Wildfire and Biodiversity Meta-Analysis Dataset (European Forests)

Overview Here is the dataset, code, and results of a meta-analysis examining the effects of fires on species abundances within various forest ecological groups. The analysis focuses on how different taxa respond to fire events of varying severity, providing quantitative estimates of these responses using effect sizes. The dataset provides 2016 unique effect sizes from 1068 unique species reported in 36 studies investigating wild or prescirbed fires. The study sites covered four European forest biomes (Mediterranean Forests, woodlands and scrubs; Temperate broadleaf and mixed forests; Temperate Coniferous Forest, and Boreal forests/Taiga) and 15 ecoregions (Olson et al., 2001). The main CSV file contains all the input and output data. The primary input data contains data from published studies examining fire effects on various taxonomic groups. Each row represents a unique effect size from a comparison between burned (treatment) and unburned (control) conditions. A second CSV file contains the citation information for the 36 included studies the main dataset. The R script contains code to (1) estimate effect sizes from the input data, and (2) run the meta-analytical model. Variable description for the main dataset Input Data Variables: Metadata study_id: Unique identifier for each study author_year: Author name(s) and publication year biome_wwf: World Wildlife Fund biome classification biome_category_wwf: Category of WWF biome ecoregion_wwf: WWF ecoregion classification country: Country where study was conducted Input Data Variables: Fire fire_type: Type of fire (e.g., prescribed, wildfire) fire_severity: Original fire severity classificationfire_severity_binned: Binned fire severity (simplified categories, years) fire_extent_km2: Extent of fire in square kilometerstime_since_fire_years: Time since fire event in years Input Data Variables: Biological group: General taxonomic group pollinator: Indicates if study examined pollinators taxa: Specific taxonomic classification (Genus or species where possible) mean_control: Mean value for control (unburned) group sd_control: Standard deviation for control group n_control: Sample size for control group mean_treatment: Mean value for treatment (burned) group sd_treatment: Standard deviation for treatment group n_treatment: Sample size for treatment group mean_control_adj: Adjusted mean for control group mean_treatment_adj: Adjusted mean for treatment group Output Data Variables: Calculated/Additional Variables ES_ID: Unique identifier for each effect size observation cv_control: Coefficient of variation for control group cv_treatment: Coefficient of variation for treatment group cv2_cont_new: Squared coefficient of variation for control group cv2_treatment_new: Squared coefficient of variation for treatment group lnrr_laj: Log response ratio (effect size) calculated using Lajeunesse method v_lnrr_1A: Variance of the log response ratio Predicted: Predicted effect size from meta-regression model Residuals: Difference between observed and predicted effect size b_CV2_1: Between-study coefficient of variation squared for control b_CV2_2: Between-study coefficient of variation squared for treatment Interpretation Notes Effect sizes (lnrr_laj) represent the natural log of the ratio between treatment (burned) and control (unburned) means Positive values indicate higher values in burned areas compared to unburned areas Negative values indicate lower values in burned areas compared to unburned areas The analysis accounts for between-study and within-study heterogeneity through the multilevel structure Software This analysis was conducted using R with the following packages: metafor (for meta-analysis) tidyverse (for data manipulation) ggplot2 (for visualization) dmetar (for heterogeneity assessment) Analysis Methods The meta-analysis was conducted using the following approach: 1. Data Preparation: Data was cleaned and prepared for analysis. For the input data, where there were observations with a zero for either the control or the treatment (i.e., an abundance of zero), to address caveat (1), we used a commonly used adjustment factor of 0.5 (Schwarzer, 2007; Weber et al., 2020) added to the treatment and control mean if one or the other was zero. For data where the study reported both mean and standard deviation, effect sizes were calculated using log response ratios (lnrr) following Lajeunesse's method (Lajeunesse, 2015). Where SDs were not reported from each study, they were estimated these using the pooled CVs from the subset of studies that do report SDs and applied the ‘missing cases’ method (Nakagawa et al., 2023) to calculate ESs and sampling variances. Outliers were robustly identified and removed (observations with residuals > |3|). 2. Meta-Analysis Models: Multilevel mixed-effects meta-analysis was conducted using the `metafor` package in R. Random effects structure included nested random effects (effect sizes nested within studies). Main model included interaction between taxonomic group (group) and fire severity (fire_severity_binned). Additional models examined effects by taxonomic group alone and fire severity alone. Post-fire recovery time was also explored as a moderator (0-5 years vs. 5-10 years) 3. Model Diagnostics: Residual analysis was performed to check model assumptions. Heterogeneity was assessed using multilevel I² statistics. Predictions were generated for each combination of taxa group and fire severity. References Lajeunesse, M.J., 2015. Bias and correction for the log response ratio in ecological meta‐analysis. Ecology 96, 2056–2063. https://doi.org/10.1890/14-2402.1 Nakagawa, S., Noble, D.W.A., Lagisz, M., Spake, R., Viechtbauer, W., Senior, A.M., 2023. A robust and readily implementable method for the meta‐analysis of response ratios with and without missing standard deviations. Ecology Letters 26, 232–244. https://doi.org/10.1111/ele.14144 Olson, D.M., Dinerstein, E., Wikramanayake, E.D., Burgess, N.D., Powell, G.V.N., Allnut, T.F., Ricketts, T.H., Kura, Y., Lamoreux, J.F., Wettengel, W.W., Hedao, P., Kassem, K.R., 2001. Terrestrial ecoregions of the world: a new map of life on Earth. BioScience 51, 933–938. Schwarzer, G., 2007. meta: An R package for meta-analysis. R News 7, 40–45. Weber, F., Knapp, G., Ickstadt, K., Kundt, G., Glass, Ä., 2020. Zero‐cell corrections in random‐effects meta‐analyses. Research Synthesis Methods 11, 913–919. https://doi.org/10.1002/jrsm.1460数据集概览 本数据集包含一项元分析（meta-analysis）的相关数据、代码与结果，该元分析旨在探究火烧对不同森林生态类群物种丰度的影响。本分析聚焦于不同分类群对不同烈度火烧事件的响应，并通过效应量（effect sizes）对这些响应进行定量估算。本数据集源自36项针对野火或计划火烧（prescribed fire）的研究，包含1068个不同物种的2016个独立效应量。研究样地覆盖4种欧洲森林生物群区：地中海森林、灌丛与灌草丛；温带阔叶混交林；温带针叶林；寒温带针叶林（泰加林），以及15个生态区（Olson等，2001）。 主CSV文件包含所有输入与输出数据。核心输入数据源自探讨不同分类群火烧效应的已发表研究，每一行对应一次火烧（处理组）与未火烧（对照组）对照所得到的独立效应量。 第二个CSV文件包含本数据集所纳入的36项研究的引文信息。 本数据集附带的R脚本可实现两项功能：（1）基于输入数据计算效应量；（2）运行元分析模型。 主数据集变量说明 输入数据变量：元数据 study_id：每项研究的唯一标识符 author_year：作者姓名与发表年份 biome_wwf：世界自然基金会（World Wildlife Fund, WWF）生物群区分类 biome_category_wwf：WWF生物群区类别 ecoregion_wwf：WWF生态区分类 country：研究开展所在国家 输入数据变量：火烧相关 fire_type：火烧类型，如计划火烧、野火 fire_severity：原始火烧烈度分类 fire_severity_binned：分箱火烧烈度（简化分类） fire_extent_km2：火烧面积，单位：平方千米 time_since_fire_years：火烧后时长，单位：年 输入数据变量：生物学相关 group：广义分类群 pollinator：传粉者标识，用于标识研究是否涉及传粉者类群 taxa：具体分类群，尽可能精确至属或种 mean_control：对照组（未火烧）均值 sd_control：对照组标准差 n_control：对照组样本量 mean_treatment：处理组（火烧）均值 sd_treatment：处理组标准差 n_treatment：处理组样本量 mean_control_adj：对照组校正均值 mean_treatment_adj：处理组校正均值 输出数据变量：计算所得/附加变量 ES_ID：效应量观测唯一标识符 cv_control：对照组变异系数 cv_treatment：处理组变异系数 cv2_cont_new：对照组变异系数平方 cv2_treatment_new：处理组变异系数平方 lnrr_laj：对数响应比（log response ratio，效应量），采用Lajeunesse法计算所得 v_lnrr_1A：对数响应比方差 Predicted：预测效应量，源自元回归模型 Residuals：残差，观测效应量与预测效应量的差值 b_CV2_1：对照组研究间变异系数平方 b_CV2_2：处理组研究间变异系数平方 结果解读说明 效应量（lnrr_laj）为处理组（火烧）与对照组（未火烧）均值比值的自然对数 正值代表火烧区域的物种丰度高于未火烧区域 负值代表火烧区域的物种丰度低于未火烧区域 本分析通过多层结构控制研究间与研究内异质性 分析软件 本分析基于R语言开展，使用的扩展包包括： metafor：用于元分析 tidyverse：用于数据处理 ggplot2：用于可视化 dmetar：用于异质性评估 分析方法 本元分析采用如下流程开展： 1. 数据预处理：对数据进行清洗与整理。针对对照组或处理组中存在丰度为0的观测值的情况，采用通用的0.5校正系数（Schwarzer, 2007; Weber等, 2020），即当两组均值任一为0时，向两组均值同时添加0.5以解决零值问题。对于同时报告了均值与标准差的研究，采用Lajeunesse法（Lajeunesse, 2015）计算对数响应比（lnrr）作为效应量。对于未报告标准差的研究，从已报告标准差的研究子集合并得到变异系数（CV）以估算标准差，并采用“缺失值处理法”（Nakagawa等, 2023）计算效应量与抽样方差。通过残差绝对值大于3的标准稳健识别并移除异常值。 2. 元分析模型：基于R语言的metafor包构建多层混合效应元分析模型。随机效应结构采用嵌套随机效应（效应量嵌套于研究内）。主模型纳入分类群（group）与分箱火烧烈度（fire_severity_binned）的交互项。此外还构建了仅以分类群或仅以火烧烈度为调节变量的模型。同时探究了火烧后恢复时长（0~5年 vs. 5~10年）作为调节变量的效应。 3. 模型诊断：通过残差分析检验模型假设，采用多层I²统计量评估异质性，并针对分类群与火烧烈度的所有组合生成预测效应量。 参考文献 Lajeunesse, M.J., 2015. Bias and correction for the log response ratio in ecological meta‐analysis. Ecology 96, 2056–2063. https://doi.org/10.1890/14-2402.1 Nakagawa, S., Noble, D.W.A., Lagisz, M., Spake, R., Viechtbauer, W., Senior, A.M., 2023. A robust and readily implementable method for the meta‐analysis of response ratios with and without missing standard deviations. Ecology Letters 26, 232–244. https://doi.org/10.1111/ele.14144 Olson, D.M., Dinerstein, E., Wikramanayake, E.D., Burgess, N.D., Powell, G.V.N., Allnut, T.F., Ricketts, T.H., Kura, Y., Lamoreux, J.F., Wettengel, W.W., Hedao, P., Kassem, K.R., 2001. Terrestrial ecoregions of the world: a new map of life on Earth. BioScience 51, 933–938. Schwarzer, G., 2007. meta: An R package for meta-analysis. R News 7, 40–45. Weber, F., Knapp, G., Ickstadt, K., Kundt, G., Glass, Ä., 2020. Zero‐cell corrections in random‐effects meta‐analyses. Research Synthesis Methods 11, 913–919. https://doi.org/10.1002/jrsm.1460