Permafrost degradation and nitrogen cycling in Arctic rivers: Insights from 1 stable nitrogen isotope studies 2

12 Across the Arctic, vast areas of permafrost are being degraded by climate change, which has the 13 potential to release substantial quantities of nutrients, including nitrogen into large Arctic rivers. 14 These rivers heavily influence the biogeochemistry of the Arctic Ocean, so it is important to 15 understand the potential changes to rivers from permafrost degradation. This study utilised 16 dissolved nitrogen species (nitrate and dissolved organic nitrogen (DON)) along with nitrogen 17 isotope values ( δ 15 N-NO 3-and δ 15 N-DON) of samples collected from permafrost sites in the 18 Kolyma River and the six largest Arctic rivers. Large inputs of DON and nitrate with a unique 19 isotopically heavy δ 15 N signature were documented in the Kolyma, suggesting the occurrence of 20 denitrification and highly invigorated nitrogen cycling in the Yedoma permafrost thaw zones along 21 the Kolyma. We show evidence for permafrost derived DON being recycled to nitrate as it passes 22 through the river, transferring the high 15 N signature to nitrate. However, the potential to observe 23 these thaw signals at the mouths of rivers depends on the spatial scale of thaw sites, permafrost 24 degradation and recycling mechanisms. In contrast with the Kolyma, with near 100% continuous 25 permafrost extent, the Ob’ River, draining large areas of discontinuous and sporadic permafrost, 26 shows large seasonal changes in both nitrate and DON isotopic signatures

through the river, transferring the high 15 N signature to nitrate. However, the potential to observe 23 these thaw signals at the mouths of rivers depends on the spatial scale of thaw sites, permafrost 24 degradation and recycling mechanisms. In contrast with the Kolyma, with near 100% continuous 25 permafrost extent, the Ob' River, draining large areas of discontinuous and sporadic permafrost, 26 shows large seasonal changes in both nitrate and DON isotopic signatures. During winter months, 27 water percolating through peat soils records isotopically heavy denitrification signals in contrast 28 with the lighter summer values when surface flow dominates. This early year denitrification signal 29 was present to a degree in the Kolyma but the ability to relate seasonal nitrogen signals across 30 Arctic Rivers to permafrost degradation could not be shown with this study. Other large rivers in 31 the Arctic show different seasonal nitrogen trends. Based on nitrogen isotope values, the vast 32 majority of nitrogen fluxes in the Arctic rivers is from fresh DON sourced from surface runoff 33 through organic-rich top-soil and not from permafrost degradation. However, with future 34 permafrost thaw, other Arctic rivers may begin to show nitrogen trends similar to the Ob'. Our 35 study demonstrates that nitrogen inputs from permafrost thaw can be identified through nitrogen 36 isotopes, but only on small spatial scales. Overall, nitrogen isotopes show potential for revealing 37 integrated catchment wide nitrogen cycling processes. 38 39 1 Introduction 40 The Arctic Ocean contains ~1% of global ocean volume but receives greater than 10% of the total 41 global riverine discharge (Frey and McClelland, 2009). This disproportionate influence of rivers 42 means that any changes in riverine inputs will likely have significant implications on marine 43 chemical, physical and biological processes . River biogeochemistry and 44 discharge also integrate catchment wide processes, making them potentially sensitive indicators of 45 change to the terrestrial environment (Holmes et al., 2000). With diminishing sea ice and opening 46 of surface waters to light, Arctic productivity is sensitive to riverine nutrient inputs and particularly 47 nitrogen which is the limiting nutrient in coastal areas (Thibodeau et al., 2017). 48 Biologically available nitrogen can exist as dissolved inorganic nitrogen (DIN) in forms of nitrate, 49 nitrite and ammonium. DIN is calculated as the sum of these three forms (DIN = NO3 -+ NO2 -+ 50 NH4 + ) (McCrackin et al., 2014) and can be taken up by primary producers (Tank et al., 2012). 51 Nitrite and ammonium are highly biologically labile and so only persist for a short time before 52 being converted into nitrate or assimilated. Nitrogen can also exist as dissolved organic nitrogen 53 (DON) but these forms generally need to be broken down (remineralised) into DIN before uptake 54 can occur (Tank et al., 2012). DON is calculated as the difference between total dissolved nitrogen 55 (TDN) and DIN: (DON = TDN -DIN) (Frey et al., 2007). Nitrate is expected to be the dominant 56 species so a simplification can be made to DON = TDN -NO3 -. As part of the nitrogen cycle, 57 exchange between these pools occurs in riverine and coastal areas depending on environmental 58 conditions. In oxic conditions, assimilation and nitrification occur, while denitrification can be 59 dominant in anoxic conditions (Voigt et al., 2017). 60 Extensive areas of permafrost influence most of the riverine inputs to the Arctic Ocean. Permafrost 61 is defined as 'any subsurface material that remains below 0 o C for at least two consecutive years' 62 (Van Everdingen, 1998). It is defined exclusively on the basis of temperature, not whether ice is 63 present. Permafrost can stabilise ancient soils, preventing breakdown of soil organic matter and is 64 classified based on its spatial extent and thickness. Continuous permafrost has 90-100% aerial 65 extent and is 100-800m thick, while discontinuous has 50-90% extent and is 25-100m thick 66 (Anisimov and Reneva, 2006). 67 Permafrost undergoes degradation through different mechanisms. The most common is active layer 68 deepening, where the top layer of soil that degrades and refreezes each year becomes deeper due to 69 increased summer temperatures and the influx of precipitation (Nelson et al., 1997). This increases 70 the depth of permafrost, allowing the active layer to penetrate previously frozen soil. Permafrost 71 can also degrade through riverbank or coastal erosion, cutting through deep horizons of permafrost 72 promoting rapid and often catastrophic degradation (Streletskiy et al., 2015). These mechanisms all 73 lead to increases in soil microbial activity that release dissolved nitrogen from previously frozen 74 organic matter (Beermann et al., 2017). The proportion of the released nitrogen species vary 75 depending on the degree and mechanism of degradation. 76 Climate change is causing annual surface air temperatures within the Arctic to increase at almost 77 twice the rate of the global average (Hassol, 2004). In 2010, air temperatures in the Arctic were 4 o C 78 warmer than the reference period of 1968-1996(NOAA, 2014. A further 4 to 7 o C increase is 79 expected by the end of the century (Hassol, 2004). These dramatic temperature changes will result 80 in the Arctic experiencing unprecedented impacts on its environments. Over the whole pan-Arctic 81 watershed, river discharge is increasing by an estimated 5.6km 3 y -1 each year based on observations 82 from 1964-2000(McClelland et al., 2006. Some recent studies have revealed even greater rates 83 occurring and predicted into the future but some uncertainty exists due to substantial variation 84 across basins and permafrost regimes (Feng et al., 2021). Discharge has already increased by ~10% 85 in Russian rivers compared to this reference period (Peterson et al., 2002). Permafrost is at high 86 risk of degradation with climate change with estimates that 10% of permafrost in the northern 87 hemisphere has disappeared in the last 100 years (NSIDC, 2018 In contrast, where shallow peat exists, warming and underlying permafrost degradation can cause 110 the active layer to deepen into mineral horizons with low C:N ratios. This can lead to flow paths of 111 groundwater being directed through these mineral horizons leading to an increased adsorption of 112 DON and release of nitrate through subsequent mineralization and nitrification (Harms, 2013). This 113 process can occur to a lesser extent on a seasonal cycle with groundwater influx from mineral 114 horizons in the winter and surface runoff from organic horizons in spring and summer. Extensive 115 future permafrost degradation in catchments with active layer deepening occurring is expected to 116 increase the seasonal groundwater contribution leading to decreased DON concentrations and 117 increased nitrate concentrations in streams and rivers (Walvoord and Striegl, 2007). 118 These studies focus on gradual active layer deepening processes. Other more rapid permafrost 119 degradation processes such as riverine and coastal erosion are more spatially limited but could be 120 responsible for moving nitrogen species rapidly and directly from terrestrial permafrost to riverine 121 or coastal environments (Berhe et al., 2007). This mechanism is understudied so the resulting 122 nitrogen export is still relatively unknown. 123 The processing and cycling of nitrogen that occurs in-stream and in near-shore coastal areas after 124 release from permafrost is also largely unknown. DON represents a 5x greater influx to Arctic shelf 125 waters from rivers than nitrate across the whole Arctic but 70% of the DON is removed in shelf 126 waters before reaching the open marine environment (Thibodeau et al., 2017 degradation mechanisms, where the peat depth determines which layer is exposed after degradation 192 (Harms, 2013). Radiocarbon dating of DOC from a fluid mud stream draining from the degrading 193 permafrost yielded an age of 20,000 years at this site. This organic matter is highly biolabile after 194 thawing occurs and can be assimilated rapidly by aquatic microorganisms after mineralisation 195 ( and Mackenzie (Figure 1(b)). Together, the proportion of continuous and discontinuous permafrost 214 within these catchments is 48%, similar to the proportions across the whole pan-Arctic catchments 215 (52%) (Tank et al., 2012). Thus, these rivers represent overall pan-Arctic conditions. These 216 catchments also cover transitions from continuous permafrost zones of the Arctic to permafrost 217 free, capturing the variability that occurs across the pan-Arctic (    observed between discharge normalised DON concentrations and permafrost extent in any of the 283 plots (Figure 2 (b)). Given that DON is the dominant form of nitrogen released from soil, the 284 increase in nitrate concentrations but not DON suggests that cycling of organic nitrogen to inorganic 285 forms in soils and/or upstream rivers may be promoted with decreasing permafrost extents. Figure  286 2 displays the variability from the extent of continuous permafrost, but not from active permafrost 287 degradation. Local scale measurements of degradation sites from the Kolyma River were used to 288 address if active permafrost degradation releases nitrogen and identify cycling processes involved. 289 Seasonal trends were also used to see when each of the species become dominant and to help 290 determine catchment-scale processes. 291  concentrations. This observation that more permafrost leads to less release of nitrate supports 313   The signals observed in the permafrost degradation site were rapidly lost in the main stem of the 333 river and into the estuary. In the river, δ 15 N-NO3was ~5‰ and δ 15 N-DON was 2 to 5‰, while 334 for δ 15 N moving downstream (similar to the concentration data) and into the estuary, suggesting 337 minimal alterations to dominant processing cycles of nitrogen in the main river stem. 338 In summary, a unique signal representing inputs from degrading Yedoma permafrost was detected 339 using concentrations and isotopic signatures of nitrogen species (Figure 3 and Figure 4). From the 340 site at Duvannyi Yar, extensive permafrost degradation brings water to the Kolyma River with very 341 high DON concentrations (272µM) and high δ 15 N-DON (6.7‰). In addition, nitrate concentrations 342 were high (40µM) with very high δ 15 N-NO3 -(11.5 ± 0.26‰) but very low δ 18 O-NO3 -(-19.2 ± 343 0.37‰). 344

Explanation of signals observed and likely processing 345
During degradation, permafrost releases large amounts of organic matter and organic nitrogen 346 (DON) from the soil and ice (272µM in this study, Figure 5). This undergoes rapid mineralization, 347 firstly to highly reactive ammonium then to nitrate via nitrification (Voigt et al., 2017  Additionally, the fact that DON is not isotopically heavier than nitrate is also expected as the 364 primary source of DON is from decaying organic matter preserved in the permafrost and this 365 process releases organic matter with a low δ 15 N to start with that forms DON also with a lower δ 15 N 366 (Sipler and Bronk, 2015). This is supplemented with a smaller contribution from DON formed from 367 the recycled heavy nitrate. Therefore, nitrogen processing in the permafrost degradation zone not 368 only involves active release of DON by heterotrophic remineralisation with anaerobic processes 369 such as denitrification, but also exchange between nitrogen pools. Oxygen isotopes of nitrate 370 provide further evidence for this recycling. 371 372 During denitrification, fractionation of nitrogen and oxygen is 1:1; therefore, oxygen isotopes 373 should behave similarly to nitrogen isotopes and become isotopically heavy (high signal) in the Yedoma permafrost degradation enter the main stem of the river. Processing in the main stem can 394 then alter these signals, explaining the signal observed at the river mouth. 395 Organic matter from this permafrost site is very biolabile and ancient permafrost DOC is the most 396 biolabile source of DOC in riverine Arctic systems due to lack of processing and survival of bacteria 397 (Vonk et al., 2013). Up to 50% of permafrost DOC can be lost in less than seven days in the Kolyma 398 River (Spencer et al., 2015). This time period is equal to water residence times between headwater 399 streams in degradation zones and the river mouth (3 days) ( (Figure 2 and Figure 3). From the permafrost zone to the main stem, DON 419 concentrations decreased and nitrate concentrations increased slightly (even with dilution effects). 420 The isotopic signature of the main stem nitrate was also significantly heavier than that of permafrost 421 and main stem DON. The concentration trends suggest that recycling of DON to nitrate was 422 occurring in the river and, when combined with the isotopic trends, some of the isotopically heavy 423 nitrogen from DON and nitrate originating in the degradation site may have contributed to these 424 recycling processes. This allows the heavy nitrogen signal from the degrading Yedoma permafrost 425 to be transferred and retained in the main stem nitrate. This was assisted by the negligible diluting 426 nitrate inputs from much of the permafrost-covered catchment. Importantly, this also suggests that 427 a significant nitrate pool in the main stem is produced from DON recycling rather than from direct 428 nitrate inputs to the river. 429 However, δ 18 O-NO3values in the main stem were higher than the degradation site (-3.4 to -4.7‰). 430 These values were much greater than would be expected from nitrogen recycling and nitrate/DON 431 exchange occurring in the main stem. Determining the cause for these high signals is difficult due 432 to a multitude of possible factors influencing the isotopic signatures. Co-occurrence of partial 433 nitrate uptake and nitrification in the main stem which decouples the nitrogen and oxygen isotopes 434 (Sigman et al., 2009)  A similar mechanism may also operate for DON where the sources in spring and summer are not 475 derived as strongly from permafrost but from surface soils that experience minimal nitrogen 476 processing. Therefore, permafrost signals and associated processes should be considered on spatial 477 (inter-catchment) and temporal scales. 478 479 and Yenisey show nitrate concentrations decreasing to almost zero in the summer months during 493 peak discharge. 494

Seasonal nitrogen species trends in rivers
In general, Figure 6 confirms that DON is the dominant form of nitrogen released from these soils 495 and transported in these rivers, due to its high concentration during the high discharge periods of 496 the spring freshet. This DON source is likely derived from surface runoff through organic rich top 497 soil (Harms, 2013). Following the local scale Kolyma section, seasonal stable isotopes trends are 498 used next to detect (1) permafrost degradation signals and (2) any in-stream processing of nitrogen. 499 500 spring and summer when modern DOC sources dominate export. A similar mechanism may also 512 operate for DON where the sources in spring and summer are not derived as strongly from 513 permafrost but from surface soils that experience minimum nitrogen processing (Harms, 2013). 514 δ 15 N-NO3and δ 15 N-DON values of the Kolyma and Ob' in late winter and early spring are high 515 before becoming lower in spring/summer and returning to high values at the end of the year ( Figure  516 7) (this is also seen in the Yukon to a lesser extent). It is notable that the Kolyma and Ob' have the 517 highest and lowest continuous permafrost extent respectively among the large Arctic rivers. We 518 evaluate the seasonal trend further in the Ob' River (with comparison to the Kolyma), which has 519 the largest seasonal isotopic shift out of all the rivers (Figure 7). 520 The Ob' has the greatest seasonal isotopic shifts with very heavy winter δ 15 N-NO3values of 12 to 523 14‰ occurring over winter and early spring but decreasing to 2‰ in summer and a change from 8 524 to 2.5‰ for δ 15 N-DON. The δ 18 O-NO3trend for the Ob' River follows a similar pattern to the δ 15 N-525 NO3 -(i.e. they are coupled). However, for the Kolyma, the two isotopes are decoupled and show 526 strong opposing trends, though this trend could be influenced by the anomalously high δ 18 O-NO3 -527 value in June and may not represent true conditions. 528 The peak δ 15 N-NO3values in the Ob' river are similar to the signal for denitrification in high-529 latitude permafrost regions (Harms, 2013)  can be adsorbed and mineralised to nitrate (Harms, 2013). Denitrification of this remineralised 540 nitrate due to the waterlogging of the soil in these large wetlands would also lead to the high isotopic 541 signatures observed. It is important to note that these denitrification processes occur without 542 permafrost degradation influence in the Ob' whereas the denitrification signal observed in the 543 Kolyma Yedoma degradation site was likely due to the permafrost degradation. Denitrification 544 signals are much more influential in the Ob' than the Kolyma where the permafrost extent is very 545 low. The high nitrate concentrations show that a substantial amount of denitrified nitrate is added 546 to the rivers and the Ob' River is displaying a source-dominated signal, with instream processes 547 possibly less influential. 548 The coupling of δ 15 N-NO3and δ 15 N-DON throughout the year suggests the same source for both 549 nitrogen species. However, some DON may also be oxidised into nitrate in the main stem and allow 550 the heavy δ 15 N signal to be transferred from the DON to the nitrate. This would also reduce the 551 DON concentrations as observed. 552 553 The observed variability of nitrate isotopes in the Ob' River can be approximated to changes 554 between two dominant sources as outlined in Figure 8. recycling could explain the decoupling throughout the year. This decoupling also suggests that 583 nitrate uptake is low and the small contribution of nitrate due to the high continuous permafrost 584 extent is likely to drive nitrate limitation in this river, despite DON remineralisation (Figure 2). 585 This similar but suppressed trend suggests that the denitrification signal is less influential and was 586 diluted, similar to local-scale observations (section 3.2). The greater coverage of permafrost in the 587 Kolyma catchment compared to the Ob' may reduce the seasonal change in nitrogen species signals, 588 especially nitrate (as observed in Figure 2) by restricting flow-paths to minimal contact with mineral 589 horizons and reducing groundwater flow. This can also explain the observed mixing line and the 590 surface source dominance throughout the year shown in Figure 8. 591 592 The local scale permafrost degradation signals observed from the Yedoma permafrost degradation 593 in the Kolyma may be visible in the seasonal trends due to similar main stem DON and nitrate 594 signals in the early season, possibly assisted by the lack of other nitrate inputs and DON recycling 595 to nitrate. However, it is not possible to observe any permafrost degradation signals in the Ob' 596 catchment or to compare trends with previous local scale findings due to the dominance of the 597 groundwater derived denitrification signal and different catchment conditions. 598

Explanations for times series trends in other Arctic rivers 599
The Mackenzie and Yukon show δ 15 N-NO3trends peaking in the summer months (Figure 7). This 600 was an opposite trend to the nitrate concentration, and more closely follows the discharge trends. 601 The Yukon had the most prolonged δ 15 N-NO3peak out of all the rivers. influenced mainly by instream processes (Harms, 2013) due to assimilation or uptake of nitrate by 608 phytoplankton in summer. The smaller isotopic shift between seasons could also signify 609 assimilation rather than denitrification (Struck, 2012). This process would be assisted by the large 610 area of lakes in the Mackenzie catchment where water residence times are increased allowing 611 extensive primary productivity (Janjua and Tallman, 2015). 612

Surface sources
The Yukon followed similar trends to the Mackenzie (for δ 15  Irrespective of N species released and the degradation mechanism, nitrogen fluxes are likely to 648 increase with permafrost degradation causing significant impact to the coastal zones. Any increases 649 in nitrogen loading to coastal Arctic areas will have large impacts on productivity since these zones 650 are heavily nitrogen limited (Thibodeau et al., 2017). Currently, productivity peaks over a short 651 period in summer when light is not limiting. However, permafrost degradation and greater nitrogen 652 fluxes may increase the magnitude of these productivity peaks inducing possible algal blooms. Yet, 653 light limitation will still control productivity later in the year. Overall, the cycling of these nitrogen 654 species in coastal zones is essential to understand further to make robust predictions of future 655 change. 656

Conclusions 657
Overall, catchment permafrost coverage seems to control main stem nitrate concentrations but not 658 DON, with large extents of continuous permafrost leading to low concentrations of nitrate in Arctic 659 rivers. In local Kolyma degradation sites, Yedoma permafrost degradation was characterised by 660 high DON and nitrate concentrations, high δ 15 N-DON and δ 15 N-NO3and very low δ 18 O-NO3 -.

661
These signatures indicate rapid recycling and exchange between nitrogen pools resulting in the 662 entire system becoming isotopically heavy for nitrogen. Upon release to the main river stem, this 663 signature is greatly diluted but evidence for recycling of degradation derived DON to nitrate, 664 transferring the heavy isotopic signature to nitrate, was observed. This DON recycling could be the 665 main source of nitrate in catchments with extensive permafrost coverage and few nitrate inputs. 666 However, these input signals from Yedoma degradation are unlikely to be observed strongly at the 667 river mouth unless degradation zones are more spatially extensive. 668 δ 15 N of nitrate, TDN and DON during summer and spring freshets generally exhibit values around 669 2 to 4‰, DON dominates the nitrogen export within these rivers, in the form of fresh DON derived 670 from surface runoff through modern, organic rich topsoil. However, Arctic rivers all have different 671 nitrogen dynamics based on their catchment characteristics. The Ob' catchment, with its lowest 672 extent of permafrost coverage and extensive peatland area demonstrates a strong denitrification 673 signal, however this cannot be linked to the degradation induced denitrification signal observed in 674 the Kolyma. The Ob' isotopic signal is strongly seasonal and influenced by the changing soil flow 675 paths that arise throughout the year. The Kolyma had a similar seasonal trend but with reduced 676 magnitude and showed evidence of differing processes occurring compared to the Ob' but were 677 similar to local scale observations. A diluted denitrification signal, DON recycling to nitrate and 678 low nitrate uptake were all possibly assisted by the lack of other nitrate inputs and high permafrost 679 coverage. In other Arctic river catchments, different factors can mask any fresh permafrost 680 degradation signals. Lacustrine nitrogen assimilation and uptake are dominant in the Mackenzie 681 and seasonal changes in water sources are important for the Yukon catchment while large freshet 682 discharges in the Lena and Yenisey likely inundate the catchments with runoff-derived nitrogen. 683 It is possible that with future decreases in catchment permafrost coverage, seasonal nitrogen 684 dynamics in Arctic rivers could begin to resemble that of the Ob' catchment. In general, increased 685 fluxes of nitrogen are expected as a result of degradation which would have impacts on coastal 686 environments and ecosystems, as well as in rivers with nitrogen limitation. However, the extent of 687 this is unclear at present. Further studies are required to explore more local scale and coastal 688 nitrogen cycling and the impacts of permafrost degradation on riverine and coastal environments. 689 This study shows how nitrogen isotopes can be used to integrate catchment wide processes in Arctic 690 rivers as well as showcasing small scale nitrogen dynamics within permafrost degradation zones. 691 Utilising this technique across further sites in the Arctic will help to further our understanding of 692 current processes and future changes in Arctic nitrogen cycling. 693 5 Data Availability 694 Data will be made available on a public repository upon final publication. 695