Hugelius et al. (2020) reported comprehensive estimates for the N stock of northern peatland (north of 30°N) as 10 ± 7 Gt N. Due to the lack of synthesis of N stocks in tropical and southern peatlands [mainly refer to Patagonian peatlands (Yu et al., 2010)], N stocks there were derived by dividing C stocks with the mean values of C/N mass ratio. Table 1 and Table 2 summarize area, C stocks, N stocks and C/N ratios of global peatlands from literature. The C stock of tropical peatlands from Ribeiro et al. (2020) was used here. For the C/N ratio of tropical peat, we compiled a dataset of 160 records (for details of all records, see Supplementary Table S1), which consists of two types of peatlands (defined by their degradation status): natural (labeled as “undisturbed”; n = 71 records), and disturbed (drained or logged, n = 75) peatlands. For undisturbed tropical peat, the oxic, above water-table active layer of peat is approximately 10–30 cm deep (Melling et al., 2005; Girkin et al., 2020). To avoid the overrepresentation of sampling from surface soil, we divided the C/N ratios into two depth groups: surface or near surface peat (“surface peat”, n = 78) and deeper peat (“lower peat”, n = 51). Here, we simply defined surface or near surface as sampling depth less than 25 cm, to match the availability of data (Supplementary Table S1). Furthermore, tropical peatlands could also be classified by two climate types: alpine or montane, and lowland. In southern peatlands, the C stock and C/N ratio (median, in order to eliminate the impact of outliers in surface peat) were estimated from Loisel and Yu (2013a) and Knorr et al. (2015), respectively.

The long-term (apparent) rate of N accumulation (LORNA) was calculated as the cumulative N mass (g m⁻²) of a peat core divided by the core’s basal age (Tolonen and Turunen, 1996). In boreal peatlands, values of LORNA were collected from literature (Limpens et al., 2006; Loisel et al., 2014; Wang et al., 2015; Schillereff et al., 2016) as shown in Table 2. In tropical and southern peatlands, values of LORNA have seldom been reported, so we estimated them from the previously reported long-term carbon accumulated rate (LORCA) and C/N ratio (Supplementary Table S1). We compiled 26 records of LORCA of tropical peat, distributed in South America (n = 19), Southeast Asia (n = 6) and Africa (n = 1). Note that some records contain more than one peat core (For example, the Congo record in Africa is mean of 61 cores). Basal ages of these tropical peat cores range from >26,000 to 1,975 cal year BP.

Recent (apparent) rate of N accumulation (RERNA) was calculated as the cumulative N mass from the peat surface to a depth corresponding to a given date, divided by its age (Tolonen and Turunen, 1996; León and Oliván, 2014). In this paper, we defined “recent” as younger than 500 years since the oldest RERNA reported in literature was about 400 years old (León and Oliván, 2014). 75 records of northern peatlands were used to investigate the RERNA (Table 3). Specifically, records of RERNA in northern peatlands were divided into two groups: 1) Group A: RERNA calculated from peat core properties with the database provided by Loisel et al. (2014) (records are available via Pangaea). In this case, multiple cores taken from the same peatland were considered as independent records since these cores were not designed as replicates (Loisel et al., 2014). In total, group A consists of 19 records, distributed in Canada (n = 14), United States (n = 1), Sweden (n = 1), Scotland (n = 2) and Russia (n = 1). 2) Group B: data collected from literature other than Loisel et al. (2014) and Li et al. (2018), providing only RERNA but without information of properties of peat cores. Group B consisted of 56 records, distributed in Canada (n = 45), United States (n = 6) and Sweden (n = 5). The data from Li et al. (2018) were discussed in Recent Rate of Nitrogen Accumulation.

We collected information about BNF in peatlands from papers and books following the criteria as below:

Rate of BNF must have been measured in peatlands. Peatland was defined as where the surface soil layer (a minimum thickness of 30 cm) consists of at least 65% organic content (Page et al., 2011; Dargie et al., 2017). We believed in the authors’ scientific judgement on the peatlands even without definite peat properties, i.e., once they claimed their study was carried out in peatlands (including statement as mire, fen, bog or swamp), the paper was included for our meta-analysis. However, if the studies were carried out in wetland when the author(s) did not indicate the existence of peat, they were not included in this analysis. In the case where a study contained multiple sites or types of peatlands, each site or type of peatland was considered as an independent record.

In total, we obtained 21 records that reported an annual N₂ fixation rate, and 8 records that reported short-term (i.e., daily or hourly) nitrogenase activities in peatlands (Table 4). Significance tests were performed with Student’s t-test. Linear correlation analyses were performed on IBM^® SPSS^® Statistics 26.

In addition, climate data were derived from database CRU TS v4.04 (Harris et al., 2020). Mean temperature during the growing season and mean annual precipitation were obtained during the period of 1981–2010. The growing season was defined as early June to end of August (Vile et al., 2014; Rousk et al., 2015).

Of all the 160 records of C/N ratio in tropical peatlands, 78 records belong to surface peat, 51 records lower peat, and 31 records unknown. The C/N ratio ranges from 10 to 85.6, with a mean of 35.0 ± 18.4 (±standard deviation) for all records. For all the 129 records of known depth, 60 were from natural/undisturbed peatlands and 69 were from disturbed or degraded peatlands. For undisturbed peats, the C/N ratio is 25.9 ± 8.2 for surface peat (n = 31) and 33.0 ± 20.0 for lower peat (n = 29), however, the difference between surface and lower peat is not statistically significant (p > 0.05). For disturbed peats, the C/N ratio of lower peat (57.6 ± 21.6, n = 22) is significantly higher than that of surface peat (35.3 ± 16.0; n = 47) (p < 0.01). The disturbed peatlands tend to have a lower C/N ratio in the surface peat than in the lower peat. This suggests different loss rates of C and N in the disturbed surface layer, the C loss being faster than that of N under degradation of tropical peat. In addition, although the difference between C/N ratios of surface and lower peats in undisturbed tropical peatlands is nonsignificant, they show a contrary trend compared to northern peatlands. In non-permafrost northern peatlands, C/N ratios decline with increasing depth as a result of slow, but continuous, anaerobic decay remobilizing the carbon (Malmer and Holm, 1984; Kuhry and Vitt, 1996; Sannel and Kuhry, 2009; Wang et al., 2014).

For lower peats, C/N ratios show little dependence on the sampling depth (r = -0.21, p = 0.141). This indicates that the C/N ratio in deeper layers is more or less randomly distributed along the core depth despite that the C/N ratio in surface peat, or approximately active layer, is significantly lower than the deeper, less active layer. Thus, the C/N ratio for all tropical peat weighted by depth is 42.9 ± 24.5 (n = 114, including both disturbed and undisturbed peatlands), with the provision that the surface and lower peat have average sampling depths of 7.5 and 125 cm, respectively. The C/N ratio for undisturbed peatlands (32.6 ± 19.5) is slightly higher than the previous estimate of tropical peats of 29.7 ± 17.8 (Leifeld and Menichetti, 2018) since we included more records from lower peat that has higher C/N ratio. The C/N ratio for undisturbed tropical peatlands is in range of that of northern peatlands (30–55, Limpens et al., 2006).

We also noticed that C/N ratios of alpine or montane peatlands in tropical regions (19.9 ± 8.3, n = 23) and tropical lowland peatlands (37.2 ± 18.3, n = 125) show significant difference (p < 0.001). This is probably due to the lower C content in those alpine or montane peatlands in tropical regions (p < 0.001, compared with C% in tropical lowland peatland), suggesting the materials (leaf and root litters etc.) that formed peat are different from that in tropical lowland peatlands, or it is a bias caused by the small sampling set in alpine or montane peatlands.

By simply adopting the depth-weighted mean of C/N ratio of 43 (33 ± 20 for undisturbed and 56 ± 23 for disturbed peatlands) and C stock of 152–288 Gt C (Ribeiro et al., 2020), as C stocks for disturbed and undisturbed are not available yet, we estimated at nitrogen pool of 2.7–8.7 Gt N for tropical peatlands by taking the uncertainty of C/N ratio for disturbed and undisturbed peatlands. In southern peatlands, the N stock has not been reported yet. Since the peat formation condition in Patagonia is similar to that in boreal regions (Loisel and Yu, 2013a), we adopted a C stock of 7.6 Gt C (Loisel and Yu, 2013a) and a C/N ratio of 45 (Loisel et al., 2014), and estimated the N stock of Patagonia at 0.17 Gt N. With northern peatlands storing 10 ± 7 Gt N (Hugelius et al., 2020), summing up the peatland N stocks in the above three regions (i.e., 10 ± 7, 2.7–8.7 and 0.17 Gt N, respectively), the global peatland N stock was estimated at 5.9–25.9 Gt N, which is 4–19% of global total N stocks in the top 100 cm of soil layer [i.e., 133–140 Gt N (Vitousek et al., 1997)].

Generally, there are two approaches used to estimate nitrogen stock in peatlands. The first approach estimates the peatland N stock by dividing the C stock with a mean value of C/N ratio (Loisel et al., 2014; Wang et al., 2015; Leifeld and Menichetti, 2018). The second approach builds a linear relationship between peat depth and N stock with known data of peat cores, with which N stock could be predicted by peat depth (Hugelius et al., 2020). Through the first approach, Loisel et al. (2014) estimated the N stock of northern peatlands at 9.7 Gt N by assuming a mean C/N ratio of 45 based on N content data of 40 peat cores. Through the same approach, Wang et al. (2015) analyzed more than 400 peat cores from Ontario, Canada, and showed that C/N ratio ranged from 22.8 ± 6.1 to 33.0 ± 0.5 (n = ∼1,600), thus they estimated northern peatlands had accumulated 18.5 Gt N since deglaciation. There were two key differences between the above two studies: Wang et al. (2015) collected information on more peat cores while the cores were distributed in a much more extended area in Loisel et al. (2014). For tropical peatlands, Leifeld and Menichetti (2018) estimated the N stock at 4 Gt N based on a median C/N ratio of 29.7 and a C stock of 119.2 Gt C.

Through the second approach, the N stock of northern peatlands was estimated at 10 ± 7 Gt N, and the large uncertainty came from the high variability of depth and total N content (Hugelius et al., 2020). There are several factors that may have large effects on N content and C/N ratio. Vegetation types that form peat are different, with Sphagnum peat having significantly lower N content and higher C/N ratio than woody and herbaceous peat (Loisel et al., 2014). Histories of peatland formation and the evolution of N content and the C/N ratio could be largely driven by climate change in Holocene (Yu, 2011; Zhao et al., 2014).

Decomposition and degradation history also affect N stock of a peatland (Anshari et al., 2010; Yu, 2012). Overall, the estimated regional N stocks still have large uncertainty because of the large spatial heterogeneity of peat C/N ratio and N stock as well as the quite limited data in both of above two approaches.

As summarized in Table 2, the estimation of LORNA of northern peatlands ranges from 0.42 to 0.85 g N m⁻² yr⁻¹ (Limpens et al., 2006; Loisel et al., 2014; Wang et al., 2015; Schillereff et al., 2016). In tropical peatlands, the average LORCA was 41.0 ± 20.9 g C m⁻² yr⁻¹ (n = 26; Supplementary Data). By applying a depth-weighted mean of the C/N ratio as 33 for undisturbed tropical peatlands (see C/N Ratio of Tropical Peatland and Nitrogen Stock), the average LORNA of tropical undisturbed peatlands was estimated at 1.24 ± 0.63 g N m⁻² yr⁻¹. The LORNA in tropical peatlands is faster than that in northern peatlands, which could be mainly explained by the higher LORCA in tropical peatlands than northern peatlands (34 g C m⁻² yr⁻¹, Hugelius et al., 2020). In southern peatlands, the LORNA in Patagonia was estimated at 0.49 g N m⁻² yr⁻¹ by adopting a LORCA of 22 g C m⁻² yr⁻¹ (Yu et al., 2010) and an average C/N ratio of 45.

A more representative dataset of peat properties would help improve the estimation of LORNAs. For example, by estimating the time-weighted C/N ratio (55 ± 33) from 40 peat cores across North America and northern Eurasia, Loisel et al. (2014) indicated that northern peatlands had accumulated N at a rate of 0.5 ± 0.04 g N m⁻² yr⁻¹ during the Holocene. Moreover, based on the data of more than 400 peat cores from Ontario, Canada, Wang et al. (2015) obtained the mean of C/N ratio as 27, and estimated LORNA of northern peatlands at 0.85 g N m⁻² yr⁻¹ by adopting the LORCA of 23 g C m⁻² yr⁻¹ from Loisel et al. (2014). However, it is unknown whether the C/N ratio of 27 or 55 ± 33 is more representative for northern peatlands. Thus, more peat cores containing C and N content are needed to help improve both spatial and temporal resolution for the estimation of N accumulation.

Since LORNA is the average accumulation rate during the whole life-time of a peat-core, it is difficult to reflect the transient or short-term behaviors of N accumulation as well as its change with time. Figure 1 illustrates the changes of a 500-year N accumulation rate from 11,000 years ago to present, which was derived from the dataset of Loisel et al. (2014) that contains 19 peat core profiles with a high-resolution of age and N content data. Although the LORNA was estimated at 0.59 g N m⁻² yr⁻¹, the highest N accumulation rate was observed to be between 9.5 and 8 kyr BP, equivalent to 1.3 (mean) g N m⁻² yr⁻¹, and the lowest was observed to be between 3 and 2 kyr BP, equivalent to 0.32 g N m⁻² yr⁻¹. The variation of the N accumulation rate in peat cores reflected the possible peatlands’ development history—in the early stage of peatland development, peatlands were covered by more herbaceous vegetation characterized by low C/N ratio and high N content, but in the latest stage, they were shifted into Sphagnum-dominated bog characterized by high C/N ratio and low N content (Larmola et al., 2014; Loisel et al., 2014). By calculating the apparent rate of N accumulation from surface to bottom horizon of a peat profile, it was shown that, across the 19 peat cores, the apparent rate had a much higher variability in the most recent period but a lower variability during the Holocene. This pattern is possibly related to large spatial variation of N deposition and/or other N inputs across the peat cores for the most recent period.

Most of the RERNA data were reported in northern peatlands, only one data of RERNA was reported in southern peatlands (0.97 ± 0.68 g N m⁻² yr⁻¹ in Isla Grande de Chiloé, Chile, León and Oliván, 2014) and no data had yet been reported in tropical peatlands. Therefore, the analysis of RERNA mainly focused on that in northern peatlands, as summarized in Table 3.

In northern peatlands, RERNA of Group A was estimated at 0.6 ± 0.4 g N m⁻² yr⁻¹ (n = 19), which was 1.2 times that of LORNA (i.e., 0.5 ± 0.04 g N m⁻² yr⁻¹). In Group B, RERNA ranged from 0.94 to 2.32 g N m⁻² yr⁻¹, with a mean of 1.63 ± 0.43 g N m⁻² yr⁻¹ (n = 56) (Malmer and Holm, 1984; Urban and Eisenreich, 1988; Turunen et al., 2004; Moore et al., 2005; Vile et al., 2014; van Bellen et al., 2020). The mean of all records from Group A and Group B was 1.35 ± 0.56 g N m⁻² yr⁻¹, which was ∼2.7 times of LORNA. The largest RERNA was reported in the Zoige Plateau, China, as 13.0 ± 7.4 g N m⁻² yr⁻¹, corresponding to a rapid recent rate of C accumulation (RERCA) of 259 ± 137 g C m⁻² yr⁻¹ (Li et al., 2018). This rapid RERCA was much larger compared to other northern peatlands where RERCAs varied between 40 and 120 g C m⁻² yr⁻¹ in the recent 200 years (Turunen et al., 2004; Loisel and Yu, 2013b; van Bellen et al., 2020). Peat cores from other parts of China also showed rapid RERCAs e.g., 184∼376 g C m⁻² yr⁻¹ in the recent 100 years in Greater Khingan Range and Sanjiang Plain (Liu et al., 2019) and 124∼293 g C m⁻² yr⁻¹ in the recent 200 years in Changbai Mountains (Bao et al., 2010). Moreover, it is still unclear whether all peatlands in China accumulate C (and N as well) at such rapid rates in recent centuries, and an explanation remains to be found. Thus, as the area of Chinese peatlands accounts only a small part for northern peatlands (∼5%, Xu et al., 2018), we did not take this abnormal rate in our global upscaling of RERNA in the following discussion.

The apparent accumulation rate of N shows both high spatial and temporal variations (Figure 1, Figure 2). For example, in Group A, three out of the 19 profiles show the largest apparent accumulation rate in the most recent 500 years, six profiles in the most recent 2,000 years, and the other 10 profiles show the highest accumulation rate between 11 and 7 kyr BP (Loisel et al., 2014). This suggests that accumulated rate of N could depend on the developing periods and the locations with different input and output rates of N.

Most of the BNF studies were performed in northern peatlands, and these available data on peatland BNF are quite limited and with large uncertainty (Table 3). Annual BNF rates were between 0.006–11.5 g N m⁻² yr⁻¹ in northern peatlands, with an arithmetic mean of 1.9 ± 2.7 g N m⁻² yr⁻¹ (n = 21, Table 3). In different types of peatland, the mean rates of BNF were 2.7 ± 3.9 g N m⁻² yr⁻¹ in fens and 0.8 ± 1.1 g N m⁻² yr⁻¹ in bogs, but the difference was non-significant (p = 0.13). Previous studies suggested that added phosphorus (P) could stimulate N fixing to meet increased N demand from plant (Larmola et al., 2014; Kox et al., 2016), and molybdenum (Mo) availability limited nitrogenase activity (Rousk and Michelsen, 2017; Warren et al., 2017). Both elements (P and Mo) should be better supplied in fens than in bogs and, indeed, a mesotrophic fen had more BNF than any other stage of peatland development of the same succession in Finland (Larmola et al., 2014). However, our results reflect large uncertainty in peatland BNF.

Despite that BNF rate was reported to be positively correlated with temperature in previous studies (Basilier et al., 1978; Schwintzer, 1983; Urban and Eisenreich, 1988), we found a weak correlation between annual BNF rate and the mean growing season temperature in northern peatlands (Figure 3A, r = −0.26, p = 0.276). Moreover, we also observed a weak and non-significant correlation between annual BNF rate and mean annual precipitation (Figure 3B, r = −0.20, p = 0.423). Our results are consistent with a recent review which found that the correlations between BNF and annual temperature as well as precipitation were weak across the globe (Davies-Barnard and Friedlingstein, 2020). But it should be noticed that our result did not take tropical peatlands into account due to the lack of data. In Long-Term Rate of Nitrogen Accumulation, the LORNA in tropical peatlands was approximately 2 times that in northern peatlands; thus, we suggest that the BNF in tropical peatlands might be much higher than that in northern peatlands, and more observations are needed to determine this point.

The method used to determine the BNF rate might introduce large uncertainty (Flett et al., 1975; Knorr et al., 2015; Saiz et al., 2019). The most common method, ARA, is based on the versatility of the nitrogenase enzyme: its preference for performing a reaction of reducing C₂H₂ to C₂H₄ is higher than that of reducing N₂ (Schollhorn and Burris, 1967). Based on electron transfer equivalents, a theoretical conversion factor (CF) of 4:1 (moles of C₂H₄ produced per mole of N₂ fixed) was considered in vitro and 3:1 in vivo (Hardy et al., 1971). These theoretical CFs were used in many studies (Basilier et al., 1978; Markham, 2009; Rousk et al., 2018). However, ARA could either underestimate or overestimate nitrogenase activity (Flett et al., 1975; Witty, 1979; Vile et al., 2014). For example, C₂H₂ inhibits the nitrogenase activity of N₂-fixing methanotrophs, by inhibiting methane oxidation (King, 1996) which provides ATP needed in the reduction of C₂H₂ (Hardy et al., 1968). N₂-fixing methanotrophs also transform C₂H₂ into compounds undetectable for ARA, thus making ARA unreliable (Flett et al., 1975). Overestimation happens when the amount of endogenously produced ethylene is comparable to the ethylene produced by nitrogenase where BNF rates are less than 1 g N ha⁻¹ d⁻¹ (Witty, 1979). Early studies indicated the bias of CF from theoretical values and suggested site-specific calibration of the CF (Granhall and Selander, 1973; Schwintzer, 1983). Recent studies obtained CF ranging from 0.26 to 1.8 by calibration with the ¹⁵N₂ method (Vile et al., 2014; Knorr et al., 2015; Warren et al., 2017). However, an elaborate study found that CF was highly variable—from 0.001 to 5.363—across different Sphagnum species, site scales and over time, indicating accurate calibration may not be practical in peatlands (Saiz et al., 2019).

Generally, BNF is performed by microbes, mainly rhizobium in many ecosystems (Cleveland et al., 1999). However, legumes are usually rare in peatland ecosystems (Schwintzer, 1983; Vitt, 2006; Borken et al., 2016; Laine et al., 2021) and the main N₂-fixers in peatlands remain controversial. Early studies reported that the nitrogen-fixing organisms in peatlands were cyanobacteria for Sphagnum and actinomycete symbionts for dwarf shrubs (Basilier et al., 1978; Schwintzer, 1983). However, recent studies suggested that methanotrophs introduced most BNF in early stage of peatland development (Larmola et al., 2014; Vile et al., 2014; Rousk et al., 2015). By quantifying the genetic expression of 16S rRNA and nitrogenase-encoding nifH, Vile et al. (2014) showed that BNF rates were mainly originated from methanotrophs rather than cyanobacteria in pristine bogs in boreal Alberta, Canada. Leppanen et al. (2015) confirmed that most of nifH sequences were assigned to the class Alphaproteobacteria (82%), and only a few were assigned to Cyanobacteria (5%). However, elevated CH₄ concentration did not enhance BNF (Leppanen et al., 2015), indicating that BNF may not be solely controlled by methanotroph activity. Kox et al. (2018) had similar conclusion that most of the 16S rRNA genes were assigned to Alphaproteobacteria, and only 0.1% were bona fide methane-oxidizing taxa and addition of methane did not simulate incorporation of ¹⁵N-nitrogen into biomass whereas oxygen depletion increased the activity of the nitrogen-fixing community. Another candidate of N-fixing organisms is Azospirillum, which co-exists with methanotrophic bacteria (Doroshenko et al., 2007) and symbiotic bacteria (i.e., Frankia (Actinomycetaceae) and Beijerinckiaceae (Rhizobiales)) (HussDanell, 1997; Huguet et al., 2001; Vile et al., 2014; Borken et al., 2016). Nevertheless, the activity of N₂-fixing organisms could be limited by other nutrients, especially phosphorus, despite sufficient C supply (Limpens et al., 2004; Larmola et al., 2014; Ho and Bodelier, 2015). Therefore, more efforts are needed to understand the mechanisms and contributions of different N-fixing organisms on BNF in peatlands.

Since BNF rates in boreal peatlands are weakly correlated with temperature, precipitation and nutrient condition (ombrotrophic bog or minerotrophic fen), the BNF could not be predicted from these factors. The method of BNF measurement and the knowledge of N₂-fixing microbes also decrease the accuracy of determination of BNF rates. Based on what we know, we simply assessed the BNF rate of global peatlands. The annual BNF of northern peatlands was estimated at 6.5 Tg N yr⁻¹, by adopting a mean rate of 1.9 g N m⁻² yr⁻¹ from measurements in boreal peatlands, and an area of 342 Mha of northern peatlands (Hugelius et al., 2020). The BNF of global peatlands was estimated at 8.0 Tg N yr⁻¹ by adopting the same rate as that in northern peatlands, and applying a global peatland area of 423 Mha (Xu et al., 2018). This number accounts for ∼14% of the pre-industrial BNF rate of global terrestrial ecosystems (58 Tg N yr⁻¹) (Vitousek et al., 2013), and is at the same order of reactive N produced by lightning (5 ± 3 Tg N yr⁻¹) (Schumann and Huntrieser, 2007).

In addition to N inputs as discussed above, the N budget of peatlands also includes several major N outputs such as denitrification and runoff/discharge (Limpens et al., 2006). Denitrification involves a series of complicated biochemical processes where microbes convert nitrate (NO₃ ⁻) and nitrite (NO₂ ⁻) to nitric oxide (NO), nitrous oxide (N₂O) and dinitrogen (N₂) in gas form (Seitzinger et al., 2006; Groffman and Peter, 2012). Denitrification is difficult to quantify because of the challenge of distinguishing its end product N₂ from ambient atmospheric N₂, and the lack of a suitable method for upscaling the site scale measurements to regional scale due to high spatial and temporal variability (van Groenigen et al., 2015). Limpens et al. (2006) suggested a denitrification (in the form of emission of N₂O) rate of 0.2 g N m⁻² yr⁻¹ by reviewing previous studies. Hill et al. (2016) reported a denitrification rate of <0.1 g N m⁻² yr⁻¹ in central peats in a bog and a fen in Minnesota, United States, and a denitrification rate of 0.36 g N m⁻² yr⁻¹ in the lag/transition part of the bog. However, Wray and Bayley (2007) reported much larger denitrification rates of 11 g N m⁻² yr⁻¹ in marshes and 24 g N m⁻² yr⁻¹ in fens in Alberta, Canada. N outflow through runoff was estimated at 0.3 (0.15–0.63) g N m⁻² yr⁻¹ (Limpens et al., 2006), but this flux is highly site-dependent, ranging from 0.01 g N m⁻² yr⁻¹ in boreal oligotrophic bogs (Hill et al., 2016) to 1.9 ± 0.3 g N m⁻² yr⁻¹ in a slope mire in Germany (Tauchnitz et al., 2010). Overall, the outputs of N are highly variant and uncertain, and limited observations critically limit global estimation.

For peatlands with measured annual BNF rates (Table 4), the mean of atmospheric N deposition was 0.5 ± 0.4 g N m⁻² yr⁻¹ (n = 11), the mean of BNF was 1.9 ± 2.7 g N m⁻² yr⁻¹ (n = 21), and average RERNA of northern peatlands was 1.35 ± 0.56 g N m⁻² yr⁻¹ (n = 80; Recent Rate of Nitrogen Accumulation). Assuming inflow and outflow of boreal oligotrophic bogs are negligible (Moore and Bubier, 2020), BNF and atmospheric deposition are the largest and second largest sources of N, which account for ∼75 and 25%, respectively, of the total input. Denitrification, the only N output to balance the N budget, is 1.0 g N m⁻² yr⁻¹. A conceptual diagram (Figure 4) shows the N cycle with large uncertainties based on literature reports of peatland N budget.