During the 2018-2019 austral summer, field surveys were conducted across 20 x 20 m plots established in several streams across the Fryxell Basin of Taylor Valley, Antarctica. Two plots were established in Bowles Creek, McKnight Creek, and Crescent Stream. Three plots were established in Canada Stream and a relict channel, but only the first and third plot in the relict channel was surveyed because plot 2 was outside the channel. Plots were assigned before the field season based on areas with homogenous NDVI signatures. Field surveys involved point transect and quadrat coverage estimates. Point transect measurements were collected by establishing cardinal (N, E, S, W), and intercardinal (NE, NW, SE, SW) transects across the 20 x 20 m plots. Every 25 cm along all transects, surface types were categorized as “dry” and “exposed” or “inundated” and “black mat, orange mat, red mat, black/orange mat, black/orange/red mat, orange/red mat, moss, cobbles/pebbles, sand, or rock.” Results were then averaged across the 20 x 20 m plots as abundances. For quadrat surveying, 1 x 1 m quadrats were randomly generated in the Esri geographic information system software, ArcGIS, within the 20 x 20 m plots. Ten quadrats per 20 x 20 m plot were established for most stream plots. However, due to field logistics and time constraints, some plots only had 6, 8, or 9 quadrats surveyed. Once in the field, 20 x 20 m general plot overview descriptions were recorded (weather, flow conditions, vegetation coverage, any notable features or changes since previous visit), and then each 1 x 1 m quadrat was surveyed. Observations of mat or moss type (black mat, orange mat, red mat, green mat, sandy orange mat, black/orange mat, black/orange/red mat, black/red mat, orange/red mat, moss, moss with black mat on top, black mat with moss on top, black mat and moss with orange mat on top), coverage, water levels (dry, moist, saturated, inundated), and substrate type were recorded. All coverages were then averaged for each quadrat and applied to the entire 20 x 20 m plot (multiple by two if ten quadrats). Further methodological details can be found in Spectral and biological characteristics of microbial mats and mosses across Fryxell Basin, Taylor Valley, Antarctica (2018-2019).

For our remote sensing techniques, we first atmospherically corrected a WorldView-2 (WV2) image of Fryxell basin taken on Dec. 12, 2018 using the same methods as Salvatore et al. 2021. We then used multispectral data, found in “Spectral and biological characteristics of microbial mats and mosses across Fryxell Basin, Taylor Valley, Antarctica (2018-2019),” as endmember spectra in linear spectral mixing models in the program, Davinci. Model outputs included averaged abundances of mat and moss within each 20 x 20 m plot and the measured and modeled WV2 spectrum of the 20 x 20 m plot. After merging all coverage datasets, averaged abundances of mat and moss within each 20 x 20 m plot were statistically compared with field survey abundances to determine how well our models were fitting measured in situ data (R², p-value, RMSE, AIC, and reduced chi-squared were calculated from the results). Other modeled results of mat and moss abundances for entire stream channels were outputted as geotiffs, which were then uploaded to the Harris geospatial deep learning software, ENVI, for further analysis. In ENVI, the reassigned WV2 bands displayed the different coverages of moss and each mat type. Using the statistics tool, percent coverages of mat and moss were found for entire stream channels. Percent coverages were converted to areal coverages by dividing the percentages by the total area of the stream channels.