Logo_Icesi

Maestría en Ciencia de Datos

URI permanente para esta colecciónhttp://hdl.handle.net/10906/66933

Examinar

Examinando Maestría en Ciencia de Datos por Autor "Arias Sinisterra, Diana Carolina"

Seleccione resultados tecleando las primeras letras
Mostrando 1 - 1 de 1
  • Miniatura
    Ítem
    Desarrollo de una metodología para la predicción estacional de déficits y excesos hídricos en los departamentos de Quindío, Risaralda y Caldas, mediante técnicas de machine learning
    (Universidad Icesi, 2024-12-10) Arias Sinisterra, Diana Carolina; Estrada Vargas, Oscar Hernan; Agudelo, Diego Fernando; Barrios Perez, Camilo
    This study presents the development of a methodology to forecast excess or deficit water conditions in the Colombian coffee region (departments of Quindío, Risaralda, and Caldas), using the Standardized Precipitation-Evapotranspiration Index (SPEI) as the main indicator. The first phase of the research focused on the consolidation and homogenization of climatic data, the characterization of the region's water conditions, and the construction of SPEI-3, which estimates the water balance using precipitation and evapotranspiration data from the last 3 months, and SPEI-6, which does so with data from the last 6 months, to understand short and medium-term variations. Firstly, a compilation and homogenization of data from various climatic sources were carried out, adjusting them to a uniform resolution for proper analysis. Subsequently, the study area was characterized, identifying its climatic particularities. In addition, a comparison of the SPEI indices with historical periods of El Niño and La Niña phenomena was performed to highlight SPEI's capacity to reflect the climatic reality of the study area. It was observed that SPEI values coincide with the seasons in which these phenomena occurred in Colombia, thus validating its usefulness as an indicator of droughts and water excesses. CPT software was used to generate the SPEI-3 and SPEI-6 forecasts for March 2024. The second phase of the project consisted of testing other predictors to perform the prediction using CCA and a machine learning model, to compare the results obtained by both methods. Finally, the importance of forecasting SPEI with greater accuracy is highlighted, as this would not only reflect the climatic reality more precisely but would also provide a valuable tool for planning and decision-making in industrial and agricultural sectors.