Maus, Victor; Giljum, Stefan; Gutschlhofer, Jakob; da Silva, Dieison M; Probst, Michael; Gass, Sidnei L B; Luckeneder, Sebastian; Lieber, Mirko; McCallum, Ian (2020): Global-scale mining polygons (Version 1) [dataset]. PANGAEA, https://doi.org/10.1594/PANGAEA.910894

This data set provides spatially explicit estimates of the area directly used for surface mining on a global scale. It contains more than 21,000 polygons of activities related to mining, mainly of coal and metal ores. Several data sources were compiled to identify the approximate location of mines active at any time between the years 2000 to 2017. This data set does not cover all existing mining locations across the globe. The polygons were delineated by experts using Sentinel-2 cloudless (https://s2maps.eu by EOX IT Services GmbH (contains modified Copernicus Sentinel data 2017 & 2018)) and very high-resolution satellite images available from Google Satellite and Bing Imagery. The derived polygons cover the direct land used by mining activities, including open cuts, tailing dams, waste rock dumps, water ponds, and processing infrastructure. The main data set consists of a GeoPackage (GPKG) file, including the following variables: ISO3_CODE<string>, COUNTRY_NAME<string>, AREA<double> in squared kilometres, FID<integer> with the feature ID, and geom<polygon> 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<string>, COUNTRY_NAME<string>, AREA<double> in squared kilometers, and N_FEATURES<integer> 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 a 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<string>, REFERENCE<string>, FID<integer> with the feature ID, and geom<point> in geographical coordinates WGS84. The overall accuracy calculated from the control points was 88.4%, other accuracy metrics are shown below.

Confusion Matrix and Statistics

Reference

Prediction Mine No-mine

Mine 394 106

No-mine 10 490

Accuracy : 0.884

95% CI : (0.8625, 0.9032)

No Information Rate : 0.596

P-Value [Acc > NIR] : < 2.2e-16

Kappa : 0.768

Mcnemar's Test P-Value : < 2.2e-16

Sensitivity : 0.9752

Specificity : 0.8221

Pos Pred Value : 0.7880

Neg Pred Value : 0.9800

Precision : 0.7880

Recall : 0.9752

F1 : 0.8717

Prevalence : 0.4040

Detection Rate : 0.3940

Detection Prevalence : 0.5000

Balanced Accuracy : 0.8987

This work was supported by the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme grant number 725525 FINEPRINT project (https://www.fineprint.global/).