Machine learning for anticipating sea level variability from regional temperature shifts

• Autores: Cristina Radin Franco
• Directores de la Tesis: Verónica Úrsula Nieves Calatrava (dir. tes.), Ana Belén Ruescas Orient (codir. tes.)
• Lectura: En la Universitat de València ( España ) en 2024
• Idioma: inglés
• Número de páginas: 174
• Tribunal Calificador de la Tesis: Marta Isabel Marcos Moreno (presid.), Jordi Muñoz Marí (secret.), Josep Lluís Pelegrí Llopart (voc.)
• Programa de doctorado: Programa de Doctorado en Teledetección por la Universitat de València (Estudi General)
• Materias:
  • Matemáticas
    • Ciencia de los ordenadores
      • Inteligencia artificial
    • Estadística
      • Análisis de datos
  • Ciencias de la tierra y del espacio
    • Climatología
      • Climatología regional
    • Geología
      • Teledetección (geología)
• Enlaces
  • Tesis en acceso abierto en: RODERIC
• Resumen

  • In recent decades, oceans have been steadily warming, leading to rising sea levels. The need for new methods to accurately predict sea level changes, considering their nonlinear nature, is critical. This issue is particularly urgent for coastal regions due to its environmental and socioeconomic impacts, where significant populations reside.

    This Thesis explores the application of specifically designed Machine Learning (ML) methods to examine regional coastal temperature changes across different depth layers and their implications for coastal sea level variability worldwide, contributing to future sea level predictions. Our research has developed a series of ML models tailored to forecast sea levels on various timescales and reconstruct historical coastal sea levels, providing valuable insights into the state of our oceans. Additionally, the automatic identification of regional-scale climate variability modes globally has been achieved, highlighting well-known climatic regions and offering new insights into unexplored areas, enhancing region-specific studies. The methodology applied in the Mediterranean Sea has proven effective for sub-basin identification of climatic zones and regional sea level predictions.

    These powerful and versatile models have the potential to support informed decision-making regarding regional and coastal climate-related changes.This work has resulted in four peer-reviewed publications that successfully addressed our goals, along with additional contributions derived from related projects and initiatives.