This study was conducted in a subtropical climate receiving 813–929 mm total precipitation during the study period from August 2019 to May 2020 with mean air temperatures ranging from 21°C to 22°C.¹ Agroecosystem biocrusts were assessed in two perennial crops: 2-month-old Vitis rotundifolia (muscadine grape) located at the University of Florida Plant Science Research and Education Unit in Citra, Florida (referred to as ‘Grape’, 29.407195, –82.139980) and a 2-year-old Citrus sinensis (orange) orchard located at the University of Florida Citrus Research and Education Center in Lake Alfred, Florida (referred to as ‘Citrus’, 28.115496, –81.713458). Soils of both sites were classified as excessively drained Entisols of the Candler series, sandy soil formed from eolian and loamy marine deposits with 6.0–6.5 pH as measured by Jalpa et al. (2020) and Nevins et al. (2020). Citrus was irrigated daily through micro-sprinklers, while Grape was irrigated daily through a drip system. Each Grapevine received at least 3.8 L of water per day from May to June, and then this amount was reduced by half during the other months. Each Citrus tree received approximately 34 L of water per day, and this amount was reduced by half during the winter. Grape had 11 N kg ha^–1 10–10–10 NPK granulated fertilizer applied in June 2019, July 2019, and March 2020, while Citrus was fertigated (5–0–7 NPK or 7–2–7 NPK) weekly and received 29 N kg ha^–1 12–4–8 NPK controlled release fertilizer in July 2019 (Supplementary Table 1).

Samples were collected at Citrus in September 2019 (Summer), November 2019 (Fall), January 2020 (Winter), and May 2020 (Spring); and at Grape in August 2019 (Summer), October 2019 (Fall), January 2020 (Winter), and May 2020 (Spring). Samples were collected from six plots at each location. Each plot was randomly located on either side of the crop trunk or vine and contained an intact biocrust and adjacent bare soil (no more than 10 cm away from the biocrust) located within the crop row (Supplementary Figure 1). There were minimal weeds in each plot due to herbicide control using glyphosate, and plots were located 122 cm away from the trunk or vine.

Biocrusts and bare soils were randomly sampled at each plot and ranged in area from 1045 cm² up to 6427 cm². Plots were at least 2 m apart from each other within a site. After each sampling point, the sample collection plot locations were shifted to the nearest intact biocrusts (no more than 2 m away from the original plot location) because not enough material remained for repeated sampling.

Biocrusts were identified by field observations and referencing the visual development scale (Belnap et al., 2008). The bare soil for each plot had no visible surface roughness or darkening (Belnap et al., 2008). To further characterize biocrusts, two replicates of biocrusts from sampling times during which they exhibited the highest N₂-fixation rates from both sites (Grape: August 2019; Citrus: May 2020) were examined for the presence of cyanobacteria and algae using an inverted microscope, Nikon Eclipse Ti2 (Nikon Instrument Inc., Japan).

Soil surface temperature and light intensity were measured at each plot (n = 6) when samples were collected. Temperatures were measured at the surface without plant shading using a thermocouple attached to a DIGI-SENSE 20250-02 temperature meter (Cole Parmer, Vernon Hills, IL, United States). Light intensity measurements were recorded using a LI-250A light meter (LI-COR Biosciences, Lincoln, NE United States). Additional data about precipitation and solar irradiance were obtained from the Florida Automated Weather Network¹ (Supplementary Table 1).

Three subreplicates each of biocrust and adjacent bare soil were collected intact from each plot to 0.5 cm depth using a 3 cm diameter corer (Supplementary Figure 1). Cores were placed in airtight jars for N₂-fixation assays and subsequent measurements of soil moisture, microbial biomass, and extractable nutrients. Additional subreplicates for ¹⁵N enrichment incubations and analysis were collected during the summer season in Grape (biocrusts: n = 3, bare soils: n = 3) and during the fall season in Citrus (biocrusts: n = 2). Bare soil samples from Citrus were not collected for ¹⁵N² enrichment incubations.

Both sites were irrigated daily in the morning, including the morning before sample collection. However, at the January and May 2020 sample collections, the Grape soil was very dry, therefore, on these dates, the soil was saturated with deionized water before soil core collection. No deionized water was added during other collection times.

Biocrust and bare soil N₂-fixation was measured under field conditions immediately after collection using an adapted version of the acetylene reduction assay (ARA) (Stewart and Bergersen, 1980; Inglett, 2013). Measurements were made using a 2-h field incubation in airtight 138 mL glass jars with 10% acetylene headspace. Non-acetylated sample blanks were simultaneously incubated for biocrusts and bare soils. Intact soil and biocrust core samples were placed on the inside lid of inverted glass jars to allow for increased ambient light access to potentially N₂-fixing phototrophic organisms (Supplementary Figure 2). The lids had a butyl rubber septum installed into a drilled hole for gas injections. To maintain field soil temperatures and prevent overheating, jars were incubated in a shallow water bath monitored with a thermocouple. After incubation, a 5 mL headspace gas sample was collected into a 3.5 mL exetainer after vigorously shaking each jar for 4 s. The jars were then opened and aerated in the field for at least 10 min before closing and storing them at 4°C for transport to the University of Florida Wetlands Biogeochemistry Laboratory in Gainesville, FL. Gas samples from ARA measurements were stored at 25°C for later ethylene analysis by gas chromatography.

To identify the conversion ratio from acetylene reduction to N₂ fixed, separate incubations with ¹⁵N₂ gas were conducted simultaneously with ARA on two biocrust replicates from Citrus in the fall, three biocrust replicates from Grape in the summer, and three bare soil replicates from Grape in the summer season following the method of Inglett (2013). These incubations were identical to those of ARA, but without injection of acetylene. Briefly, 20 mL of 98% ¹⁵N₂ was injected into the jar headspace of samples and into three empty jars. Gas samples were collected from these jars after a 2-h incubation to determine the headspace ¹⁵N₂ concentration. The jars of ¹⁵N₂ enriched samples were then opened and aerated for 10 min before being placed on ice to stop the incubation. These samples were used for N isotopic determination. Acetylated samples from the same plots where ¹⁵N₂-incubated samples were collected served as unenriched controls. ¹⁵N₂-enriched biocrust subreplicates from ¹⁵N₂-enriched jars and non-enriched control biocrust subreplicates from acetylated jars were separated from loose soil particles with a 0.5 mm sieve. The three sieved enriched biocrust, non-enriched biocrusts, enriched bare soil, and unenriched bare soil subreplicates from each plot were pooled, homogenized, and dried at 70°C.

Biocrust samples from acetylated jars were separated from loose soil particles with a 0.5 mm sieve. The three sieved biocrust and bare soil subreplicates from each plot were pooled, homogenized, and then subsampled for microbial biomass, extractable nutrients, and moisture determination (n = 6). The soil moisture content of biocrusts and bare soils was determined gravimetrically after drying in the laboratory oven for 72 h at 70°C to avoid additional mass loss due to organic matter volatilization and to allow for subsequent N analyses (Susha Lekshmi et al., 2014). Biocrust and bare soil moisture measurements from each plot were averaged together for soil moisture comparison between seasons across sites because no significant difference was detected between biocrust and bare soil moisture (n = 12). However, only biocrust soil moistures were averaged together for principal component analysis (PCA).

Microbial biomass C (MBC), N (MBN), and P (MBP) were measured using the fumigation extraction approach (Liao et al., 2014). Briefly, 1 g of sample was incubated for 24 h either in the presence of chloroform (fumigated) or without exposure to chloroform (non-fumigated) at room temperature and then extracted with 0.5 M potassium sulfate (MBC and MBN) or 0.5 M sodium bicarbonate (MBP). The C and N extracts were analyzed using a Shimadzu TOC-5050 (Japan, Tokyo) analyzer with an N module. The P extracts were digested with sulfuric acid and potassium persulfate, resuspended in double deionized water, and analyzed on the Shimadzu UV-160 spectrophotometer (Kyoto, Japan) using the molybdenum blue method (EPA Method 365.3).

Microbial biomass C, MBN, and MBP were determined by calculating the differences between fumigated and non-fumigated sample pairs with 0.37 adjustment factor (extraction efficiency) for MBC, 0.54 for MBN, and no adjustment factor for MBP following McLaughlin et al. (1986). Non-fumigated C and N fractions were quantified as extractable C (EC) and extractable N (EN), while unfumigated P extract was quantified as extractable P (EP) (Olsen et al., 1954). EC and EN are equivalent to KCl-extractable C and N, respectively, while EP is considered to contain available P in both organic and inorganic forms (Liao et al., 2014).

Gas samples from ARA measurements were stored at 25°C and analyzed for ethylene within 2 weeks of collection using a Shimadzu GC-8A gas chromatograph equipped with a flame ionization detector (FID) and HayeSep N column (2 m). The operating temperatures for the column and injection ports were 80 and 110°C, respectively. A 100 ppm standard C₂H₄ gas (Airgas, Radnor Township, PA, United States) was used for calibration, and results were reported as micromoles of C₂H₄ per square meter of soil core surface area per hour (μmol m^–2h^–1).

¹⁵N₂-enriched biocrust subreplicates from ¹⁵N₂-enriched jars and non-enriched control biocrust subreplicates from acetylated jars were analyzed for isotopic N, and total N. Atom% ¹⁵N was measured using a Thermo Finnigan Delta Plus XL isotope ratio mass spectrometer with a ConFlo III preparation system at the UF/IFAS Soil and Water Sciences Elemental Analysis Laboratory, Gainesville, FL, United States. Total N was simultaneously determined using a Costech ECS 4010 CHNS-O elemental analyzer. Atom% ¹⁵N of headspace N₂ was calculated by subtracting ¹⁵N/¹⁴N atom% of air from ¹⁵N/¹⁴N atom% of gas blanks. Atom% excess biomass was divided by atom% excess in headspace to calculate the fraction of N₂-fixation derived N according to the equation adapted from Inglett (2013).

Total N in the enriched biocrust samples was then used to calculate the N₂-fixation rate in nmols of N-N₂ g^–1 DW h^–1. The conversion factor from acetylene reduction to N₂-fixation was calculated by dividing the average biocrust acetylene reduction rate by the average enriched biocrust N₂-fixation rate. The conversion factor was obtained separately for the summer season Grape and the fall season Citrus samples. Analogous estimates were also made using the theoretical conversion factor of 3 (Howarth et al., 1988).

All data analysis was performed in R statistical software (R Core Team, 2019). First, biocrust measurements from acetylene reduction rates, microbial biomass, and nutrient measurements were compared across sites and seasons to determine the influence of interactions using general linear mixed model analysis with emmeans (Lenth, 2019) and nlme (Pinheiro et al., 2019) packages. Second, biocrusts measurements from acetylene reduction rates, microbial biomass, and nutrients were compared to bare soils within each site and season using general linear mixed model analysis with emmeans (Lenth, 2019) and nlme (Pinheiro et al., 2019) packages. A random effect for paired bare soil and biocrust samples was added to the statistical model. Biocrust and bare soil moisture, field measured light intensity, and field measured soil temperature were compared across seasons and sites using a Two-Way ANOVA with an HSD post hoc test. Biocrust microbial biomass, nutrients, and nutrient ratios were compared across seasons and sites using a Two-Way ANOVA with an HSD post hoc test. Normality was tested by the Shapiro–Wilk test for distribution and homogeneity of residuals. Non-normal data containing zeros were square root transformed, while non-normal data without zeros were log transformed. The results for general linear mixed-model analysis were reported as significant when p < 0.05 according to Tukey post hoc test. Plots were created using the ggplot2 package in R (Wickham, 2016).

To determine variables influencing N₂-fixation rates of biocrusts, PCA was performed with the princomp function (stats 4.0.3). Each site was analyzed separately and only measurements from biocrusts were included. Prior to analysis, the data were preprocessed by filtering out zeros (Grape n = 21; Citrus n = 23), testing for multivariate normality using the mvn function from the MVN package (Korkmaz et al., 2014), and for Mardia’s multivariate skewness and kurtosis. Grape N₂-fixation rates were log transformed, while other variables were left untransformed. For Citrus, all nutrient ratios were log transformed, moisture was root square transformed, and the rest of the variables were not transformed. Following transformations, the data were z-score standardized. The elbow plot and latent root criteria using base R were used to determine the number of principal components that best explained the data variation. Bootstrapped eigenvectors and loadings of at least 0.3 were used to determine the significance of loadings at the 0.01 significance level (Peres-Neto et al., 2003). PerMANOVA was performed to determine if samples were significantly separated by season using the adonis function from the vegan 2.5-7 package (Oksanen et al., 2013). Pairwise PerMANOVA was performed to determine which pairs of seasons were significantly separated from each other using wrapper function pairwise.adonis for multilevel pairwise comparison using adonis from package ‘vegan’ (Martinez Arbizu, 2020). PCA for each site was plotted with the ggbiplot function from the devtools package (currently in development by Vincent Q Vu) by only including variables with significant loadings as determined by bootstrapping. When vectors with significant loadings were not visible in the PCA plot, only one representative vector was shown.

Biocrust visual development scale, as qualitatively determined by surface coloration and roughness (Belnap et al., 2008) was at least 5 or 6 (Figures 1A,D). Microscopic inspections identified that Grape and Citrus biocrusts were both dominated by heterocystous and non-heterocystous cyanobacteria. Grape biocrusts also had filamentous algae, whereas Citrus biocrusts also contained mosses and single-celled algae (Figures 1D,E).

Soils at both sites received the highest total precipitation 24 h before sampling during the summer season (Table 1), which exceeded other seasons by 2.3 mm (Citrus) and 26 mm (Grape). At the Grape site, biocrust and bare soil moisture ranged from 0.004 to 78%, with the highest average temperature in summer and the lowest average in spring, while at the Citrus site, it ranged from 0.002 to 22%, with the highest average in the winter and lowest average in fall (Table 1). At the Grape site, biocrust and bare soil moisture were significantly greater during the summer by at least 44% compared to fall, winter, and spring, even with the inclusion of added water during winter and spring collections (Table 1, p < 0.001). At the Grape site, biocrust and bare soil temperature ranged from 27.0 to 41.9°C, with the highest average temperature in spring and the lowest average in winter, while at the Citrus site, it ranged from 23.1 to 41.8°C, with the highest average in the summer and lowest average in winter (Table 1). At the Grape site, the light intensity on the soil surface ranged from 691 to 1741 μmol m^–2 s^–1 with the highest average intensity in summer and the lowest average in fall, while at the Citrus site it ranged from 504 to 1920 μmol m^–2 s^–1 with the highest average in summer and lowest average in fall (Table 1). Light intensity was not measured during winter at both sites.

There were significant interactions between season, site, and sample type (biocrust or bare soil) for acetylene reduction rates (Supplementary Table 2). In Grape, average biocrust acetylene reduction rates were significantly greater in biocrusts (98%) than in bare soils during the summer, fall, and winter. In Citrus, average biocrust acetylene reduction rates were significantly greater in biocrusts (98%) during the fall, winter, and spring (Figure 2).

Grape biocrusts had higher average acetylene reduction rates than Citrus biocrusts. Grape biocrust acetylene reduction rates ranged from undetectable to 401 μmol m^–2 h^–1 with the highest average rate in summer (260 μmol m^–2 h^–1) and the lowest average rate in spring (4.19 μmol m^–2 h^–1) (Figure 2). Acetylene reduction rates of Citrus biocrusts ranged from undetectable to 326 μmol m^–2 h^–1 with the highest rate in winter (78 μmol m^–2 h^–1) and the lowest in summer (13 μmol m^–2 h^–1).

Average Grape biocrust acetylene reduction rates decreased from summer to spring, while average Citrus biocrust acetylene reduction rates increased from summer to spring. Grape biocrust acetylene reduction rates were significantly greater during the summer when compared to Grape biocrust acetylene reduction rates during fall, winter, and spring (Figure 2). While not significant, there was a trend of increasing acetylene reduction rates from summer to spring in Citrus (Figure 2).

The C₂H₄:N₂ conversion factor was lower in Citrus (1.22 ± 0.23, standard deviation) than in Grape (2.69 ± 0.92), leading to higher average annual N input estimates from N₂-fixation in Citrus. Estimates of biocrust N₂-fixation rates using both the experimental and theoretical 3:1 ratios are shown in Table 2. Assuming a constant C₂H₄:N₂ conversion factor and using experimental ratios, the N₂-fixation rates in Grape ranged from 1.87 × 10^–6 to 4.18 × 10^–3 g N m^–2 h^–1, while in Citrus it ranged from 2.05 × 10^–6 to 7.49 × 10^–3 g N m^–2 h^–1. Based on the assumption that phototrophic diazotrophs were engaged in N₂-fixation for 10 light hours, we estimated the daily N₂-fixation rate of these agroecosystem biocrusts to be 6–32 mg of N per day. Averaging N₂-fixation inputs separately for each season and site, and then adding for an annual estimate with the assumption that biocrust agroecosystems had 12.5% biocrust soil surface area coverage, Citrus biocrusts were estimated to provide 8.1 kg N ha^–1 year^–1, while Grape could provide 4.9 kg N ha^–1 year^–1.

Microbial and extractable carbon was significantly greater in Grape biocrusts compared to bare soil only during the fall, while at the Citrus, MBC was significantly greater in biocrusts compared to bare soil during the fall, winter, and spring seasons (Supplementary Figure 3). Grape biocrust MBC ranged from 19 to 4036 mg kg^–1 with the highest concentration in fall (2563 mg kg^–1) and lowest in spring (1122 mg kg^–1). Citrus biocrust MBC ranged from 232 to 3998 mg kg^–1 with the highest concentration in fall (2149 mg kg^–1) and lowest in summer (1578 mg kg^–1) (Supplementary Figure 3).

Grape MBN was significantly greater in biocrusts compared to bare soil in the fall and winter, while MBN was not significantly different between Citrus biocrusts and bare soils in all seasons (Supplementary Figure 3). Grape biocrust MBN ranged from 2 to 520 mg kg^–1 with the highest concentration in winter (121 mg kg^–1) and lowest in spring (36 mg kg^–1). Citrus biocrust MBN ranged from 16 to 480 mg kg^–1 with the highest concentration in summer (189 mg kg^–1) and lowest in fall (120 mg kg^–1). There was also a significant interaction between site and season for biocrust MBN (Supplementary Table 1), where Grape biocrust MBN was significantly greater in the fall compared to the spring (Table 1).

Grape biocrust MBP ranged from undetectable to 83 mg kg^–1 with the highest concentration in summer (29 mg kg^–1) and lowest in winter (10 mg kg^–1). Citrus biocrust MBP ranged from undetectable to 68 mg kg^–1 with the highest concentration in spring (29.5 mg kg^–1) and lowest in winter (3 mg kg^–1) (Supplementary Figure 3). Due to the high variability and MBP being below the detection limit in 30% of samples, it was not analyzed for significant differences.

There were different seasonal patterns in MBC:MBN ratios in Grape and Citrus. Mass-based MBN:MBP and MBC:MBP ratios were excluded because 30% of MBP values were below the detection limit. Grape biocrust MBC:MBN ranged from 6 to 44, while Citrus biocrust MBC:MBN ranged from 7 to 30. The highest biocrust MBC:MBN in Grape was during the summer (18.1 ± 5.52), while in Citrus it was during the fall (18.2 ± 2.53). The lowest biocrust MBC:MBN in Grape was in fall (9.2 ± 1.08), while in Citrus it was during the summer (9.82 ± 2.54) (Table 1).

In general, EC, EN, and EP were higher in biocrusts than in bare soils in Grape and Citrus. EC was significantly greater in biocrusts than in bare soils during the summer (Grape), fall (Grape and Citrus), and spring (Citrus) (Supplementary Figure 4). Grape biocrust EC ranged from 106 to 780 mg kg^–1 with a high in fall (512 mg kg^–1) and a low in spring (292 mg kg^–1). Citrus biocrust EC ranged from 83 to 1107 mg kg^–1 with a high in spring (419 mg kg^–1) and a low in summer (174 mg kg^–1).

EN was significantly greater in biocrusts than in bare soils during the summer (Citrus), fall (Grape), winter (Citrus), and spring (Citrus). Grape biocrust EN ranged from 7 to 255 mg kg^–1 with a high in spring (78 mg kg^–1) and a low in winter (28 mg kg^–1). Citrus biocrust EN ranged from 9 to 132 mg kg^–1 with a high in spring (58 mg kg^–1) and a low in winter (32 mg kg^–1).

There was a significant interaction between site and sample type for EP (Supplementary Table 2). EP tended to be higher in Grape biocrusts than in Citrus biocrusts. EP was significantly greater in biocrusts than in bare soils during the summer (Grape), fall (Citrus), and winter (Grape). EP in Grape biocrusts decreased from summer to spring, while in Citrus there was no significant seasonal trend (Table 1). Grape biocrust EP ranged from 36 to 268 mg kg^–1 with a high in summer (156 mg kg^–1) and a low in spring (80 mg kg^–1). Citrus biocrust EP ranged from 12 to 67 mg kg^–1 with a high in summer (48 mg kg^–1) and a low in winter (26 mg kg^–1).

No strong seasonal patterns or significant differences between sites, biocrusts and bare soils, or seasons were detected in extractable nutrient ratios. Due to undetected EN and EP, certain ratios were excluded from the following comparisons. Grape biocrust EC:EN ranged from 3 to 18, while Grape bare soil EC:EN ranged from 7 to 15. Citrus biocrust EC:EN ranged from 3 to 13, while Citrus bare soil EC:EN ranged from 1 to 10. Grape biocrust EN:EP ranged from 0.05 to 3, while Grape bare soil EN:EP ranged from 0.2 to 0.6. Citrus biocrust EN:EP ranged from 0.2 to 4, while Citrus bare soil EN:EP ranged from 0.2 to 2. Highest average Grape biocrust EC:EN was in winter (11.2 ± 1.89), while the lowest Grape biocrust EC:EN was in summer (4.9 ± 1.07). Highest average Citrus biocrust EC:EN was in fall (9.1 ± 1.98), while the lowest Citrus biocrust EC:EN was in summer (4.86 ± 1.07). Highest average Grape biocrust EN:EP was in spring (0.9 ± 1.12), while the lowest Grape biocrust EN:EP was in summer (0.4 ± 0.22). Highest average Citrus biocrust EN:EP was in spring (1.8 ± 1.23), while the lowest Citrus biocrust EN:EP was in summer (0.8 ± 0.33) (Table 1).

Variation in actively N-fixing biocrusts in Grape and Citrus was explained by the variation in environmental conditions, nutrients, biomass, and acetylene reduction rates. Biotic variables (MBC, MBN, MBC:MBN), nutrient variables (EC, EN, EP, EC:EN, EN:EP, EC:EP), acetylene reduction rates, and environmental variables (soil moisture, soil temperature) measured in N₂-fixing biocrusts were included in the PCA. At Grape, MBC:MBN, extractable nutrients, extractable nutrient ratios, and soil moisture were significant drivers of variation, as indicated by the significance of loadings (Supplementary Table 3). Based on significant loading, PC1 in Grape may represent nutrients, PC2 may represent N₂-fixation activity, and PC3 may represent microbial biomass. At Citrus, microbial biomass, MBC:MBN, EC, EN, EC:EP, EN:EP EC:EN, and temperature were significant drivers of variation (Supplementary Table 2). Based on significant loading, Citrus PC1 may represent nutrients and microbial biomass carbon, PC2 may represent N₂-fixation activity, and PC3 may represent microbial biomass nitrogen, nutrients, and environmental variables.

Actively N₂-fixing biocrust samples in Grape (46%) and Citrus (29%) were significantly separated based on the season (Supplementary Table 4). N₂-fixing biocrust samples in summer significantly differed from the fall and winter seasons in both Grape and Citrus (Supplementary Table 4). Grape summer samples were significantly different from fall and winter samples mainly because of higher EC, EN, MBC:MBN, EN:EP, and EC:EP but lower EC:EN (PC1); and higher acetylene reduction rates, soil moisture, and EP, but lower EC:EN (PC2, Supplementary Table 3). Citrus summer samples were significantly different from fall and winter samples mainly because of lower MBC, EC, EN, EN:EP, and EC:EP (PC1); and lower acetylene reduction rates, MBC:MBN, and EC:EN but higher soil temperatures (PC2, Supplementary Table 3).