Global-scale mining polygons (Version 2)

This dataset updates the global-scale mining polygons (Version 1) available from https://doi.org/10.1594/PANGAEA.910894. It contains 44,929 polygon features, covering 101,583 km² of land used by the global mining industry, including large-scale and artisanal and small-scale mining. The polygons cover all ground features related to mining, .e.g open cuts, tailing dams, waste rock dumps, water ponds, processing infrastructure, and other land cover types related to the mining activities. The data was derived using a similar methodology as the first version by visual interpretation of satellite images. The study area was limited to a 10 km buffer around the 34,820 mining coordinates reported in the S&P metals and mining database. We digitalized the mining areas using the 2019 Sentinel-2 cloudless mosaic with 10 m spatial resolution (https://s2maps.eu by EOX IT Services GmbH - Contains modified Copernicus Sentinel data 2019). We also consulted Google Satellite and Microsoft Bing Imagery, but only as additional information to help identify land cover types linked to the mining activities. The main data set consists of a GeoPackage (GPKG) file, including the following variables: ISO3_CODE, COUNTRY_NAME, AREA in squared kilometres, FID with the feature ID, and geom in geographical coordinates WGS84. The summary of the mining area per country is available in comma-separated values (CSV) file, including the following variables: ISO3_CODE, COUNTRY_NAME, AREA in squared kilometres, and N_FEATURES number of mapped features. Grid data sets with the mining area per cell were derived from the polygons. The grid data is available at 30 arc-second resolution (approximately 1x1 km at the equator), 5 arc-minute (approximately 10x10 km at the equator), and 30 arc-minute resolution (approximately 55x55 km at the equator). We performed an independent validation of the mining data set using control points. For that, we draw 1,000 random samples stratified between two classes: mine and no-mine. The control points are also available as a GPKG file, including the variables: MAPPED, REFERENCE, FID with the feature ID, and geom in geographical coordinates WGS84. The overall accuracy calculated from the control points was 88.3%, Kappa 0.77, F1 score 0.87, producer's accuracy of class mine 78.9 % and user's accuracy of class mine 97.2 %.

本数据集更新了源自https://doi.org/10.1594/PANGAEA.910894的全球尺度采矿斑块（版本1）。数据集包含44929个斑块要素，覆盖全球采矿业使用的101583平方千米土地，涵盖大规模采矿以及个体与小型采矿活动。此类斑块覆盖所有与采矿活动相关的地表要素，例如露天采矿场、尾矿库、废石堆、储水池、加工基础设施，以及其他与采矿活动相关的土地覆盖类型。本数据采用与第一版类似的技术方法，通过卫星影像目视解译获取。研究范围限定于标普（S&P）金属与采矿数据库中报告的34820个采矿坐标周边10千米缓冲区内。我们采用2019年无云哨兵-2（Sentinel-2）镶嵌影像（空间分辨率10米，来源：https://s2maps.eu，由EOX IT服务有限公司提供——包含2019年修改的哥白尼哨兵数据）对采矿区域进行数字化。我们还参考了谷歌卫星影像与微软必应影像，仅作为辅助信息以辅助识别与采矿活动相关的土地覆盖类型。主数据集采用GeoPackage（GPKG）格式文件，包含以下字段：ISO3_CODE、COUNTRY_NAME、AREA（面积，单位：平方千米）、要素ID字段FID，以及geom（采用WGS84（世界大地测量系统1984）地理坐标系的几何要素）。各国采矿面积汇总信息以逗号分隔值（CSV）格式文件提供，包含以下字段：ISO3_CODE、COUNTRY_NAME、AREA（面积，单位：平方千米）以及N_FEATURES（映射要素总数量）。基于上述采矿斑块生成了按网格单元统计的采矿面积数据集。该网格数据包含三种分辨率：30角秒（赤道处约1×1千米）、5角分（赤道处约10×10千米）以及30角分（赤道处约55×55千米）。我们采用控制点对本采矿数据集进行了独立验证：随机抽取1000个样本，并按“矿区”与“非矿区”两个类别进行分层抽样。控制点同样以GeoPackage（GPKG）格式文件提供，包含以下字段：MAPPED（模型映射结果）、REFERENCE（参考真值）、要素ID字段FID，以及geom（采用WGS84地理坐标系的几何要素）。基于控制点计算得到的总体精度为88.3%，Kappa系数为0.77，F1值为0.87，矿区类别的生产者精度为78.9%，矿区类别的使用者精度为97.2%。

搜集汇总

背景与挑战

背景概述

该数据集是Global-scale mining polygons (Version 2)，包含44,929个多边形特征，覆盖101,583平方公里的全球采矿用地，包括大规模和小规模采矿活动。数据通过卫星图像视觉解释方法生成，并提供了GeoPackage和CSV格式的文件，以及不同分辨率的网格数据集，总体准确率为88.3%。