five

Continental Europe surface lithology based on EGDI / OneGeology map at 1:1M scale

收藏
NIAID Data Ecosystem2026-05-02 收录
下载链接:
https://zenodo.org/record/4787631
下载链接
链接失效反馈
官方服务:
资源简介:
Continental Europe surface lithology based on EGDI / OneGeology map at 1:1M scale produced by GEOZS, Slovenia. European datasets harvested from national WFS for geologic units or national geological units datasets, based on OneGeology and INSPIRE Lithology and Geochronologic Era URI codelists. Layers include: EGDI_GE_GeologicUnit_EN_1M_Surface_LithologyPolygon_v2_250m_epsg.3035.tif = original EGDI surface lithology map; dtm_surface.lithology_egdi.1m_c_250m_s_20000101_20221231_eu_epsg.3035_v20240530.tif = gap filled surface lithology map; Missing values in the original EGDI lithology map have been imputed by training a random forest classifier model based on parameters derived from DTM and soil regions map from Die Bundesanstalt für Geowissenschaften und Rohstoffe (BGR). By generating 1 million random points, geographically balanced over the whole pan-EU land area, each class in the map was covered properly. Classes whose number of samples is less than 10 were discarded from the model training. The hyperparameter tuning of the model was carried out via a Bayesian approach with a criteria to maximize accuracy of 5k-fold cross validation. The tuned random forest model achieved an accuracy of 47% (Kappa=0.43) for the testing data, 20% of the generated sample points. The lithology of Turkey, on the other hand, was digitised from the available geology map produced by the General Directorate of Mineral Research and Exploration (MTA). The available raster map was post-processed and classified as 20 lithology classes using the k-means algorithm. These classes were harmonized with the classes in the EGDI lithology map. Acknowledgment: GEOZS, Continental Shelf Department at the Ministry for Transport and Infrastructure.

本数据集为基于斯洛文尼亚GEOZS制作的1:100万比例尺EGDI/OneGeology地图构建的欧洲大陆地表岩性数据集。欧洲数据集源自各国地质单元网络要素服务(WFS)或国家级地质单元数据集,基于OneGeology与INSPIRE岩性及地质年代URI编码列表构建。 包含以下图层: EGDI_GE_GeologicUnit_EN_1M_Surface_LithologyPolygon_v2_250m_epsg.3035.tif = 原始EGDI地表岩性地图; dtm_surface.lithology_egdi.1m_c_250m_s_20000101_20221231_eu_epsg.3035_v20240530.tif = 缺失值填补后的地表岩性地图; 原始EGDI地表岩性地图中的缺失值通过训练随机森林分类器进行插补,模型特征参数源自德国联邦地球科学与自然资源研究所(BGR)的数字地形模型(DTM)与土壤区域地图。通过在整个泛欧陆地区域生成100万个地理分布均衡的随机采样点,确保地图中每个岩性类别均被充分覆盖。模型训练阶段剔除了样本量不足10的类别。模型的超参数调优采用贝叶斯方法,以5折交叉验证的准确率最大化为优化目标。经调优后的随机森林模型在占采样点20%的测试集上取得了47%的准确率(Kappa系数=0.43)。 另一方面,土耳其的岩性数据源自土耳其矿产研究与勘探总局(MTA)发布的现有地质图,经数字化得到栅格地图后,通过k-means聚类算法划分为20个岩性类别,最终与EGDI岩性地图的类别体系完成对齐。 致谢:GEOZS、斯洛文尼亚交通与基础设施部大陆架司。
创建时间:
2024-07-01
二维码
社区交流群
二维码
科研交流群
商业服务