A Method for Measuring the Incremental Information Contributed from Non-Stationary Spatio-Temporal Data to be Fused

Complex hydrologic problems, such as estimating stream discharge (flow), require the utilization (fusion) of multiple sets of measurements (features). Due to the computational cost of incorporating all available features, it is desirable to use a reduced set of the most informative ones. A method is presented for determining the information gain of different feature subsets in a Bayesian network. The method is applied to the problem of estimating flow in a Spatio-Temporal Bayesian Network (STBN), under the constraint that the features retain their original physical meaning.