Nutrient fluxes from an Arctic seabird colony to the adjacent coastal marine ecosystem

Seabirds are important vectors for nutrient transfer across ecosystem boundaries. In this seasonal study, we evaluate the impact of an Arctic colony (Alkhornet, Svalbard) of Black-legged Kittiwakes ( Rissa tridactyla ) and Brünnich’s Guillemots ( Uria lomvia ) on stream nutrient concentrations and fluxes, as well as utilization by coastal biota. Water samples from seabird-impacted and control streams were collected regularly throughout the melt season (June–September) for nutrient and organic carbon analysis. Stable carbon and nitrogen isotope analysis (δ 13 C and δ 15 N) was used to assess whether seabird-derived nitrogen (N) could be traced into filamentous stream algae and marine algae as well as consumers (amphipods). Concentrations of nitrate (NO 3− ) and nitrite (NO 2− ) peaked in July at 9200 µg N L −1 in seabird-impacted streams, 70 times higher than for control streams. Mean concentrations of phosphate (PO 43− ) in seabird-impacted streams were 21.9 µg P L −1 , tenfold higher than in controls. Areal fluxes from seabird-impacted study catchments of NO 3− + NO 2− and PO 43− had estimated ranges of 400–2100 kg N km −2 and 15–70 kg P km −2 , respectively. Higher δ 15 N was found in all biota collected from seabird-impacted sites, indicating utilization of seabird-derived nitrogen. Acrosiphonia sp. from seabird-impacted sites had higher δ 15 N values (20–23‰ vs. 3–6‰) and lower C:N ratios (10.9 vs. 14.3) than specimens collected from control sites, indicating reliance on seabird-derived nitrogen sources and potentially higher N-availability at seabird-impacted nearshore sites. Our study demonstrates how marine nutrients brought onshore by seabirds also can return to the ocean and be utilized by nearshore primary producers and consumers.


Introduction
Nutrient availability is a strong regulator of marine net primary production (Field 1998).Through fluxes of animals and nutrients, ecosystems are connected, acting as sources and sinks in a dynamic relationship (Polis et al. 1997;Loreau et al. 2003).While passive, abiotic fluxes are limited to transporting nutrients along gradients in potential This article belongs to the special issue on the "Pathways and impacts of biotically-mediated marine and other stored nutrient transfer between polar ecosystems", coordinated by Peter Convey, Katarzyna Zmudczyńska-Skarbek, and Stef Bokhorst.
energy, e.g. with river flow or ocean currents, animal-mediated fluxes can go in the opposite direction (McInturf et al. 2019).There are many examples of animals functioning as vectors for nutrient transport, where these fluxes can be of a similar or higher magnitude than passive transport (Varpe et al. 2005;Subalusky and Post 2019).One such example is colonial seabirds, which are responsible for a large part of the animal-mediated nutrient fluxes across the land-sea ecotone (Otero et al. 2018).Seabird droppings, eggs and dead birds scattered around these colonies can be an important nutrient point source for adjacent terrestrial (Anderson and Polis 1999;Breuning-Madsen et al. 2010;Leblans et al. 2014), limnic (Harding et al. 2004;Keatley et al. 2009) and marine ecosystems (Gagnon et al. 2013;Graham et al. 2018;Savage 2019).Seabird-derived nutrients have even been shown to fuel multiple trophic interactions, spanning both marine and terrestrial ecosystems in the Baltic Sea (Hentati-Sundberg et al. 2020).However, many of these interactions can risk weakening, or even disappear, as multiple seabird populations around the world experience declining populations trends (Paleczny et al. 2015;CAFF 2017).Moreover, a recent estimate attributes > 44% of seabird deposited nitrogen (N) and phosphorus (P) to come from species with declining populations (Plazas-Jiménez and Cianciaruso 2020).
The effect of seabird-mediated nutrient fluxes is particularly evident in terrestrial ecosystems surrounding breeding colonies.These areas stand out with intense greenness from increased primary production and change in plant community structure (Eurola and Hakala 1977;González-Bergonzoni et al. 2017;Richter et al. 2017;Duda et al. 2018).Herbivores are attracted to these productivity hotspots which contribute to further nutrient cycling and turnover (Jakubas et al. 2008).Although not as extensively studied in the Arctic, guano deposition from seabird colonies situated close to the shore could be expected to have similar impact on adjacent coastal ecosystems, both through direct deposition in the marine environment, and through runoff of nutrientrich water from colonies to the coast (Shatova et al. 2017;Bischof et al. 2019).
With large and distinct seabird colonies and a total of around 3 million breeding pairs in the Svalbard area (Anker-Nilssen et al. 2015), combined with a comprehensive understanding of local fjord-and coastal ecosystems (Hop et al. 2002;Bischof et al. 2019), the archipelago is well-suited for studying nutrient fluxes from seabird colonies to the marine ecosystem.In order to understand potential impacts of seabird colonies on coastal ecosystems, there is a need to consider the seasonal processes that may affect the connections between seabird colonies and the coast.Excess nutrients accumulated in the marine system during winter, coupled with a rapid increase in irradiance during spring leads to a large increase in marine primary production.As the summer proceeds, primary production typically becomes limited by nutrient availability, particularly dissolved inorganic forms of nitrogen (DIN), silicate (SiO 2 ) and to some extent phosphate (PO 4 3− ) (Rysgaard et al. 1999;Hop et al. 2002;Sakshaug et al. 2009).Meanwhile, the often highly synchronised seabird breeding season (Burr et al. 2016) paired with increased runoff during the summer melt season suggests that the highest intensity of the seabird driven nutrient fluxes may coincide with the onset of the marine nutrient depleted state by June and July (Hodal et al. 2012).Thus, it has been postulated that seabirds can increase primary production in coastal ecosystems in Svalbard, in particular fjord systems where seabirds are numerous (Zmudczyńska-Skarbek et al. 2015;Shatova et al. 2017).In a study comparing Hornsund and Kongsfjorden, two fjords on the west coast of Spitsbergen, Svalbard, it was argued that the higher number of seabirds breeding in Hornsund contributed to a higher annual primary production compared to Kongsfjorden (Smoła et al. 2017).
While several studies (including studies from Svalbard) have estimated seabird-mediated nutrient fluxes from land to sea, the degree to which these nutrients are retained in the terrestrial catchment, released to the atmosphere, or transported to the coast via runoff has received considerably less attention.Some generalisations can be made about the cycling and fate of nitrogen from seabird guano.Loss through runoff during precipitation events and volatilisation of ammonia (NH 3 ) from uric acid and other nitrogencompounds in guano are important pathways of nitrogen removal from seabird colonies (Lindeboom 1984).NH 3 emission rates are however dependent on temperature, pH, moisture and substrate type (Zhu et al. 2011).NH 3 typically has a short atmospheric lifetime and will be redeposited within a few hundred meters to a few km (Pitcairn et al. 2002).With increasing distance from the colony, a gradient in seabird impact on terrestrial and limnic environments can be observed, manifested in altered species abundance, composition, and physical characteristics, as well as stoichiometry of soil and water (González-Bergonzoni et al. 2017;Richter et al. 2017;Duda et al. 2018).
Isotope fractionation is widely used in analysis of trophic position and food webs dynamics (Peterson and Fry 1987).The light isotope 14 N is preferentially metabolised and excreted relative to the heavier 15 N, leading to higher 15 N in an organism's tissue relative to its diet.Since seabird species are feeding at high trophic levels in the marine environment, their guano also typically has a much higher δ 15 N levels than atmospheric deposited nitrogen and marine DIN.These elevated δ 15 N values in guano have been used to trace seabird-derived N into and through impacted ecosystems (Zwolicki et al. 2013;Kazama 2019).
In this paper, we study a seabird colony of Black-legged Kittiwakes (Rissa tridactyla) and Brünnich's Guillemots (Uria lomvia) in Svalbard, and the interaction between the breeding birds and the adjacent coastal marine ecosystems, with two objectives: Estimating nutrient fluxes from the colony and assessing uptake in the nearby coastal ecosystem.We test three main hypotheses: (1) The runoff from seabird-impacted streams will have higher concentrations of bioavailable nutrients and organic carbon than control streams; (2) the nutrient and organic carbon concentrations in the seabird-impacted streams will exhibit seasonal variation, with higher concentrations when the seabirds are present in the colony; and (3) The algae in seabird-impacted streams and coastal organisms in adjacent coastal waters will show evidence of reliance on seabird-derived nutrients (i.e.elevated δ 15 N values).

Study area
The bird cliff Alkhornet (Fig. 1) on the west coast of Spitsbergen, Svalbard, was visited five times between June and September 2018 (see Online Resource 1, Table S1, for dates).Alkhornet supports a mixed colony of about 14 000 breeding pairs of Black-legged Kittiwakes (Rissa tridactyla) and 15 000 breeding pairs of Brünnich's Guillemots (Uria lomvia) (Sebastien Descamps, pers. com.).This can be considered a medium size colony for Svalbard (Anker-Nilssen et al. 2018).The bird cliffs are situated 700 to 900 m from the shore and consist mainly of marble and calcareous phyllite from Neoproterozoic (Dallmann 2015).Cochlearia groenlandica and other nitrophilous species can be found in abundance near and on the rock wall (Fig. 1, I).On the terrace between the rock wall and the coast (Fig. 1, II and III) we find graminoid-moss tundra.The peat here can be 1.5 m deep, which is unusually thick for Svalbard (Låg 1990).

Water chemistry
Water was collected for chemical analysis from three streams impacted by the seabird colony (referred to as B1-B3) and in two control streams (C2 and C3), with little seabird presence.A third control stream (C1) was sampled only once, in July, as there was little surface streamflow on the other sampling occasions.The C1 site will therefore not be included in analyses with a seasonal component.In addition to total concentrations of nitrogen (TN), phosphorous (TP) and organic carbon (TOC), concentrations of SiO 2 , PO 4 3− , NH 4 + and NO 3 − + NO 2 − were analysed (n = 26).Analyses of water chemistry parameters were carried out using standard and accredited methods (as described in Kaste et al. 2018).A summary of the analytical methods can be found in Online Resource 1, Table S2.Stream water turbidity was measured during the four first visits to the study area (n = 21) Fig. 1 The study area at Alkhornet, Central-West Spitsbergen in Svalbard (overview map, © Norwegian Polar Institute).Station names correspond to the study streams and marine sampling sites (coordinates in Online Resource 1, Table S1).C1 (only sampled once), C2 and C3 are controls, B1, B2 and B3 are seabird-impacted.A picture is included of the main wall (I) at Alkhornet looking north-westerly, where the main colony is located.Beneath the wall is a scree area (II) before the terrain flattens out to peatlands (III).Dotted lines corre-spond to the catchment area of each stream (B1 = 0.24 km 2 , B2 = 0.14 km 2 , B3 = 0.22 km 2 , C1 = 0.30 km 2 , C2 = 0.73 km 2 , C3 = 3.74 km 2 ).Approximately 10% of C3 is glaciated, while the other catchment areas are non-glaciated.Stream water were sampled less than 50 m from the shore, at a distance of 0.7-0.8km from the colony centre for seabird-impacted streams and 2.0-3.5 km from the colony for control streams with a Eutech TN-100 handheld turbidity meter (Thermo Scientific).

Biotic sampling, stable isotope analysis (SIA), and C:N ratios
Kittiwake guano (n = 3) was either collected from a piece of plastic (50 × 50 cm) placed under the colony for 1 h (n = 2) or scraped off a rock (n = 1).The sampling was done at the accessible margins of the bird cliff where there is a preponderance of Kittiwake nests, and hence, we assume that Guillemot guano was not obtained.Filamentous, epilithic algae in streams were hand-picked and rinsed in deionized water, while in the adjacent coastal ecosystem, the macroalgae Acrosiphonia sp. and Ectocarpus sp. were collected in the tidal zone and rinsed with deionized water.Finally, amphipods (Gammarus setosus and species from the family Gammarellidae) were hand-picked and stored alive in filtered seawater (filtered through 5 µm pore size) for 24 h in order to clear their guts.Particulate organic matter samples were also collected for streams by filtering stream water onto precombusted glass fibre filters (Whatman GF/F; nominal pore size of 0.7 µm).All samples were stored frozen (− 20 °C) until further processing.All biotic samples were freeze-dried and homogenized prior to stable C (δ 13 C) and N (δ 15 N) isotope analysis, while filters were freeze dried and analysed for only δ 15 N.For accurate determination of amphipod and guano δ 13 C, inorganic carbonates were removed prior to analysis through acidification of homogenized subsamples with 1 M HCl (Brodie et al. 2011;Søreide and Nygård 2012), while δ 15 N values were determined through analysis of unacidified subsamples.Analysis of δ 13 C and δ 15 N was carried out at the UC Davis Stable Isotope Facility (SIF) with a continuous flow Isotope Ratio Mass Spectrometer.Long-term standard deviation at UC Davis is ± 0.2‰ for 13 C and ± 0.3‰ for 15 N. Stable isotope analysis also yielded data on C and N content for analysed samples, which were used to calculate molar C:N ratios for all biotic samples, and particulate N concentrations in stream water.

Runoff estimation
A digital elevation model (DEM) with a ground pixel size of 5 × 5 m was used for delineation of catchment area (Norwegian Polar Institute 2014).A drainage direction raster was made in QGIS 3.4 with GRASS GIS tools (GRASS Development Team 2017; QGIS Development Team 2019) and from this; the catchments draining towards the coordinates of each station were delineated.Estimation of runoff was made from weather data from the nearby meteorological station at Isfjord Radio (Norwegian Meteorological Institute (MET), http:// sharki.oslo.dnmi.no/, accessed January 2019) and the Arctic limited AROME (Application of Research to Operations at Mesoscale) weather model using median values of a 3 × 3 cell matrix surrounding the study area.AROME Arctic is METs operational forecasting model for Svalbard with a 2.5 km horizontal resolution (Müller et al. 2017).
A total of 190 mm precipitation was recorded at Isfjord Radio during the sampling period, while 221 mm was recorded from November 2017 to the first sampling in June 2018, and was assumed to be stored in the catchment as snow until the melt-season onset in June.Corresponding numbers based on AROME were higher, at 295 and 400 mm respectively.Precipitation measurements from Isfjord Radio, similarly to other weather stations in Svalbard, are known to often underestimate the true precipitation during winter (Førland and Hanssen-Bauer 2000).Additionally, its location at sea level typically equate to lower precipitation than the higher elevations that constitute much of the catchment basins (Killingtveit et al. 2003).Meanwhile, the AROME weather model accounts for elevation increase, which may explain the deviation between modelled precipitation and measurements at Isfjord Radio.
In order to estimate runoff, we used data on total annual surface runoff from London river (Blomstrandøya) close to Ny-Ålesund, where the ratio of surface runoff to annual precipitation was 0.69 (Killingtveit et al. 2003).Based on this ratio, the total estimated runoff in summer 2018 was 284 mm based on Isfjord Radio data and 479 mm based on AROME data.We used the area under the curve (AUC) function in the package "MESS" in R 3.5.1 (Ekstrøm 2019; R Core Team 2019) with the spline function (Yeh and Kwan 1978) to estimate nutrient and TOC fluxes from the study streams based on the AUC-derived runoff and concentration data for the individual sampling dates.AUC-derived total annual runoff estimates were matched to the ratio-based estimates (284 and 479 mm) by first estimating runoff (excluding snowmelt) for each sampling day by multiplying the 3-day average precipitation prior to the sampling days by 0.69 (Killingtveit et al. 2003).We then adjusted sampling day runoff estimates to account for snowmelt contribution by adding additional runoff for the June and early July sampling days in order to achieve an AUC-derived total annual runoff estimate matching the ratio-based estimate (weighted as two thirds and one third for June and July respectively, based on observations of streamflow and snow cover during fieldwork).Groundwater transport was assumed to be negligible as this only occurs in the permafrost active layer, which is ~ 2 m deep at other coastal sites in Svalbard (Ny-Ålesund and Isfjord Radio; Christiansen et al. 2018).Given the many uncertainties connected to this approach, runoff estimates are given as intervals with the likely underestimation from Isfjord Radio as a low estimate and AROME as a high estimate.Unfortunately, these estimates could not be verified with other catchment areas on Svalbard, as The Norwegian Water Resources and Energy Directorate had no adequate measurements of waterflow for the 2018 season (Songe, pers.comm.).By pairing catchment areas and runoff estimates with water chemistry data, seasonal yields of the different compounds can be approximated.Seasonal yields were estimated using AUC, following the same method as the runoff calculations.

Normalised difference vegetation index (NDVI) in catchments
The normalised difference vegetation index (NDVI) was used as a measure of seabird impact on the terrestrial vegetation.The index is calculated as follows; NDVI = [(NIR − red) × (NIR + red) −1 ], where NIR is near infrared spectral bands (700-805 nm) and red is red bands (635-675 nm).An 8-bit aerial orthophoto from 2009 by the Norwegian Polar Institute with a ground pixel size of 0.4 × 0.4 m taken with an UltraCam X sensor was used.The NDVI calculation was done with the QGIS "GRASS" package (GRASS Development Team 2017; QGIS Development Team 2019).

Ordination
To identify variation in stream water chemistry among sites and sampling days, we used principal component analysis (PCA).A subset of key parameters was selected for inclusion in the PCA to avoid missing values and to reduce dependence between parameters due to their intrinsic properties, e.g. between TN and NO 3 − + NO 2 − .A correlation plot of all parameters can be found in Online Resource 1, Fig. S1.The filtered subset contained 6 variables; TP, NH 4 + , NO 3 − + NO 2 − , SiO 2 and δ 15 N and particulate N collected on GF/F filters.All parameters were standardised to a mean of 0 and standard deviation of 1.To explain causes of variation in the matrix, the following external environmental variables were available; catchment sizes, NDVI medians of the catchments, the two runoff estimates, sample date, air temperature the last three days and distances from stream sampling points to centre of the colony.Redundancy analysis (RDA) was used to assess explanatory power of these variables.A best model was made using forward selection of a null model, with the pseudo-F value from 10 4 permutations as a selection criterion when p values < 0.05.The best environmental predictors were subsequently fitted to the PCA ordination using the "envfit" function in the "vegan" R package (Oksanen et al. 2019).

Stream water chemistry
N, P and organic C concentrations were generally much higher in seabird-impacted streams than in control streams (Fig. 2).Particulate matter in seabird-impacted streams had a mean δ 15 N value of 14.8‰ ± 3.9‰ (n = 15), while controls had a mean of 3.7‰ ± 1.2‰ (n = 11).The majority of dissolved N was present as NO 3 − + NO 2 − , with mean concentrations of 3334 µg L −1 ± 2665 µg L −1 (n = 15) in seabird-impacted streams and 90 µg L −1 ± 43 µg L −1 (n = 11) in control streams.In the seabird-impacted streams, NO 3 − + NO 2 − displayed a seasonal pattern with a clear increase from below 1700 µg N L −1 just after snowmelt in June to peak values of up to 9198 µg N L −1 at the end of July.Seasonal changes were also observed for TP concentrations, with highest values recorded in June and July.While both stream groups had TP concentrations up to 127 µg L −1 , most of the P was dissolved as PO 4 3− in seabird-impacted streams, as compared to control streams which never had PO 4 3− values above 4 µg P L −1 .SiO 2 concentrations were similar between seabird-impacted streams and control, and all streams had low concentrations when the streamflow was high in July.The C1 catchment, which includes a small, glaciated area, had particularly high turbidity linked to the high flow in July (see Online Resource 1, Table S1, for all values).
The variance in stream water chemistry is summarised in the PCA ordination diagram (Fig. 3), highlighting the differences between seabird-impacted and control stations.Median NDVI in the catchment and the estimated discharge were found to be the best environmental predictors for stream water chemistry (permutation test of RDA with 10 4 permutations, pseudo-F 1,23 = 18.4,p value < 0.0001 and pseudo-F 1,23 = 5.0, p value = 0.0029, respectively (see Online Resource 1, Fig. S2 for an RDA diagram)).The two runoff estimates based on AROME and Isfjord Radio had negligible difference as environmental predictors.

Flux estimates
Using the 2018 runoff estimate based on Isfjord Radio precipitation measurements and AROME Arctic model output, integrated summer NO 3 − + NO 2 − areal flux in the seabird-impacted streams was approximated to range from 430 to 2500 kg N km −2 , and PO 4 3− from 5.1 to 21 kg P km −2 .For control streams, the corresponding estimates were 19-31 kg N km −2 and 0.6-1.7 kg P km −2 , respectively (see Table 1 for all estimates).Except for a particularly high value for B3 in July, TOC concentrations remain stable through the season and the integrated summer flux of TOC is at similar magnitude to the NO 3 − + NO 2 − -N flux for the seabird impacted streams.This contrasts with C3 where the flux of TOC (760-1300 kg) is one order of magnitude higher than the NO 3 − + NO 2 − -N flux.The catchment areas B1-B3 represent approximately 50% of the total area around the colony with similarly high NDVI values.Furthermore, the Kittiwake guano collected had a N-content between 16 and 28% of the dry weight.By using previous studies of seabirds in Svalbard we have constructed an approximate N and P budget based on the number of seabirds breeding in the Alkhornet colony.From experiments using the doubly-labelled water method to determine field metabolic rates (FMR) and water turnover rates, previous studies have estimated food consumption and assimilation efficiency for Kittiwakes and Guillemots (Gabrielsen et al. 1987;Brekke and Gabrielsen 1994;Mehlum and Gabrielsen 1995;Gabrielsen 1996;Ellis and Gabrielsen 2001).Combining these estimates with time spent in the colony for adults and chicks (Brekke and Gabrielsen 1994;Barrett et al. 2002), a total guano deposition of 118 000 kg dry weight was estimated.The values and equations used for this calculation can be found in Online Resource 1, Table S3.In total, the Kittiwakes contribute with 37 000 kg and Brünnich's Guillemots with 81 000 kg during the breeding season.

Biota
An ornithogenic signal (higher δ 15 N) was found in all biota collected from the seabird-impacted sites (Fig. 4 and Table 2).The discrepancy in δ 15 N values is particularly high for Acrosiphonia sp.(see Table 2.) Furthermore, this macroalgae had an atomic C:N ratio that was lower at seabird-impacted sites than controls.The much higher C:N ratio and higher δ 13 C value of the Ectocarpus sp.sample at B1 is likely to reflect taxonomic differences in C:N ratios compared to Acrosiphonia sp.Epilithic algae from seabird-impacted streams B2 and B3 had δ 13 C values below − 40‰, potentially indicating a biogenic (and therefore low δ 13 C) source of CO 2 in these carbon-and nutrient-rich catchments.Amphipods δ 15 N values were higher at seabirdimpacted sites (8.4‰) than at control sites (7.0‰) (Welch's t test, t 4.49 = 10.23,p = 0.0003), while δ 13 C values did not differ significantly (see Online Resource 1, Table S4, for all values).

Discussion
Seabirds play an important role in Arctic nutrient cycling by creating nutrient hotspots around their colonies.While previous research has largely focused on seabird-mediated nutrient transport from sea to land and the impact on terrestrial ecosystems, we show that seabird colonies near the shores can also act as point sources of nutrients to adjacent coastal ecosystems.Despite the presence of a large terrestrial sink in the study site, our results demonstrate substantial seabird-derived nutrient fluxes back to coastal waters, and use of ornithogenic nitrogen in littoral and nearshore coastal organisms.

Stream geochemistry and main drivers
Our first hypothesis proposes elevated nutrient concentrations in streams near the seabird colony.This is well supported by our results, where nutrient concentrations, and especially DIN (dissolved inorganic nitrogen) and DIP (dissolved inorganic phosphorous), were higher in seabirdimpacted streams than control streams, and differences between study streams were best explained by the median NDVI values in the catchments (Fig. 3).Maximum NDVIvalues in the Isfjorden-area as a whole have increased by 0.015 between 2009 and 2015, which is in accordance with the long term trend  of an NDVI increase of 0.0017 year −1 (Vickers et al. 2016), thus we assume the NDVI-change to be negligible between 2009, when the aerial orthophoto was taken, to 2018 when these samples were collected.We also hypothesized that there would be a distinct seasonality in stream nutrient concentrations linked to seabird presence.While we did observe a high degree of seasonality in nutrient concentrations across all streams, much of this seasonality was linked to fluctuations in runoff rather than the timing of seabird presence in the colony. DN concentrations in our control streams were similar to those reported by Tye and Heaton (2007) for small non-glaciated catchments in Svalbard, suggesting that the concentrations observed in these streams are within the expected range for non-seabird impacted Svalbard streams.Furthermore, Tye and Heaton (2007) report δ 15 N values of NO 3 − in their runoffs to lie between − 1 and 5‰, which is in congruence with the δ 15 N values of epilithic stream algae sampled in our + to the streams early in the melt season, as seen in the vicinity of penguin rookeries in Antarctica (Wodehouse and Parker 2013).As the thaw depth increases in July, NO 3 − in soil becomes mobilised which can explain the high values seen here (Yano et al. 2010).Furthermore, we can expect microbial activity to be much greater in the productive and soil-rich surroundings of the bird cliff, which can lead to oxidization of NH 4 + (Banerjee and Siciliano 2012) or assimilation by vegetation (Yano et al. 2010).Highly different TOC levels between seabird-impacted and control sites are also evident.The elevated TOC concentrations in the seabird impacted streams likely reflects the deep organic-rich soils and abundant vegetation present in the guano-impacted catchments, with the highest concentrations observed early in the season when the flow pathways are shallow and through organic-rich layers (McGovern et al. 2020).
In the present study, Kittiwake guano had δ 15 N values higher than 10‰, as expected from previous work (Wainright et al. 1998).We also expect guano from Brünnich's Guillemots to have similar δ 15 N values, as Guillemots occupy the same trophic position as Kittiwakes in the Barents Sea (Hop et al. 2002).However, the extreme δ 15 N values in POM and epilithic algae from the seabird-impacted streams reached 20‰.This suggests that further fractioning has occurred in the deposited N. Isotopic fractionation during NH 3 volatilisation from the guano will enrich 15 N content of residual nitrogen (Frank et al. 2004).The nitrogen isotopes can be further fractionated in microbial processes e.g.nitrification and denitrification, and after assimilation by autotrophs.Previous work on nitrogen flux through Arctic tundra has shown how efficiently nitrogen is assimilated by vegetation (Odasz 1994;Choudhary et al. 2016), and that most of the N losses through leaching are due to decomposition of organic matter (Yano et al. 2010).The role of soil as a transition step between seabird deposition and release to aquatic ecosystems have been observed in other systems as well (Harding et al. 2009).Although the δ 15 N values in the streams were taken from particulate matter, likely reflecting eroded N-enriched catchment soils, the δ 15 N of dissolved nitrogen in the seabird-impacted streams is likely similarly high, as evidenced by the extremely high δ 15 N in epilithic stream algae relative to the δ 15 N values for guano.
Although TP-concentrations were positively associated with streamflow across all study sites, the main drivers for seasonality in observed concentrations are likely different between controls and seabird-impacted catchments.In control streams, TP values were strongly correlated with turbidity, with highest concentrations in C3, where glacial influence and a large catchment area led to elevated turbidity and transport of glacial silt throughout the melt season.Although TP-concentrations were high in this stream, PO 4 3− was only a minor contributor to TP concentrations, likely due to a strong contribution of particulate P, which is much less available for uptake by primary producers (Reynolds and Davies 2001).In contrast, PO 4 3− was the dominant form of P in seabird-impacted streams, suggesting an important source of bioavailable phosphorus to the coastal environment.

Fluxes and terrestrial retention
Marine-derived nutrients deposited in and around the bird cliff represent a strong point source of nutrients to terrestrial and coastal ecosystems, and originate primarily from the Fig. 4 δ 15 N and δ 13 C biplot of collected amphipods, macroalgae and Kittiwake guano.Amphipod and guano δ 13 C values are from acidified samples, while values for algae are from unacidified samples.All δ 15 N values come from unacidified samples.A discrepancy in δ 15 N values were found between bird and control stations for all biota, with the algae samples displaying an extreme difference.For both the marine and stream algae, station B1 had the lowest δ 15 N value of the bird stations.The terrestrial signal of stream algae is evident as much lower δ 13 C values than the other samples pelagic zone where the surface-feeding Black-legged Kittiwakes and the diving Brünnich's Guillemots hunt.This difference in foraging behaviour impacts their choice of foraging grounds, e.g. with Kittiwakes tending to be more attracted to marine-terminating glacier fronts than Guillemots (Urbanski et al. 2017;Nishizawa et al. 2020).Both species will forage relatively nearby their colonies during the breeding season, commonly within a radius of 20 km from the colony and rarely leaving on foraging trips more than 50 km away from the colony (Mehlum et al. 2001;Falk et al. 2002;Bertrand et al. 2021).The diet of Kittiwakes in Svalbard is dominated by fish, with polar cod (Boreogadus saida) being an important food item (Lønne and Gabrielsen 1992), along with crustaceans and other marine invertebrates (Vihtakari et al. 2018).Brünnich's Guillemots prey on fish, with pelagic crustaceans also making up a large portion of their diet (Lønne and Gabrielsen 1992;Anker-Nilssen et al. 2000).
The composition of the guano varies with prey item and quality.The N-content of the Kittiwake guano samples analysed in the present study was in congruence with Brekke and Gabrielsen (1994), who report a mean value of 21% for Svalbard Kittiwakes and Brünnich's Guillemots.Meanwhile, phosphorus concentrations have previously been estimated as approximately 40% of the N content, i.e. 8.8% of guano dry weight (Sakshaug et al. 1994).This translates to an annual seabird-driven flux to Alkhornet of 25 000 kg N and 10 000 kg P, of which 67% originates from the Guillemot colony and 33% from the Kittiwakes.The larger body mass and high wing loading of guillemots contributing to a higher energy expenditure (Gabrielsen 1996) compared to Kittiwakes (Gabrielsen et al. 1987) explains the high guano contribution from this species, despite earlier departure from the colony by Guillemot chicks (fledging at about 25 days of age) and similar number of breeding pairs at Alkhornet (Brekke and Gabrielsen 1994).Arctic ecosystems are typically oligotrophic, with a mean annual wet deposition of only 74 kg N km −2 over the past 20 years in Ny-Ålesund, Svalbard, well below deposition rates in more densely populated areas in Europe (Kühnel et al. 2011).The difference between our estimated fluxes and background N deposition rates in Svalbard highlight the potential role of seabird colonies as important nutrient sources to the terrestrial environment.
Based on our estimated nutrient and organic carbon fluxes to the coast, areal fluxes of DIN, DIP and TOC from the seabird-impacted catchment areas were much higher than the controls.While TOC fluxes from seabird streams were elevated relative to fluxes for control streams, the effect of the seabird colony was much more pronounced for dissolved inorganic N and P, highlighting the potential role of these streams as a high-quality nutrient source for downstream ecosystems.
When comparing the estimated fluxes of DIN and DIP with the estimated N and P input from the seabird colony to the terrestrial environment, we see that the majority of estimated nutrient inputs were retained on land (or potentially evaded).The seabird-fertilized, deep peat at Alkhornet is likely to act as a large nutrient sink.Moss communities have been shown to retain virtually all added N in oligotrophic Arctic tundra (Yano et al. 2010).Considering the area of vegetation clearly influenced by seabird fertilization is about twice as large as the B1-B3 catchments, the runoff flux of NO 3 − + NO 2 − -N to the coast represents about 4-8% of annual guano-deposition N. Correspondingly, the flux of NH 4 + -N represents 0.05-0.09% of guano-deposited N, and the flux of PO 4 3− -P represents 0.1-0.2% of guano-deposited P.
Both concentrations and areal fluxes of DIN and DIP in the control catchment areas are in accordance with other fluvial flux estimates reported from the Arctic (Tye and Heaton 2007;Holmes et al. 2012;Blaen et al. 2014;Wadham et al. 2016).However, the numbers presented in the present study should be interpreted with caution since the stream flow was not measured directly.In our coarse estimates of runoff, variation between catchment areas, such as differences in water holding capacity, evapotranspiration and the rate of snowmelt were not accounted for.Additionally, the fraction of glacial ice in catchment C3 is likely to sustain elevated discharge through the melt season.However, with only five sampling dates, the glacial influence is unlikely to represent a main source of uncertainty for these flux estimates.Although highly uncertain, our goal was to generate preliminary estimates to give insight into the potential scale of the seabird effect on nutrient fluxes.

Coastal impact
Our third hypothesis proposed that impact of seabirds would be detectable in the nearby coastal ecosystem.The Acrosiphonia sp.samples show clear signs of utilising seabirdderived N, with high δ 15 N values and low C:N molar ratios at seabird-impacted sites compared to control sites.Many macroalgal species, Acrosiphonia sp.included, have adapted to the seasonal nutrient availability in the Arctic by endogenic storage of NO 3 − when it is abundant in winter and spring.As the waters become NO 3 − depleted over summer, this reserve is used to maintain photosynthetic activity and a constant tissue C:N ratio (Hodal et al. 2012).Thus, the added NO 3 − flux from the seabird colony might have a larger influence on microalgae which have higher nutrient turnover.Nonetheless, the extremely high δ 15 N values from Acrosiphina sp.close to Alkhornet and the lower C:N ratios found at the control sites implies a beneficial effect of the seabird colony on this species.Zmudczyńska-Skarbek et al. (2015) found no ornithogenic signal in the coastal benthic food web close to another mixed seabird colony in Isfjorden.However, their sampling was done 50 m offshore at depths from 5 to 15 m, where the seabird impact can be expected to be more diluted than in the intertidal zone and closer to shore.Furthermore, brown algae such as the Saccharina latissima sampled in their study are often more reliant on internally stored nutrients than nitrophilic green macroalgae such as Acrosiphonia sp.(Hurd et al. 2014), thus showing a limited response to nutrient enrichment in summer (Gordillo et al. 2006).Still, the lack of a clear signal of seabird-derived nutrients in this earlier study (Zmudczyńska-Skarbek et al. 2015) points towards a limited spatial extent of this nutrient subsidy.
The δ 15 N values of the amphipods also differed between seabird-impacted stations and controls, but were within the range of values reported for Gammarus setosus collected from other areas in Isfjorden during summer 2018 (Skogsberg 2019).The Gammarus setosus and Gammarellidae amphipods collected in the present study can feed opportunistically and on a range of food types (Legeżyńska et al. 2012), which, paired with seasonal changes in diet and metabolism, can complicate interpretation of stable isotope values.Different strategies such as phytodetrivory, scavenging and deposit feeding will influence their trophic position and exposure to ornithogenic influence, and the observed variation in δ 13 C supports such high diversity in feeding strategies.The low variance within the values from both seabird-impacted and control stations, combined with the apparent similarity of habitat, give support to a possible seabird signal in amphipods from influenced sites.

Concluding remarks
The stream catchments investigated in the present study are small compared to the main river systems on Svalbard, and this will generally be the case for other Arctic seabird colonies as well.However, the extremely nutrient rich water draining from these seabird colonies suggests that these colonies can still be important point sources of DIN and DIP to adjacent coastal waters.At the same time, these fluxes are not constant.The size of the breeding seabird population will vary between years.Several seabird species in Svalbard have experienced population decline in recent years (Anker-Nilssen et al. 2015, 2018;CAFF 2017).Climate change in Svalbard, with longer and wetter summers, is likely to have important implications for both seabird-driven fluxes as well as other terrestrial impacts on marine systems (Bilt et al. 2019;McGovern et al. 2020).Increased terrestrial production has the potential to assimilate more of the nutrient input from seabird colonies, while increased microbial activity and runoff, including increased frequency of large rainfall events even during winter (Hansen et al. 2014), could enhance fluxes to the coast.While the present study focuses on fluvial transport of ornithogenic nutrients to the coast, feathers and droppings from overflying seabirds and other bird activity in the coastal environment constitutes a direct pathway of coastal nutrient enrichment.However, we interpret this direct effect to be of minor importance for this site.The tidal zone in front of Alkhornet is narrow and not heavily used by Guillemots and Kittiwakes.The extremely high δ 15 N values in the collected marine algae (and epilithic stream algae) relative to guano also indicates that further fractionation of ornithogenic N has taken place, most likely in the terrestrial environment (Hayashi et al. 2018).
There is also a need to further investigate the spatial extent of influence from seabird-derived nitrogen on marine biota, as well as the factors governing the distribution and fate of these nutrients in the marine environment.The extent of this effect will have ecological implications, not only for coastal species but also for species using the coastal environment for part of their life cycle.Still, the work by Zmudczyńska-Skarbek and Balazy (2017) points towards a rather localized effect in most systems.Nonetheless, with 3 million breeding pairs of seabirds in Svalbard (Anker-Nilssen et al. 2015), their combined impact might still be substantial.To generalise our findings to other seabird colonies, there is a need to consider geophysical and geochemical characteristics of the sites.The main colony on Alkhornet is situated approximately 700 m from the coastline, thus reducing direct guano inputs from the colony to coastal waters and leading to increased importance of terrestrial retention and transformation for determining nutrient fluxes from the colony to adjacent coastal ecosystems.Other seabird colonies on Svalbard are located on steep cliffs directly adjacent to coastal waters, where direct nutrient inputs to coastal waters are likely to be larger, and more closely linked to seasonality in bird presence at the cliff.Furthermore, Alkhornet is an exposed site, where marine currents, swells and waves mix the water column efficiently and we would expect rapid dilution and offshore transport of nutrients delivered from land.We would therefore expect to see stronger nutrient retention in more sheltered sites.
This study adds to a new appreciation for the importance of nutrient fluxes from land to sea in the Arctic and their importance for marine primary production (Terhaar et al. 2021).We show that colonial seabirds, particularly larger species, can create coastal nutrient hotspots with ecological effects that may propagate through the food web.However, with declining seabird populations, climate change and other changes in coastal and terrestrial ecosystems, these complex and dynamic nutrient fluxes are also changing (Hansen et al. 2019; Plazas-Jiménez and Cianciaruso 2020).There is a need for further studies to unravel the role of the terrestrial system in the release of seabird-derived nutrients to the coast, and the potential impacts of these nutrient inputs for coastal ecosystem structure and function.There is also a need for a better understanding of the seasonality of these nutrient fluxes, including the effect of variable seabird presence in colonies and runoff phenology, as well as how the timing of these inputs aligns with key ecological processes in the marine environment (e.g.phytoplankton bloom timing and periods of nutrient limitation).This is particularly relevant in the Arctic, where seabird colonies are common along the coast, and may provide an important 'pulse' of nutrients to nutrient-limited systems during the brief summer growth season.

Fig. 3
Fig. 3 PCA ordination plot of SiO 2 , NO 3 − + NO 2 − -N, NH 4 + -N, TP, δ 15 N amount of particulate carbon nitrogen (PartN) on GF/F filters.The notation of the sample stations corresponds to the name and sampling date (roman numbers from I to V).The red being bird stations B1-B3 and blue being controls C2-C3.C1 is included but does not have a roman number as it was only sampled once.NDVI median of catchment basins and a runoff estimate based on Isfjord Radio weather data were fitted as environmental predictors (green arrows)