Unexpectedly minor nitrous oxide emissions from fluvial networks draining permafrost catchments of the East Qinghai-Tibet Plateau

Variability of N₂O concentrations and fluxes

All sampled streams and rivers were supersaturated on all dates (117.9–242.5%, n = 342 samples from 114 site visits) in N₂O with respect to the atmosphere. Dissolved N₂O concentrations fluctuated between 10.2 and 18.9 nmol L⁻¹ with an average of 12.4 ± 1.7 nmol L⁻¹, which is one-third of the global average³ (37.5 nmol L⁻¹; Supplementary Table 3). Significantly higher N₂O concentrations were observed in spring (P < 0.001), followed by fall and summer (Supplementary Fig. 2a). Despite differences in catchment attributes including permafrost fraction and population densities (Supplementary Table 2), N₂O concentrations were not significantly different among the four river systems (Supplementary Fig. 2a).

Diffusive N₂O fluxes from EQTP rivers were predominantly positive (to the atmosphere), ranging from −14.0 to 40.6 µmol m⁻² d⁻¹ with an average of 9.4 ± 6.2 µmol m⁻² d⁻¹ (n = 436 samples from 114 site visits). This mean flux is an order of magnitude lower than the global average³ (94.3 µmol m⁻² d⁻¹; Supplementary Table 3). Diffusive N₂O fluxes were similar in summer and fall, and significantly higher than those in spring (P < 0.05; Supplementary Fig. 2b). The asynchronous seasonal patterns between concentrations and fluxes are likely caused by water temperature and precipitation (Supplementary Discussion 1). As with concentration, no significant differences were found for N₂O diffusion among the four rivers (Supplementary Fig. 2b).

N₂O ebullition has seldom been documented in lotic ecosystems, although it can coincide with CH₄ bubble release²¹. Because of the presence of large organic reserves in surrounding permafrost, shallow water depths, and their exposure to low barometric pressure, streams and rivers on the EQTP are notable hotspots of CH₄ ebullition¹¹. Thus, we had hypothesized that N₂O would also be entrained during the widespread processes of CH₄ bubble release. The mean N₂O ebullition rate from EQTP rivers was 0.74 ± 2.47 µmol m⁻² d⁻¹, and accounted for 4.1 ± 11.9% of total N₂O fluxes (diffusion + ebullition) across all sites. Despite this potential for high ebullition rates, accounting for this flux had a little overall effect, as the average total N₂O flux (10.2 ± 7.1 µmol m⁻² d⁻¹) was still nine-fold lower than the global mean diffusive N₂O.

Terrestrial processes modulating N₂O dynamics

The availability of inorganic N is often the primary determinant of rates of N₂O production and emission in both permafrost-affected soils¹² and fluvial networks². In the QTP, dissolved N released from both thawing permafrost soils and animal manure (Supplementary Discussion 2) is biologically available for plant uptake^22,23,24 or instead may be exported to river channels. Because plant growth is N-limited in this region²⁴, we predicted that an increase in vegetation cover would result in greater plant uptake of terrestrial N, and consequently lower inputs to, and concentrations of N in streams and rivers, up to a point. Indeed, riverine dissolved inorganic N (DIN) concentrations decreased with increasing vegetation cover for all sites (Fig. 2a). However, mechanisms causing the observed decline in riverine DIN cannot be elucidated from vegetation coverage alone, as productivity and greenness per unit plant cover decline at higher elevations²⁵. Thus, we also examined normalized difference vegetation index (NDVI, a measure of productivity and greenness) to indicate plant N uptake. Sites with high NDVI had low riverine DIN in non-continuous (namely discontinuous, sporadic, and isolated) permafrost zones (Fig. 2b), in line with the hypothesis of greater plant influence on riverine N availability when and where vegetation productivity and greenness were higher. In contrast, in areas with continuous permafrost, sites with high NDVI had high riverine DIN (Fig. 2b), suggesting that terrestrial N was sufficient to support plant growth and associated productivity/greening regardless of season, and indeed likely exceeded the N demand of plant community²⁶. If so, then the surplus DIN can be exported to river corridors. Terrestrial N₂O can also be transported along with DIN to surrounding watercourses, and higher N₂O concentrations in some sites draining permafrost areas could be supported by transport of terrestrial N₂O and/or greater in situ N₂O production supported by more DIN inputs. Even so, EQTP streams and rivers collectively received reduced terrestrial N after plant uptake. DIN concentrations (0.54 ± 0.30 mgN L⁻¹) in EQTP waterways were at the lower end of the range reported for global streams and rivers³ (0.002–21.2 mgN L⁻¹) and constrained within a relatively narrow range. The low N₂O concentrations and fluxes are consistent with the low N availability in these rivers.

Biogeochemical processes regulating N₂O dynamics

Riverine N₂O concentrations could not be effectively predicted by simple linear regressions with environmental variables [R² ≤ 0.1 in most cases, including dissolved oxygen (DO, P > 0.05, R² = 0.004) and NH₄⁺ concentrations (P < 0.001, R² = 0.1); Supplementary Table 4]. However, we found a strong positive relationship between NO₃⁻ and N₂O concentrations when DO saturation (%O₂) was undersaturated (<100%) in the water column (Fig. 3a). This result was validated by a regression tree analysis that identified %O₂ as the primary control on N₂O concentration, and that higher N₂O concentrations occurred when %O₂ < 100% and NO₃⁻ ≥ 0.58 mgN L⁻¹ (Fig. 3b). The tree had a greater explanatory power (R² = 0.56) than the simple linear regression of NO₃⁻ and N₂O concentrations (R² = 0.23; Supplementary Table 4). N₂O concentrations were uniformly low and weakly related to NO₃⁻ when %O₂ ≥ 100% (Fig. 3a), suggesting that N₂O present at these sites was derived from rare surface sediment patches in the channel that maintain hypoxic-anoxic conditions despite abundant O₂ in the water column or from external sources including inputs of dissolved N₂O from permafrost soils, intermediate runoff (Supplementary Fig. 3), and upwelling groundwater²⁷ to maintain such low to modest but supersaturated N₂O concentrations. Regardless of the specific source, this result means that the widespread occurrence of well-oxygenated overlying waters of EQTP rivers (139 of 227 samples) limits the extent of hypoxic-anoxic regimes needed for N₂O generation via denitrification.

The emission factor EF_5-r (EF_5-r = N₂O-N/NO₃⁻-N) is a surrogate for the conversion of riverine NO₃^– to dissolved N₂O²⁸. Mean EF_5-r for EQTP rivers (0.17%) was lower than both the global average (0.22%; Supplementary Table 3) and Intergovernmental Panel on Climate Change (IPCC) default value (0.26%) and is indicative of a small portion of riverine NO₃^– being converted to dissolved N₂O in EQTP rivers. According to our regression tree analysis, the discrepancies between EF_5-r for EQTP rivers and IPCC estimate corroborate past studies that a simple linear NO₃⁻ model does not adequately predict actual N₂O^29,30, because N₂O concentrations will not necessarily increase with NO₃^– loads in oxic environments. We therefore recommend that the IPCC methodology should be revised to consider nonlinear relationships or interactions among multiple environmental variables.

Microbial processes underlying N₂O dynamics

N₂O yield [ΔN₂O/(ΔN₂O + ΔN₂) × 100%] is a useful metric of relative N₂O generation²⁷, and in EQTP rivers, the N₂O yield (0.003–0.87%, average 0.23%) was 11 times lower than has been reported for lotic settings (0.01–53.8%, average 2.47%; Supplementary Table 5). This small percentage indicates that N processing in EQTP streams and rivers predominantly generates dinitrogen (N₂) instead of N₂O. Laboratory determination of benthic N₂O production rates confirmed the consistent conversion of N₂O to N₂, as these rates were negative (N₂O was consumed) for two-thirds of the sites (Supplementary Table 6). Low N₂O yields have been associated with the availability of ample OC to support the complete reduction of NO₃⁻ to N₂², and in addition to supplying N, thawing permafrost is also a source of biolabile OC to EQTP streams and rivers¹¹ (Fig. 4a).

Examination of relative gene abundances involved in N₂O production and consumption provides further insights regarding the reason for the low N₂O yield in EQTP rivers. The key enzymes for N₂O production are two types of nitrite reductase (nirS and nirK)³¹, and N₂O consumption is mediated by clade I and II nitrous oxide reductase (nosZ) which catalyzes N₂O reduction to N₂³². A high ratio of (nirS + nirK)/nosZ indicates an amplified capacity for N₂O production relative to its loss, leading to high N₂O concentrations in the water column. This ratio for EQTP riverbed sediments (average 1.96) was far below values reported from other lotic settings worldwide that vary from 2.16 to 3.24 × 10⁶ (average 19.8; Supplementary Table 7), providing compelling evidence for a molecular basis for the low N₂O yield in EQTP rivers. Unexpectedly, we also found a negative correlation between the (nirS + nirK)/nosZ ratio and %O₂ (Fig. 4b; Supplementary Discussion 3), illustrating that the microbial community increasingly favors the reduction of N₂O to N₂ as DO saturation increased³³.

Physicochemical processes governing N₂O dynamics

To understand potential controls on N₂O fluxes (({{{{{{rm{F}}}}}}}_{{{{{{{rm{N}}}}}}}_{2}{{{{{rm{O}}}}}}})), we used stepwise regression to assess the relationships between ({{{{{{rm{F}}}}}}}_{{{{{{{rm{N}}}}}}}_{2}{{{{{rm{O}}}}}}}) and multiple environmental variables known to influence ({{{{{{rm{F}}}}}}}_{{{{{{{rm{N}}}}}}}_{2}{{{{{rm{O}}}}}}}). The analysis showed that %O₂, pH, water temperature, total phosphorus, and NO₃⁻ all had significant but weak relationships with ({{{{{{rm{F}}}}}}}_{{{{{{{rm{N}}}}}}}_{2}{{{{{rm{O}}}}}}}) (P < 0.001, R² = 0.14; Supplementary Table 8).

Local hydrogeomorphology is, at least qualitatively, a reliable predictor of ({{{{{{rm{F}}}}}}}_{{{{{{{rm{N}}}}}}}_{2}{{{{{rm{O}}}}}}}) downstream trends². Along the longitudinal continuum, ({{{{{{rm{F}}}}}}}_{{{{{{{rm{N}}}}}}}_{2}{{{{{rm{O}}}}}}}) was highest in 3^rd-order (headwater) streams, declined in 4^th– and 5^th-order (medium-sized) rivers, and was slightly elevated in 6^th– and 7^th-order (large) rivers (Fig. 3c). The decline in flux from 3^rd– to 5^th-order streams may reflect the reduced perimeter-to-surface-area ratio (ratio of wetted perimeter to cross-sectional area) and hyporheic exchange rates (exchange rates of dissolved substances between surface water and groundwater beneath and alongside the river channel) with increasing stream order^34,35,36, while the increase in 6^th– and 7^th-order sites might be due to increasing riverine DIN concentrations² (Supplementary Fig. 4). Furthermore, increased suspended sediment loads can enhance N₂O generation in larger turbid channels, as suspended particles provide micro-niches that support N transformations³⁷, and thus facilitate N₂O production in the water column^2,36. Suspended sediment concentration increased with stream order for these rivers¹¹, lending support to the hypothesis that this mechanism contributes to higher fluxes observed in 6^th– and 7^th-order channels.

Regional and global implications

Based on our flux measurements, we estimated that EQTP 3^rd– to 7^th-order streams and rivers emitted 0.206 Gg N₂O-N yr⁻¹ (5–95^th percentiles: 0.129–0.291 Gg N₂O-N yr⁻¹, 2603 km² of river channel area). Our upscaling did not include 1^st– and 2^nd-order streams, which can contribute disproportionately high areal fluxes to overall fluvial N₂O emissions^36,38. Low-order streams are always well connected to continuous permafrost and hence should receive high N inputs while having reduced N₂O solubility owing to high altitudes, and these conditions are expected to lead to high N₂O fluxes. Based on this logic, we estimated a total emission of 0.275 Gg N₂O-N yr⁻¹ from 1^st– to 7^th-order streams and rivers (5–95^th percentiles: 0.162–0.400 Gg N₂O-N yr⁻¹, 3049 km²) by extrapolating the relationship in Fig. 3c to include 1^st– and 2^nd-order streams.

Despite large uncertainties due to a lack of observational data from 1^st– and 2^nd-order streams (Supplementary Discussion 4), the upscaling exercise enables us to place our estimates in a broader context of both regional N₂O budgets and fluvial emissions at the global scale. The percentage of N₂O in total GHG (CO₂ + CH₄ + N₂O) emissions expressed as CO₂ equivalents corresponded to 1.0% for EQTP 3^rd– to 7^th-order drainages, then dwindled to 0.4% for EQTP 1^st– to 7^th-order streams and rivers, falling within the range of pristine rivers (0.2–1.2%)^39,40. These values contrast those from human-impacted fluvial networks, where N₂O percentages are generally much higher (2.8–13.9%, average 6.8%; Supplementary Table 9) due to often elevated inputs of fertilizer- or sewage-derived N that boost N₂O emissions. Expressed as per unit stream/river and basin area, EQTP 1^st– to 7^th-order streams and rivers released a total of 0.08 t N₂O-N km⁻² yr⁻¹ and 0.32 kg N₂O-N km⁻² yr⁻¹, respectively, to the atmosphere, which are one order of magnitude lower than those from lotic systems worldwide [0.65 (range: 0.08–2.55) t N₂O-N km⁻² yr⁻¹ and 2.44 (range: 0.55–5.78) kg N₂O-N km⁻² yr⁻¹, respectively; Supplementary Table 9]. Applying the emission rate per unit stream/river and basin area to the entire QTP 1^st– to 7^th-order drainage networks, we obtained a riverine N₂O emission of 0.432–0.463 Gg N₂O-N yr⁻¹, which is minor (~0.15%) given their areal contribution (0.7%) to global streams and rivers⁴¹. In addition, these N₂O estimates are probably overestimated. The emission of N₂O accumulated underneath the ice during winter was estimated to be 15% of the annual emission³⁹. This ice-melt outgassing of winter N₂O was included in the above annual N₂O emissions; however, this flux may be very limited in permafrost-affected systems due to minimum N inputs from frozen soils in winter⁴². These alpine permafrost waterways emit large amounts of CH₄¹¹, but fortunately they are currently small contributors of N₂O delivery to the atmosphere, demonstrating CH₄ and N₂O dynamics are uncoupled within these systems.

Although QTP fluxes were small, existing global estimates do not effectively capture this natural fluvial source of N₂O, nor are the major drivers of N₂O dynamics well known for these systems. Our study is a step forward in quantifying fluvial N₂O evasion from a cryospheric biome, and highlights the unique dynamic nature of N₂O concentrations and fluxes in high-altitude environments. Our finding that oxygen saturation was the first and primary correlate of N₂O concentrations may also have broader implications for aquatic N₂O dynamics in other high-altitudinal streams.

Future fluvial N₂O emissions under warming climate

Temperatures are rising faster in high altitudes and latitudes than in other regions⁴³. As warming continues, permafrost thaw is expected to increase, liberating substantial amounts of dissolved N¹². As this process progresses into deeper soil layers below the rhizosphere, diminished plant uptake²⁴ should favor greater export of N to streamflow via deep flow paths^14,44,45. Meanwhile, warmer water temperatures reduce gas solubility and enhance hypoxia and denitrification at the expense of anammox⁴⁶, directing more N towards denitrification, and concomitant N₂O production and evasion to the atmosphere (Fig. 5). Moreover, the duration of the ice-free season across the cryosphere is rapidly increasing and will continue to increase⁴⁷, suggesting a potential proxy for riverine N₂O release. Furthermore, human perturbations may bring an extra N burden to the cryosphere and exacerbate these impacts. Taken together, these processes might render streams and rivers draining permafrost catchments across the globe to become hotspots of N₂O to the atmosphere in the future, leading to positive non-carbon climate feedback of currently unanticipated magnitude because of an increase in fluvial N₂O production following the development of climate change and escalation of anthropogenic influence.

The high degree of spatiotemporal variability in riverine N₂O observed here is likely to exist in other unexplored cryospheres. Further progress in understanding how aquatic ecosystems in these climate-sensitive regions will respond to ongoing global warming would benefit greatly from more N₂O measurements from glacial and permafrost-affected lotic and lentic systems at high spatial and temporal resolution. Alongside N₂O measurements, capturing the fate of thawed N in cryospheric aquatic systems are indispensable to stitching together pathways and processes into a holistic framework. It is time to improve our understanding of cryospheric aquatic N₂O emissions in shaping the global N₂O budget, and how this contribution might be altered in progressively warming high altitudes and latitudes.

Source: Ecology - nature.com