Identification of Atmospheric Variable Using Deep Gaussian Processes

Abstract

Mathematical and physical modelling only provide an approximate description of the true nature of a dynamic system. The higher the accuracy of the model, the more likely it becomes analytically intractable; therefore, empirical models or black box models are used. When dynamic systems are considered as black box models, almost no prior knowledge about the system is considered. Deep Gaussian Processes, which use hierarchical structure to provide adequate identification of very complex systems, can be used to identify the mapping between the system input and output values. With the given mapping function, we can provide one-step ahead prediction of the system output values together with its uncertainty, which can be used advantageously. In this paper, we use deep Gaussian Processes to identify a dynamic system and evaluate the method empirically. In the illustrative case, we study one-step-ahead prediction of air temperature in the atmospheric surface layer.