%.

Tremblay et al. (2009) conducted a control-impact watershed study at the Ruisseau Des-Eaux-Volées experimental watershed, 100 km north of Québec City. They compared subwatersheds that were submitted to clear-cuts equivalent to 50% of the drainage area with a pristine control watershed. Twenty-metre buffer strips were left untouched along perennial streams. They monitored a number of water quality variables, including NO₃^–, which was found to increase very significantly (> 600% increase compared to pre-logging concentrations) in the receiving water after logging. Other increases in concentrations were observed for K⁺ (300%), Mg⁺⁺ (18%) and Fe_total (71%).

The sylvicultural industry has been using herbicides for a number of years to control unwanted vegetation. High runoff associated with floods contributes in mobilizing herbicides. The rate of mobilization of these contaminants will also depend on the method of application (e.g. direct aerial application vs. drift from application in adjacent areas; NCASI 2009).

Beneficial management practices and modeling

As with agriculture, BMPs associated with logging focus in large part on riparian zone management. Regulations and guidelines have been promoted and promulgated by a number of Canadian studies in different jurisdictions (e.g. Luke et al. 2007; Manitoba Conservation and Water Stewardship 2008). Many other beneficial management strategies exist, such as limiting clear-cut areas, minimizing the number of gullies created by machinery, etc. Neary et al. (2009) suggest the following criteria for selecting and implementing BMPs in forested watersheds: (1) avoid soil compaction and complete harvesting to bare ground; (2) maintain buffer strips between bare ground and water courses; (3) do not apply herbicides near water courses; (4) limit harvesting and soil disturbance in high-slope areas; (5) limit road construction and river crossings to a minimum; (6) minimize disturbance of soil hydraulic conductivity.

Both statistical and deterministic models have been used to simulate or predict water quality in Canadian forested catchments. In recent years, an increase in the water temperature modelling effort has been noted in western (e.g. Moore et al. 2013) and eastern (e.g. Caissie et al. 2005; Chenard and Caissie 2008) watersheds. Computational methods to estimate extreme suspended sediment concentrations and loads have been developed or adapted to the Canadian context (e.g. Tramblay et al. 2007; Araujo et al. 2012). Nutrient models have also been used in the context of exploring BMPs for small forested watersheds (e.g. Nour et al. 2006).

Climate change

Climate change is potentially a major stressor of forest ecosystems in Canada and elsewhere. Boisvert-Marsh et al. (2014) have documented shifts in tree species distribution in eastern North America, with an average northward shift of distribution of 3 km for 11 northern species. With these changes in populations and territories, water yields in forested catchments are also expected to change, with both conifer and deciduous forest catchments more susceptible to changes in the ratio of potential evapotranspiration to precipitation than mixed forest watersheds (Creed et al. 2014). This is likely to impact water quality, although there are fewer studies modelling water quality than flow for future scenarios. Delpha and Rodriguez (2014) modelled turbidity and fecal coliforms in 24 eastern Canadian watersheds, 16 of which were mostly forested. Their linear mixed effects model used cumulative precipitation as a proxy for water yield. A marginal increase in turbidity (2–4%) was predicted by most scenarios. Fecal coliforms were predicted to increase by approximately 2% in the summer, while they would decrease by 1 to 2% in the winter. Daigle et al. (2014) have simulated future (2050 horizon) water temperature in the Ouelle, Ste-Marguerite and Little Southwest Miramichi rivers using multi-layer perceptrons fed by precipitation and air temperatures from five different climate change scenarios. Their simulations show increases in warm water spells (e.g. water temperature >24 and 28°C, which are stressful and lethal limits for salmonid fish).

Mixed land use

Many watersheds in southern Canada are characterized by heterogeneous or mixed land use. Significant urban, agricultural and forested areas are often found in the same watershed. For instance, Neary et al. (2009) state that many North American urban areas rely on forested sub-watersheds as potable water sources. Few studies have attempted to investigate the within-basin variability of water quality as a function of land use variability in Canada. In one such study, Rousseau et al. (2002) developed, using an integrated modeling system and a mixed land use watershed, a risk-based environmental load allocations (ELAs) framework that links wet (nonpoint/diffuse) and dry (point) weather sources to probabilities of exceeding WQGs. The paper focused on determining whether WQGs defining recreational uses of water were achievable through two management scenarios: (1) treatment of a municipal waste water effluent, and (2) implementation of various nutrient management plans. Dry weather sources were assumed to solely contribute to bacteriological impairment of water. Meanwhile, both wet and dry weather sources were assumed to contribute to aesthetic impairment. Simulation results showed that treating the municipal effluent while reducing the agricultural diffuse loads by 27% allowed, on average, for attainment of the bacteriological WQG 100% of the summer time while lowering the probability of exceeding the aesthetic WQG from 0.32 to 0.19. These results clearly illustrated the benefits of independently assessing the impacts of point and nonpoint sources on the attainability of a designated water use. The proposed framework should prove to be useful to communicate with stakeholders and start up a pollution trading discussion in impaired, mixed land use watersheds. In another study, this time in Kentucky (USA), Coulter et al. (2004) based their investigation on the premise that the major land use types (urban, agricultural, forested) differ in their contribution to non-point source pollution. The East Hickman Creek sub-watershed comprised three regions: urban (in which 99% of the drainage area is urban), agricultural and mixed (in which 43% of the drainage area is agricultural, 57% is urban). They showed that nitrate fluxes were not significantly different between sites. Total P concentrations changed by season for the mixed region, but it is unclear how much of this change is attributable to flow variability.

Given that watershed degradation and water quality deterioration are often associated with multiple stressors that persist over time (Rousseau et al. 2005; Fluharty 2011), more research on the complex interactions of multiple land use, namely at the urban–rural interface, and their impacts on floods and water quality is required.

Conclusions

This paper presented a review of recent research on the impacts of urbanization, agriculture and forestry on water quality. The contribution of the Canadian research community to this field of research has been important but could benefit from greater integration, given that many watersheds are characterized by a mosaic of different land uses. BMPs are often associated with flow retention, especially in urban settings. Using riparian zones as filters is still one of the main BMPs for load attenuation in forested and agricultural watersheds. Canadian climate poses a particular challenge to water quality managers. The aforementioned mitigation measures, and those that were not listed (e.g. infiltration structures), can have varying seasonal performances. For example, Nigel et al. (2013) and Ratté-Fortin et al. (2015) have shown that, in early spring, riparian vegetative filters are not a very efficient BMP in agricultural watersheds unless their widths are designed to account for concentrated flow conditions (e.g. Gumiere et al. 2013, Rousseau et al. 2014). Urbanization will likely continue in Canada over the next decades. This means that the pressure on urban rivers will also likely increase. Climate change will also prove to be a major challenge. As the changes in temperature and precipitation regimes occur, the likelihood of extreme flood events will also evolve. The physico-chemical processes linking high-flow events to altered water quality will therefore bring about changes in the latter, and adaptation strategies will need to be developed. In order to be prepared for these future challenges, water quality monitoring efforts should be enhanced. Given the importance of flow as the key driver and pathway of pollutants, flow-dependent sampling should be considered and deployed in vulnerable urban, agricultural and forested watersheds.