Ensemble interpolation methods for spatio-temporal data modelling
Real time weather forecasting is a highly influential tool in decision making for agriculture. Geographic Information Systems (GIS) can be built to provide information about topographic data such as elevation and distance to oceans or water reservoirs. This data has begun to have increased availability, providing easier access for developing new applications. By using geographic information together with terrestrial measurements from weather stations, the spatial and temporal scales of the climatic variables can be analyzed by interpolation and forecasting. Most of the interpolation methods provided in common GIS tools are only related to the spatial domain, limiting its use in numerical modelling and prediction of climatic states. However, by adopting a Bayesian approach, it appears possible to estimate the dynamic behaviour of the unobserved climate pattern using a state-space representation. Using this framework, the ensemble Kalman filter or a more general sequential Monte Carlo method could be used for the estimation procedure. A wireless sensor network providing continuous data to populate such a model is described here for potential application of this approach.