UAV_Field Inventory Data

This dataset comprises analysis-ready data (ARD) designed to support the estimation of Above-Ground Biomass (AGB) in Miombo woodlands using UAV-derived structural and spectral metrics in combination with machine learning models. The data were collected from 98 georeferenced forest plots located within Kilosa and Kitulangalo forests, situated in Morogoro Region, Tanzania. The UAV flights and corresponding ground forest inventory were conducted during the same temporal window, between February and March 4, 2024.Each plot record includes 23 predictor variables and 1 response variable (AGB). The predictor variables are grouped as follows:Height percentiles (P10 to P100): UAV-derived canopy height percentiles in meters, capturing vertical forest structure.Statistical height metrics: Height standard deviation (Hsd), skewness (Hsk), and kurtosis (Hkurt).Spectral variables: Mean and standard deviation of Red (R), Green (G), and Blue (B) bands captured from UAV RGB imagery.Band ratios: R/G, R/B, G/B used to enhance spectral feature differentiation.Structural complexity: Point Cloud Density (PCD), indicating vegetation density and texture.The response variable, AGB (in Mg/ha), was calculated from field inventory data and aligned with UAV measurements per plot to facilitate supervised learning.The dataset includes 98 rows and 24 columns, saved in CSV format with a file size of approximately 0.02 MB. All variables are fully populated, with no missing data. Column headers are descriptive of the contents, and units are included or implied based on standard practice (e.g., meters for height metrics, Mg/ha for AGB).This dataset is ready for direct import into statistical or geospatial analysis environments (e.g., R, Python, QGIS) for biomass modelling and validation. It provides a high-quality, spatially-explicit dataset for researchers exploring the integration of UAV-based remote sensing and machine learning in tropical dry forest biomass estimation.

本数据集包含分析就绪数据（Analysis-Ready Data, ARD），旨在支持结合无人机（Unmanned Aerial Vehicle, UAV）提取的结构与光谱指标及机器学习模型，开展米翁博林地（Miombo woodlands）的地上生物量（Above-Ground Biomass, AGB）估算工作。
数据采集自坦桑尼亚莫罗戈罗地区基洛萨（Kilosa）与基图兰加洛（Kitulangalo）林区内的98个带地理坐标的森林样地。无人机航飞与配套地面森林调查于同期开展，时间范围为2024年2月至3月4日。
每个样地记录包含23个预测变量与1个响应变量（AGB）。预测变量可分为以下类别：
1. 高度百分位数（P10至P100）：无人机提取的冠层高度百分位数（单位：米），用于表征森林垂直结构；
2. 统计高度指标：高度标准差（Hsd）、偏度（Hsk）与峰度（Hkurt）；
3. 光谱变量：无人机RGB影像提取的红（R）、绿（G）、蓝（B）波段的均值与标准差；
4. 波段比值：R/G、R/B、G/B，用于增强光谱特征区分度；
5. 结构复杂度：点云密度（Point Cloud Density, PCD），用于表征植被密度与纹理特征。
响应变量AGB（单位：Mg/ha）通过野外调查数据计算得到，并与每个样地的无人机测量数据对齐，以支撑监督学习任务。
本数据集共包含98行与24列，以CSV格式存储，文件大小约为0.02 MB。所有变量均完整填充，无缺失值。列标题清晰描述了对应内容，单位符合行业通用规范（如高度指标单位为米，AGB单位为Mg/ha）。
本数据集可直接导入统计或地理空间分析环境（如R、Python、QGIS）用于生物量建模与验证，为探索基于无人机遥感与机器学习结合的热带干旱森林生物量估算研究的科研人员提供了高质量、空间明确的数据集。