Análisis y pronósticos de escenarios climáticos para la provincia de Tungurahua

Abstract

The objective of this research work is to determine the forecast models for climate change scenarios in the cantons of the province of Tungurahua, through ARIMA models and derivatives so that with this knowledge correct decisions can be made for the climate scenarios for the province of Tungurahua. To accomplish the research work, the monthly data for the period 2013-2021 provided by the Honorable Provincial Government of Tungurahua was used, in which the variables analyzed were: wind direction, maximum, average, and minimum humidity, precipitation, maximum, average and minimum temperature, wind speed which are registered in the meteorological stations of the (HGPT) of Tungurahua. Through an exploratory analysis of the data, missing data was found which was imputed to perform a better modeling, atypical data was detected in the variables under study, in addition a decomposition of the time series and it was observed that there is seasonality. Using the Box Jenkins methodology, the best models for the time series were found and for the validation of the proposed models, the assumptions of normality, homoscedasticity and independence to errors were tested. Using the Akaike and Bayesian information criteria, the best of the proposed models for the variables under study were chosen. For the variable Maximum Temperature, an ARIMA (0,0,0) (2,0,0) [12] is presented as the best model, which meets the validation assumptions. Finally, predictions were obtained using the models proposed for the meteorological variables under study in the Baños and Patate cantons. It is concluded that the Box Jenkins technique is the best technique for predictions. It is recommended in the study of forecasts to use and update techniques such as Machine Learning for better forecast accuracy.