Abstract ID: 119

Spatial distribution of mean crown size in Amazon rainforests assessed using Fourier transform textural ordination of VHR panchromatic imagery.

Crown size is rarely considered in protocols for field inventory in rainforests, yet it can be argued as an important component in the estimation of above-ground live biomass (AGLB). Earth observation data could prove useful, at least to approach crown sizes of the dominant strata, which are known to comprise a large share (up to 50 %) of AGLB. In rainforests, Fourier textural ordination (FOTO) of very high resolution panchromatic images has been shown to be an efficient and promising approach to infer biophysical parameters, and more particularly crown size. Here we show how the method can be applied at a regional scale using data freely available in virtual globes (e.g. Google EarthTM). The effect of acquisition parameters variability is discussed using a light interaction model (DART). Less than 8% of error in the prediction of tree crown size can be attributed to random effects and/or instrumental error, prior to any correction of the latter. This error reduces down to 4 % when instrumental parameters are taken into account. A map of canopy texture across the Amazon basin is produced, showing a pattern of variability very different from previous maps of biophysical parameters in the Amazon.

Session:  Carbon - Scaling carbon fluxes to the region from measurements in plots, towers, and aircraft.

Presentation Type:  Poster