Mining Pits at Písečný vrch

Mining Pits at Písečný vrch

Submitted by Mgr. Jan Fišer on
Podnadpis
Pit Inventory for Recency Index Modelling at Písečný vrch, Bečov, Czech Republic
Creators and Contributors
Creator
Jan Fišer ; Czech University of Life Sciences Prague , Department of Spatial Sciences jan.fiser@uhk.cz
Contributor
Supervisor: Vítězslav Moudrý ; Czech University of Life Sciences Prague , Department of Spatial Sciences moudry@fzp.czu.cz
Publisher
Masaryk University
Description
Účel a kontext vzniku datasetu

This dataset was created to document pit-like mining relics at Písečný vrch near Bečov, Czech Republic. It contains an inventory of mapped features used for the non-destructive assessment of a complex mining palimpsest affected by prehistoric exploitation and recent extraction. The dataset supports Recency Index modelling, prioritisation of features for future archaeological verification, and heritage protection of the site.

Metody

The methodology combined systematic field survey, LiDAR-derived terrain analysis, historical aerial photograph interpretation and historical land-use reconstruction. All information was integrated at the level of individual mapped pits and used to support non-destructive Recency Index modelling based on a MaxEnt presence-background approach.

Obsah a struktura datasetu
  • data_table_with_predictions.csv – Structured tabular dataset containing unique pit identifiers, derived attributes, model prediction values, associated evaluation metrics, and geographic coordinates for each mapped pit.
  • maxent_analysis.R – R script used to run the MaxEnt presence-background analysis on the input dataset, generate Recency Index predictions and calculate associated model evaluation metrics.
  • readme.txt – Documentation file describing the structure and content of the dataset. It includes the dataset title, coordinate reference system, definitions of all variables, explanations of Recency Index prediction and uncertainty metrics, and notes on data provenance and third-party data restrictions.
Technické informace

The dataset is provided as UTF-8 encoded CSV files with semicolon-separated values. Spatial coordinates are referenced in the S-JTSK / Krovak East North coordinate reference system (EPSG:5514). The accompanying analysis script, maxent_analysis.R, was written in R and contains the full Recency Index modelling workflow. It requires the R packages tidyverse, maxnet, pROC, patchwork, dotwhisker, car and corrr. The script uses data_table.csv as the input file and exports data_table_with_predictions.csv with Recency Index predictions, threshold-based classification, leave-one-out cross-validation predictions and bootstrap-derived uncertainty metrics.

The script also performs predictor preparation, multicollinearity assessment using Spearman correlations and VIF, AICc-based MaxEnt model tuning, model fitting, ROC/AUC evaluation, Youden Index thresholding, predictor importance analysis, response curve generation and prediction uncertainty assessment. A single random seed is set for reproducibility, and 500 bootstrap replicates are used for resampling-based uncertainty estimates. The script additionally produces supplementary outputs, including the AICc tuning table and figures used for model validation, predictor interpretation and uncertainty visualisation.

Výzkumné otázky

The dataset was created to address the following research questions: How are pit-like mining relics distributed across the Písečný vrch mining landscape? Which mapped pits can be independently associated with recent extraction? Which pits of undetermined age are less similar to confirmed recent mining features and therefore represent higher-priority candidates for future archaeological verification? How can field observations, LiDAR-derived terrain variables, historical aerial photographs and historical land-use information be integrated to support the non-destructive interpretation of a complex mining palimpsest?

Vhodné způsoby znovupoužití datasetu

The dataset is appropriate for reuse in archaeological spatial analysis, non-destructive prospection, heritage-management planning, methodological comparison and teaching. It can be used to examine mapped pit-like mining features at Písečný vrch, reproduce or evaluate the Recency Index modelling workflow, and support comparative studies of mining palimpsests and feature prioritisation. Reuse should acknowledge that the Recency Index is a relative decision-support measure, not a definitive chronological classification. The dataset contains derived tabular data only; raw aerial photographs, LiDAR point clouds and source cartographic data are not redistributed due to third-party rights.

Omezení datasetu a nevhodné způsoby znovupoužití

This dataset is not appropriate for definitive dating of individual pits or for confirming prehistoric origin without further archaeological verification. The Recency Index is a relative decision-support measure, not a substitute for excavation, direct dating or stratigraphic interpretation. The dataset is also not suitable for analyses requiring raw LiDAR point clouds, original aerial photographs, source cartographic data, high-resolution terrain modelling or precise volumetric reconstruction. Because it was created for object-level analysis of already mapped features at Písečný vrch, it should not be used directly for landscape-scale prediction of unknown archaeological sites without methodological adaptation.

Time Reference
Období
Palaeolithic
Neolithic
Neolithic
Bronze Age
Modern–Industrial Period
unspecified dating
Datum sběru
2021
Datum vytvoření
2026-07-01
Rok vydání
2026
Subject

Klíčová slova

Typ posudku
other expert analysis
Druh objektu
mining feature
Jazyk datasetu
English
Licence
Creative Commons Attribution 4.0 International
1 - 3 z 3 souborů v datasetu
Název Velikost souboru
readme.csv 5.1 KB
data_table_with_predictions.csv 140 bytes
maxent_analysis.R 7.33 KB

Náhled

Name
readme.csv
readme_0.csv
README
1) Dataset:
Mining Pits at Písečný vrch
Pit Inventory for Recency Index Modelling at Písečný vrch, Bečov, Czech Republic
Description:
This dataset contains tabular data used to classify mining pits at the Písečný vrch site. The dataset was used in a MaxEnt-based modelling workflow to estimate the relative probability that individual pits represent recent mining features (“Recency Index”).
Coordinate reference system:
S-JTSK / Krovak East North (EPSG:5514)
2) data_table_with_predictions.csv
ID:
Unique identifier of each mapped pit.
x_5514:
X coordinate in S-JTSK / EPSG:5514.
y_5514:
Y coordinate in S-JTSK / EPSG:5514.
z_5514:
Elevation value in S-JTSK / EPSG:5514.
r:
Pit diameter in metres measured in the field.
deep:
Binary variable in metres expressing pit depth:
1 = deep pit (>0.5 m)
0 = shallow pit (<0.5 m)
quartz:
Visible quartzite in the pit wall:
1 = present
0 = absent
historical_land_use:
Category of historical land use derived from georeferenced Stable Cadastre maps:
arable_land, common_pasture, pasture, pasture_with_fruit_trees
r1958:
Feature identified at the pit location in the 1958 aerial photograph:
n = no visible feature
j = pit visible
k = quartzite / quarry trace visible
v = vegetation visible
r1963:
Feature identified at the pit location in the 1963 aerial photograph.
n = no visible feature
j = pit visible
k = quartzite / quarry trace visible
v = vegetation visible
r1975:
Feature identified at the pit location in the 1975 aerial photograph.
n = no visible feature
j = pit visible
k = quartzite / quarry trace visible
v = vegetation visible
recent_prosp:
Recent pit identified only during field survey:
1 = pit with visible recent traces recorded during field survey
0 = no visible recent traces recorded during field survey
slope:
Slope value trend at the pit location(degree).
surface_category:
Interpretation of recent terrain disturbance based on DEM analysis:
1 = undisturbed terrain
0 = recently disturbed terrain
nearest_neighbour:
Distance in metres to the nearest neighbouring pit.
difference_raster_value:
Value extracted from the DEM of Difference (DoD), expressing local topographic change in metres.
recent:
Final binary classification of recentness based on field survey and interpretation of historical aerial photographs.
1 = recent pit
0 = potentially older / non-recent pit
prediction:
Fitted Recency Index from the full model (all pits), on a 0-1 logistic scale. Higher values indicate a greater estimated likelihood of recent (20th-century) origin.
1 = pits with recency index above treshold (>0.441)
0 = pits with recency index under treshold (<0.441)
prediction_category:
Binary classification (1 = recent, 0 = non-recent) obtained by applying the Youden-optimal threshold, derived from leave-one-out cross-validated predictions, to prediction.
pred_lwr:
Lower bound of the 95% bootstrap confidence interval for the Recency Index, based on the 2.5th percentile of predictions across 500 bootstrap-refitted models.
pred_upr:
Upper bound of the 95% bootstrap confidence interval for the Recency Index, based on the 97.5th percentile of predictions across 500 bootstrap-refitted models.
pred_ci_width:
Width of the 95% bootstrap confidence interval (pred_upr minus pred_lwr). Larger values indicate greater uncertainty in the predicted Recency Index for that pit.
pred_stability:
Proportion of the 500 bootstrap replicate models whose classification (recent / non-recent, using the Youden-optimal threshold) agrees with the classification obtained from the full model for that pit. Values near 1 indicate a classification that is reproduced consistently under resampling; values near 0.5 indicate a classification that is essentially arbitrary under resampling and should be treated with caution.
loocv_prediction:
Predicted Recency Index for the pit obtained under leave-one-out cross-validation, i.e., from a model refit with that pit excluded from the training data. Comparing this value to prediction indicates how much a pit's own inclusion in the training data influences its predicted score; large discrepancies flag potentially influential or unstable observations.
3) maxent_analysis.R
R script containing the full analysis pipeline used for Recency Index modelling. The script loads the input pit inventory table, defines the predictor set, assesses multicollinearity using Spearman correlations and VIF, performs MaxEnt hyperparameter tuning using AICc, fits the selected MaxEnt presence-background model, generates Recency Index predictions, evaluates model performance using AUC and leave-one-out cross-validation, determines the Youden Index threshold, calculates predictor importance and response curves, quantifies prediction uncertainty using 500 bootstrap replicates, and exports the final table with predictions and uncertainty metrics.
Notes:
This repository contains only derived tabular data used in the published analysis. Raw aerial photographs, lidar point clouds, and source cartographic data are not redistributed because they are subject to third-party rights.
MIME type
text/csv
Velikost souboru
5226
Je součástí
Mining Pits at Písečný vrch
Způsob užití
Original File
Má hlavičku
Ne
Oddělovač
^
Is documentation
True