Bibliographic citations
Flores, A., (2024). Predicción diaria de temperaturas extremas y precipitación en la región Áncash mediante posprocesamiento estadístico del modelo GFS [Universidad Nacional Agraria La Molina]. https://hdl.handle.net/20.500.12996/6308
Flores, A., Predicción diaria de temperaturas extremas y precipitación en la región Áncash mediante posprocesamiento estadístico del modelo GFS []. PE: Universidad Nacional Agraria La Molina; 2024. https://hdl.handle.net/20.500.12996/6308
@misc{renati/1116848,
title = "Predicción diaria de temperaturas extremas y precipitación en la región Áncash mediante posprocesamiento estadístico del modelo GFS",
author = "Flores Aguilar, Andy Jean Pierre",
publisher = "Universidad Nacional Agraria La Molina",
year = "2024"
}
In this research, forecast models were developed for extreme temperatures and precipitation probability in the area of Huascarán National Park and its surroundings, in the Áncash region. Gridded data from PISCOp were used to calibrate the precipitation models and daily extreme temperature data collected from 10 stations in the study area were used to calibrate the temperature models. The Model Output Statistics (MOS) method was applied together with specific techniques for each type of prediction. The predictors for these models correspond to forecasts from the Global Forecast System (GFS) numerical model with a resolution of 0.25°. For the daily extreme temperature forecasts, the dimensionality reduction technique known as principal component analysis (PCA) was used, together with multiple linear regression to obtain the forecast equations. In the case of the precipitation probability models, both PCA and Random Forests were used for classification tasks. The models developed for extreme temperatures outperformed the non-post-processed predictions of the GFS model in the study area and also outperformed the predictions by persistence. Higher accuracy was observed in the prediction of minimum temperatures compared to maximum temperatures. However, certain limitations were identified in the model's ability to predict the magnitude of extreme temperature events. In relation to the probability of precipitation, results varied depending on the threshold set. The models for thresholds of 0.1 mm/day and 1 mm/day proved to be effective in predictions, while the model for 5 mm/day generated a high number of false alarms. These findings indicate that the developed models are promising, but also point to areas that could be improved to increase their accuracy, especially in the identification of extreme events.
This item is licensed under a Creative Commons License