This study was conducted in 3 streams in Michigan and 4 streams in Idaho that encompassed a range of N and P concentrations (detailed site information in Appendix S1: Tables S1 and S2). The Michigan streams were all located in northern hardwood forests on the Upper Peninsula, while the Idaho streams were in the high elevation Great Basin Desert. Inclusion of sites across these two different climatic regions allowed a wider range in nutrient concentrations, nutrient ratios, and environmental conditions across which to test our hypotheses than would have been possible if we had focused our study in one of these regions. Previous research has documented the potential for both denitrification and N2 fixation in streams in both regions (Eberhard et al. 2018, Nevorski and Marcarelli 2022).
On the first sampling day at each stream, a rough habitat map was made for an ~50-80 m long reach. Reach length differed among sites based on stream width and the number of pools or riffles it took to reach approximately 20 sampling patches. Grid paper and measuring tape were used to create scaled, patch-level maps of each stream reach (see Appendix S1: Figures S1-S7). Pools and riffles were mapped first based on visual observation and measurement of width and length, then patches were assigned based on substrate type (rock, sediment, wood, and/or macrophyte) within each pool and riffle. Patch area was later calculated by counting each cell of grid paper occupied by a given patch and then multiplying it by the cell area in m2.
N2 fixation and denitrification rates were measured by acetylene reduction and acetylene block, respectively, mid-day during peak hours of sunlight. Detailed methods for the assays are reported in Appendix S2 and followed the approach of Eberhard et al. 2018 and Dodds et al. 2017. For denitrification, we measured nutrient-amended, potential rates because most previous stream studies have used this method and we wanted to be able to compare estimates to these studies, and because this method is quick and easy to conduct with many replicates to estimate rate variability. Moreover, the acetylene block method inhibits nitrification, so measuring without amendments of nitrate can underestimate denitrification rates (Dodds et al. 2017). Chloramphenicol was used to suppress additional protein synthesis during the incubation in all chambers. However, the chambers were not sparged with nitrogen or helium to create anoxic conditions. Each stream was sampled over two days due to the number of patches in each stream. Both process rates were measured on the same sampling day for each patch. The two sampling days per stream were typically back-to-back, but in some cases were not due to inclement weather.
For each patch there were 1-4 sample chambers and 1-4 blank chambers, with each sample chamber having a paired blank chamber (chambers described in Supporting Information Appendix S2). The number of sample and blank chambers depended on the size of a patch in a stream. Some streams had very large patches of a single substrate type, so we collected replicate samples in these larger patches. For example, one sediment patch in Diggie Creek was 390 m2 so we collected 2 replicates, while the largest sediment patch in South Fork Mink Creek was 59 m2 and had 1 replicate (see Appendix S1: Figures S1 and S2). Blank chambers were set up to simulate an environment with minimal N2 fixation or denitrification to control for chamber effects and changes in temperature during the incubations. Rocks found on the shore near the stream were used for blanks for stream rocks, and stream water was used as a blank for sediment, wood, and macrophytes. At the end of the incubations, substrates from chambers were removed, measured for surface area and volume, and organic matter content of sediment substrates were measured as described in Appendix S2.
Biofilm (from rock substrate) and substrate samples (sediment, wood, or macrophyte) were collected from each sample chamber in every stream patch and placed in a sterile 15 mL falcon tube. Rocks were taken out of the chamber and scrubbed, and 12 mL of scrub water was collected for biofilm samples. 12-mL sediment cores were collected from sediment chambers using a 10 mL syringe. Wood biofilms were sampled by using a pocketknife to cut off ~ 4 surface shavings from each stick in a chamber and storing with chamber water. Macrophytes were sampled by tearing off a small part of the macrophyte and collecting it with chamber water. All 15 mL falcon tubes were placed in a mobile -20 C freezer after collection and in a -10 C freezer upon return to the lab. DNA from each sample was extracted using the Power-soil DNA Isolation Kit (Qiagen) and relative abundances of nifH and nirS were quantified using qPCR following methods detailed in Appendix S2. Relative abundances are reported as the Cycle threshold values (CT) of each sample determined from the automatic analysis settings. CT values were then averaged for each patch and relativized to the median value of the respective target gene. The relativized CT values were then multiplied by -1 to account for a greater negative difference between the median CT value and original CT values, which indicated higher DNA concentrations in the samples.
To test the hypothesis that patches with more light availability and lower DIN concentrations would have higher rates of N2 fixation and nifH gene abundance, while patches with more organic matter, lower hyporheic dissolved oxygen concentrations and higher DIN concentrations would have higher rates of denitrification and nirS gene abundance, we sampled a variety of stream characteristics. To measure nutrient concentrations in stream water, we collected ~40 mL water samples from each stream patch. The water was filtered using Millipore 0.45 μm nitrocellulose membrane filters into 60 mL bottles. Samples were frozen until later laboratory analysis for ammonium (NH4+), NO3-, soluble reactive phosphorus (SRP), and total dissolved phosphorus (TDP) following methods detailed in Appendix S2. We measured canopy cover (%) using a spherical densiometer (Lemmon 1956) and water velocity (m/s) using a Marsh McBirney Flo-mate attached to a wading rod to measure velocity (m/s) at 0.6 × stream depth in each patch. Peizometers were installed in each patch following Baxter et al. (2003) to measure hyporheic dissolved oxygen concentrations (DO mg/L) using a YSI ProODO probe.
