Biological soil crusts (biocrusts) are widespread soil photosynthetic communities covering about 12% of Earth’s land surface and play crucial roles in terrestrial carbon (C) and nitrogen (N) cycles, yet scalable quantifications of biocrusts and their biogeochemical contributions are notably lacking. While remote sensing has enormous potential to assess, scale, and contextualize biocrusts and their functions, the applicability of hyperspectral data in predicting C- and N-related biocrust traits remains largely unexplored. We address this issue by evaluating the potential of in situ hyperspectral data to predict C and N across a range of biocrust species and different environmental conditions. We found that in situ hyperspectral reflectance measurements can be used to predict biocrust tissue C/N ratios and N concentrations with relatively high accuracy but to a lesser extent for potential biocrust N₂ fixation rates. Critical wavelength domains included the visible region of the spectrum from roughly 490 to 600 nm, which most effectively captured variations in biocrust tissue C, and the shortwave infrared region from 1150 to 1350 nm and 1550 to 1650 nm, which most effectively captured biocrust tissue N and N₂ fixation potential. Finally, we provide evidence that multi- and hyperspectral missions with targeted band placement, such as the proposed 26-band Landsat Next, could be effective in predicting biocrust traits. This work provides a critical step in understanding how to apply data from new and upcoming satellite missions to the monitoring of biocrusts.

Biocrust sample collection

A total of 14 species were chosen for their variation in C and N cycling and their abundance in the Southwest United States and globally, with 10 replicate samples for each of the 14 target species. Biocrusts were collected from the Sand Flats Recreation Area near Moab, Utah (38.5781°N, 109.4472°W). Biocrusts were identified based on known field appearance characteristics such as color and morphology. Descriptions of the sampling site in detail are available in (Torres-Cruz et al., 2018). Two cyanolichens (C. tenax, C. coccophorum) and one darkly-pigmented cyanobacterium (likely Scytonema sp.) were selected as known N₂-fixers (N₂-fixers hereafter) and other species were chosen due to their high abundance at the Moab site and in drylands worldwide. Samples were collected by first wetting with deionized water and a spray bottle, then using a 1-inch (2.54 cm) clear plastic tube to core the target species. We included only the minimum amount of below-crust mineral soil required to maintain sample integrity. We attempted to collect species in isolation, but this proved impossible for some species due to their propensity to grow interspersed with other species rather than in a monoculture even at the 1-inch scale. The amount of non-target species in a sample was minimized to the extent possible. Samples were then allowed to dry fully in the lab before photos were taken. Species were sorted into the following groups based on common functional, morphological, and taxonomic categorizations: 1) N₂-fixers; 2) Squamulose, which includes chlorolichens with a squamulose morphology and the Myrmecia photobiont; 3) Crustose/foliose, which includes chlorolichens with a crustose or foliose morphology and the Trebouxia or Asterochloris photobiont and 4) Mosses.

Biocrust traits measurement

Functional traits associated with biocrust C and N cycling were assessed from tissue and trace-gas exchange. The biocrust samples were placed in a greenhouse where their moisture was maintained by gently spraying with deionized water for several days before being air dried. Following the final watering event, a set of photographs was taken while the samples were still wet. Tissue C and N samples were collected from air-dried biocrusts and were oven-dried at 60ºC. A subsample of the ground tissue was analyzed for total C and N on a Vario Micro Cube elemental analyzer (Elementar, Langesbold, Germany). Samples were assessed on a per mass (e.g., for tissue C and N) or per area (e.g., for potential N₂ fixation) basis. It should be noted that area-based measurements do not take into account surface relief and the area of the target species was measured using the ImageJ software based on photographs (Abràmoff et al., 2004). The area-based measurements are intended to represent overhead views, as would be obtained during typical field surveys of biocrust community cover and composition.

For the potential N₂ fixation rate (NF_pot) trait assessment, there are two incubation methods commonly used; the Acetylene Reduction Assay (ARA) (Hardy et al., 1968), which takes advantage of a lack of substrate specificity when measuring the reduction of acetylene (C₂H₂) into ethylene (C₂H₄) by nitrogenase, and second, the addition of an enriched ¹⁵N₂ headspace and the subsequent measurement of ¹⁵N in N₂-fixer tissue. There are drawbacks to both methods. ARA is not a true measurement of N₂ fixation, but rather an assessment of nitrogenase activity. In turn, the ¹⁵N technique is challenging for assessments of whole communities, which biocrusts are (i.e., multiple N₂-fixing species living together across extremely small scales). N₂-fixer tissue cannot be isolated from other tissues or soil (e.g., heterotrophic N₂-fixers in soil) and thus whole samples must be used, which are often too diluted for observing the ¹⁵N signal. To minimize the drawbacks of both methods, we assessed a subset of samples using both ARA and ¹⁵N₂, while the remaining samples were assessed using only ARA. Samples were given a 10% headspace of either acetylene or ¹⁵N₂ before being incubated for eight hours in the greenhouse with ambient daylight and temperature. Following the incubation, a 4 ml headspace sample was collected from the ARA samples and analyzed for ethylene on a Shimadzu GC-14A with a flame ionization detector (FID) (Shimadzu, Kyoto, Japan). All the samples were then uncapped and the ¹⁵N₂ samples were homogenized by ball mill before subsamples were wrapped in tin capsules and sent to the Central Appalachian Stable Isotope Facility for δ¹⁵N analysis. The ANOVA analysis followed by Tukey's Honestly Significant Difference (HSD) test was performed for each of the four traits to determine if there is a significant difference between each of two the four biocrust groups.

Hyperspectral reflectance measurement

In this study, in situ spectral measurement refers to measuring the reflectance of biocrust samples in a laboratory environment with a portable spectroradiometer. Specifically, hyperspectral reflectance within the wavelength range of 350-2500nm was measured at a 1 nm spectral resolution for all biocrust samples using an ASD FieldSpec 3 spectroradiometer (Analytical Spectral Devices, Boulder, CO, USA). The reflectance measurements were made in a non-destructive manner using a plant probe assembly consisting of a leaf clip and a contact probe (see Figure S1 in supporting information S1) (Udelhoven et al., 2003). The contact probe, with a diameter of approximately 25mm, has an internal light source of a halogen bulb with a color temperature of 2900K. The leaf clip has a white reference panel and a black background panel, which are used when checking white reference and during the actual reflectance measurement, respectively.

We strictly followed the procedures used in previous studies when operating the spectroradiometer, including turning on the spectroradiometer at least 30 minutes before collecting any spectra optimizing the spectroradiometer, and checking the white reference using the ASD ViewSpec Pro software (Malvern Panalytical, 2014) every five minutes or whenever spectra exhibited abrupt discontinuities (Young and Reed, 2017). Each sample was held against the contact probe using the leaf clip’s dark background panel. We collected five spectra from each sample when the samples were wet. The biocrust samples were then air-dried before we collected another five spectra for each sample. We also collected spectra of dry mineral soil from a sampling site with similar characteristics to quantify the main differences between spectra of biocrusts and mineral soil (Phillips et al., 2022). The collected spectra were saved as binary files, which were imported into R (R Core Team, 2022) using the ‘asdreader’ package (Roudier and Lalibert, 2017). We excluded the reflectance within 350-400 nm and 2450-2500 nm from the following analyses due to the noises within these wavelength ranges as reported in previous studies (Barnes et al., 2017; Loozen et al., 2019).