Date of Award

January 2026

Document Type

Thesis

Degree Name

Master of Public Health (MPH)

Department

School of Public Health

First Advisor

Kaveh Khoshnood

Second Advisor

Sunil Parikh

Abstract

Introduction: Artisanal and small-scale mining (ASM) is a labor-intensive method of informal mineral extraction. ASM has a significant presence in Burkina Faso. This study aims to use Sentinel-2 satellite imagery to identify the location of ASM sites in Burkina Faso. There is a lack of comprehensive countrywide data on the location and prevalence of ASM sites in the country. While previous researchers have used higher-resolution data to identify mines, there is limited research on the effectiveness of Sentinel-2 imagery for ASM site identification and its potential limitations.

Methods: ASM sites were classified using a deep learning model in ArcGIS Pro and using Sentinel-2 imagery from Google Earth Engine. The model was trained using ground truth data of areas that were positive and negative for ASM sites. Positive ASM samples were collected from an existing dataset of sites in the western region of Burkina Faso. Negative samples were categorized into four different classifications: vegetation, bare soil, water, and urban areas. ArcGIS Pro deep learning tools and a U-Net architecture were used to classify pixels.

Results: The model’s validation and training curves converged. The model reported an accuracy of 0.963 and a Dice similarity coefficient of 0.062. The training model sample results show errors in classification. There is overclassification of vegetation and underclassification of bare soil and urban areas. Several lakes and bodies of water are misclassified as ASM sites.

Discussion: While the training was successful and showed that the model had converged, the accuracy of the final classification of the satellite imagery was severely limited due to inadequate spatial resolution from Sentinel-2 imagery and a lack of appropriate training data.

Comments

This thesis is restricted to Yale network users only. It will be made publicly available on 09/16/2027

Download

Share

COinS