Comprehensive intercomparison and evaluation of global upscaled foliar trait maps

Global trait maps of specific leaf area (SLA), leaf nitrogen (N) and phosphorus (P) contents have been generated using a wide range of data-driven upscaling approaches. We comprehensively studied their consistency and agreement with sPlotOpen data at 0.5 degee grid cells. For this, we developed approaches to separate the maps into their plant functional type (PFT) components by taking into account within-grid-cell heterogeneity and stratified sPlotOpen data by PFT.

We found that despite many differences in the upscaling approaches, the maps fall into two groups: One group using remote sensing based, fractional PFT cover  in the upscaling, while the other did not. Spatially, the main differences between the two groups are located in areas of high within-grid-cell trait heterogeneity and these areas dominate global trait variations due to the combined effects of fractional PFT cover and trait differences between-PFTs.

The agreement of upscaled maps with sPlotOpen data strongly depends on the way sPlotOpen data are scaled to the grid cell. When using a similar scaling approach as the upscaling approaches a similar level of agreement can be observed for both groups of maps. However, only the maps that used PFT information could capture main features of between-PFT differences, especially the low values of SLA and N in evergreen needleleaf forests. Within-PFT trait variations of upscaled maps partly showed similar patterns as sPlotOpen data when aggregated to latitudinal averages but considerable differences remain and the evaluation is challenging without having the original maps per PFT. 

We conclude that fractional PFT cover contains essential information for capturing global, top-of-canopy trait patterns using upscaling approaches at moderate to coarse spatial resolution.