Regresar

Kriging Kalman combined filter to estimate and predict the evolution of climatic states in some weather stations in Ecuador

Abstract:

This article proposes a methodology that involves the Universal Kriging filter (UKF) and the Kalman filter (KF) to study temporal-space dynamic models. Spatial prediction surfaces are constructed using the UKF and the KF is used to estimate the temporal effects. The UKF provides a successful estimation approach from the point of view of spatial statistics, while the KF describes a well-established recursive procedure to estimate the states and parameters in these models. The methodology is illustrated using 30-year time series of 3 meteorological stations in Ecuador. The model allows to make predictions on temperature, precipitation and humidity, obtaining estimates of unknown states very similar when compared to original series. The root mean square error was used as a measure of goodness of fit to measure the estimation quality of the algorithm, obtaining satisfactory results.