Gaussian Bayesian Network Modeling to Improve Spatial Growth Estimates of Heterogeneous Forests

An approach is presented for improving the spatial estimation of leaf area index (LAI) of a heterogeneous forest by integrating the Physiological Principles Predicting Growth (3-PG) model output with the Moderate Resolution Imaging Spectroradiometer (MODIS) satellite imagery. LAI was chosen as the variable of interest because leaf area is the exchange surface between the photosynthetically active component of the canopy and the atmosphere. A novel inference strategy within the Gaussian Bayesian networks (GBNs) has been developed in this work to take care of the different structures of the inaccuracies in the two data sources. The Linear Mixture Model (LMM) was used to decompose MODIS pixels using class fraction derived from an aerial and an Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) images. In this way spatially heterogeneous output was produced. Results showed that the spatial output obtained with the GBN was more accurate than both the spatial 3-PG model output and the satellite estimate, as the root mean square error reduced from 1.21 to 0.57, and the relative error from 20.25% to 7.73%. In this work, we conclude that the GBNs can improve the spatial estimation of the LAI values of a heterogeneous forest by integrating a spatial 3-PG forest growth model with satellite data.