Mapping Vegetation Structure Gradients in Colombia With Spaceborne Lidar

Abstract:

Climate, topography, and land use are among the main drivers of vegetation structure in the tropics. Until recently, our primary source of information on tropical vegetation structure at broad geographic scales has been multispectral satellite imagery. Though invaluable for providing estimates of vegetation cover, such imagery cannot provide measurements of vegetation height or vertical distribution of vegetation in canopies. Therefore, an outstanding question in biodiversity research is how elements of vegetation structure vary and co-vary geographically across environmental gradients. Here we use machine learning with spaceborne lidar data from the Global Ecosystem Dynamics Investigation (GEDI), multispectral satellite imagery, elevation, and climate data to map how different elements of vegetation structure respond to environmental and land use gradients in Colombia. We find that different vegetation structure elements respond in different ways across gradients, leading to distinct combinations of climate, land use, and vegetation structure in different parts of the country. In addition, we find that there are specific values of environmental variables where vegetation structure changes more rapidly. Unsupervised classification of combinations of vegetation structure and environmental variables can be viewed as structurally informed vegetation types. In some places, these types align closely with traditional ecoregion delineations and in other places they diverge significantly. The continuous and classified map outputs can be used to inform biodiversity research, assess representativeness of protected areas, and understand human impacts on vegetation structure.