Observation Impact Assessment on the Prediction of the Earth System Dynamics Using the Adjoint-Based Method

Abstract

Mathematical models of the Earth system and its components represent one of the most powerful and effective instruments applied to explore the Earth system's behaviour in the past and present, and to predict its future state considering external influence. These models are critically reliant on a large number of various observations (in situ and remotely sensed) since the prediction accuracy is determined by, amongst other things, the accuracy of the initial state of the system in question, which, in turn, is defined by observational data provided by many different instrument types. The development of an observing network is very costly, hence the estimation of the effectiveness of existing observation network and the design of a prospective one, is very important. The objectives of this paper are (1) to present the adjoint-based approach that allows us to estimate the impact of various observations on the accuracy of prediction of the Earth system and its components, and (2) to illustrate the application of this approach to two coupled low-order chaotic dynamical systems and to the ACCESS (Australian Community Climate and Earth System Simulator) global model used operationally in the Australian Bureau of Meteorology. The results of numerical experiments show that by using the adjoint-based method it is possible to rank the observations by the degree of their importance and also to estimate the influence of target observations on the quality of predictions.