lucas-mega
收藏资源简介:
LUCAS-MEGA是一个大规模多模态土壤环境系统数据集,通过融合欧洲异构土壤和环境数据与LUCAS土壤调查数据构建而成。该数据集包含72,000多个土壤样本和1,000多个融合特征,整合了68个ESDAC源数据集,涵盖数值(标量和矢量)、分类、文本和视觉数据。LUCAS-MEGA专为土壤科学中的表示学习设计,提供了一个统一的样本-特征空间,使模型能够学习土壤、气候、地形、土地利用、水文和环境变量之间的关系。数据集遵循MEGA原则:多模态(M)、端到端机器学习就绪(E)、高质量(G)和可访问(A)。数据集发布包括融合表示(主推版本)和标准化表示(中间层),分别适用于模型训练、土壤环境分析和源数据检查。主要文件包括平面数据表、层次化JSON字典、完整元数据和地理管理层次结构等。
LUCAS-MEGA is a large-scale multimodal soil environmental system dataset constructed by integrating heterogeneous European soil and environmental data with LUCAS soil survey data. This dataset contains over 72,000 soil samples and more than 1,000 fused features, integrating 68 ESDAC source datasets, and covers numerical (scalar and vector), categorical, textual, and visual data. Designed specifically for representation learning in soil science, LUCAS-MEGA provides a unified sample-feature space, enabling models to learn the relationships among soil, climate, topography, land use, hydrology, and environmental variables. The dataset adheres to the MEGA principles: Multimodal (M), End-to-End machine learning-ready (E), High-quality (G), and Accessible (A). The dataset release includes fused representations (the recommended primary version) and standardized representations (the intermediate layer), which are suitable for model training, soil environmental analysis, and source data inspection respectively. Its main files include flat data tables, hierarchical JSON dictionaries, complete metadata, and geospatial management hierarchies, among others.





