When wolves aren’t enough: Revisiting trophic cascades in northern Wisconsin
Abstract
Elimination of top predators in many terrestrial ecosystems has transformed food webs and ecosystem functions, and predator restoration is hoped to reverse the negative effects of this trophic downgrading. We use detailed wolf colonization records and vegetation surveys from the 1950s and 2000s to assess whether the return of wolves to northern Wisconsin in the 1970s restored a trophic cascade protecting native species from excessive deer browse. Surprisingly, understory plant diversity, richness, and frequency decreased more in areas with longer wolf presence, providing no evidence that wolves protected forests. Instead, our data suggest a “bottom-up” process wherein deer concentrate in areas where their preferred plant resources are abundant, and wolves track deer. Disentangling top-down from bottom-up trophic processes requires explicit tests of these competing hypotheses.
Dataset DOI: 10.5061/dryad.bvq83bkmk
Description of the data and file structure
We used quantitative surveys of Wisconsin plant communities spaced 50 years apart to assess whether wolf recolonization affected diversity and frequency of 241 different plant taxa. Eighty-three sites initially surveyed between 1942 and 1956 were resurveyed between 2000 and 2005 by recording the presence of all vascular plants across 20 1-m2 quadrats. We used the documented history of wolf dispersal across northern Wisconsin and maps of wolf occurrences based on tracking, aerial observations, and radio-collar data to quantify wolf occupancy time from 1980 to 2005. We then tested for an association between wolf occupancy time and changes in plant communities.
Files and variables
File: Data.zip
Description: This zip file contains all the data necessary to reproduce the results in Brice et al., "When wolves aren’t enough: Revisiting trophic cascades in northern Wisconsin." The file includes the following spreadsheets:
DeerForage_df.csv
Description: data on the frequency of plants that deer highly prefer and avoid at sites in 1950 and 2005, and covariates used in the final model of plant frequency. NAs are indicated with blank cells. Columns include:
- Site: a unique numerical ID for each of 83 study sites across northern Wisconsin
- Species: plant species name
- LRR: log response ratio for the change in plant frequency from the 1950s to 2000s (log(2000s response/1950s response))
- Freq_1950: the frequency of each plant species at each site when first surveyed in the 1950s. Frequency is the number of quadrats out of a maximum of 20 at which the species was found
- Freq_2005: the frequency of each plant species at each site when surveyed again in the early 2000s. Frequency is the number of quadrats out of a maximum of 20 at which the species was found
- Palate.cat: a categorical variable for how palatable each plant species is to deer, with "high" indicating that deer have a high preference for that species, and "low" indicating deer have a low preference for that species
- Palate.cont: a continuous variable for how palatable each plant species is to deer, with 1 denoting high preference and 0 denoting low preference
- Wolf_Years: the number of years that a wolf pack was present at each site since wolves began recolonizing Wisconsin
- Abund_1950_forage: the abundance of plant species that deer highly prefer to eat at each site in the 1950s, calculated by summing the frequencies of 21 high deer preference species at each site
- Precip_summer: average summer precipitation at each site, calculated by taking the sum of precipitation during June, July, and August at each site each year and averaging across 1980-2000. Measured in mm
- Tmax: average maximum temperature at each site, calculated by averaging the maximum daily temperature each month, taking the maximum monthly temperature each year, and averaging across 1980-2000. Measured in degrees Celsius
- AvgBA_1950: average 1950s basal area at each site. Basal area was measured at breast height and in square centimeters for various tree species at each site, and we took the average across species
- Soil_PC1: a summary metric derived as the first axis of a principal components analysis (PCA) of soil properties (pH, % organic matter, % Sand, % Silt, % Clay, % N, P (ppm), K (ppm), Ca (ppm), and Mg (ppm)) within the first 10-cm measured at sites in the 2000s. These data were only available for 78 sites.
- Hunting: a binary variable indicating whether hunting was allowed (1) or not allowed (0) at each site.
- Dist_road_m: the distance from each site to the closest road of any kind (MTFCC Codes S1100, S1200, S1400, S1500), measured in meters
Diversity_df.csv
Description: data on the change in diversity and richness of plant species at 83 sites in 1950 and 2005, and covariates used in the final models of plant diversity and richness. NAs are indicated with blank cells. Columns include:
- Site: a unique numerical ID for each of 83 study sites across northern Wisconsin
- LRR_Diversity: log response ratio for the change in plant species diversity (Hill 1) from the 1950s to 2000s (log(2000s response/1950s response))
- LRR_Richness: log response ratio for the change in plant species richness (Hill 0) from the 1950s to 2000s (log(2000s response/1950s response))
- Wolf_Years: the number of years that a wolf pack was present at each site since wolves began recolonizing Wisconsin
- Abund_1950_forage: the abundance of plant species that deer highly prefer to eat at each site in the 1950s, calculated by summing the frequencies of 21 high deer preference species at each site
- Precip_summer: average summer precipitation at each site, calculated by taking the sum of precipitation during June, July, and August at each site each year and averaging across 1980-2000. Measured in mm
- Tmax: average maximum temperature at each site, calculated by averaging the maximum daily temperature each month, taking the maximum monthly temperature each year, and averaging across 1980-2000. Measured in degrees Celsius
- AvgBA_1950: average 1950s basal area at each site. Basal area was measured at breast height and in square centimeters for various tree species at each site, and we took the average across species
- Soil_PC1: a summary metric derived as the first axis of a principal components analysis (PCA) of soil properties (pH, % organic matter, % Sand, % Silt, % Clay, % N, P (ppm), K (ppm), Ca (ppm), and Mg (ppm)) within the first 10-cm measured at sites in the 2000s. These data were only available for 78 sites.
- Hunting: a binary variable indicating whether hunting was allowed (1) or not allowed (0) at each site.
- Dist_road_m: the distance from each site to the closest road of any kind (MTFCC Codes S1100, S1200, S1400, S1500), measured in meters
Diversity_SiteValues.csv
Description: data on the diversity and richness of plant species at 83 sites in 1950 and 2005. Columns include:
- Site: a unique numerical ID for each of 83 study sites across northern Wisconsin
- Diversity_1950: plant species diversity, measured as Hill number 1, at each site in the 1950s
- Diversity_2000: plant species diversity, measured as Hill number 1, at each site in the 2000s
- Richness_1950: plant species richness, measured as Hill number 0, at each site in the 1950s
- Richness_2000: plant species richness, measured as Hill number 0, at each site in the 2000s
Frequency_df.csv
Description: data on the frequency of all understory plants identified at sites in 1950 and 2005, and covariates used in the final model of plant frequency. NAs are indicated with blank cells. Columns include:
- Site: a unique numerical ID for each of 83 study sites across northern Wisconsin
- Species: plant species name
- LRR: log response ratio for the change in plant frequency from the 1950s to 2000s (log(2000s response/1950s response))
- Freq_1950: the frequency of each plant species at each site when first surveyed in the 1950s. Frequency is the number of quadrats out of a maximum of 20 at which the species was found
- Freq_2005: the frequency of each plant species at each site when surveyed again in the early 2000s. Frequency is the number of quadrats out of a maximum of 20 at which the species was found
- Wolf_Years: the number of years that a wolf pack was present at each site since wolves began recolonizing Wisconsin
- Abund_1950_forage: the abundance of plant species that deer highly prefer to eat at each site in the 1950s, calculated by summing the frequencies of 21 high deer preference species at each site
- Precip_summer: average summer precipitation at each site, calculated by taking the sum of precipitation during June, July, and August at each site each year and averaging across 1980-2000. Measured in mm
- Tmax: average maximum temperature at each site, calculated by averaging the maximum daily temperature each month, taking the maximum monthly temperature each year, and averaging across 1980-2000. Measured in degrees Celsius
- AvgBA_1950: average 1950s basal area at each site. Basal area was measured at breast height and in square centimeters for various tree species at each site, and we took the average across species
- Soil_PC1: a summary metric derived as the first axis of a principal components analysis (PCA) of soil properties (pH, % organic matter, % Sand, % Silt, % Clay, % N, P (ppm), K (ppm), Ca (ppm), and Mg (ppm)) within the first 10-cm measured at sites in the 2000s. These data were only available for 78 sites.
- Hunting: a binary variable indicating whether hunting was allowed (1) or not allowed (0) at each site.
- Dist_road_m: the distance from each site to the closest road of any kind (MTFCC Codes S1100, S1200, S1400, S1500), measured in meters
GrowthForm_df.csv
Description: data on the frequency of understory plants with 5 distinct growth forms (forb, shrub, tree, graminoid, fern), identified at sites in 1950 and 2005, and covariates used in the final model of plant frequency. NAs are indicated with blank cells. Columns include:
- Site: a unique numerical ID for each of 83 study sites across northern Wisconsin
- Species: plant species name
- LRR: log response ratio for the change in plant frequency from the 1950s to 2000s (log(2000s response/1950s response))
- Freq_1950: the frequency of each plant species at each site when first surveyed in the 1950s. Frequency is the number of quadrats out of a maximum of 20 at which the species was found
- Freq_2005: the frequency of each plant species at each site when surveyed again in the early 2000s. Frequency is the number of quadrats out of a maximum of 20 at which the species was found
- Growth.form: which of five distinct growth forms the species belongs to (forb, shrub, tree, graminoid, fern)
- Wolf_Years: the number of years that a wolf pack was present at each site since wolves began recolonizing Wisconsin
- Abund_1950_forage: the abundance of plant species that deer highly prefer to eat at each site in the 1950s, calculated by summing the frequencies of 21 high deer preference species at each site
- Precip_summer: average summer precipitation at each site, calculated by taking the sum of precipitation during June, July, and August at each site each year and averaging across 1980-2000. Measured in mm
- Tmax: average maximum temperature at each site, calculated by averaging the maximum daily temperature each month, taking the maximum monthly temperature each year, and averaging across 1980-2000. Measured in degrees Celsius
- AvgBA_1950: average 1950s basal area at each site. Basal area was measured at breast height and in square centimeters for various tree species at each site, and we took the average across species
- Soil_PC1: a summary metric derived as the first axis of a principal components analysis (PCA) of soil properties (pH, % organic matter, % Sand, % Silt, % Clay, % N, P (ppm), K (ppm), Ca (ppm), and Mg (ppm)) within the first 10-cm measured at sites in the 2000s. These data were only available for 78 sites.
- Hunting: a binary variable indicating whether hunting was allowed (1) or not allowed (0) at each site.
- Dist_road_m: the distance from each site to the closest road of any kind (MTFCC Codes S1100, S1200, S1400, S1500), measured in meters
WolfResponse_df.csv
Description: data on the number of years a wolf pack occupied a site since they began to recolonize Wisconsin, and covariates used in the final model of wolf occupancy time. Columns include:
- Site: a unique numerical ID for each of 83 study sites across northern Wisconsin
- Wolf_Years: the number of years that a wolf pack was present at each site since wolves began recolonizing Wisconsin
- Roads_100k: road density at each site, measured as kilometers of roads within 100 square kilometers
- SWE: average snow-water equivalent (SWE) at each site as the average of mean annual SWE from 1980-2000, measured in kilograms per square meter
- Dist_Unpaved_road_km: the distance from each site to the closest unpaved road (MTFCC Code S1500), measured in kilometers
- Abund_1950_forage: the abundance of plant species that deer highly prefer to eat at each site in the 1950s, calculated by summing the frequencies of 21 high deer preference species at each site
- Hunting: a binary variable indicating whether hunting was allowed (1) or not allowed (0) at each site.
Code/software
We used the software R (v.4.3.1) to conduct all analyses and make all tables and figures included in our manuscript and supplemental files. The file "WisconsinCascade_Code.R" includes all the code necessary to replicate our results. We used the following packages: tidyverse for data manipulation and figures, rstatix for descriptive statistics, lme4 for mixed models, lmerTest for summaries of mixed models, car to check VIF, DHARMa to check model fit, and emmeans for post-hoc tests and to make predictions.
Access information
Data for some covariates was derived from the following sources:
- Daymet for climatic variables (precipitation, temperature, snow-water equivalent)
- TIGER for shapefiles of roads