Zonas susceptibles a incendios forestales basado en algoritmos de aprendizaje automático (Machine Learning) en el cantón Macará, Loja – Ecuador
Forest fires represent a significant threat, especially in areas with dense vegetation and adverse climatic conditions such as dry forests, and can destroy forests, reduce biodiversity, degrade soil and increase carbon dioxide in the atmosphere, aggravating climate change. This study focuses on iden...
Zapisane w:
| 1. autor: | |
|---|---|
| Format: | bachelorThesis |
| Język: | spa |
| Wydane: |
2025
|
| Hasła przedmiotowe: | |
| Dostęp online: | https://dspace.unl.edu.ec/jspui/handle/123456789/31746 |
| Etykiety: |
Dodaj etykietę
Nie ma etykietki, Dołącz pierwszą etykiete!
|
Dodaj etykietę | Streszczenie: | Forest fires represent a significant threat, especially in areas with dense vegetation and adverse climatic conditions such as dry forests, and can destroy forests, reduce biodiversity, degrade soil and increase carbon dioxide in the atmosphere, aggravating climate change. This study focuses on identifying areas susceptible to forest fires in the Macará canton, Loja, Ecuador, using machine learning algorithms to develop a predictive model for the management and prevention of these events. Topographic variables, vegetation indices and anthropogenic factors were analyzed using Sentinel-2A data and hot spots from the VIIRS sensor (2018-2021). Subsequently, the information was extracted to create a database. Machine learning algorithms such as logistic regression, decision trees and multivariate adaptive regression splines (MARS) were used. The MARS model showed the best performance with an AUC of 0.97 and a Kappa index of 0.91, indicating high accuracy in predicting areas susceptible to forest fires. The areas with the highest susceptibility were found in zones with high vegetation density, high accessibility and significant human interaction. This model allows the creation of a forest fire susceptibility map, crucial for the planning and implementation of prevention and mitigation strategies by local authorities. In conclusion, this study provides a detailed analysis of the factors that influence forest fires in Macará using the MARS model highlighting the importance of distance to anthropogenic areas, rivers and Moisture, in addition to providing a map of susceptibility to forest fires, which is a practical tool for their management, contributing to the conservation of ecosystems and the welfare of local communities. Limitations include the need for meteorological data and cloud cover, which limits data collection. |
|---|