Diagnostic Up-scaling of GPP from Eddy Covariance to Global Estimates

The uptake of atmospheric CO2 by plant photosynthesis is the largest global carbon flux and drives all terrestrial carbon cycle processes. While the photosynthesis processes at the leaf and canopy levels are quite well understood, so far only very crude estimates of its global integral, the Gross Primary Production (GPP) can be found in the literature. Existing estimates have been lacking sound empirical basis. Reasons for such limitations lie in the absence of direct estimates of ecosystem-level GPP and methodological difficulties in scaling local carbon flux measurements to global scale across heterogeneous vegetation. Here, we present global estimates of GPP based on different diagnostic approaches. All these up-scaling schemes integrated high-resolution remote sensing products, such as land cover, the fraction of photosynthetically active radiation (fAPAR) and leaf-area index, with carbon flux measurements from the global network of eddy covariance stations (FLUXNET). In addition, meteorological datasets from diverse sources and river runoff observations were used. All the above-mentioned approaches were also capable of estimating uncertainties. With six novel or newly parameterized and highly diverse up-scaling schemes we consistently estimated a global GPP of 122 Pg C y-1. This value is 5 % higher than estimates from inversions of 18O and CO2 atmospheric concentration. In the quantification of the total uncertainties, we considered uncertainties arising from the measurement technique and data processing (i.e. partitioning into GPP and respiration). Furthermore, we accounted for the uncertainties of drivers and the structural uncertainties of the extrapolation approach. The total propagation led to a global uncertainty of 15 % of the mean value. Although our mean GPP estimate of 122 Pg C y-1 is similar to the previous postulate by Intergovernmental Panel on Climate Change in 2001, we estimated a different variability among ecoregions. The tropics accounted for 32 % of GPP showing a greater importance of tropical ecosystems for the global carbon cycle than previously thought. On the other hand, the total GPP of natural temperate ecosystems and in shrublands was estimated smaller than by applying a NPP/GPP ratio of 0.5 to inventory-based NPP. We estimated higher GPP values also for croplands. Our thorough estimation of spatially distributed GPP provides a new dataset extremely valuable for benchmarking ecosystem process models.