SCB-Net

Spatially Constrained Bayesian Network (SCB-Net): An Approach to Obtaining Field Data-Constrained Predictive Maps with Uncertainty Assessment. In this project, we have developed an innovative approach to ensure that predicted lithologies accurately reflect the samples collected in the field. Moreover, our model leverages multisource remote sensing data to make predictions in areas where no direct samples are available. Our study focuses on the Churchill Province, located in Quebec, Canada.

Predictive Lithological Map displaying 16 lithologic units

(instability = uncertain predictions)

Authors: Victor S. Santos (INRS & NRCan), Erwan Gloaguen (INRS), and Shiva Tirdad (NRCan).

Link to Preprint - ArXiv

INRS: Institut National de la Recherche Scientifique

NRCan: Natural Resources Canada

Inputs

Northeast area

Probability masks

Purpose	File name
Training	train_prob_mask_bs10_400_code_r3.tif
Validation	val_prob_mask_bs10_400_code_r3.tif

Remote sensing layers

Type	File(s)
Multispectral	sentinel2_multispec_east_qc_100m.tif
RADAR	ALOS_PALSAR_RADAR_MOSAIC_QC_100m.tif
Magnetic	MAG_QC_LOWRES_RESMAG_4269_epsg.tif; MAGRES_QC_LOWRES_AS_4269_epsg.tif; MAGRES_QC_LOWRES_DV1_4269_epsg.tif
DEM	alos_elev_east_qc_100m.tif

North area

Probability masks

Purpose	File name
Training	train_prob_mask_bs10_400_code_r2_north.tif
Validation	val_prob_mask_bs10_400_code_r2_north.tif

Remote sensing layers

Type	File(s)
Multispectral	sentinel2_multispec_75m.tif
RADAR	ALOS_SAR_NORD_QC_75m.tif
Magnetic	MAG_QC_LOWRES_RESMAG_4269_epsg.tif; MAGRES_QC_LOWRES_AS_4269_epsg.tif; MAGRES_QC_LOWRES_DV1_4269_epsg.tif
DEM	alos_elevation_75m.tif

Outputs

Predictions

Variable	Description
pred_grid	Deterministic predictions
pred_grid_mean	Mean of predictions
pred_grid_var	Variance of predictions
cat	Map of the most probable class per pixel

Access to data

Link to remotely sensed data, probability masks, and weights of the models.

Requirements

• python>=3.7
• numpy>=1.22
• pandas>=1.4
• xarray>=2022.03
• rioxarray>=0.10
• rasterio>=1.3
• geopandas>=0.11
• shapely>=1.8
• scipy>=1.8
• tqdm>=4.64
• matplotlib>=3.5
• scikit-learn>=1.1
• textdistance>=4.2
• translate>=3.6
• tensorflow>=2.8
• opencv-python>=4.5
• pyproj>=3.3

License

This project is licensed under the Creative Commons Attribution 4.0 International License - see the LICENSE file for details.