Data from: Reduced fire severity offers near-term buffer to climate-driven declines in conifer resilience across the western United States

This archive includes field data, post-fire recruitment models, and spatial projections of post-fire recruitment probability under different climate and fire severity scenarios produced for Davis et al. (2023).

Description of the Data and file structure

Individual datasets in the Dryad archive include the following:

  1. Field plot data. “Davis_et_al_regen_plot_data.csv” This data includes post-fire regeneration density for eight conifer species from the western US that were surveyed in plots 2-30 years following fire. Predictors that were used in the manuscript “Reduced fire severity offers near-term buffer to climate-driven declines in conifer resilience across the western United States” are included with each plot including climate data, fire severity, heat insolation load index, surrounding tree cover, and distance to nearest live seed source. The dataset is a compilation of many datasets. The field methods performed to collect the data varied by dataset/study and are described in detail in each individual study. For a list of the publications that produced each individual dataset please see the supplemental information Table S1 from Davis et al. 2023, the provided .csv file called “data_dryad_contributor_key.csv” or the definitions for each of the values of the contributor id column in the metadata.

  2. Statistical models. The final generalized linear mixed models for recruitment probability for each species and for all species combined used to make spatial projections of post-fire recruitment probability in Davis et al. 2023 are included as .rds files. These files can be opened and used in the free R Statistical Software (https://www.r-project.org). There are seven final models including: one for all study species combined (“all”), one for each individual species: Abies lasiocarpa (subalpine fir; abla), Picea engelmannii (Engelmann spruce; pien), Pinus contorta (lodgepole pine; pico), and Pseudotsuga menziesii (Douglas-fir; psme); one for Pinus ponderosa (ponderosa pine; pipo) and Pinus jeffreyi (Jeffrey pine; pije) combined, and one for Abies concolor (white fir; abco) and Abies grandis (grand fir; abgr) combined. Model files are named with the four-digit PLANTS species codes listed above in parentheses as follows: “[species]_glmm_binomial_model_final.rds”.

  3. Spatial projections. Projections of recruitment probability made with the final models for 10-years post-fire under four climate scenarios (1981-2000, 2001-2020, 2031-2050 RCP 4.5, 2031-2050 RCP 8.5) and two fire severity scenarios (low severity: 10 m to a seed source, 30% surrounding live tree cover within 300-m radius of plot, relativized burn ratio (RBR) of 100; high severity: 150 m to a seed source, 10% surrounding live tree cover within 300-m radius of plot, RBR of 400). Spatial projections are made for all study species combined (“all”) and for each individual species: Abies lasiocarpa (subalpine fir; abla), Picea engelmannii (Engelmann spruce; pien), Pinus contorta (lodgepole pine; pico), and Pseudotsuga menziesii (Douglas-fir; psme). Projections for Pinus ponderosa (ponderosa pine; pipo) and Pinus jeffreyi (Jeffrey pine; pije) are made with one model due to difficulties differentiating seedlings of these species in the field. Projections for Abies concolor (white fir; abco) and Abies grandis (grand fir; abgr) are also made with one model due to the hybridization of these two species in parts of their ranges. Data were produced by first fitting generalized linear mixed models to predict values of recruitment probability in field plot locations (Davis_et_al_regen_plot_data.csv) using a range of spatially explicit predictor variables. Fitted model objects (.rds files) were then used in conjunction with spatial datasets (not included in the current version of this archive due to large file sizes) to develop regional maps. All layers use EPSG:4326 (WGS84/longlat) as the crs. Individual files were saved using the four-digit PLANTS species codes listed above in parentheses. For each species, the file naming convention is as follows: “[species]_[severity]_[time period].tif”. For example, “pico_high_sev_1981_2000.tif” is the projection of recruitment probability for Pinus contorta under the high severity fire scenario with the climate conditions from 1981-2000. Projections with future climate (2031-2050) are the mean of projections made with climate data from five different global climate models (GCMs; see Table S3 in Davis et al. 2023). We made projections under two carbon emission scenarios (RCP 4.5 and RCP 8.5). Future projection files specify the RCP scenario in the file name as follows: “[species]_[severity]_mean_[RCP]_[time period].tif”. For example, “pico_high_sev_mean_rcp45_2031_2050.tif” is the mean of the five individual projections of Pinus contorta recruitment probability under the high severity fire scenario with future climate data from each of the five GCMs with the RCP 4.5 emissions scenario. Plot size in projections is set to 100 m2 so the projections can be interpreted as the probability of at least one seedling regenerating by 10 years post-fire in a 100 m2 plot under the given climate and fire severity scenario, which is equivalent to a density of 100 trees per ha or around 40 trees per acre. Please note that the threshold probability at which recruitment is considered likely varies between species due to differences in the models. Therefore when interpreting the probabilities it is not appropriate to compare raw probabilities between species. It is best to interpret the probabilities in light of the threshold probabilities above which recruitment is most likely, which are provided below, in the metadata, and in Table S9 in Davis et al 2023. There are several ways to calculate the threshold probability above which recruitment is most likely. We provide the thresholds using two widely used methods. “Threshold.ss” refers to the probability threshold used to categorize recruitment as likely or unlikely which maximizes the sum of specificity and sensitivity. “Threshold.k” refers to the probability threshold used to categorize recruitment as likely or unlikely which maximizes Cohen’s kappa. Because these two different methods yield different thresholds for some species, it is best to interpret probability values between the two threshold values as indicative of higher uncertainty and sites where post-fire recruitment may or may not occur under the given scenario. Thresholds for each species model and the all-species model are as follows:

Species Threshold.ss Threshold.k
Abies concolor/A. grandis 0.35 0.36
Abies lasiocarpa 0.16 0.37
Picea engelmannii 0.50 0.50
Pinus contorta 0.59 0.46
Pinus ponderosa/P. jeffreyi 0.22 0.28
Pseudotsuga menziesii 0.39 0.44
All species 0.64 0.54

Please see the metadata file for each dataset (“Davis_et_al_postfire_recruitment_plot_data_2022.xml” and “Davis_et_al_postfire_recruitment_projections_2022.xml”) for detailed descriptions of the datasets and their components.

Sharing/access Information

Links to other publicly accessible locations of the data: This compiled dataset currently only available on Datadryad.org. Parts of the field data have been published elsewhere and these publications and links are listed in the “data_contributors_key.csv” file, the metadata, and Table S1 in Davis et al. 2023.

Was data derived from another source? Yes If yes, list source(s): Pre-existing spatial data used as input data for the post-fire recruitment models and projections are described in detail in the metadata.