Identifying Significant Determinants for Canopy Development on an Alpine Test Site by means of Artificial Neural Networks

摘要

We develop a methodical approach to model structural canopy development (biomass, leaf area) of different grassland stands in the North Italian landscape. Classic punctual field measures are linked to temporal-spatial high-resolution remote sensing data using artificial neural networks in order to generate large-scale forecasts for canopy development. We find empirical evidence to support our claim that RGB (red, green, blue) colour values can contribute to a better understanding of canopy development over time and in space. We provide graphical and statistical measures to identify the form and the importance of influence factors on canopy development. This approach allows us to scale up the plot-level measurements to landscape-level measurements (e.g. from biomass data to a biomass map).