Introduction

The increased use of fertilizers in agriculture, the need for fossil fuels for transportation and power generation, as well as increases in waste products from animal and human populations have caused major shifts in the way nitrogen cycles globally (Galloway et al. 2008). The changes in the N cycle are reflected in atmospheric reactive nitrogen (Nr) increases which cause the formation of smog and particulate matter which affect human health, and as eutrophication of terrestrial and aquatic ecosystems (Sutton et al. 2011).

One of the ways of better understanding the movement and potential effect of Nr on the planet is to trace its path using a budget approach. Large scale global, as well as national, regional and ecosystem N budgets have been constructed and reported in the literature (e.g. Gruber and Galloway 2008) which allow scientists and policy makers an appreciation of the importance of various sources and ultimate fates of N on the planet and to assess how best to reduce leakages into the environment. These large scale budgets in turn are dependent on in-depth sectoral studies which describe N pathways and pools which quantify interactions between agricultural, urban and industrial activities to the terrestrial, atmospheric and marine sub-components of the planet (Howarth et al. 2012).

An important component of Nr global budgets is its transfer from terrestrial systems to oceans. Howarth et al. (2006) described freshwater N export as a function of total terrestrial ecosystem processes which also included climate and deposition. They estimated that an average of 20–25 % of anthropogenic N deposition to landscapes was exported in rivers, with the remaining 75–80 % retained in plants and soils or denitrified and returned to the atmosphere as N2 or N2O gases.

The present-day flux of nitrogen from large rivers in North America and Europe to the North Atlantic was calculated by Boyer et al. (2006a), Boyer and Howarth (2008) to be 4.1 Tg year−1, which they estimated to be a quadrupling of pre-industrial values. These increases in riverine N exports have created important ramifications for estuaries causing eutrophication and anoxia in heavily affected areas, not only in North America and Europe, but also elsewhere on the planet (Diaz and Rosenberg 2008; Swaney et al. 2012).

In recent years, more work has been done to improve models designed to estimate terrestrial nitrogen (Boyer et al. 2006b; Kroeze et al. 2012) and carbon (Schlünz and Schneider 2000) losses to estuaries and to attempt to develop models which link these fluxes to landscape and ecosystem processes (Vörösmarty et al. 2000; Schlesinger et al. 2006; Wollheim et al. 2008). This is needed in order to better understand how to reduce the influence of society in modifying the nitrogen cycle which is causing increasing eutrophication in fresh waters and estuaries, as well as the carbon cycle which is the main driver of climate change. A consistent theme in these papers (e.g. Howarth et al. 1996; Green et al. 2004; Kroeze et al. 2012) is the need to obtain better high quality, well documented river N export data with which to test the models, especially from more isolated, less polluted regions. Data from these environments are important as they can provide an estimate of background or pre-pollution conditions which are necessary in order to understand what controls anthropogenic contributions (Howarth et al. 2006; Kroeze et al. 2012).

Until recently, the amount of nitrogen data available from Canadian rivers was relatively scarce due to the isolated nature of many of its northern rivers, as well as due to the difficulty of accessing databases scattered in government, hydropower companies and research papers. As the northern regions of Canada drain significant amounts of land and are relatively unpolluted by nitrogen, we therefore felt it important to assemble and interpret newly available data from this region, as well as the more affected southern regions in order better understand terrestrial-aquatic N cycling interactions, as well as its contribution to oceanic nitrogen budgets in the northern portion of North America.

Another important element exported by rivers from the landscape to estuaries is organic carbon (OC). Approximately 430 × 1012 g of OC is lost globally from terrestrial ecosystems to rivers and eventually estuaries (Schlünz and Schneider 2000). Though this transfer is small compared to direct atmospheric fluxes between land and the atmosphere, it is nevertheless important as it is modified by temperature and hydrology (i.e. climate) as well as anthropogenic activities such as land use. Much of the exported OC is remineralized to CO2 in both lakes and estuaries, with the remaining amounts precipitating in sediments which affects their geochemical and ecosystem functioning (Cole et al. 2007). The problems with data availability and quality which we faced with Nr were also reflected with OC. Clair et al. (1999) estimated a freshwater OC budget for Canada, but they were limited in their conclusions and accuracy by the sparseness of the data available to them at the time. For the earlier work, Clair et al. (1999) attempted to incorporate N in their study of river exports, but there was then not enough data available to produce even rudimentary estimates.

For this study, we assembled river N and OC concentration data which were collected mostly since 1990, from catchments draining the Canadian landscape and which empty into marine environments, with the goal of quantifying N and C exports from the landscape to estuaries. With the measured estimates available, we were then able to develop simple statistical models which included geographical, hydrological and land use characteristics which we used to predict export estimates for regions where no data existed. We then assumed that analysis of the model inputs would allow us to identify the catchment characteristics which were most important in explaining the controls of N and C export in Canadian rivers. We then combined the results from measured and estimated catchments to produce estimates of total freshwater N and C exports from the terrestrial to marine ecosystems for all of Canada.

The relationship between dissolved N and C in rivers should provide an indication of water pathway and soil ecosystem conditions which prevail in catchments. Higher N inputs generated from agricultural or urban runoff or from precipitation in more populated and agricultural areas should be reflected in higher N/C ratios of dissolved elements (Khalili et al. 2010). We therefore tested for the hypothesis that the ratios of exported N and C would be greater in rivers draining the catchments in the southern portions of the country where more N pollution occurs.

Study area

Canada occupies 10,216,590 km2 of the northern part of North America (Fig. 1). Because of its great size, geographical diversity and location, Canada has a large number of ecosystem types, from Arctic Tundra in the north, to wet, cool, temperate forests on both coasts, as well as semi-arid regions in the Prairie region located in the center of the country. The country is ideally situated to allow a better understanding of N exports from rivers and thus terrestrial N cycling, as it contains a wide range of atmospheric deposition amounts, population densities, as well as agricultural and industrial activities.

Fig. 1
figure 1

Location of Canada and its major catchments. Light colored areas are where water chemistry data existed and darker one where exports were estimated. Numbers for the measured catchments are related to Table 1 where names and other characteristics are listed

Full size image

The country’s large, northern watersheds receive low N deposition and have little human activity, while the large southern rivers drain the Great Lakes region and southern British Columbia catchments which are influenced by relatively high population densities and significant agricultural activity. All Canadian rivers studied in this work eventually discharge into either the Atlantic, Arctic or Pacific Oceans. Only one river, the Milk, located in southern Alberta eventually discharges into the Mississippi drainage. However, as this river originates in the US state of Montana and eventually returns there after traversing a small part of Canada, we decided not to include it in our Canadian estimate.

N deposition in Canada is highest in the south eastern part of the country, with values exceeding 1,000 kgN km−2year−1 in the southern Ontario and Québec region, much of which originates in the US Midwest and Eastern Seaboard (Vet et al. 2005). There is also a region of higher N deposition in the petroleum industry rich central Alberta (~300 kgN km−2year−1, southern portion of #2). The far northern portion of Canada receives <100 kgN km−2year−1 (Moran et al. 2008). There is intensive agriculture in southern Ontario and Québec, the southern portions of the Prairies Provinces (Manitoba, Saskatchewan and Alberta, catchment #30), and in the lower Fraser Valley near Vancouver on the west coast.

Methods

Data acquisition and export calculations

Overall, the catchments for which we had water chemistry values drained approximately 7,868,251 km2 or almost 76 % of the Canadian landmass. Aquatic chemistry monitoring has been done on a regular basis for most of the large southern rivers in Canada, with water sampling usually co-located at gauge sites allowing easy calculation of N and C exports. We identified 30 catchments or regions where water chemistry data were collected by Environment Canada and Provincial government partners, as well as university and industry collaborators over varying periods of time (Fig. 1; Table 1).

Table 1 Characteristics and locations of river catchments with measured values
Full size table

Due to the isolated nature of many of northern catchments and the reduced information demands for ecosystem management or protection, sampling in northern Canada is spotty and what little information exists can be difficult to access. An important source of water chemistry data we used originated with hydro power generation companies who maintain facilities in the Boreal Forest or have generated data in the course of fulfilling environmental impacts evaluations. Another source came from individual university researchers who conducted studies in the region. Most of the data we used were collected at river mouths above estuarine effects, though the Churchill and Nelson Rivers in Manitoba (Fig. 1; Map #’s 21 and 30) were sampled at power dam sites several hundred km from Hudson Bay, though these sites nevertheless included more than 90 % of the catchments’ surface areas (Fig. 1).

We only had rudimentary water chemistry and runoff estimates for the Arctic Archipelago (Fig. 1; #3) as the only water chemistry information available was collected in the ice-free season in 2007 and 2008, at 66 Arctic streams and rivers which drained catchments ranging in size from 0.6 to 7,434 km2 (Dr. J. Culp, Environment Canada, pers. Com.). As there are no gauged rivers in this large region, we used modeled runoff values from Spence and Burke (2008) for the Canadian far north which we combined with the available concentration data to provide estimates of N and C exports.

As the Arctic archipelago estimate only used summer water chemistry values collected in a large range of catchment sizes, it did not take into account the seasonality inherent in the snow and ice melt period in late spring and early summer (Thomas et al. 2011), so it is difficult to accurately evaluate annual export values or the potential error in the estimate. However, as will be seen below, the range of TN and TOC values measured was relatively narrow and we think that the estimate is useful as a first approximation to situate potential values for the far north.

Table 1 shows values from the catchments characteristics where exports were measured from water chemistry data, as well as the number of samples to generate N and C exports. Because of irregularity in sampling frequencies between studies, we used a basic method for estimating annual elemental exports. Where both water chemistry and discharge were available, mean monthly N and C exports were calculated for the period of record available by multiplying concentration with mean flow for each month of the record and monthly values were averaged over the record period. Monthly averaged data were then summed for each catchment to produce an annual export estimate. Sampling on the larger rivers (St Lawrence, Fraser, Columbia, Mackenzie) and the smaller ones in the southern part of the country was usually done monthly or at least seasonally as can be seen by the sampling frequency (Table 1), so that the export estimates we present usually take into account the seasonality of river flows.

There were a number of catchments for which data quality was not as high as hoped for. The Saint John River for example, had only 34 samples over an 11 year period collected near the river mouth, and runoff had to be estimated from regional runoff maps because the closest gauge site was 100 km upriver and only took in only 75 % of the catchment. Many of the rivers draining into Hudson Bay (Grande Baleine, Broadback, Rupert, Koksoak, Eastmain, Nottaway) were sampled monthly in the late 1970s or early 1980s for only 1 year. This frequency did not allow for any estimation of inter-annual variability though they were sampled over a complete seasonal cycle.

Sample analytical methods

For quality control reasons, we preferred using data collected since the year 2000, but where only earlier data was available, such as in northern Québec, these were nevertheless used to provide spatial coverage (see Table 1 for sampling dates). The total nitrogen (Nt) values found were almost always measured using persulphate/UV digestion or similar wet oxidation methods. Because of this consistency between the various sampling programs, the discussion will be focused on this parameter as we have the greatest confidence in its comparability throughout Canada.

NO3 − and NH4 + are usually measured in Canadian freshwaters, though analytical methods and detection limits have changed over time. Since 1990, most NO3 − and NH4 + were measured using ion chromatography, while pre-1990 samples were mostly done using colorimetric methods. In waters draining non-urban and/or non-agricultural catchments, NH4 + and NO3 − are usually very low and often at undetectable levels which we report as ‘no data’ (ND). Though we show inorganic N mean concentrations where they are available to give a relative idea of its importance, we only calculated Nt exports with the understanding that they include both inorganic and organic N.

Dissolved organic or total organic carbon (DOC, TOC) analytical approaches suffer from many of the same issues as for N. Samples analyzed before the mid-1990s were usually analyzed using wet oxidation methods which were later superseded by high temperature combustion methods which provided results approximately 25 % higher than wet oxidation (Koprivnjak et al. 1995). However, as we mostly kept to post-2000 data (except for northern Québec), this issue is not an important factor in our overall analysis.

Organic nitrogen and carbon analyses are usually reported either as dissolved from filtered samples or as total from unfiltered. Though water samples collected from urban or agricultural areas are usually filtered to remove particulate matter, samples from remote or undisturbed catchments rarely need to be, so that both total and dissolved values are commonly found in databases. We assume that sample collection and processing by individual programs or researchers was suited to local conditions and will assume that they provide equivalent results. We therefore report all organic N and C data as total nitrogen (Nt) and total organic carbon (TOC), even though not all samples were unfiltered.

Spatial analysis and regression models

Data sources for spatial analyses of population, agriculture, elevation came in a variety of formats including image, raster and vector types. All were converted into 1 km raster data and analyzed using ESRI Zonal statistics tool with a drainage basin vector layer. Centroid values were calculated for each drainage basin polygon using ETGeoWizards Tools in conjunction with ESRI ArcView. Data sources for GIS analysis are listed in “Appendix”.

Nitrogen deposition measurements have not been done for the whole of Canada, though eastern Canada has been well covered by the Canadian Air and Precipitation Monitoring Network (CAPMoN) (Vet et al. 2005). The only estimates of N deposition for the whole country are modeled using the “A Unified Regional Air-quality Modeling System” (AURAMS) (Moran et al. 2008). A shortcoming of this model, from the point of view of our study, is that it only predicts total N deposition to approximately 60°N, so that it is not possible to quantitatively compare N inputs to outputs for a large portion of the country. We therefore use the information from the AURAMS output in our discussion of results, but did not use it in statistical model development.

In order to fill catchment export calculation gaps for the 24 % of the country which wasn’t covered by sampling, we explored a number of linear and non-linear statistical models using catchment centroid latitude (a surrogate for anthropogenic influences) and longitude, runoff, percent area under agriculture, population and catchment sizes and mean slopes as input variables to develop predictive equations which could then be used to estimate export values for the catchments where no data existed.

Results and discussion

Nt exports

Nitrogen concentrations in all Canadian rivers we studied were relatively low (Table 2; Fig. 2) compared to more polluted water courses such as the Mississippi (1,500 to more than 3,100 kg N km2year−1 in the upper portions of the catchment, Goolsby & Battaglin 2001) or rivers draining the western European landmass (>1,000 kg N km2year−1; Billen et al. 2011). Moreover unlike from more heavily populated regions, organic nitrogen dominated Nt concentrations at all of our rivers (Table 2), a phenomenon also noted by Scott et al. (2007) for a number of rivers in the US. The highest mean Nt concentrations in the study were from the Nelson River (0.76 mg l−1) and the St Lawrence (0.65 mg l−1), followed by the Hayes (0.49 mg l−1), Harricanaw (0.49 mg l−1) and Saint John (0.41 mg l−1) Rivers (Table 2). The Nelson, St Lawrence and Saint John all drain agricultural regions, while the Hayes and Harricanaw have extensive wetlands within their drainages.

Table 2 Mean NH4–N, NO3–N and Nt concentrations, area normalized measured N and C exports and Nt and TOC exports from the measured catchments
Full size table
Fig. 2
figure 2

Measured and estimated Nt export rates from the Canadian landscape

Full size image

The largest catchment, the St Lawrence River (Fig. 2; Map # 33) exported 235.6 kt year−1 Nt, 3.5× as much as the next highest river, the Nelson which empties into Hudson Bay (Table 2; Fig. 2). As the St Lawrence catchment contains the largest population centers in Canada and a number of significant ones on the US side of the Great Lakes, as well as a great deal of agriculture in both countries, this finding is not surprising.

The mean area normalized export value for the St Lawrence (211 kg km−2year−1) was low compared to estimates from eastern US rivers (Boyer et al. 2006a; Goolsby & Battaglin 2001; Howarth et al. 2006) and Europe (Billen et al. 2011), especially considering the concentration of population, industry and agriculture in this watershed. Though the heavily populated and farmed portion of the St Lawrence catchment can receive >1,000 kg km−2year−1 N deposition plus agricultural and urban runoff, the undeveloped and forested western and northern portions only receive ~200 kg km−2year−1 N (Moran et al. 2008), much of which is either taken up by catchment plants or denitrified in wet soils and lake sediments.

Our measured St Lawrence export value is 1/3 of that estimated by Howarth et al. (1996) (660 kt year−1) and 70 % of that modeled by Boyer et al. (2006a) (340 kt year−1). Howarth et al. (1996) estimates were derived from a 1978 report which suffered from the inadequate data which was available at the time and we feel that our database is more trustworthy than the earlier unpublished work. The Boyer et al. (2006a) value is modeled, which we suspect underestimates denitrification in Great Lake sediments and littoral wetland areas. Moreover, biological uptake of N by algae with subsequent precipitation into sediments of the Great Lakes may be other reasons for the lower than expected area normalized exports from this watershed.

The reason for the large differences between our data and the other two studies are not likely due to improvements in sewage treatment as Holeton et al. (2011) show a slight increase in N discharges from Canadian sewage treatment plants in the last decade. Legislated changes in both Canada and the US have managed to reduce atmospheric reactive N emissions to the Great Lakes region (Zbieranowski and Aherne 2011), but those small reductions cannot be the source of the large discrepancies between our values and previous ones.

The second highest N exporting river in Canada is the Nelson (Fig. 2; Map # 30), which produces 66 kt year−1 though it’s N export rate of 58.4 kg km−2year−1 is 27 % of the St Lawrence’s (Fig. 2; Table 2) and much lower than the values reported for the upper Mississippi watersheds which abut on this drainage (Goolsby and Battaglin 2001). The Nelson drains a large portion of Canada’s Prairie Provinces which are the source of much of Canada’s grain growing and animal husbandry, as well as portions of North Dakota and Minnesota in the USA. Though N deposition in this catchment is low, (between 200–400 kg km−2year−1 N deposition) we feel that the intense agricultural activity in the southern half should nevertheless cause larger N normalized export rates.

The Mackenzie River (Fig. 2; #2) has the largest catchment of any Canadian river and exports the third largest N amount (50 kt year−1) from the landscape into the Arctic Ocean. Much of the oil and gas production and as well as associated petrochemical industries in Canada are found in the catchment’s upstream portion, locally causing Nr deposition amounts of up to 340 kg N km−2year−1 (Moran et al. 2008). The deposition levels in the south are not translated into high N exports at its Arctic Ocean estuary however, as the catchment shows the lowest area normalized export value (29 kg km−2year−1) of any other drainage area measured, even lower than the Arctic Archipelago (39.4 kg km2 year−1) which we found surprising. Our export estimate is lower than the 60 kt year−1 reported by Holmes et al. (2012), though the difference is not great considering the difficulties and inaccuracies involved in such calculations. Our study used a similar sampling frequency to theirs (~4.2X year−1), though we had data from a longer sampling period (10 vs 4 years).

One consistent factor which seems to be affecting export rates from the St Lawrence, Nelson and Mackenzie catchments seems to be the role of denitrification. Though we have not quantified N loadings into these catchments, it is evident that a densely populated and agriculturally intense catchment such as the Great Lakes should be exporting higher amounts than are measured at its outlet. The heavily agricultural Nelson catchment should also be exporting more than 58.4 kg km2 year−1. We would also expect that the Mackenzie catchment which has major petrochemical industries in its southern portion would generate more N per unit area than the high Arctic region.

We suspect denitrification has a major influence reducing N exports from these catchments for a number of reasons which have been discussed by Boyer et al. (2006b); Seitzinger et al. (2006); Behrendt & Opitz (2000). These studies suggests that denitrification is enhanced with increased residence time of waters in lakes and rivers due to a number of factors related to hydrology and watercourse physical characteristics. Residence time of the Great Lakes, the largest freshwater system in the world allows much greater opportunities for contact with sediments and anaerobic conditions which can lead to denitrification. Much of the Nelson River flow passes through Lake Winnipeg. This lake is 24,514 km2 in area with a shallow mean depth of ~12 m and has been identified as major source of N loss in the catchment due to probable denitrification and loss to sediments (Scott et al. 2011).

The Mackenzie River also contains large lakes in its drainage, and Emmerton et al. (2007) show that over 11,200 km2 of the catchment is composed of flooded vegetation surfaces in the spring flood period which should provide good conditions for denitrification processes. Moreover, the location of the industrial areas are near the headwaters of the Mackenzie, so that waters receiving the greater N deposition have a higher residence time in the catchment, leading to greater denitrification opportunities. The physical and hydrological characteristics of these large catchments therefore clearly point to the potential of denitrification being an important player in N dynamics and show a need for further research.

Our data that the Yukon River exports 52.6 kg N km−2year−1 in its Canadian portion, which is lower than the value of 81 kgN km−2year−1 reported by Holmes et al. (2012) for the whole catchment. A number of reasons may be responsible for the discrepancy. First, the Canadian upstream portion of the Yukon catchment occupies 38 % of the total catchment and receives less precipitation and generates lower runoff than the downstream USA portion. Moreover, our sampling frequency was semi-monthly for eight years, while the Holmes et al. (2012) information was collected seasonally for only 3 years. So the difference in estimates may be due a combination of difference in data quality and climatic conditions.

The Saint John River catchment in the Maritime Provinces of eastern Canada (Map #45) had the highest unit area Nt export of any Canadian river studied at 300 kg km−2year−1. This river receives >200 kgN km−2year−1 in deposition (Moran et al. 2008) and is almost 10 % agricultural land (potato cultivation in large part) which contributes N runoff to the catchment. This export value compares well with rivers from the nearby US state of Maine where Howarth et al. (2006) reported values between 320–400 KgN km−2year−1, while Cronan (2012) showed normalized export values near 200 kgN km−2year−1. These two USA estimates, though somewhat different, at least provide us with some certitude that our data are not unreasonable.

The next highest normalized export values are from the Maritime Provinces (Map #49, N deposition ~90 kgN km−2year−1) and the Island of Newfoundland (NF) (Map # 70, N deposition ~50 kgN km−2year−1) both regions emptying into the Atlantic Ocean. The Maritime catchments are small, near higher sources of N emissions in the US and Canada and contain agriculture which explain the higher Nt export. Newfoundland catchments on the other hand, receive low N deposition amounts, being further downwind of eastern North American emission sources and support little agriculture. These high export values are affected by the contribution of organic matter from wetlands which are controlled by terrain conditions and high runoff amounts (Clair et al. 1994).

The rivers of Labrador which drain into the Atlantic Ocean (Map #48) and those from northern Quebec which drain into southern Hudson Bay and the Gulf of St Lawrence have N export rates as high as for the St Lawrence despite having much lower atmospheric deposition and low population or agricultural pressures. However, they are also relatively flat, short and contain high wetland amounts which then produce organic N which is exported. The very high N export into Hudson Bay by the Harricanaw River (Fig. 2; #68) is also most likely explained by the high prevalence of peatlands in its catchment (Barnett et al. 2011).

On the west coast of Canada, the Fraser River which empties into the Pacific Ocean (Map #19) and the Columbia which drains into Washington State of the US (Map #36) are influenced by agricultural activities, but because of the large portion of their catchments which are in wilderness areas with low N deposition (up to 150 kgN km−2year−1; Table 1), Nt export rates are low at 72 and 97 kgN km−2year−1 (Table 2). A number of other coastal catchments export high N amounts mostly because of high runoff in these coastal regions. Not surprisingly, these Nt export values are between 50 to 100 % lower than those reported by Schaefer et al. (2009) for the US portion of the west coast which receive greater atmospheric and land use inputs.

The Arctic Archipelago stream Nt concentrations were low (inorganic N was undetectable using standard EC methods), with a mean value from 66 samples of 0.19 mg l−1 Nt (0.86 max, 0.03 min, 0.13 median). As the average annual runoff for the Arctic Archipelago was estimated at 0.3 m (Spence and Burke 2008), we calculated normalized export for this region to be approximately 39 kg km−2year−1. As our estimate does not take into account the seasonality inherent in this region’s snow and ice melt period in late spring and early summer (Thomas et al. 2011), it is difficult to accurately evaluate annual export values, but this estimate can be used as a first approximation to situate potential N export values for the far north of Canada.

Our Arctic value was 20 % of the model prediction of Green et al. (2004) who suggested that current Nt exports from polar regions should be in the range of 200 kgN km−2year−1, a value which seems to be mostly dependent on theoretical N fixation estimates. Our data show that this assumption needs to be adjusted to better reflect the reality of N fixation by the impoverished lichen fauna in this region.

TOC exports

Measured TOC export rates ranged from 8,223 kg km−2year−1 in the Nottaway River (Map # 42; Table 2), to a low of 1,812 kg km−2year−1 from the Yukon River (Map # 1), a dry, mountainous region which drains into Alaska and eventually into the Pacific Ocean (Fig. 3). These values all fall within the range of observations assembled by Alvarez-Cobelas et al. (2012) from 550 catchments located across the globe.

Fig. 3
figure 3

Measured and estimated TOC export rates from the Canadian landscape

Full size image

Mean annual TOC concentrations were highest from four rivers draining into Hudson Bay (Table 2). As mentioned above, this region is characterized by very shallow catchment slopes which generate extensive wetlands. The high concentrations lead to the highest area normalized TOC export rates, especially in the southern portion of Hudson Bay. Interestingly, the more northern and western HB catchments (Nelson, Churchill and Hayes) which produce lower runoff, export some of the lowest TOC amounts.

The relatively high TOC exports from catchments located in the eastern and southern shores of Hudson Bay and were compared to TOC export rates estimated by Rosa et al. (2012) as part of a study on regional cation weathering potential. For the seven catchments our study had in common with theirs (Grande Baleine, Broadback, La Grande, Rupert, Koksoak, Harricanaw, Nelson), differences in export values ranged between 60 and 380 %. However, their values were only estimated from a spring and autumn sampling for 1 year, so that their sampling regime was not adequate for dealing with these highly seasonal systems.

Our TOC export estimate for the Yukon is similar to that of Holmes et al. (2012) (1,812 vs. 1,771 kg km−2year−1) and Raymond et al. (2007; 1,770 kg km−2year−1) but somewhat higher than Striegl et al. (2007, 1,388 kg km−2year−1). Our Mackenzie value (1,384.5 kg km−2year−1) is very similar to that of Raymond et al. (2007, 1,400 kg km−2year−1) but considerably higher than the Holmes et al. (2012) value of 820 kg km−2year−1. All of these studies report on data collected seasonally, though ours used data collected over a 10 year period compared to two and four years for theirs. We suspect that our data set was able to take in a larger range of conditions which included more of the natural variability which could be expected over this large area. Striegl et al. (2007) also estimated an export rate of 1,623 which was close to ours (1,695 kg km−2year−1) for the Porcupine River which drains into the Yukon.

Our study’s mean TOC concentration for the St Lawrence River was 3.7 mg l−1 for data collected between 2000 and 2007 which was close to that of Pocklington & Tan (1987) who estimated a value of 3.8 mg l−1 DOC with a 3–14 % addition for particulate C from samples collected at our site from 1981 to 1985. Though our sampling frequency for the Saint John River catchment was not optimal, our TOC export estimate of 5,540 kg km−2year−1 was near to that from the nearby Penobscot River in Maine (Cronan 2012) which has a similar distribution of forestry-agricultural usage (5,830 kg km−2year−1).

We applied the same approach as for Nt, to estimate TOC exports from the Arctic Archipelago region using measured stream chemistry data and estimated runoff. The mean TOC value measured in the Arctic Archipelago from 66 stream samples was 1.8 mg l−1 (5.8 max, 0.3 min, 1.7 median). Multiplying the mean value with the estimated runoff of 0.3 m for this region provided an annual export estimate of 495 kg km−2year−1 which we then applied to all catchments north of the 63rd latitude.

Estimating Nt and TOC exports for all of Canada

In order to estimate exports for the 24 % of the country for which we had no data we used a simplified, statistical approach which only needed rudimentary data available from geographical and hydrological databases. We considered using the more sophisticated “net anthropogenic nitrogen inputs” (NANI) approach described by Howarth et al. (2012), however as all the unsampled catchments were in the northern portion of the Canadian mainland, there was little agricultural or other human activity and there was no atmospheric deposition data available, so we felt that the approach was not suitable for the conditions we dealt with.

Based in part on the observation that Nt exports were roughly highest in the south and in wetter regions of the coasts (Fig. 4 mid and bottom), we developed a statistical model using geographical parameters (catchment centroid latitudes and longitudes, surface area, mean slope) as well as hydrology and population and agricultural intensities. We attempted linear and non-linear regression approaches but found that the best approach was the use of a simple multilinear regression which explained 58 % of the variability:

$$ {\text{N}}_{\text{t}} \exp \left( {{\text{kg}}\, {\text{km}}^{ - 2} {\text{y}}^{ - 1} } \right) = 259 - 3.9\,\left( {\text{lat}} \right) + 2.5\,\left( {\text{pop}} \right) + 0.02\,\left( {\% \,{\text{agric}}} \right) - 5.7\,\left( {\text{slope}} \right) + 133\,\left( {\text{ro}} \right)\,r^{2} = 0.58,\,n = 30,p < 0.001 $$
(1)

where latitude (lat) is the catchment centroid value, population (pop) is in average people km−2, %agric is the percent of the catchment under agriculture, slope is the catchment median slope in m km−1, and runoff (ro) is in meters. Latitude in this case is not indicative of any particular process or group of processes, but is a surrogate for a combination of anthropogenic activities causing high Nr emissions and nutrient runoff. Runoff is an indication of water flow through the system and in smaller systems has a positive influence on N export which has also been noted by Lewis et al. (1999).

Fig. 4
figure 4

Relationship between Nt export and size (top), latitude (mid) and runoff (bottom) of study catchments

Full size image

The negative sign on the catchment slope factor suggests flatter catchments have reduced exports more than could be expected, which confirms the concept that denitrification is greater in low slope conditions than in steeper regions. The Maritimes, Newfoundland and Harricanaw regions, are outliers to this trend due to high wetland contributions of organic N. The interactions between competing catchment and hydrological factors fit in well with the synthesis produced by Seitzinger et al. (2006) in describing the factors controlling N dynamics in aquatic ecosystems.

When we assessed the TOC data from our measured catchments using the same approach as for Nt, we found that our best export prediction came from an equation which only took into account runoff and catchment slope:

$$ {\text{TOC Export}}\,\left( {{\text{kg}} {\text{km}}^{ - 2} {\text{y}}^{ - 1} } \right) = 1706 - 507.6\,\left( {\text{slope}} \right) + 4865\,\left( {\text{ro}} \right)r^{2} = 0.42,\,n = 30,p < 0.001 $$
(2)

where slope is in m km−1, and runoff (ro) is in m year−1. The regression which could only explain 42 % of the variability could not be significantly improved by the addition of other variables or by log-normalizing the data, and is probably the best outcome that could be achieved, as Alvarez-Cobelas et al. (2012) concluded that there could not be a single approach for developing OC export models due to the wide range of local conditions which can affect the export outcome. Our result is similar to that of Lauerwald et al. (2012) who also included land cover and wetland factors in their models, though our correlation coefficients were worse than theirs (0.42 this study, 0.55 and 0.60 for their regression from small and large catchments), most likely because we had fewer catchments in our study (30 vs their 246 small and 207 large catchments) as well as fewer sites with very low TOC exports.

Clair et al. (1994) have shown that flat landscapes correspond well with the occurrence of wetlands and thus high TOC exports in Canada. Under relatively high precipitation and suitable temperature conditions for the formation and subsequent senescence of mosses and other wetland species, more water flow will cause greater export of decayed organic plant matter into receiving water courses. On the other hand, catchments with low runoff, such as the Yukon, Nelson, Mackenzie, Fraser and Columbia (Table 1) had considerably lower TOC exports, due to more oxidation of soil plant matter under drier conditions.

We combined the calculated export values with the measured ones to produce estimates of N and C riverine exports from the whole of the Canadian landmass (Table 3). Our combined measured and estimated results (Table 3) suggest that the Canadian landscape exports 884.1 kton Nt per year, most of it in organic form, with 65 % of this amount being captured by our data sources and the remaining 35 % being estimated.

Table 3 Summary of measured and estimated carbon and nitrogen exports from the major Canadian regions
Full size table

Compared to the N budgets done by Howarth et al. (1996) and Boyer et al. (2006a), our total value (Table 3) which includes exports from the Arctic Ocean, Hudson Bay, the St Lawrence River, as well as the Atlantic coast of Canada (801 kton year−1) is lower than their estimate of 960 kton year−1 for the east coast of Canada. The most important difference comes from our Gulf of St Lawrence-Atlantic coast contribution which is lower than their estimate. This is discussed above and the evidence seems to support the fact that the main cause for the change is improved data quality especially with the St Lawrence River exports.

Our TOC export value for all of Canada (18,210 kton year−1) was compared to Clair et al. (1999) estimate of 14,250 kton year−1 though the earlier value did not include the Arctic Archipelago. The difference between the two (3,960 kton year−1) is due to the addition of the archipelago data, so that we feel confident that the newer value is relatively accurate.

Nitrogen/carbon relationships

We hypothesized that the N/C ratio of the exported organic matter would decrease with increases in latitude as atmospheric deposition and runoff from agriculture and urban influences would also decrease as shown by Khalili et al. (2010). Our data supported this hypothesis (Fig. 5a) and pointed out another interesting phenomenon. The highest N/C ratios were from the Columbia (0.13), the Nelson (0.1) and the St Lawrence (0.09) with the next highest being Maritime Province catchments (0.06), the Fraser (0.055) and the St John (0.054), all of which have significant agricultural activity. The catchments with the lowest N/C values on the other hand, are all heavily influenced by wetlands. We plotted % agricultural area against N/C ratios (Fig. 5b) and estimated a relationship described by a linear equation:

Fig. 5
figure 5

Relationship between aquatic N/C ratios and catchment centroid latitude (top) and % area under agriculture (bottom)

Full size image
$$ {\text{N}}/{\text{C ratio}} = 0.0373 + 0.0014\,\left( {\% \,{\text{agric}}.} \right)\,r^{2} = 0.33,\,p = 0.001 $$
(3)

The relationship only predicts 33 % of the variability in the data, but is nevertheless highly significant. The higher N/C ratio in agricultural catchments is not surprising and is due to leakage from fertilizer use or manure storage piles which are more likely to be leached into catchment streams. Atmospheric deposition seems less important to the N/C ratio as it will be more evenly distributed across the landscape and thus more likely to be incorporated into plant and soil matter. The poor predictability of this equation is not surprising as a number of factors will influence the N/C ratio at the point of sampling, including the presence of wetlands in the catchment (Gergel et al. 1999), as well as the location of the agricultural land in relation to the water sampling site.

Conclusions

In this study, we produced river N and C export data from pristine as well as from polluted catchments located in the northern portion of North America. We have improved N export estimates from the St Lawrence River, the second largest N exporting river in North America as well as from a number of other rivers in southern Canada and have identified the Nelson River, a previously ignored catchment, as an important contributor of reactive N to Hudson Bay and thus the eastern Arctic Ocean.

We found that catchments with shallow slopes and high runoff (>0.5 m year−1), such as are found in Newfoundland and eastern Hudson Bay export higher than expected levels of N due to the influence of wetland organic matter, though they are composed of organic matter with low N/C ratios. High runoff and steep slopes in low N deposition areas such as in the Pacific Coast rainforest also export larger than expected N amounts due to high water volumes through the systems.

However, shallow catchments where wetlands are not as prevalent, such as the St Lawrence, or where runoff is lower than 0.5 m year−1 such as the Mackenzie, show lower than expected N exports most likely due to in-catchment denitrification. Slow moving rivers such as the Mackenzie underscore the importance of denitrification in reducing catchment N exports, as values from this site which experiences higher N deposition levels in its southern portion, are even lower than exports from high Arctic catchments which are far from anthropogenic sources.

Organic carbon exports are in large part determined by the presence or absence of wetlands and runoff levels, as flatter catchments with high runoff will produce the greatest TOC export. Though TOC fluxes on a local level are obviously affected by land-use and water control factors, our data show that on a large scale and in relatively undisturbed regions, catchment characteristics and the hydrological cycle are the main determinants of organic carbon as was also found by Alvarez-Cobelas et al. (2012). Finally our data show that unsurprisingly, N/C ratios are strongly determined by the importance of agriculture in catchments.