Data from: Tamm Review: a meta-analysis of thinning, prescribed fire, and wildfire effects on subsequent wildfire severity in conifer dominated forests of the Western US
Abstract
Increased understanding of how active forest management (i.e., mechanical thinning, prescribed burning, and managed wildfire) affects subsequent wildfire severity is urgently needed as people and forests face a growing wildfire crisis. In response, we reviewed scientific literature for the US West and completed a meta-analysis that answered three questions: (1) How much do treatments reduce wildfire severity within treated areas? (2) How do the effects vary with treatment type, treatment age, and forest type? (3) How does fire weather moderate the effects of treatments? We found overwhelming evidence that mechanical thinning with prescribed burning, mechanical thinning with pile burning, and prescribed burning only are effective at reducing subsequent wildfire severity, resulting in reductions in severity from 62% to 72% relative to untreated areas. In comparison, thinning only was less effective – underscoring the importance of treating surface fuels when mitigating wildfire severity is the management goal. The efficacy of these treatments did not vary among forest types assessed in this study and was high across a range of fire weather conditions. Prior wildfire had more complex impacts on subsequent wildfire severity, which varied with forest type and initial wildfire severity. Across treatment types, we found that effectiveness of treatments declined over time, with the mean reduction in wildfire severity decreasing nearly threefold when wildfire occurred greater than 10 years after initial treatment. Our meta-analysis provides up-to-date information on the extent to which active forest management reduces wildfire severity and facilitates better outcomes for people and forests during future wildfire events.
README
Data from: Tamm Review: A meta-analysis of thinning, prescribed fire, and wildfire effects on subsequent wildfire severity in conifer dominated forests of the Western US
https://doi.org/10.5061/dryad.zcrjdfnkp
This data was extracted from existing publications (through 2022) to use in analyses for the publication "Tamm Review: A meta-analysis of thinning, prescribed fire, and wildfire effects on subsequent wildfire severity in conifer dominated forests of the Western US" (available here: https://doi.org/10.1016/j.foreco.2024.121885). This paper is a meta-analysis of the effects of thinning, prescribed fire, and wildfire on subsequent wildfire severity. Studies included in this dataset had to meet the following criteria: (1) Data was collected in conifer-dominated forests in the western US; (2) Empirical data on wildfire severity in both treated and untreated areas (i.e., controls) was collected (no modeling results are included); (3) Wildfire occurred after the treatment was completed; (4) The variables measured after the wildfire needed to include some measure of wildfire severity, including: bole char height, crown scorch height, percent tree or basal area mortality, percent canopy cover change, percent crown scorch, percent crown consumption, wildfire severity derived from satellite imagery (RBR, dNBR, correlated CBI, or RdNBR), and/or percent of area burned at high severity; and (5) treatments examined included one or more of the following: prescribed fire alone (“prescribed burn”), thinning alone (“thin only”), thinning in combination with slash/activity fuel removal (largely through pile burning, but at two sites slash was removed from the site; “thin and pile burn”), thinning in combination with prescribed burning (“thin and prescribed burn”), or prior wildfire (i.e., studies that examined short-interval fires). The data includes study information, treatment information, fire information, fire weather when available, mean fire severity in treated and untreated areas, standard deviation of fire severity in treated and untreated areas, the measure of fire severity used in the study, and sample size in treated and untreated areas when available.
Description of the Data and file structure
The dataset in the Dryad archive include the following:
- Davis_et_al_forest_treatment_metaanalysis.csv: This dataset includes all the data used in our analysis. It includes fire severity in treated and untreated areas that were subsequently burned by a wildfire. Information is also provided on the study the data was extracted from, attributes of the wildfire, and treatment type and age. The dataset is a compilation of information from 40 studies. The methods used to collect the data varied by study and are described in detail in each individual study. For a list of the publications associated with each individual study please see the supplemental information Table A1 from Davis et al. 2024. Missing values are given by "NA."
Please see the metadata file ("Davis_et_al_forest_treatment_metaanalysis.xml") for detailed descriptions of the dataset and its components. A brief description of each column in the dataset follows:
- Authors: Authors of the original study we extracted data from
- Study_year: Year the original study was published
- Journal: Outlet of original publication
- Title: Title of original publication
- Study_id: Unique identifier for each study created to be used as a random effect in the meta-analysis models
- Pub_Type: Type of publication
- Latitude: Approximate latitude of study location in a geographic coordinate system with datum WGS 1984 in decimal degrees. Note that many of these coordinates are approximate based on maps provided in the study and/or a point taken from the center of the wildfire in which treatments occurred. They do not represent exact field plot locations.
- Approximate longitude of study location in a geographic coordinate system with datum WGS 1984 in decimal degrees. Note that many of these coordinates are approximate based on maps provided in the study and/or a point taken from the center of the wildfire in which treatments occurred. They do not represent exact field plot locations.
- Ecoregion: Ecoregion in which the study occurred (https://geospatial.tnc.org/datasets/b1636d640ede4d6ca8f5e369f2dc368b/about)
- Forest_type: Broad forest type group based on site descriptions in each study. Categorical variable with below categories:
- CA mixed con: Forest type dominated by some combination of ponderosa pine (Pinus ponderosa), Jeffrey pine (Pinus jeffreyi), white fir (Abies concolor), red fir (Abies magnifica), incense cedar (Calocedrus decurrens), sugar pine (Pinus lambertiana), black oak (Quercus kelloggii), and Douglas-fir (Pseudotsuga menziesii). This forest type was found across the Sierra Nevada and Klamath ecoregions, and not all species were present at all sites.
- Int mixed con: Forest type dominated by ponderosa pine (Pinus ponderosa) and Douglas-fir (Pseudotsuga menziesii), with additional components of lodgepole pine (Pinus contorta), western larch (Larix occidentalis), grand fir (Abies grandis), white fir (Abies concolor), southwestern white pine (Pinus strobiformis), blue spruce (Picea pungens), and/or Engelmann spruce (Picea engelmannii) depending on geographical location. This forest type spanned the interior Northwest through the Rocky Mountains and Southwest.
- Lodgepole: Forest type dominated primarily by lodgepole pine (Pinus contorta; rather than sites that included lodgepole as a component of mixed conifer or subalpine forests). These sites were constrained to the Klamath ecoregion but were not combined with CA mixed conifer due to the different historical fire regimes and tree species traits associated with these two forest types.
- Ponderosa_Jeffrey: Forest type of almost pure ponderosa (Pinus ponderosa) and/or Jeffrey pine (Pinus jeffreyi), with only occasional components of other species such as Gambel’s oak (Quercus gambelii), Douglas-fir (Pseudotsuga menziesii), or incense cedar (Calocedrus decurrens). These sites were broadly distributed including in the interior Northwest, Sierra Nevada and Southwest.
- Subalpine: Forest type dominated by subalpine fir (Abies lasiocarpa), Engelmann spruce (Picea engelmannii), and lodgepole pine (Pinus contorta). Sites with this forest type were located in the Middle Rockies – Blue Mountains, Okanagan, and UT-WY Rockies ecoregions.
- Treatment_type: Type of treatment that occurred in study location. Categorical variable with below categories:
- Rx_burn: Prescribed fire/broadcast burn without prior thinning
- Thin_PB: Thinning treatment and activity fuels/slash were either piled and burned or removed from the site
- Thin_Rx_burn: Thinning treatment and a prescribed fire/broadcast burn following the thinning
- Thin: Thinning treatment only, without subsequent pile or broadcast burning
- Wildfire: Area burned in a prior wildfire
- Prior_fire_severity: In the case when treatment was a wildfire, severity of the prior wildfire (if reported) as a categorical value (low, moderate, high)
- Treatment_age_reported: Age of treatments when they burned. Either exact age in years is provided or the range of ages is provided (e.g. "<9" is treatments were <9 years old when they burned). When two numbers are separated by an underscore ("_") this indicates the treatment age ranged between those two numbers (e.g., "2_9" is treatments were between 2 and 9 years old when they burned).
- Treatment_age_category: Age of treatment when burned by a subsequent wildfire grouped into two categories for analysis.
- "≤10": All treatments occurred 10 years or less prior to being burned by wildfire.
- ">10": Treatments were greater than 10 years old when burned by wildfire. In the case that treatments younger and older than 10 years were grouped and only one effect size was reported, then we included those observations in this category. For example, if some of the treated area was completed 12 years before affected by wildfire and some was treated 7 years before the wildfire, we would categorize that observation as “>10.”
- Time_since_trt_exact: Age of treatment (years) when burned by a subsequent wildfire for studies that provided the exact age of treatments and not a range of values.
- Max_wind_speed: Maximum wind speed (kmh) during the wildfire event or during the portion of the wildfire when the study sites burned as reported in the initial study
- Min_relative_humidity: Minimum relative humidity (%) during the wildfire event or during the portion of the wildfire when the study sites burned as reported in the initial study
- Max_temp: Maximum temperature (degrees Celsius) during the wildfire event or during the portion of the wildfire when the study sites burned as reported in the initial study
- ERC: Energy release component (ERC) during the wildfire event or during the portion of the wildfire when the study sites burned as reported in the initial study
- Fuel_moisture_10_hr: 10-hour fuel moisture (%) during the wildfire event or during the portion of the wildfire when the study sites burned as reported in the initial study
- Fire_name: The name of the wildfire event that burned each treatment. In the case that a study examined multiple wildfires and combined them into one observation this field will be "various."
- Fire_year: Year the the wildfire which burned through treated areas occurred.
- Site_ID: Unique site id created for each site (not reflective of original site names in original studies) used as a random effect in the meta-analysis models because some sites had multiple observations (e.g. with different metrics of fire severity). Some studies had multiple sites.
- Mean_severity_control: Mean fire severity in the control/untreated area. The way in which severity is measured varies and is indicated in "Severity_type" and "Severity_unit" columns.
- Mean_severity_treated: Mean fire severity in the treated area. The way in which severity is measured varies and is indicated in "Severity_type" and "Severity_unit" columns.
- SD_control: Standard deviation of fire severity in the control/untreated area. The way in which severity is measured varies and is indicated in "Severity_type" and "Severity_unit" columns.
- SD_treated: Standard deviation of fire severity in the treated area. The way in which severity is measured varies and is indicated in "Severity_type" and "Severity_unit" columns.
- Severity_type: Measure of fire severity used in this study, which is recorded in the "mean_severity" columns and the units of which are described in the "Severity_unit" column.
- Severity_unit: The units for the "Mean_severity_control/treated" columns for the measure of fire severity recorded in the "Severity_type" column.Potential values include:
- CBI_from_RdNBR: Composite Burn Index calculated from the relative differenced Normalized Burn Ratio
- dNBR: differenced Normalized Burn Ratio derived from satellite imagery
- m: meter
- Percent
- RdNBR: relative differenced Normalized Burn Ratio derived from satellite imagery
- Severity_group: Measure of wildfire severity used in the study grouped into five groups and used as a random effect in the meta-analytic models. Potential values include:
- per_high_severity_effect: Percent area that burned at high severity (based on satellite-derived severity metrics)
- satellite_severity_effect: Satellite-derived (e.g., Landsat) fire severity measured as RBR, dNBR, CBI (modelled from RdNBR), or RdNBR
- tree_mort_ba_loss_effect: Percent tree mortality or tree basal area/canopy cover loss
- crown_scorch_torch_effect: Percent crown volume scorched or torched (consumed)
- char_scorch_height_effect: Bole char height or scorch height
- Sample_size_control: Sample size of the control/untreated area; units varied by study (e.g. plots, treatment units, trees, or pixels).
- Sample_size_treated: Sample size of the treated area; units varied by study (e.g. plots, treatment units, trees, or pixels).
- Mean_method: For studies that reported medians (including with boxplots) this column indicates the method used to estimate the mean. We estimated means and standard deviations using the method for unknown non-normal distributions (MLN) in the estmeansd R package (McGrath et al., 2023). In the case that the first quartile was < 0 (2.6% of observations), we used the estimator for the mean from Luo et al. (2018) and the estimator for the standard deviation from Wan et al. (2014) because the MLN method requires positive values. Potential values include:
- manuscript: The mean was provided in the manuscript and thus not estimated.
- mln: We used the method for unknown non-normal distributions (MLN) in the estmeansd R package (McGrath et al., 2023) to estimate the mean from the median and interquartile range.
- luo_wan: We used the estimator for the mean from Luo et al. (2018) because the first quartile was <0.
- SD_method: For studies that reported medians and interquartile ranges (including with boxplots) this column indicates the method used to estimate the standard deviation of the mean. We estimated means and standard deviations using the method for unknown non-normal distributions (MLN) in the estmeansd R package (McGrath et al., 2023). In the case that the first quartile was < 0 (2.6% of observations), we used the estimator for the mean from Luo et al. (2018) and the estimator for the standard deviation from Wan et al. (2014) because the MLN method requires positive values. Potential values include:
- manuscript: The standard deviation or standard error and sample size were provided in the manuscript and thus we did not need to estimate them.
- mln: We used the method for unknown non-normal distributions (MLN) in the estmeansd R package (McGrath et al., 2023) to estimate the standard deviation from the median and interquartile range.
- to_impute: Not enough information was provided to estimate the standard deviation so we imputed the standard deviation with predictive mean matching and 100 imputations for use in the meta-analytic models.
- luo_wan: We used the estimator for the standard deviation from Wan et al. (2014) because the first quartile was <0.
Sharing/access Information
Links to other publicly accessible locations of the data:
This compiled dataset is currently only available on Datadryad.org.
All data were extracted from existing publications listed in Table A1 in Davis et al. 2024.
Was data derived from another source? Yes
If yes, list source(s):
All data were extracted from existing publications listed in Table A1 in Davis et al. 2024 and described in more detail in the metadata file ("Davis_et_al_forest_treatment_metaanalysis.xml").