Construction and Experiment of Hierarchical Bayesian Network in Data Assimilation

A Hierarchical Bayesian Network Algorithm (HBN) is developed for data assimilation and tested with an instance of soil moisture assimilation from hydrological model and ground observations. In essence, HBN is a framework that can statistically describe Bayesian models and capture the dependences in the models more realistically than non-hierarchical Bayesian models. In this work, data assimilation separates into data level, process level and parameter level, and conditional probability models are defined for each level. The data model mainly deals with the scale differences between multiple data, while the process model is designed to take account of non-stationary process. Soil moisture from Soil Moisture Experiment in 2003 and Variable Infiltration Capacity Model is sequentially assimilated with HBN. The result shows that the assimilation with HBN provides spatial and temporal distribution information of soil moisture and the assimilation result agrees well with the ground observations. In summary, the HBN is a good algorithm together with data, process and parameter model, which shows great potential for data assimilation development.