A model for regional-scale oak savanna management: the roles of fire, canopy, and soils for understory plant diversity
Data files
Sep 17, 2025 version files 54.55 KB
-
GLRI_ms1_data_with_metadata.xlsx
23.48 KB
-
GLRI_ms1_data.csv
9.44 KB
-
GLRI_OSE_forDryad.R
16.81 KB
-
README.md
4.83 KB
Abstract
We used these data to analyze direct effects of prescribed fire and mechanical thinning on oak savanna groundlayer plant diversity, indirect effects via changes to ecosystem structure (canopy openness and litter depth), and how these effects depended on soil productivity context and the spatial scale at with diversity was analyzed (mean 1-m2 plant species richness vs. total plot 1000-m2 plant species richness).
The dataset is comprised of those two scales of richness, which are the primary response variables. We analyzed the effects of three primary management variables: fire frequency (number of fires since 2000), canopy thinning (the proportion of woody stems > 5 cm DBH removed), and shrub/sapling thinning (the proportion of woody stems < 5 cm DBH removed). Indirect effects via management effects on ecosystem structure (% canopy openness and mean litter depth) were modeled using structural equation modeling, and context-dependence was analyzed using interaction terms between soil productivity (in the form of a PCA axis) and each management variable in mixed models.
Dataset DOI: 10.5061/dryad.wpzgmsc28
Description of the data and file structure
We assessed how variation in management history in oak savannas affected groundlayer plant communities by sampling 100 sites within the U.S. Great Lakes region of Illinois, Indiana, Ohio, Michigan, and Wisconsin. We sampled sites that historically supported oak savannas, are currently undergoing restoration for oak savanna using a variety of management approaches, and that represent a range of soil resource conditions. We identified geographically-proximal clusters of two to seven edaphically similar sites that represented a variety of management histories, differing in the frequency and intensity of canopy thinning, shrub and sapling thinning, and prescribed fire. Most clusters included at least one site with each of the following histories – one or more prescribed fires, mechanical canopy or shrub and sapling thinning, a combination of burning and thinning, and no management with fire or thinning in the 40 years prior to sampling. Our goal in identifying a range of management histories within clusters was to control for measured (e.g., soil characteristics) and unmeasured (e.g., land-use history) environmental gradients, while testing for differences among management approaches.
Files and variables
File: GLRI_ms1_data.csv
Description: The dataset used in analyses.
Variables
- cluster - Cluster name: name of site grouping containing 2-7 plots in similar landscape
- site.name - Site name: Name for site containing one 0.1 ha plot
- year - Year of sampling (If spring and fall sampling occurred in different year, this is year of fall sampling)
- burns.since.2000 - Fire frequency: Number of prescribed fires conducted between 2000 and the year of sampling
- pct.sand - Percent soil sand content: Mean of 40 soil cores (4 taken every 5 meters along transect)
- pct.water - Percent soil water holding capacity: Proportion of oven dried weight to saturated wet weight, pooled site-level sample
- soilPC1.inv - Larger values = higher organic matter, silt and clay content, and water holding capacity: Soils principle component axis (Includes soil organic matter, soil texture (percent sand, silt, and clay content), Mehlich-III phosphorus, Bray-II phosphorus, pH, total ion exchange capacity, and water holding capacity. The first PC axis described 60% of variation in soil variables and was positively related to percent silt, percent clay, soil organic matter, total exchange capacity, and water holding capacity and negatively related to percent sand.)
- mean.dens.cor - Canopy openness: Mean of 40 densiometer readings (4 taken every 5 meters along transect)(Proportion of 96 cells not covered by canopy X 1.04)
- mean.litdp - Leaf litter depth: Mean of 40 depth measurements (4 taken every 5 meters along transect)
- thinning.canopy - Canopy thinning intensity: total DBH of cut stumps/(total DBH of cut stumps + total DBH of standing trees)(Trees = 5 + cm DBH)
- thinning.shrub - Shrub thinning intesnity: # cut shrub and sapling stems/(# cut shrub and sapling stems + # live shrub and sapling stems)
- mean.plotSR - Average quadrat groundlayer species richness: Mean of SR in 10, 1X1 m quadrats
- bigplotSR - Total plot species richness: All species observed in 0.1 ha plot
File: GLRI_ms1_data_with_metadata.xlsx
Description: The dataset used in analyses, along with a metadata tab describing variables.
Variables: The same as in "GLRI_ms1_data.csv"
File: GLRI_OSE_forDryad.R
Description: R code for all analyses and figures included in manuscript, including Appendices.
Code/software
We conducted analyses in R v.4.0.3 (R Core Team 2020), using the following packages: lavaan, MuMIn, car, visreg, lme4.
Code is organized by the questions addressed in the manuscript, and all variables are contained in the file "GLRI_ms1_data.csv". The file ""GLRI_ms1_data_with_metadata.xlsx" is provided for a quick reference of variable description. Question 1, "How do management and soil influence richness directly and via structure?", is analyzed with SEM (lavaan package); Question 2, "How do management effects vary across a soil productivity gradient?", is analyzed with generalized linear mixed models (lme4 package); Question 3, "Are effects on richness scale-dependent?", is analyzed by conducting analyses under Questions 1 and 2 on both small (1-m2) and large (1000-m2) richness.
Code for visualizing data also included below primary analyses. Figures 3a-d (partial regression plots) represent individual SEM paths constructed from multiple regression models with structure similar to SEMs.