Africa MATRIX: Demographic compensation across biomes stabilizes Africa’s forest structure under climate change

OverviewThis Figshare deposit provides the processed datasets, trained Random Forest model objects, validation outputs, and climate covariates supporting the Africa MATRIX forest demographic modeling framework.These materials enable reproduction of model training, evaluation, and simulation results reported in the associated manuscript, including demographic rate estimation (mortality, upgrowth, recruitment), cross-validation, cross-sample interval validation (Inventory 1 → Inventory 2), and long-term forest projections (2010–2100) under constant and scenario-based climate forcing (e.g., SSP1–2.6 and SSP5–8.5).All simulations were implemented in R and executed on Linux-based high-performance computing systems. The complete preprocessing, training, validation, and projection scripts are available separately via the companion GitHub repository.Contents of this depositThis repository includes the following components:Processed training and simulation datasetsPlot- and state-space–level inputs used for model training and projection, including species-group × DBH-class abundance matrices and structural and diversity metrics.Trained Random Forest modelsSerialized R model objects (.rds) for mortality, upgrowth, and recruitment, including feature lists and tuned hyperparameters.Validation outputsOut-of-fold predictions from 5-fold cross-validation and outputs from forward (Inventory 1 → Inventory 2) simulation used to assess emergent stand dynamics.Climate covariatesStatic baseline climate inputs and annual plot-level climate time series for SSP1–2.6 and SSP5–8.5 projection scenarios.Relationship to codeAll R scripts used to generate the contents of this deposit—including preprocessing, model training, cross-validation, forward validation, and long-term projections—are publicly available in the associated GitHub repository and are not duplicated here.LicenseAll files in this deposit are released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license unless otherwise noted.CitationPlease cite this Figshare deposit when using these data or model objects:DOI will appear here after publication.ContactDesmond Sosu MensahDepartment of Forestry and Natural Resources, Purdue UniversityEmail: [e-mail] / [e-mail]

## 概览
本Figshare存档提供了支撑非洲MATRIX森林人口统计建模框架的处理后数据集、训练好的随机森林（Random Forest）模型对象、验证输出与气候协变量。本批次材料可复现关联论文中报道的模型训练、评估与模拟结果，涵盖人口统计率估算（死亡率、径级生长、种群补充）、交叉验证、跨样本间隔验证（样地调查1→样地调查2），以及恒定气候强迫与基于情景的气候强迫（如SSP1–2.6与SSP5–8.5）下的2010–2100年长期森林投影。
所有模拟均通过R语言实现，并运行于基于Linux的高性能计算系统。完整的预处理、模型训练、验证与投影脚本可通过配套的GitHub代码仓库单独获取。

## 本存档内容
本存档包含以下组件：
1. **处理后的训练与模拟数据集**：用于模型训练与投影的样地水平与状态空间水平输入数据，包括物种组×胸径级多度矩阵以及结构与多样性指标。
2. **训练好的随机森林模型**：针对死亡率、径级生长与种群补充的序列化R模型对象（.rds格式），包含特征列表与调优后的超参数。
3. **验证输出**：5折交叉验证的折外预测结果，以及用于评估林分动态涌现特征的正向（样地调查1→样地调查2）模拟输出。
4. **气候协变量**：针对SSP1–2.6与SSP5–8.5投影情景的静态基准气候输入数据与年度样地水平气候时间序列。

## 与代码的关联
用于生成本存档内容的所有R脚本（涵盖预处理、模型训练、交叉验证、正向验证与长期投影）均已公开上传至关联的GitHub代码仓库，本存档未重复包含这些脚本。

## 授权协议
除非另有说明，本存档内所有文件均采用知识共享署名4.0国际许可（CC BY 4.0）协议发布。

## 引用方式
当使用本数据集或模型对象时，请引用此Figshare存档：
出版后将提供DOI编号。

## 联系方式
联系人：德斯蒙德·索苏·门萨（Desmond Sosu Mensah）
普渡大学林学与自然资源系
邮箱：[e-mail] / [e-mail]