Downstream Modification of Mercury in Diverse River Systems Underscores the Role of Local Conditions in Fish Bioaccumulation

Fish consumption advisories for mercury (Hg) are common in rivers, highlighting connections between landscape sources of Hg and downstream fluvial ecosystems. Though watershed conditions can influence concentrations of Hg in smaller streams, how Hg changes downstream through larger rivers and how these changes associate with Hg concentrations in fish is not well understood. Here we present a continuum of concentrations and yields of total mercury (THg) and methylmercury (MeHg) from small tributary systems draining diverse western Canadian headwater landscapes through to major transboundary rivers. We associate these downstream patterns with THg concentrations in tissues of resident fish in major rivers. Mean concentrations and yields of unfiltered THg from over 80 monitored tributaries and major rivers were highly variable in space ranging from 0.28 to 120 ng L−1 and 0.39 to 170 µg ha−1 d−1, respectively. Using spatial data and a hierarchical cluster analysis, we identified three broad categories of tributary catchment conditions. Linear mixed modeling analysis with water quality variables revealed significantly lower THg concentrations in tributaries draining cordillera-foothills (geometric mean: 0.76 ng L−1) regions relative to those draining forested (1.5 ng L−1) and agriculturalized landscapes (2.4 ng L−1), suggesting that sources and mobility of THg in soils and surface waters were different between landscapes. However, these concentration differences were not sustained downstream in major rivers as local sources and sinks of THg in river channels smoothed differences between landscape types. Extensive fish tissue monitoring in major rivers and ANCOVA analysis found that site-specific, river water THg and MeHg concentrations and local catchment conditions were stronger associates of THg concentrations in fish than broader trends in rivers within and across landscape classes. Consequently, site-specific, targeted monitoring of THg and MeHg concentrations in water and fish is a preferred study design when assessing regional-level patterns in fish tissue concentrations.

ABSTRACT Fish consumption advisories for mercury (Hg) are common in rivers, highlighting connections between landscape sources of Hg and downstream fluvial ecosystems. Though watershed conditions can influence concentrations of Hg in smaller streams, how Hg changes downstream through larger rivers and how these changes associate with Hg concentrations in fish is not well understood. Here we present a continuum of concentrations and yields of total mercury (THg) and methylmercury (MeHg) from small tributary systems draining diverse western Canadian headwater landscapes through to major transboundary rivers. We associate these downstream patterns with THg concentrations in tissues of resident fish in major rivers. Mean concentrations and yields of unfiltered THg from over 80 monitored tributaries and major rivers were highly variable in space ranging from 0.28 to 120 ng L -1 and 0.39 to 170 lg ha -1 d -1 , respectively. Using spatial data and a hierarchical cluster analysis, we identified three broad categories of tributary catchment conditions. Linear mixed modeling analysis with water quality variables revealed significantly lower THg concentrations in tributaries draining cordillera-foothills (geometric mean: 0.76 ng L -1 ) regions relative to those draining forested (1.5 ng L -1 ) and agriculturalized landscapes (2.4 ng L -1 ), suggesting that sources and mobility of THg in soils and surface waters were different between landscapes. However, these concentration differences were not sustained downstream in major rivers as local sources and sinks of THg in river channels smoothed differences between landscape types. Extensive fish tissue monitoring in major rivers and ANCOVA analysis found that site-specific, river water THg and MeHg concentrations and local catchment conditions were stronger associates of THg concentrations in fish than broader trends in rivers within and across landscape classes. Consequently, site-specific, targeted monitoring of THg and MeHg concentrations in water and fish is a preferred study design when assessing regional-level patterns in fish tissue concentrations.

INTRODUCTION
Rivers serve as primary conduits of erodible and soluble terrestrial material to downstream lacustrine and marine environments. During riverine transit, this material may be deposited, re-suspended, biochemically transformed, supplemented or diluted by external sources, or simply delivered downstream (Schlesinger and others 2006). Mercury (Hg) is subjected to these fluvial processes and may be transformed to its toxic organic form methylmercury (MeHg) in aqueous environments and biomagnify up food chains (Liu and others 2011). Despite focus on Hg delivery, cycling, and toxicity within lacustrine, wetland, and estuarine receiving environments (Watras and others 1995;Branfireun and others 2001;Jiang and others 2018;Rudd and others 2018), Hg concentrations in fish collected from lotic ecosystems can be higher than in those ecosystems (Kamman and others 2005). In some cases, concentrations are high enough to warrant public health advice to limit consumption of fish tissues (Kinghorn and others 2007;Bookman and others 2008; Government of Alberta 2019; U.S. Environmental Protection Agency 2020). Because Hg inputs to and concentrations in aquatic systems are influenced by upstream sources and landscape characteristics (Hurley and others 1995), but also affect MeHg concentrations in fish tissues (Harris and others 2007;Orihel and others 2007), these river-based advisories highlight biogeochemical connections along this continuum.
Mercury delivered to river systems originates from sources external and internal to upstream watersheds and may mobilize downstream. External Hg sources include atmospheric deposition of anthropogenic (for example, coal combustion) and natural (for example, wildfires) material onto terrestrial landscapes and water bodies (Grigal 2002;Eckley and others 2013). Internal Hg sources include ore deposits (Kerfoot and others 2002), and Hg can be imported to watersheds via industrial releases (Bravo and others 2013). Organic soils and other biomass serve as important reservoirs of contemporary and legacy Hg from these internal and external sources (Grigal 2002;Steffen and Morrison 2016). Mobility of Hg throughout watersheds depends on many factors including climate, hydrologic connectivity, topography, erodibility of soils, vegetation cover, and land management practices, among others (Balogh and others 2003;Haynes and Mitchell 2012;Oswald and Branfireun 2014;Liu and others 2018). Previous assessments established associations between Hg in rivers with catchment extents of agricultural and industrial development (Hurley and others 1995;Babiarz and others 1998;Lyons and others 2006), wetlands (St Louis and others 1994;Shanley and others 2005;Schuster and others 2011;Bradley and others 2013), forest coverage and associated logging and wildfire events (Kelly and others 2006;Kronberg and others 2016), and erosive landforms (Kerr and Cooke 2017;Vermilyea and others 2017). Though these landscape-river connections are evident, particularly in smaller river systems, important physical and chemical changes within rivers may occur as these waters travel downstream (Harvey and Gooseff 2015).
As lotic systems progress downstream from headwaters to higher-order rivers, these waters become less reflective of landscape biogeochemistry as runoff is aggregated from larger diverse areas, erosion and deposition patterns change, basin geomorphology flattens, and groundwater influences increase (Temnerud and Bishop 2005). These downstream changes can enhance or depress concentrations of suspended material, dissolved organic matter, and ions in river water and may influence concentrations of Hg by association. These spatial changes can be confounded by difficulties quantifying Hg concentrations and yields in rivers considering: 1. rapid changes in flow and sediment transport occur often and associate closely with Hg (Babiarz and others 1998;Riscassi and others 2011;Haynes and Mitchell 2012); 2. ultra-Downstream Modification of Mercury in Diverse River Systems clean sampling requirements are not easily incorporated into automated sampling systems (Bradley and others 2013); 3. regression-type surrogate approaches (for example, turbidity) can be problematic during high flows (Wall and others 2005); and 4. high analytical costs limit the frequency, duration and locations of Hg sampling. As such, many previous studies of Hg in rivers have been of short duration, low sampling frequency, and limited spatial replication within and between river systems. Longer-term replicated sampling in rivers may improve our understanding of how landscapes impact Hg delivery to river systems, and if they can be associated with Hg concentrations in resident fish tissues.
Here, we present an integrated downstream spatial assessment of Hg in rivers from headwater tributaries, through passage into major rivers, to accumulation in resident fish tissues across Alberta, Canada. Specifically, we associate concentrations and watershed yields of unfiltered (UF) and filtered (F) total Hg (THg; all forms of Hg in a sample) and MeHg in tributaries and major rivers with different catchment landscape characteristics. We then use fish tissue monitoring from major Alberta rivers to provide insights into connections between THg in fish, river water Hg concentrations, and upstream catchment characteristics. We hypothesized that intact, forested landscapes would produce lower Hg yields and river water concentrations relative to those draining developed catchments. By extension, we expected to observe lower THg concentrations in fish tissues collected from rivers with lower river water THg concentrations and THg loading from catchments.

Study Region
Alberta is an ecologically diverse Canadian province. Landscapes vary from high elevation Rocky Mountain cordillera landscapes (> 3,000 m a.s.l.) in the west, through adjacent forested Foothills, to lower-relief Boreal Plains forests and wetlands, parkland forests, and prairie grasslands/shrublands toward the north, east, and south, respectively ( Figure 1a). Glacial till characterizes much of the province's surficial geology downstream from the Rocky Mountains. Approximately 55% of the province is covered with intact and fragmented/ logged forests, followed by agricultural lands (28%), wetlands ( 10%) and surface waters ( 5%), barren/rocky lands (2%), and urban, industrial, and road land uses (1.5%) (ABMI 2019).
Precipitation decreases from cordillera environments (> 600 mm y -1 ), northeastward though Boreal Plains (350-500 mm y -1 ) and southeastward to agricultural prairie grasslands (< 350 mm y -1 ; 1970-2000 normals) (Environment and Climate Change Canada 2018). Catchment water yields associate closely with precipitation patterns in Alberta (Kienzle 2012). Mean daily air temperature across the province reflects a continental climate with most of the province colder than -10°C in January and not warmer than 18°C in July.
Albertan landscapes contain the headwaters of some of Canada's largest rivers including the Mackenzie, Churchill, and Nelson rivers. A tributary of America's Missouri-Mississippi river system, the Milk River, also passes through southern Alberta. Flow of most major central and southern Albertan rivers is low, consisting of glacier-fed mountain headwater rivers (for example, Bow at Calgary: 2-5 km 3 y -1 [full records]; North Saskatchewan at Edmonton: 4-12 km 3 y -1 ; South Saskatchewan upstream of Medicine Hat: 2-13 km 3 y -1 ) (Water Survey of Canada 2021). Larger Boreal rivers occur in the north, including the glacier-sourced Athabasca (at Fort McMurray: 11-32 km 3 y -1 ) and Peace (at Peace River: 29-91 km 3 y -1 ) rivers. Annual flow of most rivers is highest during spring freshet, is supplemented by summer rainfalls, and decreases substantially through autumn and winter. Most major Alberta rivers (not Athabasca or Beaver) are regulated with flows higher in winter and lower during freshet than would occur naturally, though broad expected seasonal patterns of flow still exist. Certain rivers pass through major urban areas downstream of the

River Water Quality Monitoring and Analysis
The Government of Alberta has monitored water quality monthly at several stations on its major rivers (draining 82% of provincial area) for decades ( Figure 1b). Tributaries of major rivers have been monitored sporadically over time, including annual and seasonal (March -October) monthly sampling (Figure 1a; Table S1). At each station, water was collected as near surface grab samples and analyzed for general water quality (major ions, nutrients, trace metals) at accredited laboratories. Monitoring for concentrations of trace-level THg at tributary and major river stations began in rivers as early as 2007. MeHg samples were taken from only selected tributaries and major rivers starting in 2013. During open water, most THg and MeHg samples were obtained by wading waist-deep and collecting water as surface grab samples at approximately 30 cm depth into new amber glass bottles (Environmental Sampling Supply). On larger rivers (for example, Athabasca, Peace rivers), samples were often collected at 30 cm depth at the thalweg from a boat. Winter samples, where open leads on rivers were not present, were collected by drilling a hole through river ice, clearing slush, and collecting water below the ice by hand or using a glass bottle attached to a plastic-lined sampling rod. All samples were collected using the ''clean-hands, dirty-hands'' USEPA Method 1669 for trace metals sampling (U.S. Environmental Protection Agency 1996) with some modifications for single-person sampling in smaller rivers of central and southern Alberta. An extensive quality control program consisting of distilleddeionize water blanks and replicate samples was included in regular sampling (see Supporting Information, Table S2). After collection, THg and MeHg samples were field-preserved with HCl to Figure 1. a Land cover and mercury sampling stations on tributaries of major rivers of Alberta, western Canada, including drainage areas; b land cover and mercury sampling stations on major rivers of Alberta, including drainage areas; and c hierarchical cluster classification of tributary drainage areas using land cover, topographic slope, and climate variables. The Athabasca and Peace rivers contribute water to the Mackenzie River system to the north, the Milk River to the Mississippi system to the south, and all others to the Nelson-Churchill system to the east. 0.2% and 0.4% volume, respectively, and stored at 4°C until analysis. Samples for filtered measurements were filtered using 0.45 lm cellulose nitrate filters within 24 h of collection.
All water THg analyses were performed using oxidation, purge and trap, and cold vapour atomic fluorescence spectrometry (CVAFS) via USEPA Method 1631E (U.S. Environmental Protection Agency 2002) (see Supporting Information). THg analyses were performed by two different laboratories with a method detection limit of 0.04 ng L -1 for Innotech Alberta and a reportable detection limit (RDL) of 0.06 ng L -1 for the Biogeochemical Analytical Service Laboratory at the University of Alberta (BASL; 2012-2018) (see Figure S1 for lab comparison). Innotech Alberta analyzed only THg UF samples while BASL analyzed both THg UF and THg F (RDL: 0.08 ng L -1  [TAl]), and in-river measurements (pH, specific conductivity [COND]). These supporting parameters were selected based on typical associations with Hg reported in literature (for example, Babiarz and others 2012).

Fish Collection and Analysis
THg concentrations in fish tissues from major Alberta rivers were sourced from monitoring programs operated by the Government of Alberta (province-wide), the Oil Sands Monitoring Program (Athabasca and Peace rivers), and Environment and Climate Change Canada (North and South Saskatchewan rivers). These monitoring programs spanned different years between 2006 and 2016, which overlapped in time with river water Hg sampling (Table S3). Generally, fish collection methods consisted of gill netting, angling, electrofishing, or other means. Biometric information from each fish was recorded including species, whole-fish wet weight, total and/or fork length, and sex. Fillets or whole fish (filleted later) were frozen and submitted to accredited laboratories for THg determination using USEPA 1631E and USEPA 7473 (see Supporting Information).

River Water Mercury Concentrations and Yields
Data used for numerical assessments included all monthly THg samples from tributaries with more than 16 THg UF samples over three years and five separate months (2007-2018; n = 47; Table S1), and all monthly samples from major rivers (2007-2018; n = 35 stations). At all stations, mean concentrations and other descriptive statistics were calculated, as well as F:UF and MeHg:THg concentration ratios, where available. Watershed yields were calculated at all stations for each sample by quantifying the product of THg concentrations and mean daily flow measured on the sample day at the nearest flow station (Table S4) and dividing by the catchment area upstream of the sample location (see next section). Since MeHg sampling was much less widespread amongst river stations than THg, MeHg UF and MeHg F concentrations and yields were used to support THg analyses and interpretations.

Watershed Associations with Mercury in River Water
For all catchments upstream of water quality stations, watershed characteristics were compiled using spatial data within a Geographical Information System (ArcGIS; v.10.7.1; ESRI, Redlands, USA). ArcGIS Hydro Tools and an interpolated 100 m digital elevation model were used to delineate catchment areas upstream of water quality stations. Land cover data from the Landsat 5 Thematic Mapper (30 m resolution) for Alberta and surrounding jurisdictions were obtained from Canadian and US governments (Agriculture and Agri-Food Canada 2015; Homer and others 2015). Relative catchment areas of each land cover type were calculated. Mean topographic slope was quantified for each watershed using national digital elevation models (Gesch and others 2002;Natural Resources Canada 2013). Mean annual precipitation and air temperature within each catchment were taken from climate model interpolations of western North America (ABMI 2021). Stream density, used as a supporting measure of hydrologic connectivity, was calculated using total stream lengths from 1:50,000 digital topographic maps and dividing by upstream catchment area.
To group watersheds by landscape characteristics for comparisons of Hg concentrations and yields in rivers, we entered all spatial data (described above) of 47 independent tributary catchments into a hierarchical clustering analysis (HCA) using standardized scores, Euclidean distances, and Ward's clustering method (manual cluster selection) using SPSS (v26; IBM, New York, USA). Absolute ranges of spatial data within each cluster were used to similarly classify the larger, more integrated major river watersheds post-HCA (Table S5). We then used log 10 -transformed THg concentration and yield data from all stations to assess differences between landscape types using linear mixed-effects models in SPSS. Mixed-models statistically account for model variability from repeated sampling (nonindependence) at each station while still assessing fixed effects (that is, differences between landscape clusters). In the models, we used autoregressive covariance structures for repeated data, random effects modeling of regression intercepts, and Sidak multiple comparisons to distinguish significance in Hg among landscape types. Statistical similarity and differences in geometric means between landscape types for THg concentrations and yields were subsequently reported (a = 0.05). Different linear mixed models were developed for tributaries and major rivers to assess relative changes in Hg concentrations and yields between landscapes of these smaller and larger systems. Principal components analysis (PCA) was used (Canoco v. 5.03; Biometris, The Netherlands) to group THg concentrations with general chemistry and river flow for further biogeochemical insights between river and landscape types.

Mercury in Fish of Major Rivers
Analyses of fish THg data focused on two taxonomic groups that afforded robust spatial comparisons: walleye (Sander vitreus) and suckers (family Catostomidae). Walleye were collected from all major rivers in Alberta, but were absent from upper (mountain) reaches. Using THg concentrations and wet weights, we grouped walleye data (n ‡ 5 individuals) by broad geographic region in major rivers (Table S3) and applied landscape type classifications to these groups based on the closest major river water quality station to the grouping geography. In the sucker family, species represented included longnose sucker (Catostomus catostomus), shorthead redhorse (Moxostoma macrolepidotum), and white sucker (Catostomus commersoni). Longnose sucker and white sucker have similar broad distributions in Alberta, although the former can inhabit mountain reaches of rivers more successfully than the latter. Shorthead redhorse occur in major rivers and tributaries of cen-tral and southern Alberta, but are absent from northern basins. The taxonomic grouping of these species is justified, based on comparisons among species and within distinct locations (geographic groupings) for wet weight, THg concentration, and the relationship between weight and THg concentration. As for walleye, geographic groupings were constructed for suckers (Table S3) and major river landscape classifications were coded to those groups.
Analysis of covariance (ANCOVA) was used to compare THg concentrations in walleye and sucker tissues within and among major river groupings (that is, landscape classes). ANCOVA was chosen a priori as the standard parametric test for comparing a characteristic (THg concentration) of groups (distinct population or location) while controlling for the effect of another variable (size) on that characteristic. The covariate used was wet weight, as that was the most consistently reported measure of size among and between programs. An ANCOVA with Type III sums of squares was performed for each of the walleye and sucker taxonomic groups, with JMP15 (SAS, Cary, USA; see Supporting Information).
Using linear regression, THg concentrations of fish groupings were compared to major river THg and MeHg concentrations in water. THg concentrations in walleye and suckers were represented by least-squares means of geographic groupings from the ANCOVAs. Least-squares means can be considered size-adjusted means, because the ANCOVA adjusts group means of the dependent variable (THg concentration) by controlling for the effect of the covariate (wet weight). THg and MeHg concentrations in water were represented by recorded arithmetic means (full records) from the nearest water quality monitoring stations to fish geographic groupings (Table S3).

River Water Mercury Concentrations and Yields
Concentrations of unfiltered and filtered THg from 47 small tributary systems were highly variable within and between rivers (Figure 2). Mean THg UF concentrations of sites ranged from 0.28 to 120 ng L -1 with coefficients of variation between 43 and 315%. Filtered THg concentrations were about 47% of unfiltered concentrations across all tributary sites with mean concentrations ranging from 0.19 to 18 ng L -1 . Tributaries of the lower Red Deer River basin reported the highest mean THg con-Downstream Modification of Mercury in Diverse River Systems centrations of all sites, including some sites beyond acute guidelines for protection of aquatic life (PAL; CCME 2003). Upstream tributaries in basins generally had lower THg concentrations than those downstream. MeHg UF concentrations were similarly highest in tributaries of the lower Red Deer River compared to those north or west, and MeHg F concentrations comprised about 53% of MeHg UF concentrations of all tributaries sampled for MeHg ( Figure S2).
Mean THg UF concentrations of a given major river were highest in the Red Deer (mean ± S.E.: 6.3 ± 1.1 ng L -1 ) and Peace (6.0 ± 1.2 ng L -1 ), and lowest in the Bow (1.8 ± 0.2 ng L -1 ) and Beaver (1.8 ± 0.1 ng L -1 ). Across individual stations (Figure 2), mean THg UF was below 6 ng L -1 except for two from the lower Red Deer (7.8, 17 ng L -1 ) and one from the Peace (7.7 ng L -1 ). Of all THg UF samples (n = 3189), only 1.2% were above PAL guidelines and were mostly from the Red Deer, Athabasca, and Peace rivers and collected during open water conditions. Plains-sourced rivers (Battle, Beaver) had the highest mean THg F :THg UF (39-46%) while mountain-sourced rivers had more THg bound to coarse suspended particles (THg F :THg UF : 13-26%). Concentrations of THg in major rivers generally increased downstream from west to east across the province. MeHg UF concentrations ( Figure S2) were highest in the Red Deer River (0.14 ± 0.02 ng L -1 ) and lowest in the Peace River (0.061 ± 0.013 ng L -1 ), and all samples collected province-wide (n = 629; max: 1.4 ng L -1 ) were well below national guidelines (4 ng L -1 ) (CCME 2003). MeHg F concentrations comprised 41% of MeHg UF concentrations in major rivers. Province-wide, MeHg UF :THg UF (from 21 stations) was 0.6-12.1% (mean: 3.7%) and highest in the Beaver River. Overall, MeHg concentrations in tributaries and major rivers were positively correlated with THg concentrations ( Figure S3). Similar to concentrations, watershed yields of THg UF of tributary stations were highly variable in space ranging from 0.39 to 170 lg ha -1 d -1 with THg F comprising 41% of those yields (Figure 2). In contrast with concentrations, THg UF yields were typically highest in those tributaries draining high elevation catchments in the west and decreased toward eastern regions, though tributaries of the lower Red Deer River were an exception to this pattern. THg UF yields from major rivers (3.1-130 lg ha -1 d -1 ) were similar to those from tributaries and generally decreased from western stations toward eastern stations. Yields were lowest in plains-sourced rivers relative to mountainsourced rivers, and only 22% of major river THg UF yields were attributed to THg F . Annual riverine THg yields of major rivers (0.09-1.3 lg m -2 y -1 ) were lower than province-wide total atmospheric deposition ( 6-25 lg m -2 y -1 ) (Steffen and Morrison 2016) and rivers of other jurisdictions (Table 1).

Watershed Associations with Mercury in River Water
We distinguished three broad landscape classes across Alberta using climate, topography, and land cover geospatial data from 47 independent tributary catchments (Figures 1c, 3, S4). Cordillerafoothills catchments reside in western Alberta and are colder, wetter, and steeper than other landscapes and are predominantly covered in intact forests and barren rocky surfaces. Water quality of tributaries draining these catchments is dilute and sediment-and nutrient-poor. Forested catchments occur mostly within low-relief Boreal Plains forests of northern Alberta. These landscapes are drier than cordillera-foothills catchments and are predominantly forested with wetlands and few agricultural areas. Water quality in forested catchments is of low conductivity and pH, moderate suspended sediment concentrations, and high DOC concentrations compared to other tributaries. Finally, lowrelief agriculturalized catchments are relatively warm, dry, and covered by crop and pasture lands with few forests compared to other catchments. Water quality in these smaller, sometimes ephemeral streams is concentrated with dissolved ions, DOC, nutrients, metals, and suspended sediments.
THg UF concentrations in tributaries draining cordillera-foothills regions (geometric mean: 0.76 ng L -1 ) were statically lower than those draining forested (1.5 ng L -1 ) and agriculturalized (2.4 ng L -1 ) catchments ( Figure 4, Table S6). At the alpha 0.07 level, agriculturalized catchments had higher THg UF concentrations in river water than the other two landscape types, as also observed for arithmetic mean (± SE) MeHg UF concentrations between the landscapes (forested: 0.16 ± 0.01; agriculturalized: 0.29 ± 0.06 ng L -1 ). Filtered THg concentrations draining forested (0.95 ng L -1 ) and agriculturalized (0.89 ng L -1 ) catchments also were significantly higher than cordillera-foothills (0.37 ng L -1 ) rivers (Table S6). Notably, forested tributaries had much higher F:UF concentration ratios (65%) compared to agriculturalized (37%) or cordillera-foothills (49%) tributaries. Tributary concentrations of THg in cordillera-foothills and agriculturalized landscape class were decoupled from dilution-concentration flow patterns as THg PCA scores were toward right angles to a flow-conductivity gradient in ordination space ( Figure 5). Typically larger forested tributaries showed THg concentrations associated closer to dilution-concentrations patterns in ordination space. In major rivers, THg concentrations were statistically similar between all landscape types (a = 0.05) as they increased in the cordillera-foothills regions and generally decreased in agricultural and forested regions, relative to tributaries. MeHg UF concentrations similarly decreased in agriculturalized major rivers from 0.28 to 0.16 ng L -1 . However, THg F concentrations in forested major rivers (0.68 ng L -1 ) were higher than cordillera-foothills regions (0.43 ng L -1 ) at an alpha of 0.06. Cordillera-foothills and forested major rivers showed a coupling of THg concentrations with flow along the concentration-dilution gradient, but agriculturalized major rivers were decoupled from this pattern, similar to unforested tributary classes. THg UF and THg F were close in ordination space at all sites, similar to correlations between THg and MeHg ( Figure S3).
THg yields of tributaries showed much different patterns between landscape types relative to concentrations. Both THg UF and THg F yields of cordillera-foothills and forested tributaries were statistically greater than agriculturalized catchments (Figure 4). Nearly three-quarters of THg yields from forested catchments were due to THg F compared to the other landscape types (33-43%). Interestingly, THg UF yields of cordillera-foothills major rivers (7.6 lg ha -1 d -1 ) were significantly higher than both forested and agriculturalized major rivers (1.3-2.6 lg ha -1 d -1 ) while major rivers showed similar relative landscape differences in THg F yields as tributaries, with agriculturalized catchments having lower yields than the other landscapes.

Mercury in Fish of Major Rivers
The ANCOVAs for walleye and suckers produced similar results indicating significant variation in size-adjusted THg concentrations among major rivers. For walleye and suckers, six geographical groupings met the assumption of linearity for THgwet weight regressions and had slopes significantly different from zero (Table S7). For walleye, slopes of the regression lines were not statistically different among the six geographical groupings (F 5,382 = 1.52; p = 0.182), but corresponding intercepts were statistically different among groupings (F 5,387 = 14.37; p < 0.001). A Tukey's HSD test found that the least-squares mean (± S.E.) of THg concentrations in walleye from the Peace-middle grouping (0.24 ± 0.01 lg g -1 ww; cordillera-foothills class) was significantly lower than other forested locations ( Figure 6).
For suckers, slopes of the regression lines were statistically different among the six geographical groupings (F 5,390 = 2.77; p = 0.018) and upon removal of the slope with the lowest r 2 (Athabascalower), resulted in no difference in slopes among the remaining five groupings (F 4,188 = 1.57; p = 0.185). Accordingly, we found statistical differences in intercepts among groupings (F 4,192 = 8.24; p < 0.001) and found least-squares means of THg concentration increased in fish from the Bow-upper (0.10 ± 0.02 lg g -1 ww) to the Athabasca-middle (0.18 ± 0.01 lg g -1 ww) to Red Deer-lower (0.28 ± 0.03 lg g -1 ww; Figure 6). This pattern represented a transition from a cordillera-foothills dominated station to two forested stations. A lone agriculturalized grouping in the lower South Saskatchewan River was similar in fish tissue concentrations to both forested groupings,   while the cordillera-foothills Oldman R. middle site had fish Hg tissue concentrations similar to other forested and agriculturalized sites. Linear regressions between least-squares mean fish tissue THg concentrations and Hg concentrations in river water (THg, MeHg) showed variation among rivers in the former can be described by the latter, particularly by MeHg. For walleye, there was no significant relationship between THg concentrations in fish and in water, but there were strong relationships with mean MeHg concentrations in unfiltered (n = 5; r 2 = 0.920; p = 0.010) and filtered (n = 5; r 2 = 0.655; p = 0.097) water. For suckers, there were strong relationships between least-squares means of THg concentrations and mean THg concentrations in unfiltered (n = 5; r 2 = 0.784; p = 0.046) and filtered (n = 5; r 2 = 0.585; p = 0.132) water and mean MeHg concentrations in unfiltered water (n = 4; r 2 = 0.992; p = 0.004).

Landscape Connections to Mercury Concentrations in Headwater Tributaries
The biogeochemistry of small, headwater tributaries reflects their close connections to landscapes (Alexander and others 2007). Our PCA results from tributaries support this connection as close groupings of THg UF , THg F , t-Al, and TKN in ordinal space suggests co-mobilization of particle-bound nutrients, THg, and metal oxy-hydroxides from soils and surface waters via erosion and soil flushing (Grigal 2002;Park and others 2018). Particularly in smaller headwater streams (that is, cordillera-foothills, agriculturalized) with limited in-channel sinks and sources of landscape material, these associations were de-coupled with flow, which is consistent with hysteresis-type patterns indicating off-channel, landscape contributions of chemicals to rivers (Buendia and others 2016). Considering broad seasonal changes of Hg concentrations with flow in rivers ( Figures S5-S8), inconsistent monitoring locations relative to cities (Figure 1) and industrial effluents (not shown), and human impacts on landscapes dominated by agriculture (ABMI 2019), we suggest that THg and MeHg sources and mobility at our scale of monitoring are related much more to landscape conditions, rather than local urban or industrial activities. As such, we expected broad landscape differences to be a key driver of THg mobility downstream in tributary and major river systems.
Alberta's high-relief regions present climate and topographic conditions ideal to erosively mobilize THg due to high precipitation rates upon steep, dendritic landscapes (Figure 3; Riscassi and others 2011). However, throughout the Rockies, sediment and organic matter supply to rivers is poor due to predominance of unvegetated mountains, coarse gravels and boulders, thin or patchy soils, lack of wetland development, and national protection of landscapes from widespread development. Though foothills rivers drain rolling landscapes with more developed soils, forest cover fortifies soils and reduces erosive mobilization of THg UF (Gamby and others 2015). The general absence of lakes or wetlands on higher-relief topography likely reduces export of dissolved organic material and THg F downstream (Guentzel 2009; St Pierre and others . Geometric means (± 95% confidence intervals) of unfiltered (upper panels) and filtered (lower panels) total mercury concentrations (left panels) and yields (right panels) for rivers within each of the three hierarchical clusters. Consecutive letters within each landscape grouping denote statistical differences at a = 0.05 and *denotes statistical differences at a = 0.06-0.07 with the preceding category. 2019; Staniszewska and others 2021); thus, most Hg mobility in these systems is particle-bound. Together, these conditions support the observed lower THg concentrations in cordillera-foothills tributaries relative to other landscape types. Masswasting events, however, are common in these regions and can produce pulses of sediment to rivers that are offset from flow changes (Lee and others 2013).
In agriculturalized tributary catchments, which are dominated by drier climates with convective storms, rich organic chernozemic soils (CSC 1998) are exposed to storm-related gulley erosion and downstream mobilization of soils, nutrients and organic matter (Figure 3). This process is aided by row crop management that typically results in greater mechanical interventions of soils (for example, disking, seeding) and periods when erosive runoff moves across and through soils without established vegetation cover or soil cohesiveness (Kort and others 1998;Sherriff and others 2016). Often this material carries bound Hg in coarse and colloidal forms (passing through filters) and subsequently results in relatively high THg UF and THg F concentration in rivers (Hurley and others 1995;others 1998, 2012; Balogh and others . This sediment-rich load includes bound MeHg likely produced in methylating agricultural environments such as riparian corridors, beaver-induced flooded areas, and shallow wetlands and lakes (Ullrich and others 2001). Though intense storm-driven erosion of agricultural soils can increase THg concentrations in rivers, we also observed agriculturalized landscapes with lower river water THg concentrations that are associated closely with catchments of perennial pastures and grassland cover (Figure 3). This relationship can be attributed to fortification of soils by grasses that could limit soil detachment events (Kort and others 1998) and likely partly explains the mere statistical similarity between forested and agriculturalized THg concentrations in tributaries (Figure 4).
In the low-relief Boreal Forest, landscapes store water in organic soils and wetlands that typically results in surface waters high in DOC concentrations that can bind THg F (Figures 3, 5) (St Louis and others 1994; Wasiuta and others 2019). As such, we observed the highest DOC concentrations and THg F :THg UF concentration ratios from forested tributaries compared to others. Many of these wetlands and organic soils are adjacent to tributary channels and during hydrological events may result in efficient surface and subsurface flushing and transport of DOC and THg downstream (Bushey and others 2008;Devito and others 2012). Larger tributaries of forested regions may also mean a greater role of in-channel and river bank contri- Figure 6. Least squares means (± 1 standard error) of total mercury concentrations from geographical groupings (that is, landscape classes) of fish from major Alberta rivers (left panels). Mean unfiltered and filtered total mercury and methylmercury concentrations (± SE) of nearby water quality monitoring stations relative to fish tissue concentrations of total mercury in geographical groupings of fish collected from major Alberta rivers (right panels). Regression symbol coloring matches the landscape classes in the left panel legend.
butions of legacy material from landscapes, including particulate material and bound THg, during high flows. Together, these characteristics and processes of forested tributary landscapes may explain the lack of hysteresis observed and suggests potential changes in sources and mobilization of Hg as river sizes increase. Using data from major rivers, consequently, we tested the broader question if larger rivers downstream modify these landscape differences (that is, signals) as their geomorphologies and water compositions change.

Mercury Comparisons Between Tributaries and Major Rivers
Relative to tributaries, major river environments flattened differences in THg concentrations between landscape types and effectively reduced discernable upstream landscape influences on THg. We attribute this pattern to differences in how tributaries and larger rivers process sediments and dissolved material (and hence Hg) across different landscape types. In cordillera regions dominated by particle-bound transport of THg (Figure 2), hydrological energies can be vastly different between headwater regions and major river systems due to sharp slope transitions in connected river systems. This gradient promotes sediment entrainment in tributaries (albeit in low concentrations) but longterm sediment deposition in major rivers; notably observed as braided geomorphologies in high elevation environments (Knighton 2014). In cordilleran major rivers of Alberta, these deposits would likely include glacial flour and metal oxide material from upstream headwaters, which can bind significant concentrations of THg F (Hawkings and others 2021). Thus, upon high water events, these channel deposits would represent a steady source of local suspended material and bound THg to major rivers compared to scant sediment supplies in tributaries. Our PCA results support an in-channel source of sediments and THg in cordillera-foothills major rivers as it reported ordinal space clustering of flow, THg, TSS, and t-Al concentrations.
Agriculturalized regions, alternatively, are lowrelief with impacted soils, intensive gulley erosion, and high sediment and THg concentrations in headwater rivers. Slope transitions there are less severe downstream relative to high-relief regions, thus promoting sedimentation at tributary confluences or continued entrainment of sediment, colloids, and THg downstream, particularly during high water in major rivers. Groundwater in these relatively small arid plains rivers is an important proportion of in-channel water that is often dilute in particle-bound THg during baseflow ( Figures S5-S8). Therefore, pulses of high THg UF and THg F concentrations in tributaries during storms likely mix with major river waters more dilute in THg and depress concentrations. These arguments are supported by our PCA results that showed agriculturalized major rivers with no near-field associations between flow, suspended sediment or THg, similar to observations from agriculturalized tributaries.
Forested regions showed more consistent THg UF concentrations between tributaries and major rivers, likely due to a combination of processes. Similar to tributaries, major forested rivers reported relatively high THg F :THg UF and DOC concentrations that likely reflected sustained inputs of DOCbound THg from adjacent wetland landscapes and upstream tributaries. For example, the fully Borealsourced Beaver River showed the highest THg F and DOC concentrations of all major rivers. PCA results from both river types also suggested in-channel sources of soil material and THg within these systems ( Figure 5). Major Boreal rivers (for example, Peace, Athabasca) and tributaries flow across legacy glacial till deposits and source sediment in channels and river banks that are erodible during high water events (Shrestha and Wang 2018) likely mobilizing particle-bound THg. Finally, large groundwater proportions of river flow are a key characteristic of large Boreal tributaries and rivers (Gibson and others 2016) and are typically poor in particlebound THg concentrations ( Figure S5-S8). Lowrelief topography and extensive surface water stores typical of Boreal landscapes also likely promotes storage of sediment and bound THg. Together, these synergies between river types likely explain the relatively few differences observed in Hg concentrations and associated water quality parameters downstream in forested regions.
Overall, downstream modification of THg concentrations from tributary to major river ecosystems suggests a weaker connection between landscapes and water in major rivers, due to inchannel processes that become relatively more important in larger rivers (Harvey and Gooseff 2015). Fish tissue THg concentrations collected from these rivers, alternatively, showed mixed results relative to this downstream continuum, suggesting an importance of local inputs of THg and MeHg relative to basin-scale conditions.

Landscape and Major River Connections to Mercury in Fish Tissues
Several studies have reported positive associations between habitat surface water THg and MeHg Downstream Modification of Mercury in Diverse River Systems concentrations and THg concentrations in fish tissues (for example, Mason and others 1996). In small catchments, associations have been established between watershed characteristics and fish tissue concentrations (Warner and others 2005b;Rypel 2010;Marusczak and others 2011;Mattieu and others 2013). Using this hypothesis, we would expect few differences in THg fish tissue concentrations between fish caught in major rivers of cordillera-foothills, forested, and agriculturalized watersheds. However, we observed THg concentrations in fish varied significantly within and across different landscape types (Figure 6), suggesting site-specific, river water concentrations may be of greater importance than broad trends across landscape and river types. For example, both walleye and suckers showed close positive associations between THg tissue concentrations and MeHg concentrations in river water, despite varying differences in landscape type within regressions ( Figure 6). This pattern supports MeHg concentrations and production rates in water being a key driver of bioconcentration and subsequent biomagnification of MeHg foodwebs (Mason and others 1996). For omnivorous suckers, river water THg concentrations also associated closely with tissue concentrations, which may reflect its bottom-feeding trophic position and greater non-diet exposure to THg in sediments (Rose and others 1999). Walleye, alternatively, are piscivorous species that primarily gain Hg via diet, rather than environmental exposure, and thus reported poor associations between tissue and river water THg concentrations, but higher tissue concentrations overall. Regardless, observations of THg concentrations in fish that were significantly different between landscape types (walleye and suckers), or significantly different within landscape types (suckers), point to site-specific differences driving Hg uptake in fish rather than broad regional differences (Rose and others 1999).
Several unique characteristics of specific river reaches where fish were collected may explain why THg concentrations in fish varied within and between major rivers. First, sites in the lower Red Deer River, though classed as forested in this study (Table S5), receive local runoff from exposed badlands and agriculturalized areas that can produce substantially high concentrations of THg and MeHg, particularly during summer storm events (Figure 2; S5-S8; Kerr and Cooke 2017). Donadt and others (2021) reported fish tissue concentrations in this region of the Red Deer River, consequently, are likely higher than surrounding regions. This helps explain statistically higher THg concentrations in suckers in the lower Red Deer relative to fish collected from other forested and cordillera-foothills groupings. Second, the Oldmanmiddle site reported fish with significantly higher THg concentrations than the Bow-upper site despite both groupings classified as cordillera-foothills. This difference reflects the high variability of THg concentrations within classifications as the Bow River site reported THg concentrations of only 0.51 ± 0.5 ng L -1 relative to the more locallyagriculturalized Oldman River site (2.8 ± 0.6 ng L -1 ; Table S5). This finding matches the work of Brinkmann and Rasmussen (Brinkmann and Rasmussen 2012) that reported increases in THg concentrations in aquatic biota (including suckers) along a continuum of the Oldman River from cordillera/foothills landscapes to more locally agricultural landscapes.
Finally, walleye THg concentrations at the cordillera-foothills Peace R. Middle site were statistically lower than forested sites, despite reporting relatively high river water concentrations of THg. However, MeHg concentrations were very low at this site, possibly reflecting its water source from a large, cold, DOCpoor, mountain-fed oligotrophic reservoir likely limited in methylation sites (Bodaly and others 1993;Hsu-Kim and others 2018). Together, our results suggest that site-specific THg and MeHg concentrations and local characteristics associate better with fish tissue THg concentrations rather than broad catchment characteristics, whose signals diminish from headwaters to major rivers. Similarly, mercury loads carried by rivers that are largely influenced by river and catchment size differences between stations showed poor associations with fish tissue mercury concentrations (Figure S9). These findings are supported by results from Eagles-Smith and others (Eagles-Smith and others 2016) that showed site-specific THg and MeHg concentrations in rivers may be an important covariate of observed fish tissue concentrations. Consequently, our study suggests that site-specific, targeted monitoring of THg concentrations in water and fish, rather than landscape binning of collected fish, is a more important consideration in study designs when assessing regional-level patterns in fish tissue concentrations. This finding, in part, is likely due to downstream modification of landscape signals and water quality by large rivers relative to smaller tributary systems.

Mercury Yields and Connections to Landscape and Climate Change
In contrast with concentrations and our hypothesis, THg yields from intact tributaries were larger as they received more rainfall and had more widespread surface and subsurface hydrological connections than agriculturalized regions (Figure 4). Cordillera-foothills and forested regions are characterized by relatively frequent precipitation and large snow packs, widespread hydrological connectivity, ample coverage of surface waters, and large rivers. In contrast, drier agriculturalized regions dominated by net evaporation have smaller, lower-relief rivers subject to beaver activity, stagnation and drying. Short storm events drive much of Hg transport downstream and represent only short, intensive periods of hydrological connectivity that are buffered over annual scales. This seasonal pattern may sustain high mean concentrations in these rivers, but it does not equate to maintained export of Hg downstream.
Though Hg concentrations in aquatic ecosystems are of great importance for uptake by biota, yields are important measurements when considering source water quality in rivers and how potential future changes in climate and land use may impact those concentrations. Water quantity and quality in large rivers reflects mixing of various atmospheric, catchment runoff, subsurface, and inchannel sources of water and chemicals and disturbances to these sources can impact downstream water quality. For example, the highest yields of runoff and THg occur in cordillera and forested regions that are relatively intact and present more pristine environments compared to agriculturalized catchments. Cordillera-foothills and forested regions produce typically lower Hg concentrations in runoff and comprise most of the flows in Alberta's major rivers (Water Survey of Canada 2021), leading to among the lowest observed Hg concentrations and yields of major rivers compared to other jurisdictions (Table 1). However, these regions are experiencing notable changes to their climates and landscapes with potential downstream effects on Hg content in major rivers. Precipitation and streamflow has declined measurably across the Alberta climate in recent decades (Rood and others 2005;Newton and others 2021), in particular in cordillera-foothills and Boreal regions. Declining streamflow from intact regions may shift the relative importance of runoff and Hg yields from agriculturalized regions in some major rivers, with potential negative effects on river water quality. Shortening of winter periods and increasing air temperatures projected for the Alberta climate (Newton and others 2021) further increase the risk of more widespread methylation in surface waters and more frequent and intense wildfires that can impact Hg cycling in forested landscapes (Kelly and others 2006). Land use change such as land clearing for roads, forestry, and energy activities has proceeded quickly across Alberta in cordillerafoothills and forested regions (ABMI 2019) and such changes are known to increase Hg yields downstream (Hsu-Kim and others 2018). How these changes impact downstream Hg delivery and concentrations in major rivers is an important, ongoing question in Alberta and will serve as an important bellwether for other rivers systems worldwide. Source water protection, consequently, is an important consideration in water quality management of large river systems undergoing climate and land use change, and in the case of Hg likely has long-term implications for the ultimate uptake of Hg in food webs of large rivers.

ACKNOWLEDGEMENTS
Funding for water quality and portions of the fish work was provided by the Government of Alberta through operational budgets and grants to partners. We would like to thank the numerous Alberta Environment and Parks (AEP) technicians and supervisors who collected mercury and other water quality samples since 2007. We appreciate the work of Water Survey of Canada and AEP, taking hours of manual flow measurements and data truthing annual flow records. We would also like to thank AEP and ECCC fisheries technicians for collecting and processing fish for mercury analyses. Research funding of the river water monitoring program was provided by the Government of Alberta and the Government of Canada. Funding for portions of the fish work was provided by the Oil Sands Monitoring Program co-led by the Governments of Canada and Alberta, but does not necessarily reflect the position of the Program. We acknowledge the laboratory staff from the Alberta Centre for Toxicology at the University of Calgary for analyses of the fish collected by AEP and from Innotech Alberta and the University of Alberta Biogeochemical Analytical Service Laboratory for mercury in water analyses. Also thanks to Bureau Veritas staff for analysis water samples for general chemistry. Finally, we appreciate the time and input from anonymous reviewers whose input improved our manuscript.

OP E N A C C ES S
This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit h ttp://creativecommons.org/licenses/by/4.0/.

DATA AVAILABI LI TY
Raw water quality and fish tissue quality data are available from: https://www.alberta.ca/open-gover nment-program.aspx#jumplinks-0