Pyrodiversity of boreal lake islands begets biodiversity of beetles, plants, and birds
Data files
Jan 23, 2026 version files 168.50 KB
-
alpha_diversity_ALL_islands.csv
3.44 KB
-
alpha_diversity_model_selection_19islands.csv
2.49 KB
-
Beetle_species_list.xlsx
39.22 KB
-
beetles_speciesmatrix_presenceabsence.csv
49 KB
-
BetaDiversity_Analysis_RCode.R
9.54 KB
-
borealbirds_speciesmatrix_presenceabsence.csv
4.98 KB
-
island_isolation.csv
558 B
-
island_logarea.csv
686 B
-
Model_Selection_19islands_RCode.R
12.47 KB
-
RDA_analysis.R
6.47 KB
-
RDA_environmental_variables.csv
8.59 KB
-
README.md
9.41 KB
-
time_since_fire_beetles.csv
310 B
-
time_since_fire_birds.csv
309 B
-
time_since_fire_plants.csv
308 B
-
vascularplants_speciesmatrix_presenceabsence_RDA.csv
10.41 KB
-
vascularplants_speciesmatrix_presenceabsence.csv
10.31 KB
Abstract
Global fire regimes are changing, raising concerns about the ability of fire-prone ecosystems to maintain biodiversity. We tested whether the pyrodiversity-biodiversity hypothesis (i.e., variation in post-fire characteristics promotes biodiversity) or alternative hypotheses better explain patterns of biodiversity in a true island system. Using fixed-area sampling plots in a chronosequence of 42 boreal lake islands spanning gradients in island area (1-350.4 ha), isolation (0.1-7.9 km from mainland), and fire history (1-231+ year since fire), we tested whether alpha and beta diversity of beetles, plants, and birds increased with spatial (within-island variation in burn severity) and temporal (variation in time since fire among islands) pyrodiversity, respectively. Species richness of plants and birds increased with spatial pyrodiversity, indicating that habitat heterogeneity from localized variation in burn severity supported more species in some groups. Beta diversity of all taxa increased with temporal pyrodiversity, highlighting the importance of conserving age-class variation within the boreal patch mosaic. In contrast, the habitat amount hypothesis and island biogeography theory were weak predictors of species richness across all taxa, and island area and isolation did not consistently affect beta diversity among the islands. Our findings emphasize the importance of maintaining pyrodiversity in boreal landscapes to combat biodiversity loss in the age of “megafires” and suggest leveraging the fire refugia effects of large lakes within the region to conserve vital components of temporal pyrodiversity such as old-growth forests.
https://doi.org/10.5061/dryad.tdz08kq78
These data are from a field experiment conducted in the Lac la Ronge Provincial Park region of northern Saskatchewan, Canada between 2019-2022. The experiment made use of 42 islands spanning gradients in time since fire, island area, and isolation with data collected from a single 30 x 30 m plot established on each island. The edge of each island plot was at least 15 meters from the water's edge. Information on local habitat variables as well as species data for beetles, plants, and birds that were collected from each island plot are included in the accompanying .csv and R code files.
The .csv file 'alpha_diversity_model_selection_19islands.csv' contains information that was used in the model selection analysis of our paper. Island refers to a unique two-digit island code for each island, 'year.burned' is the calendar year of the most recent burn for each island, 'fire.stdev' is the standard deviation of relativized burn ratio scores for each island that was obtained from the burn severity rasters constructed in Google Earth Engine, and 'fire.stdev.stand' is the standard deviation of relativized burn ratio scores for each island after applying a resampling procedure to account for differences in island size (see manuscript). The variables 'beetle.richness', 'bird.richness', and 'plant.richness' correspond to the number of species of these groups detected on each island. The variables 'beetle.msom.richness' and 'bird.msom.richness' are the adjusted species richness values for beetles and birds calculated using multi-species occupancy models, which accounted for species-level variation in detection. Island area is listed under the variable 'log10area' and was calculated as log10(area+1) to avoid negative values associated with small island sizes. Island isolation is listed under the variable 'buffer5000' as a proportion of land within a 5 km buffer around each island, with a value of 1 indicating an island is completely surrounded by water (no land). Time since fire (e.g. time.since.fire.beetles, time.since.fire.plants, and time.since.fire.birds) is the number of years between the age of the oldest tree at each island site and when each of the major taxa was sampled. The variables 'Lat' and 'Long' indicate the latitudinal and longitudinal coordinates (decimal degrees) for the center of each island plot. The variables 'buffer_1' and 'buffer_10' refer to the habitat amount (value of 1 indicating an island is completely surrounded by land/habitat) for each island measured at scales of 1 km (birds) and 10 km (beetles/plants, see Supplemental Material of paper).
The R code file 'Model_Selection_19islands_RCode.R' contains R code that was used to perform our model selection analysis with the data contained in the .csv file 'alpha_diversity_model_selection_19islands'. This file also includes code that was used to calculate the p-values, regression coefficients, and deviance explained by the models, and to test for spatial autocorrelation in our models.
The .csv file 'alpha_diversity_ALL_islands.csv' contains the same variables and data descriptions as the file 'alpha_diversity_model_selection_19islands' except that it also includes data for an additional 23 islands that were not included in our model selection analysis.
The .csv files containing island species community data (beetles_speciesmatrix_presenceabsence.csv, vascularplants_speciesmatrix_presenceabsence.csv, and borealbirds_speciesmatrix_presenceabsence.csv) were used to measure beta diversity (jaccard's index) across gradients of time since fire, island area, and island isolation. The information in each community matrix includes data on the presence/absence for species (columns; 466 beetle species, 101 plants species, and 54 birds species for the full set of 42 islands (rows). A cell value of 1 indicates presence and a value of 0 indicates absence. Note: the files vascularplants_speciesmatrix_presenceabsence.csv and vascularplants_speciesmatrix_presenceabsence_RDA.csv have empty cells for the island 'BS' because this island burned before vascular plant surveys could be conducted.
The .csv file 'island_isolation.csv' contains isolation values for each island under the column 'buffer5000'. The values are the proportion of land within a 5 km buffer around each island, with a value of 1 indicating an island is completely surrounded by water (no land). This file was used to calculate pairwise differences in isolation for the 42 islands used in our beta diversity analysis.
The .csv file 'island_logarea.csv' contains data on total island area for each island (log10(area+1) transformed). This file was used to calculate pairwise differences in island area for the 42 islands used in our beta diversity analysis.
The .csv files 'time_since_fire_beetles', 'time_since_fire_plants', and 'time_since_fire_birds' contain the column 'TSF' which indicates the estimated number of years. This estimate varies slightly across taxa and sites depending on when sampling occurred during our three year study.
The R code file 'BetaDiversity_Analysis_RCode.R' contains R code that was used to model pairwise dissimilarities in our species community data as a function of pairwise differences in time since fire (temporal pyrodiversity), island area, and island isolation. This code uses data contained in the species matrices ('beetles_speciesmatrix_presenceabsence.csv', 'plants_speciesmatrix_presenceabsence.csv', and 'birds_speciesmatrix_presenceabsence.csv') and the files used to construct our time since fire ('time_since_fire_beetles.csv', 'time_since_fire_plants.csv', and 'time_since_fire_birds.csv'), island area ('island_logarea.csv'), and island isolation ('island_isolation.csv') distance matrices.
The .csv file 'RDA_environmental_variables.csv' includes environmental variables that were used in our Redundancy Analysis (RDA) and step function analysis. Island refers to a unique two-digit island code for each island, 'plant.richness' is the total number of plant species detected on each island, 'log10area' is the area of each island (subject to a log10(area+1) transformation), 'buffer5000' is a measure of island isolation and reflects the proportion of land within a 5 km buffer around each island, with a value of 1 indicating an island is completely surrounded by water (no land). The variables 'time.since.fire.beetles', 'time.since.fire.plants', and 'time.since.fire.birds' is the number of years between the age of the oldest tree at each island site and when each of the major taxa was sampled on each island. The variables 'charcoal.yes' and 'charcoal.no' indicate the volume of coarse woody debris with and without charcoal, respectively, for each island (units in m3). The variables 'frugivore.abund' and 'insectivore.abund' refers to the counts of frugivorous and insectivorous birds, respectively. 'Bird.abundance' is the total number of bird counts for each island. 'Basal.area' is the cross-sectional area of living and dead trees measured at diameter breast height within a 10 m radius of the center of each plot (units in m3). 'Beetle.catch.rate' is the number of beetle specimens collected on each island divided by the total number of trap days. The variables 'deciduous.canopy.cover.percent.a.d' and 'deciduous.understory.cover.percent.a.d' refers to the % cover of deciduous vegetation measured using the line-intercept method, where overstory includes stems ≥ 4 cm diameter, > 2 m in height and understory includes stems < 4 cm diameter breast height [DBH], < 2 m in height. The variables 'conifer.canopy.cover.perecent.a.d' and 'conifer.understory.canopy.percent.a.d' is the same as above, but for conifers and the 'a.d' at the end of each variable name indicates that the data for alive and dead vegetation was combined. The variable 'total.cwd.volume' (units in m3) refers to the total volume of coarse woody debris (CWD, ≥ 4 cm diameter), estimated using the formula: V = π r2 h, where r is the radius at the middle of each piece and h is the length of each piece of CWD in each island plot. The variables 'cwd.dc1', 'cwd.dc2', etc. are the volume of CWD listed by decay class (units in m3), with decay class 1 (i.e., cwd.dc1) referring to fresh deadwood and decay class 5 being the most-decayed stage.
The .xlsx file 'Beetle_species_list.xlsx' contains the complete list of 466 beetle species (Coleoptera) sampled on 42 lake islands (1 to 231+ years since fire) in the Lac la Ronge region following the nomenclature of Bousquet et al. (2013). Beetle species marked with an asterisk (*) indicate those that were only detected on either of the two islands sampled in the year immediately following fire (islands MI and HF, see Supplemental Table 1). References for body size information are listed in the Supplemental Material.
The R code file 'RDA_analysis.R' contains R code that was used for Redundancy Analysis of our species community data. This code uses data contained in the .csv files 'beetles_speciesmatrix_presenceabsence.csv', 'bird_speciesmatrix_presenceabsence.csv', and 'plant_speciesmatrix_presenceabsence_RDA.csv' (note that the name changed for the plant data) and the environmental data in .csv file 'RDA_environmental_variables.csv'.