Application Of Machine Learning To Map Underground Fires In Hwange Colliery Using Earth Observation Data
- Author
- Nyakonda Mitchell
- Title
- Application Of Machine Learning To Map Underground Fires In Hwange Colliery Using Earth Observation Data
- Abstract
-
Underground coal fires (UCFs) present a significant challenge with far reaching environmental and safety implications globally, depicting complex spatio-temporal dynamics. This study employed a roust machine learning approach using Random Forest Algorithm to model Underground Coal Fires (UCF’s) at Hwange Colliery Mines. The methodology utilized a Semi-automatic Classification Plugin within QGIS, integrating Landsat 8’s optical and thermal bands. Analysis of historical UCF events Through Normalized Burn Ratio (NBR) and the Normalized Difference Vegetation Index (NDVI) was relevant to develop a robust predictive model capable of identifying potential UCFs and estimating extent of underground fire propagation. Furthermore, the research explored the probability of implementing monitoring systems to provide timely warning and mitigate the detrimental impacts of underground coal fires on the environment and local communities. Random Forest model detected UCFs in most parts of Hwange mining area such Wankie mine, the road linking Kamandama mine disaster memorial site and the rest Hwange town and Hwange Colliery. Spatial temporal analysis of the propagation of the UCF for the period 2020 - 2023, revealed that the year 2020 had the most UCF affected areas with 6.06% later decreasing to 3.35% by 2023. The NDVI analysis showed minimum plant growth and a majority of bare ground and fire affected areas illuminating the harsh ecological effects of underground coal fires in the region of interest. These findings inform sustainable resource management practices and enhance disaster risk reduction strategies in coal mining regions. Therefore, prevalence of UCFs in Hwange Colliery Mines underscore the urgent need for immediate action to address the environmental menace. Through the synergistic application of advanced machine learning techniques and Earth Observation data, disaster management practitioners can improve their ability to detect, monitor and respond to UCF incidents, thereby safeguarding natural resources and human well-being.
- Date
- JUNE 2024
- Publisher
- BUSE
- Keywords
- Machine Learning, Underground Fires,Hwange Colliery
- Supervisor
- Pedzisai
- Item sets
- Department of Geosciences
