We evaluate the current state of knowledge concerning the ecosystem- and community-level importance of N2 fixation in streams. We reviewed the literature reporting N2-fixation contributions to stream N budgets and compared in-stream N2-fixation rates to denitrification and dissolved inorganic N (DIN)-uptake rates. In-stream N2 fixation rarely contributed >5% of the annual N input in N budgets that explicitly measured N2 fixation, but could contribute higher proportions when considered over daily or seasonal time scales. N2-fixation rates were statistically indistinguishable from denitrification and DIN-uptake rates from the same stream reach. However, published N2-fixation rates compiled from a wide variety of streams were significantly lower than denitrification or DIN-uptake rates, which were indistinguishable from one another. The data set we compiled might be biased because the number of published N2-fixation measurements is small (9 studies reporting rates in 22 streams), the range of stream conditions (NO3–-N concentration, discharge, season) under which N2-fixation and other N-processing rates have been measured is limited, and all of the rate estimates have associated methodological artifacts. To broaden our understanding of how N2 fixation contributes to stream ecosystems, studies must measure all rates concurrently across a broad range of stream conditions. In addition, focusing on how N2 fixation supports food webs and contributes to benthic community dynamics will help us understand the full ecological ramifications of N2 fixation in streams, regardless of the magnitude of the N flux into streams from N2 fixation.
Thousands of types of bacteria fix N2 (gas) in many different aquatic and terrestrial microhabitats. In aquatic systems, N2 fixation is carried out mainly by cyanobacteria, which are specialized autotrophic prokaryotes (Whitton and Potts 2000), although heterotrophic bacteria similar to those found in terrestrial environments also might be important stream N2 fixers (Buckley and Triska 1978). Nitrogenase is the enzyme responsible for N2 fixation, and because O2 strongly inhibits nitrogenase activity, researchers initially thought that heterocysts, specialized thick-walled cells, were necessary for cyanobacteria to carry out the 2 seemingly incompatible processes of photosynthesis and N2 fixation (Walsby 1985). However, in oceans, N2 fixation also can be carried out by nonheterocystous cyanobacteria (Bergman et al. 1997) that use mechanisms such as within-cell or temporal separation to allow co-occurence of photosynthesis and N2 fixation (Giani and Krumbein 1986, Reddy et al. 1993). In streams, the dominant autotrophic N2 fixers are heterocystous cyanobacteria, such as Nostoc, Anabaena, Calothrix, and Phoridium, and unicellular cyanobacterial endosymbionts of diatoms of the order Rhopalodiales, including Epithemia and Rhopalodia (Wehr and Sheath 2003). No researchers have confirmed the existence of free-living, N2-fixing unicellular cyanobacteria in streams, but endosymbionts within Rhopalodia gibba are closely related to 2 strains of the unicellular N2-fixing cyanobacterium Cyanothece sp. (Prechtl et al. 2004), which is commonly found in ocean environments (Reddy et al. 1993).
N2 fixation represents a source of N at both the organism and ecosystem levels. In lakes, N2 fixation by heterocystous cyanobacteria influences competitive interactions (Sterner 1989) and can make up as much as 82% of annual N budgets (Howarth et al. 1988). In the oceans, N2 fixation is proportionally less important than in lakes, but recent work has shown that marine nonheterocystous cyanobacteria are significant contributors to the global N cycle (Zehr et al. 2001, Montoya et al. 2004). Despite intensive study of N2 fixation in the open ocean, estuaries, and lakes (Howarth et al. 1988), and the common presence of N2 fixers in stream benthic communities, N2-fixation rates in streams have seldom been measured. A few notable exceptions have shown that N2-fixation rates in streams can be quite high where periphyton communities are dominated by cyanobacteria (Horne and Carmiggelt 1975, Grimm and Petrone 1997) or when ambient NO3– levels are low (Grimm 1994).
N cycling is currently a broad focus of ecosystem ecology because human activities have approximately doubled the amount of N cycling globally (Vitousek et al. 1997). Increased N loads to coastal ecosystems and subsequent eutrophication and hypoxia in areas such as the Gulf of Mexico (Rabalais et al. 2002) have led stream ecologists to study how N is transported from terrestrial areas to rivers and estuaries further downstream. Stoddard (1994) proposed that increased N loading to terrestrial systems would result in predictable alterations in the amount and timing of N transport and export from watersheds. Headwater streams are thought to be important sinks of N in the landscape (Alexander et al. 2000), and the application of nutrient spiraling theory (Newbold et al. 1981, Stream Solute Workshop 1990) has led to the discovery that small headwater streams have high rates of uptake of inorganic N (Peterson et al. 2001). Intersite research programs have produced large data sets of N-uptake parameters for streams in a variety of ecosystems (e.g., Webster et al. 2003). Recently, researchers have focused on quantifying denitrification, which represents a permanent loss of N from stream ecosystems (Seitzinger 1988, Royer et al. 2004, Mulholland et al. 2004b). These new, readily available data provide an opportunity to examine the importance of N2 fixation in streams relative to other N-processing rates.
We evaluate the current state of knowledge concerning the importance of N2 fixation in streams. We use several criteria to evaluate the potential importance of N2 fixation because few authors report N2-fixation rates in these ecosystems. First, we summarize studies where N2-fixation contributions to stream N budgets have been considered. We then compare N2-fixation rates to denitrification and uptake rates of dissolved inorganic N (DIN; NH4+-N and NO3–-N) from the published literature. Comparing N2 fixation to denitrification provides insight into gaseous inputs and outputs from the stream N pool, whereas comparing N2 fixation to DIN-uptake rates provides insight into sources of N for autotrophic and heterotrophic production. We restrict our comparison to rates measured in the channel itself rather than including rates measured in riparian areas and terrestrial uplands because we know the least about N2-fixation contributions to surface water and benthic processes. This comparison identifies a major gap in the stream ecology literature. We conclude with a discussion of the potential limitations of N budgets and our rate comparison, and the importance of N2 fixation and cyanobacteria to stream communities and ecosystems.
Contribution of N2 Fixation to Stream N Budgets
At the ecosystem level, the importance of in-stream N2 fixation has been considered in several N-budget studies where N2 fixation was included as an N source. For the purposes of this comparison, we selected only N budget studies in which N2 fixation was measured directly (typically with the acetylene reduction assay) rather than studies in which it was estimated as a remainder of the N budget. Studies in which the contributions of N2 fixation to stream N budgets are measured directly are rare, and such data are available for only a few streams.
N budget studies suggest that N2 fixation might not be a large source of N to stream ecosystems and rarely contributes >5% of the N input on an annual basis. N2 fixation contributed only 0.01% of the annual N input in Bear Brook, a small headwater forested stream (Meyer et al. 1981). Bear Brook is heavily shaded, and when that N budget was constructed, periphyton were essentially absent from the stream biota (Fisher and Likens 1973). N2 fixation contributed 4.2% of the N annually to a riffle in a 2nd-order Quebec stream, but when a similar reach was dammed by beaver, N2 fixation by sediment microbes contributed 68% of the annual N budget (Naiman and Melillo 1984). This increase did not occur because of a difference in N2-fixation activity between the 2 habitat types, but rather because of the greater sediment area available for microbial colonization in the beaver pond compared with in the riffle reaches (Francis et al. 1985). N2 fixation supplied 5% of the annual N input, compared with 73% from upstream (hydrological inputs as NO3–-N and dissolved organic N [DON]) and 22% from terrestrial organic matter, in another small forested stream (Triska et al. 1984). N2 fixation contributed 4%, an amount similar to the input from atmospheric deposition to the pond surface, of the annual N budget in an oligotrophic, streamlike pond with significant water flow and abundant Nostoc pruniforme (Dodds and Castenholz 1988). Overall, these contributions are generally greater than or within the range of contributions of N2 fixation observed in mesotrophic lakes (0.1–0.3% of annual budget), but less than in the budgets of eutrophic lakes (5–82%) (Howarth et al. 1988).
These annual budgets suggest that, overall, N2 fixation contributes less N to stream reaches than do hydrologic or litter inputs, but they do not take into account seasonal or successional variations in N flux and N2 fixation. For example, N2 fixation in a 3rd-order montane stream reach was far less than the annual hydrologic input of total N or NO3–-N; however, N2 fixation was greater than the NO3–-N flux during late summer, when discharge and NO3–-N concentration were low and biological activity was high (Marcarelli 2006). Annual rates of N2 fixation were very high (8.0–12.5 g/m2) in Sycamore Creek, a desert stream, and were comparable with rates measured in eutrophic lakes and rice fields (Grimm and Petrone 1997). However, daily contributions of N2 fixation to the total N input to stream benthos ranged from 0 to 85% depending on the season and abundance of cyanobacteria (Grimm and Petrone 1997). In Sycamore Creek, contributions from N2 fixation to the N budget were controlled directly by DIN availability, which was controlled by algal community composition and biomass, which were, in turn, controlled by floods that scoured the periphyton community (Fisher et al. 1982, Grimm 1987).
When considering the contribution of N2 fixation to annual N budgets, N2 fixation is often compared with hydrologic and litter N inputs without considering how N inputs are assimilated by stream biota. In marine systems, ⅔ of the N obtained by cyanobacteria via N2 fixation is assimilated directly into cellular material (Mulholland et al. 2004a), and the remaining ⅓ is released from the cell as DON, which can be used by the surrounding algal and bacterial community (Brookshire et al. 2005). Therefore, it is likely that most N introduced into a stream via N2 fixation is stored at least temporarily in benthic biomass either by direct incorporation or by DON assimilation. In contrast, the fate of hydrologic N cannot be quantified once it enters a stream reach; it might pass through the stream reach unaltered, be retained temporarily through cycling by the biota, be transformed and exported as dissolved or particulate organic N, or be lost permanently through denitrification.
Last, nutrient budgets are susceptible to problems related to stream size. Nutrient inputs, such as N2 fixation, that are measured per unit stream area are dependent on the overall area of the stream bottom and are not directly comparable with hydrologic or linear inputs, such as NO3–-N transported from upstream or laterally from the riparian zone, which are independent of stream size (Cummins et al. 1983). Some studies have avoided this problem by estimating nutrient budgets for the entire watershed (e.g., Triska et al. 1984). However, this approach is not feasible for large watersheds or studies with a more limited research scope. Comparison of N2 fixation to N-transformation rates, such as denitrification and DIN uptake, that also are measured on an areal basis might provide more insight than comparisons with hydrologic or linear input processes when trying to evaluate the importance of N2 fixation in stream ecosystems.
Comparing N2-fixation, DIN-uptake, and Denitrification Rates in Streams
We compared whole-stream biological DIN-uptake rates with in-stream rates of N2 fixation and denitrification. These comparisons are appropriate because DIN uptake, N2 fixation, and denitrification are all measured on an areal basis and are independent of system size. From an ecosystem perspective, N2 fixation and denitrification are an important source and sink, respectively, for N, whereas NH4+-N and NO3–-N uptake are transformations of N into organic form where N will be stored temporarily within a stream. Comparing N2 fixation and denitrification in streams is logical because these inverse processes convert N2 gas to inorganic form and inorganic N to N2 gas, respectively, and thus, indicate true input to and output from the available N pool. N2 fixation also can be compared appropriately to DIN uptake in streams because the N obtained via N2 fixation is incorporated into autotrophic or heterotrophic biomass, temporarily stored, and then released into the water column through mineralization in a manner similar to DIN spiraling. Therefore, comparing DIN-uptake rates to N2 fixation allows assessment of the relative contributions of N obtained from uptake of inorganic material from the water column and N obtained from fixation for biological production. In addition, a rich literature provides comparisons with N2-fixation rates because most stream N-cycling literature in the recent past has focused on measuring DIN-uptake rates and denitrification.
We surveyed the literature to find all possible reports of N2-fixation, denitrification, and DIN-uptake rates measured in the same stream. We located references with Web of Science (Thomson Scientific, Philadelphia, Pennsylvania), Water Resources Abstracts (ProQuest-CSA, Bethesda, Maryland), reference sections of other studies, and personal communication with researchers. This search identified 9 studies in which N2-fixation rates were reported for 22 stream reaches (Appendix). We then searched for NO3–-N-uptake, NH4+-N-uptake, and denitrification rates measured in the same stream reaches for which N2-fixation rates had been reported. We called these systems multirate streams. We compared N2-fixation, denitrification, NH4+-N-uptake, and NO3–-N-uptake rates with Kruskal–Wallis tests with α = 0.05 (SAS, version 9; SAS Institute, Cary, North Carolina; Zar 1999). When this test was significant, we assessed post hoc differences with pairwise comparisons on the basis of Mann–Whitney U tests. We corrected p values from the post hoc tests for multiple tests using the Dunn–Sidak method (Gotelli and Ellison 2004). The data sets we have compiled here cannot be considered a random analysis of all streams, and therefore, the conclusions we reach should be treated with caution.
Our literature review revealed that N2 fixation, denitrification, and DIN uptake are very rarely measured in the same stream system. We identified only 1 study that measured N2-fixation and DIN-uptake rates as part of a comprehensive study (Howard-Williams et al. 1989), and this study excluded denitrification. We identified 17 study reaches in which N2 fixation and ≥1 of the other rates had been measured (Table 1). These rates typically were measured in the same month or season, although for at least a few reaches, the rate estimates were made years or decades apart (e.g., Watershed 2, Oregon, and Watershed 6, New Hampshire; Appendix). All measurements were made between March and November. Rates were measured only during summer in most (15 of 17) studies and during spring, summer, and autumn in 2 studies. Study reaches in multirate streams were 1st to 4th order with discharge (Q) from 0.001 to 13.7 m3/s.
Comparisons in multirate streams indicated that N2-fixation rates are frequently similar to other N-processing rates. However, the relative importance of N2 fixation is extremely variable when considered on a stream-by-stream basis (Table 1). N2-fixation, denitrification, and DIN-uptake rates did not differ from each other in the multirate streams (Kruskal–Wallis, χ23 df = 6.03, p = 0.11). N2-fixation rates were higher than denitrification rates in 5 of the 9 streams in which both were measured and lower in 4. N2 fixation ranged from 1250× lower than denitrification in Watershed 6, New Hampshire, to 117× greater than denitrification in Sycamore Creek, Arizona (Table 1). N2-fixation rates were greater than NO3–-N uptake rates in 5 of 17 streams in which both were measured, lower in 10, and approximately equal in 2 (Sycamore Creek and Toxaway–lake outlet, Idaho). N2 fixation ranged from 8650× lower than NO3–-N uptake in Warm Spring Creek–lake inlet, Idaho, to 13.1× greater than NO3–-N uptake in Watershed 2 (Table 1). N2-fixation rates were greater than NH4+-N uptake in 2 of 6 streams where both were measured and lower in 4. N2 fixation ranged from 1100× lower than NH4+-N uptake in Warm Spring Creek–lake inlet to 4.8× greater than NH4+-N uptake in Sycamore Creek (Table 1).
We also did a wider literature search to compile N2-fixation, DIN-uptake, and denitrification rates from a wide variety of stream studies. We used N2-fixation rates from all 22 stream reaches identified in the 1st literature survey. We identified studies of whole-stream DIN-uptake rates from the review by Ensign and Doyle (2006). This data set was supplemented with rates from the multirate streams (if not included) and from several recently published studies (Appendix). We found denitrification rates by searching the Web of Science. We compiled rates from these references, references cited therein, and from the multirate stream studies (Table 1). We included rates only if they were measured directly in enclosures or at the reach scale; we excluded rates if they were estimated using reach-scale total N or NO3–-N mass-balance methods. However, we did include rates from denitrification studies that measured whole-reach N2:Ar balance using membrane-inlet mass spectrometry (e.g., Laursen and Seitzinger 2002). For comparison, we converted all rates to units of micrograms of N per square meter per hour. We excluded studies where rates could not be converted to similar units (e.g., rates given per unit biomass with no report of biomass per unit area) from the data set. For each study, we noted the method used to measure the rate, the month and year of measurement, study location, Q, stream order, and nutrient concentration when available (Appendix). We call these systems literature-review streams. We compared rates as described above in Multirate streams.
We found fewer estimates of N2 fixation (n = 22) than of any other rates (denitrification: n = 62, NH4+-N uptake: n = 67, NO3–-N uptake: n = 87). NH4+-N uptake varied across 4, N2 fixation varied across 5, denitrification varied across 6, and NO3–-N uptake varied across 7 orders of magnitude (Fig. 1). The median N2-fixation rate (10 μg N m−2 h−1) was 1 to 2 orders of magnitude lower than the median of the other 3 rates (denitrification median = 1605 μg N m−2 h−1, NH4+-N uptake median = 1300 μg N m−2 h−1, NO3–-N uptake median = 870 μg N m−2 h−1; Fig. 1). N2-fixation, denitrification, and DIN-uptake rates differed in the literature-review streams (Kruskal–Wallis, χ23 df = 25.7, p < 0.001; Fig. 1). N2-fixation rates were significantly lower than the other 3 N-processing rates, which did not differ from each other (Fig. 1).
The literature-review streams were much more variable than the multirate streams in terms of timing of studies and stream characteristics, such as Q and NO3−-N concentration. Fifty-four percent of N2-fixation, 49% of NH4+-N-uptake, and 60% of NO3−-N-uptake rate measurements were made during summer (Fig. 2A, D). In contrast, denitrification rates were measured throughout the year (Fig. 2A). Q ranged from 0.001 to 13.7 m3/s in studies of N2-fixation, from 0.0001 to 2.4 m3/s in studies of NH4+-N and NO3−-N uptake, and from 0.0004 to 13,100 m3/s in studies of denitrification rates (Appendix). Frequency analysis indicated that denitrification, N2 fixation, and NO3−-N uptake were most often measured in streams where Q was 0.1 to 1 m3/s, whereas NH4+-N uptake tended to be measured in streams where Q ranged from 0.01 to 0.1 m3/s (Fig. 2B, E). Therefore, we also analyzed rates in streams grouped by Q in the same order of magnitude (e.g., rates in streams where Q = 0.01–0.1 m3/s; Fig. 2B, E). The 4 N-processing rates were statistically indistinguishable in every group except Q = 0.1 to 1 m3/s (χ23 df = 10.7, p = 0.01). For this group, which also had the largest number of rate estimates (total n = 72), NO3−-N- and NH4+-N-uptake rates were significantly different from N2-fixation rates, and denitrification rates were not distinguishable from the other rates. This result indicates that our literature-review analysis of the relative importance of N-processing rates probably was affected by the fact that the frequency of measurement of each N-processing rate differed with Q (Fig. 2B, E). NO3−-N concentrations in the study streams varied from 1 to 16,520 μg/L, and the distribution of studies across this range also varied among the 4 N-processing rates (Fig. 2C, F). N2 fixation, NO3−-N uptake, and NH4+-N uptake were most often measured in streams where NO3−-N concentrations averaged 10 to 100 μg/L, whereas denitrification was most frequently measured in streams with NO3−-N concentrations of 1000 to 10,000 μg/L (Fig. 2C, F).
Implications and Potential Data Biases
Several biases are inherent in the N-cycling literature and should be recognized when N-processing rates and N budget summaries are compared across studies. First, studies are not randomly distributed across stream types; many were done in preselected habitats where rates were expected to be high. For example, 12 of 17 N2-fixation rate estimates from multirate streams were from our work on central Idaho streams, where we expected N2 fixation to be an important contributor to the N budget because of low ambient DIN concentrations. We would expect N2 fixation to be high in streams where communities are strongly limited by DIN availability and low in streams with high DIN availability because of the high energetic cost of N2 fixation compared with the costs of NH4+-N or NO3−-N uptake (Howarth et al. 1988). Nevertheless, rates of N2 fixation in the central Idaho streams were low compared with other rates reported in the literature (Marcarelli 2006).
Second, the denitrification rates found in our review might be skewed toward streams where denitrification was a large contributor to the N cycle because they were measured most frequently in streams with high NO3−-N concentrations (Fig. 2C). Denitrification rates can increase with increasing availability of NO3− as an oxidization substrate (Bernot and Dodds 2005). Therefore, streams with high rates of denitrification might not support high rates of N2 fixation and vice versa. In the literature-review streams, denitrification rate increased significantly with NO3−-N concentration (log10[y + 1] = 0.64 + 0.86 log10[x + 1], F1,48 = 47.4, p < 0.0001, r2 = 0.50; Fig. 3A). In contrast, N2-fixation rate was not related to NO3−-N concentration (Fig. 3B), probably because of the limited number of rates reported in the literature and the small range of NO3−-N concentrations across which N2-fixation rates were measured.
The narrow and low range of N concentrations across which N2-fixation rates have been examined in streams is in stark contrast to the range in lakes, where N2 fixation is associated (counterintuitively) with high N concentrations and eutrophic conditions. A review of N budget studies in lakes showed that N2 fixation contributes 6 to 82% of the annual N budget in eutrophic lakes (Howarth et al. 1988). The lakes with the highest N2-fixation contributions are nutrient-rich systems that support cyanobacterial blooms. However, it is unclear whether cyanobacterial dominance in these nutrient-rich lakes is the result of the ability of cyanobacteria to fix additional N to support growth and outcompete other taxa, or of competition for some other resource, such as light (e.g., Ferber et al. 2004). To our knowledge, no estimates of benthic N2-fixation rates have been made in nutrient-rich 1st- to 5th-order streams, probably because the periphyton communities in these streams are sometimes nutrient saturated (Bernot and Dodds 2005, Earl et al. 2006). The high N contribution from N2 fixation in eutrophic lakes suggests that N2 fixation in eutrophic streams should be examined more closely.
The lowest rates of denitrification in our data set were from central Idaho streams (MAB and L. Jeffs, Utah State University, unpublished data). It is possible that 0 or low rates of denitrification have been measured in other streams with low NO3−-N concentrations but not published. A low publication rate for negative results is a common problem in biological research (Csada et al. 1996), and could lead to overestimation of the importance of denitrification in streams.
Seasonal variation in denitrification is well represented in the literature, but seasonal variations in DIN uptake and N2 fixation are not; most studies have been done during summer. This focus on summer undoubtedly has biased our understanding of the relative importance of N-processing rates. For example, in subalpine streams, small increases in water temperature (3–5°C) stimulate N2 fixation (Marcarelli and Wurtsbaugh 2006). The small number of N-processing studies during seasons when Q is high, such as spring snowmelt in the western montane US, is especially troubling because most nutrients move during high-flow periods (e.g., Wurtsbaugh et al. 2005). A full understanding of N-cycling rates in streams will require more effort across the entire year and at a variety of Q values.
In every study we reviewed, N2-fixation rates were measured per unit area of substrate in enclosed containers (most frequently using the acetylene-reduction assay; Stewart et al. 1967), and then scaled to stream area. In contrast, denitrification and nutrient-uptake rates sometimes were measured with whole-stream techniques that account for spatial heterogeneity. Whole-stream techniques for measuring denitrification and nutrient uptake include both surface and hyporheic-zone processes (Findlay 1995), whereas enclosure techniques typically focus on surface processes. Development of a whole-stream N2-fixation technique would permit more direct comparisons of N2-fixation rates with whole-stream uptake and denitrification rates, could be applied to a larger range of stream and river sizes, and would eliminate some of the uncertainty concerning the effects of methods on the rates compared in our study.
Whole-stream nutrient-uptake techniques are most commonly applied in small streams (Ensign and Doyle 2006). In contrast, whole-stream denitrification techniques, particularly those based on changes in N2 gas concentrations, can be applied in small streams and large rivers (Laursen and Seitzinger 2002). Thus, denitrification has been measured in systems that are much larger than the systems used for studies of N2 fixation or whole-stream nutrient uptake. However, denitrification metrics on the basis of changes in N2 gas concentrations actually represent a balance between N2 loss via N2 fixation and N2 gain via denitrification. An assumption of the method is that N2-fixation rates will be negligible when NO3–-N concentrations are high; thus, this technique has been applied most often in N-rich large rivers. In contrast, whole-stream denitrification methods that use 15N tracers (e.g., Mulholland et al. 2004b) measure only denitrification.
Although some denitrification rates reported in our review were made with whole-stream techniques (e.g., Laursen and Seitzinger 2002, Mulholland et al. 2004b), others were made in enclosed containers with the acetylene-block technique. The acetylene-block technique can underestimate denitrification rates by 50% compared to 15N-tracer methods (Seitzinger et al. 1993) because it inhibits coupled nitrification–denitrification and can incompletely inhibit N2O production, although many studies used a modified acetylene-block technique that accounted for this inhibition (Bernot et al. 2003). We took care to exclude potential denitrification rates (e.g., rates measured with additions of NO3–-N or DON) when extracting data from denitrification studies for the literature-review data set.
NO3–-N- and NH4+-N-uptake rates reviewed in our study were made in streams that spanned a large geographic area, included a large number of estimates (compared with the number of N2-fixation estimates), and probably represented the potential range of DIN-uptake rates in small streams (<5th order). Some of the NO3–-N- and NH4+-N-uptake rates were measured with traditional enrichment injections, whereas others were measured with 15N tracers (Ensign and Doyle 2006). Enrichment experiments overestimate the nutrient uptake length (Sw) 2 to 3× compared with tracer experiments (Mulholland et al. 2002, Payn et al. 2005), and therefore, underestimate the mass-transfer coefficient (uptake velocity; mm/h). However, because they elevate nutrient concentrations, enrichment experiments also overestimate uptake rates (Dodds et al. 2002). Therefore, many of the DIN-uptake rates in our review are probably overestimates, and N2 fixation might be even more important than our analysis suggests. Future work should focus on comparing N2-fixation rates to DIN-uptake rates measured with 15N-tracer additions or the promising multilevel release technique of Payn et al. (2005).
Ecological Importance of N2 Fixation for Stream Communities
Our review examined N2 fixation relative to N budgets and N-processing rates from an ecosystem perspective, but N2 fixation by particular taxa could have important consequences at the level of stream community dynamics. Cyanobacteria probably have a competitive advantage over other periphyton taxa in N-limited streams because of their ability to fix atmospheric N2, as has been observed in lakes (e.g., Sterner 1989). This advantage can have important implications for the patch dynamics of algal community structure. For example, cyanobacterial abundance in Sycamore Creek is controlled spatially by hyporheic exchange patterns. Cyanobacteria are abundant at N-poor downwelling edges of sandbars, and taxa that do not fix N2 are abundant at N-rich upwelling edges of sandbars (Henry and Fisher 2003). In other stream studies, P enrichment increased the abundance of N2-fixing taxa (e.g., Elwood et al. 1981) and, in turn, increased N2-fixation rates (Marcarelli and Wurtsbaugh 2006, 2007). These results suggest that cyanobacteria do not become dominant only when N concentrations are low, but rather are controlled by a combination of chemical factors that includes P availability (e.g., Marcarelli and Wurtsbaugh 2007). Nutrient concentrations can vary spatially even within a nutrient-limited stream reach (Dent and Grimm 1999), and this spatial variability might affect patch-level community composition and, in turn, contributions of N2 fixation to whole-stream N budgets.
The ability of cyanobacteria to fix N2, and therefore to gain a competitive advantage in streams, also might be constrained by physical factors. For example, temperature is an important factor controlling the spatial distribution of N2-fixation rates in central Idaho streams because warm temperatures favor N2-fixing taxa in the periphyton assemblage (Marcarelli and Wurtsbaugh 2006). N2 fixation is an energetically expensive reaction, and many N2 fixers in streams are autotrophs that obtain the energy required to fix N2 through photosynthesis. Therefore, N2 fixation might be less important in shaded streams than in streams where light, and therefore autotrophic activity, is high. N2 fixation appears to be particularly important in streams in deserts, where in-stream primary production might be the predominant energy source (e.g., Minshall 1978). The importance of light for N2 fixation also is suggested by diel studies in streams, in which N2-fixation rates are greater during the day than at night (Horne 1975, Livingstone et al. 1984, Grimm and Petrone 1997).
Some stream and lake foodweb studies have questioned whether cyanobacteria are a high-quality food source for higher trophic levels because they are N rich or whether they are a poor-quality food source because they are defended against grazers. In general, experimental manipulations indicate that grazers in streams avoid feeding on cyanobacteria, and grazer avoidance can increase the relative abundance of cyanobacteria by removing competing algal taxa (Power et al. 1988, Dudley and D'Antonio 1991, Abe et al. 2006). Cyanobacteria have a variety of grazing defense mechanisms, such as mucilage that makes them difficult to ingest (Power et al. 1988, Dudley and D'Antonio 1991), toxins that deter macroinvertebrate feeding (Aboal et al. 2002), and basal trichomes that allow rapid regeneration of filaments (Power et al. 1988). Diatoms with cyanobacterial endosymbionts might be more palatable than cyanobacteria for stream grazers. Some work suggests that Epithemia can become dominant under grazed conditions because of its adnate growth form (Hill and Knight 1987), but in other systems, this genus does not appear to be particularly resistant to grazing (Peterson and Grimm 1992). In some grazing studies, cyanobacteria were not the preferred food source, but some grazers still ingested cyanobacteria (Power et al. 1988, Abe et al. 2006). If cyanobacteria are abundant they might provide a significant amount of food to higher trophic levels regardless of food preference, provided they are not toxic to grazers. Stable-isotope studies in streams where N2 fixation occurs might provide insight into this question.
Data Gaps and Research Needs
Our review of N processing in streams highlighted gaps in the N-cycling literature that could influence our understanding of these processes. First, measurements of denitrification, DIN uptake, and N2 fixation are not distributed in similar ways across streams with differing NO3–-N concentrations or Q. Measurements of rates tend to be biased toward streams with either high or low DIN concentration, depending on which state should favor the given rate; e.g., N2 fixation is more often measured in streams with low DIN concentrations, whereas denitrification is more often measured in streams with high DIN concentrations. This bias certainly hampers our ability to compare the importance of these rates across streams with varying N loads. In addition, all rates were measured more frequently in streams where Q < 1 m3/s, although denitrification also has been measured in larger rivers. This bias could cause severe limitations in our understanding of N cycling because ecosystem processes change with river size (Vannote et al. 1980). Last, studies of DIN uptake and N2 fixation are heavily biased toward summer months (June–August). In some systems this focus might be appropriate because snow cover or stream freezing might essentially stop some biotic processes during the winter. However, in other systems this bias could alter our ability to evaluate the relative importance of different N-processing rates on annual timescales.
The general importance of N2 fixation in streams is difficult to assess given our current state of knowledge. N2-fixation rates clearly are high in some streams, particularly ones with low DIN concentrations, but too little evidence is available for us to conclude why N2 fixation appears to be important in some streams and not in others. Further examination of how physical and biological characteristics such as temperature, light, nutrient concentrations, and grazing control N2 fixation might help us understand patterns of N2 fixation within and among streams.
Our analysis suggests that assessing the importance of N2 fixation as part of an annual N budget might underestimate its importance, perhaps because of scaling issues or seasonal changes in the importance of different N sources. To broaden our knowledge of how N2 fixation contributes to stream ecosystems, N2 fixation must be measured in concert with other N processes, including denitrification, hydrologic N import and export, and DIN-uptake rates across a broad range of stream conditions. Future work should compare N2-fixation rates to other inputs and losses of N that are not commonly considered in stream N budgets, such as N2 fixation by riparian organisms (Compton et al. 2002), groundwater N contributions (Wondzell and Swanson 1996), and losses of N via biogeochemical pathways other than denitrification (Burgin and Hamilton 2007). In addition, focusing on how N2 fixation supports food webs will help us understand how N2 fixation contributes to benthic community dynamics, regardless of the overall contributions of N2 fixation to stream N budgets.
Streams throughout much of the industrialized world are polluted with DIN, either directly through point- or nonpoint-source pollution or indirectly through N deposition (Vitousek et al. 1997). Our finding that N2 fixation is negatively related and denitrification is positively related to DIN concentration implies that N pollution should promote denitrification and favor a less important role for N2 fixation in streams. If true, N pollution might inherently change the N cycle in stream ecosystems by changing the balance between N2 fixation and denitrification. Without understanding how environmental conditions control N2-fixation rates in streams, we will be unable to understand how increasing N loads have altered, and will alter, the N cycle and the community and populations dynamics of stream organisms.
L. Ashkenas, W. Dodds, M. Gooseff, R. Hall, S. Johnson, and J. Meyer helped us locate the data used in this study. C. Arp, H. Bechtold, T. Edwards, K. Goodman, R. Hall, C. Luecke, H. Van Miegroet, J. Zarnetske, and 2 anonymous referees provided comments and discussion that greatly improved the quality of our manuscript. This research was supported by National Science Foundation grants to MAB and WAW (DEB 01-32983) and to AMM and WAW (Doctoral Dissertation Improvement, DEB 04-12081). AMM was also supported by the College of Natural Resources and the Ecology Center at Utah State University and by the NSF-Idaho EPSCoR program (EPS 04-47689) during literature review and manuscript preparation.
Appendix. Studies used in the literature review. Superscripts indicate the measurement method used in each study. See individual references for details. Q = discharge, R. = river, Cr. = creek, Br. = brook, Dr. = drain, DOC = dissolved organic C, nr = not recorded, — = not measured in the study stream
Table 1. Mean N2-fixation, denitrification, NO3–-N-uptake, and NH4+-N-uptake rates (range). All rates were converted to μg N m−2 h−1. See Appendix for citations for rate studies and specific measurement methods. Cr. = Creek, Br. = Brook, — = data not available
 This section of the journal is for the expression of new ideas, points of view, and comments on topics of interest to benthologists. The editorial board invites new and original papers as well as comments on items already published in J-NABS. Format and style may be less formal than conventional research papers; massive data sets are not appropriate. Speculation is welcome if it is likely to stimulate worthwhile discussion. Alternative points of view should be instructive rather than merely contradictory or argumentative. All submissions will receive the usual reviews and editorial assessments.