Pronóstico de la precipitación mensual en la cuenca del río Piura utilizando redes neuronales artificiales
Loading...
Código QR
Authors
Obregón Yataco, José Esteban
Contact Email
Abstract
El objetivo de esta tesis fue pronosticar la precipitación mensual del mes de marzo sobre la cuenca del río Piura utilizando predictores del mes de enero. Anticipar el acumulado mensual de marzo permitiría a los tomadores de decisiones gestionar de manera más eficiente los recursos hídricos en las regiones dependientes de esta cuenca. Actualmente, existen modelos estadísticos utilizando regresión logística (Rivas et al, 2024) y regresión multiple (Sulca & Takahashi, 2025) para pronosticar la precipitación en esta región de interés. Sin embargo es necesario también contar con modelos de inteligencia artificial (IA), como herramienta complementaria, tal como lo hacen, por ejemplo, el Centro Europeo de Previsiones Meteorológicas con su modelo AIFS. En este contexto, se desarrolló un modelo IA como herramienta complementaria a los modelos existentes. Este modelo fue entrenado utilizando predictores como temperatura superficial del mar, agua precipitable total, y viento zonal y meridional, para pronosticar la precipitación de marzo. Aunque todos estos predictores fueron inicialmente estadísticamente significativos, se aplicó un análisis de descubrimiento de causalidad que permitió descartar aquellos sin relación causal directa con la precipitación. Con los cuatro predictores restantes se construyó un modelo convolucional para pronosticar la precipitación espacial en marzo sobre la cuenca del río Piura. La validación del modelo se realizó para el período 1986-2017, no incluido en el entrenamiento, y mostró una buena correspondencia, especialmente en la zona costera. Finalmente se concluye que el modelo de IA que se alimento de dichos 4 predictores en enero para pronosticar la precipitación en la cuenca del río Piura en marzo tuvo un buen ajuste en su entrenamiento (periodo 1950 1985, 2018-2024) espacialmente hablando a lo largo de la cuenca. Mientras que su desempeño durante la evaluación (periodo 1986-2017) mostró buena correlación con respecto a lo observado principalmente en la zona occidental de la cuenca.
The objective of this thesis was to forecast the monthly precipitation for March over the Piura River basin using predictors from January. Anticipating the accumulated monthly precipitation for March would allow decision-makers to manage water resources more efficiently in the regions dependent on this basin. Currently, statistical models using logistic regression (Rivas et al., 2024) and multiple regression (Sulca & Takahashi, 2025) exist to forecast precipitation in this region of interest. However, it is also necessary to have artificial intelligence (AI) models as a complementary tool, as is done, for example, by the European Centre for Medium-Range Weather Forecasts with its AIFS model. In this context, an AI model was developed as a complementary tool to existing models. This model was trained using predictors such as sea surface temperature, total precipitable water, and zonal and meridional winds to forecast March precipitation. Although all these predictors were initially statistically significant, a causality discovery analysis was applied, allowing those without a direct causal relationship to precipitation to be discarded. A convolutional model was then constructed using the remaining four predictors to forecast spatial precipitation in March over the Piura River basin. The model was validated for the period 1986–2017, which was not included in the training, and showed good agreement, especially in the coastal zone. Finally, it was concluded that the AI model, which was trained using these four predictors in January to forecast precipitation in the Piura River basin in March, showed a good fit during its training (periods 1950–1985 and 2018–2024) spatially across the basin. Its performance during the evaluation period (1986–2017) showed good correlation with observations, primarily in the western part of the basin.
The objective of this thesis was to forecast the monthly precipitation for March over the Piura River basin using predictors from January. Anticipating the accumulated monthly precipitation for March would allow decision-makers to manage water resources more efficiently in the regions dependent on this basin. Currently, statistical models using logistic regression (Rivas et al., 2024) and multiple regression (Sulca & Takahashi, 2025) exist to forecast precipitation in this region of interest. However, it is also necessary to have artificial intelligence (AI) models as a complementary tool, as is done, for example, by the European Centre for Medium-Range Weather Forecasts with its AIFS model. In this context, an AI model was developed as a complementary tool to existing models. This model was trained using predictors such as sea surface temperature, total precipitable water, and zonal and meridional winds to forecast March precipitation. Although all these predictors were initially statistically significant, a causality discovery analysis was applied, allowing those without a direct causal relationship to precipitation to be discarded. A convolutional model was then constructed using the remaining four predictors to forecast spatial precipitation in March over the Piura River basin. The model was validated for the period 1986–2017, which was not included in the training, and showed good agreement, especially in the coastal zone. Finally, it was concluded that the AI model, which was trained using these four predictors in January to forecast precipitation in the Piura River basin in March, showed a good fit during its training (periods 1950–1985 and 2018–2024) spatially across the basin. Its performance during the evaluation period (1986–2017) showed good correlation with observations, primarily in the western part of the basin.
Description
Universidad Nacional Agraria La Molina. Escuela de Posgrado. Maestría en
Meteorología Aplicada
Keywords
Cuenca del Río Piura; Cuencas hidrográficas; Estaciones meteorológicas; Evaluación; Precipitación atmosférica; Pronóstico del tiempo
Citation
Date
2026
Collections
Seleccionar año de consulta:
Licencia de uso

Excepto si se señala otra cosa, la licencia del ítem se describe como info:eu-repo/semantics/openAccess

