Multivariate forest modelling and mapping using Quickbird imagery and topographic data in Chelsea, Quebec

This research determined the capability for modelling and mapping multivariate forest complexity as an indicator of forest biodiversity using remote sensing and topographic data in the municipality of Chelsea, Quebec. In 70 field plots, 37 structure and composition variables were measured. Image spectral and spatial variables were derived from Quickbird imagery, and topographic variables were derived from a DEM. Several field based structure/composition indices were developed and then modeled using multiple regression against the geospatial variables. The results were compared to an index derived directly from the geo-spatial data using Redundancy Analysis (RDA). The best models were an index derived additively from a set of ten core field variables and a 4 predictor variable RDA model. These models were then applied over the study area to obtain a map of predicted forest complexity for Chelsea, Quebec, which can be used to aid in biodiversity survey planning and conservation efforts.