Forecasting temperature time series for irrigation planning problems

Abstract

Climate change is a reality and efficient use of scarce resources is vital. The challenge of this project is to study the behaviour of humidity in the soil by mathematical/statistical modeling in order to find optimal solutions to improve the efficiency of daily water use in irrigation systems. For that, it is necessary to estimate and forecast weather variables, in this particular case daily maximum and minimum air temperature. These time series present strong trend and high-frequency seasonality. This way, we perform a state space modeling framework using exponential smoothing by incorporating Box-Cox transformations, ARMA residuals, Trend and Seasonality.