Aplicaciones de Machine Learning a la cartografía de manantiales. Clasificación y caracterización de manantiales representativos de la hidrogeología kárstica en el GMU Las Loras
Loading...
Official URL
Full text at PDC
Publication date
2024
Defense date
18/09/2024
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citation
Abstract
RESUMEN. El Geoparque Mundial de la UNESCO (GMU) Las Loras es muy diverso geológicamente en relación con la interacción del paisaje con las aguas subterráneas, contando con un destacado número de manantiales, mayoritariamente de tipo kárstico, entre los que sobresalen aquellos asociados a cascadas y a generación de formaciones tobáceas. En este Trabajo Fin de Máster se dirige en primer lugar, a desarrollar una clasificación genética de tipo de manantial, a partir de la que se aplica una caracterización de formaciones tobáceas asociadas a tres manantiales petrificantes con formación de tuf, considerados como destacados hábitats de interés comunitario. Los tres manantiales seleccionados para la caracterización de los edificios tobáceos tienen una notable importancia, no solo desde el punto de vista hidrogeológico, sino también ambiental y culturalmente. El GMU Las Loras es un territorio poco conocido lo que destaca la importancia de este estudio.
La caracterización y clasificación de los manantiales determinan la importancia de los procesos geológicos en su funcionamiento y tipología. Un tipo de manantial puede compartir características comunes de otros tipos, lo que resalta la caracterización más detallada para reconocer las variaciones entre ellos. Los tres manantiales estudiados en este trabajo se han clasificado de acuerdo con Stevens et al. (2020) como upland hillslope, aunque comparten características comunes de otros tipos de manantiales.
Considerando la importancia de los manantiales en el conjunto del GMU Las Loras, se han aplicado las técnicas de Machine Learning (ML) en la identificación de manantiales desaparecidos en el GMU Las Loras. Para ello, se han elegido 14 variables explicativas que permiten predecir el valor de una variable objetivo, con el fin de crear mapas predictivos. Estos mapas se han obtenido usado los algoritmos ABC, RFC y QDA, ya que se consideran como los más representativos para este trabajo. Aunque los resultados de ML no han sido concluyentes en el estudio de manantiales desaparecidos en la zona de estudio, se ha podido confirmar que pueden ser útiles en este tipo de trabajos.
ABSTRACT. Las Loras UNESCO Global Geopark (UGGp) is geologically diverse in terms of the interaction between the landscape and groundwater, including a significant number of springs, especially karstic, with notable examples associated with waterfalls and tuff formations. In this Master's Thesis, a genetic classification of spring types is applied first, followed by a characterization of tuff formations associated with three petrifying springs with tuff formations, considered as important habitats of community interest. The three selected springs for the characterization of tuff formations hold significant importance not only from a hydrogeological perspective but also environmentally and culturally. The UGGp Las Loras is a very unknown territory, which highlights the importance of this study. The characterization and classification determine the importance of geological processes in their functioning and typology. One type of spring can share common characteristics with other types, highlighting more detailed characterization to recognize the variations between them. The three springs studied in this study have been classified according to Stevens et al. (2020) as upland hillslope, although they have common characteristics with other types of springs. Considering the importance of springs in the UGGp Las Loras, Machine Learning (ML) techniques have been applied to identify disappeared springs in the UGGp Las Loras. In this respect, 14 explanatory variables were chosen to predict the target variable. The algorithms ABC, RFC and QDA have been used for predictive maps, as they are considered the most representable in this study. Although the ML results have not been conclusive in identifying disappeared springs in the study area, it has been confirmed that these techniques can be useful in this type of studies.
ABSTRACT. Las Loras UNESCO Global Geopark (UGGp) is geologically diverse in terms of the interaction between the landscape and groundwater, including a significant number of springs, especially karstic, with notable examples associated with waterfalls and tuff formations. In this Master's Thesis, a genetic classification of spring types is applied first, followed by a characterization of tuff formations associated with three petrifying springs with tuff formations, considered as important habitats of community interest. The three selected springs for the characterization of tuff formations hold significant importance not only from a hydrogeological perspective but also environmentally and culturally. The UGGp Las Loras is a very unknown territory, which highlights the importance of this study. The characterization and classification determine the importance of geological processes in their functioning and typology. One type of spring can share common characteristics with other types, highlighting more detailed characterization to recognize the variations between them. The three springs studied in this study have been classified according to Stevens et al. (2020) as upland hillslope, although they have common characteristics with other types of springs. Considering the importance of springs in the UGGp Las Loras, Machine Learning (ML) techniques have been applied to identify disappeared springs in the UGGp Las Loras. In this respect, 14 explanatory variables were chosen to predict the target variable. The algorithms ABC, RFC and QDA have been used for predictive maps, as they are considered the most representable in this study. Although the ML results have not been conclusive in identifying disappeared springs in the study area, it has been confirmed that these techniques can be useful in this type of studies.