We chose plant species based on abundance and their importance in field experiments in which we wanted to predict biomass from non-destructive measurements. We chose 14 species including graminoids (Aristida stricta, Rhynchospora megalocarpa), ericaceous subshrubs (Gaylussacia dumosa, blueberries: Vaccinium darrowii, V. myrsinites), palmettos (Sabal etonia, Serenoa repens), and larger shrubs including two ericads (Lyonia fruticosa, L. lucida), four scrub oaks (Quercus chapmanii, Q. geminata, Q. inopina, Q. myrtifolia), and the spindly shrub Palafoxia feayi. Nomenclature follows Wunderlin et al. (2019).
These species are dominant plants in a range of habitats at our study site, Archbold Biological Station. Archbold Biological Station (Swain 1998), located in south-central Florida, is one of the larger tracts of undisturbed Lake Wales Ridge landscape, a region renowned for its endemism (Estill and Cruzan 2001, Turner et al. 2006). Habitat types (modified from Abrahamson et al. 1984) include rosemary scrub, scrubby flatwoods, oak-dominated sand pine scrub, flatwoods, southern ridge sandhill turkey oak phase, and southern ridge sandhill, which in its hickory phase is also known as oak hickory scrub (Menges 2007). We sampled plants from four habitats: scrubby flatwoods, rosemary scrub, sandhill, and flatwoods.
Fire is the dominant ecological disturbance in Florida scrub (including rosemary scrub, scrubby flatwoods, and sand pine scrub), sandhill, and flatwoods (Menges 2007, Menges and Gordon 2010) and prescribed fire is a key management tool used to maintain fuel loads, manage for biodiversity, and promote key species (Menges et al. 2017). We collected data from sites spanning a range of times-since-fire, which we separated into in five categories (~6 months, 1-2 years, 3-6 years, 7-14 years, and 15-30 years). Because some plants were habitat restricted, we did not sample all species in every habitat type.
Specific locations for biomass sampling were chosen randomly within the time-since-fire categories and habitat types. We generated 200 random starting points and characterized points for their time-since-fire based on GIS shapefiles of fire history (Menges et al. 2017). Within each time-since-fire class, our goal was to collect aboveground biomass from 8 plants per species. In some cases, we collected additional samples to improve the regressions.
In the field, we navigated to each random starting point, making sure it was > 20 m from a burn-unit edge (to avoid potential edge effects). From this starting point, we chose a random compass direction and created a 40 m transect. Along this transect, we used random numbers to choose stratified random sampling points (eight points, one for every 5 m of the transect) for each target species present. We then selected the nearest plant (within 7 m) of each species to the sampling point.
Plants were defined as groups of stems in a clump, with no stem more than 15 cm from other stems. This definition was chosen to be consistent with research projects requiring the biomass regressions. In the field, we counted the number of stems (branching below ground level) and measured height (in cm, perpendicular to the ground, as the plant was in its natural posture), length (largest crown diameter, in cm), and width (perpendicular crown diameter, in cm). For graminoids, we counted the number of clumps within a distinct patch instead of the number of stems. For Serenoa repens and Sabal etonia, crown dimensions referred to the major clumps at the end of rhizomes, not the small axillary sprouts.
Once the plant was chosen and measured, we clipped all aboveground parts of plants and placed them into labeled paper bags. Plant material was returned to the lab, cut into small pieces (10-15 cm), placed in a drying oven, dried at 70 degrees C for at least 48 hours, and weighed to the nearest 0.01 g on an Ohaus Scout-Pro scale. We confirmed that 48 hours was sufficient drying time by drying some samples for longer and reweighing them.
Analysis.
We ran preliminary analyses including the raw number of stems per plant and several transformations of this variable. The number of stems per plant was never a significant predictor of aboveground biomass. In addition, we binned number of stems into three categories (individually for each species) and examined its predictive power in general linear models (GLMs), Again, stems had no significant predictive effects. Therefore, we built regression models without including number of stems.
We used GLMs to test effects of time-since-fire in regressions of aboveground biomass on height, canopy length, and canopy width. Based on preliminary analyses using linear measures and various transformations, we used natural log (ln) transformations for all four variables. We used natural log transformed data and provide the SE in the tables; this information provides a correction factor to avoid bias if back-transformation to non-log data is needed (Sprugel 1983).
If time-since-fire was a significant predictor, we also ran regressions for individual time-since-fire classes for that species; otherwise, we combined all plants of a species into a single regression. We also created regressions using only height and canopy length as predictors, for our projects where canopy widths were not measured.
We also tested GLMs for all species and for species grouped by relatedness (often genus; oaks, palmettos, lyonias, and blueberries) to see if species identity or its interaction with time-since-fire was a significant predictor of biomass. These analyses will determine if aboveground biomass regressions combining closely related species are appropriate.
