Our research focuses on improving the spatial assessment of aboveground biomass of woody vegetation by combining field observations, remote sensing and auxiliary data, from local case studies to global scale, with focus on the tropical region.
An improved pan-tropical biomass map
Using a novel data fusion approach (see below), we have integrated two existing large-scale biomass maps (Saatchi et al., 2011; Baccini et al., 2012) with local high-quality biomass data into an improved pan-tropical aboveground biomass map of woody vegetation at 1 km resolution for the 2000’s.
The method and results are described in Avitabile et al. (2016). The data are freely accessible, and can be viewed and downloaded here:
A first global forest biomass map
The pan-tropical biomass map was integrated with the boreal forest biomass component (Santoro et al., 2015) into the GEOCARBON global aboveground forest biomass map at 0.01° resolution. This is the first global biomass map based on field observations and remote sensing data. The map covers only forest areas, where forest are defined as areas with dominance of tree cover in the GLC2000 map (Bartholomé and Belward, 2005).
The global forest biomass map can be downloaded here. The boreal forest biomass map can be accessed at www.biomasar.org. Additional global datasets can be accessed through the GEOCARBON Data Portal:
Our approach for pan-tropical biomass mapping
Aboveground biomass plays a key role in the carbon cycle and climate processes and in the last few years there has been a large effort to improve its spatial assessment, particularly in the tropical region where uncertainties are higher (Avitabile et al., 2011; 2012).
Two large-scale maps (Saatchi et al., 2011; Baccini et al., 2012) provide wall-to-wall biomass density at moderate resolution for the tropical belt. While total biomass estimates tend to converge (Mitchard et al., 2013), the regional maps present considerable disagreement in terms of absolute values and spatial distribution of biomass. It is unclear which product is more appropriate but it is likely that map accuracy varies spatially, according to the distribution and quality of their training data.
We optimally combined these two biomass maps using independent reference data into an improved pan-tropical biomass map at 1 km resolution for the 2000’s. This study was carried out in the context of the EU FP7 GEOCARBON project.
Harmonize reference biomass data with different spatial and thematic characteristics
- A variety of high-quality independent research field observations, forest inventory plots and high-resolution biomass maps were screened and harmonized, developing specific procedures to upscale plot data and integrate datasets with different characteristics.
Assess the error structures of the regional biomass maps
- The reference dataset was used to better understand the spatial accuracy of the regional maps. Error maps were produced using additional covariates and a RandomForest modelling approach.
Integrate regional biomass maps into an improved pan-tropical biomass product
- A fusion method based on bias-removal and weighted-average of the regional maps (Ge et al., 2014) was used to produce a novel map with lower error variance and higher accuracy than the input maps (Avitabile et al., 2016).