Comparing the spatio-temporal differences of global NPP simulation data with different resolutions

<p>*Corresponding author: Xiaolu Tang (lxtt2010@163.com)</p><p>Net primary productivity (NPP) is a key parameter to characterize terrestrial ecological processes. NPP reflects the carbon sequestration capacity of vegetation to absorb atmospheric carbon dioxide, and plays an important role in mitigating atmospheric carbon dioxide content. Currently, the majority of studies focused on the model efficiency total NPP at the global scale. However, whether the model resolution of NPP affects the NPP amount at the global is still uncertainty. To fill this knowledge gap, we first collected 3307 NPP field observations from published literatures, and then model NPP using climate, soil, and vegetation variables using Random Forest (RF) to predicted global NPP at the spatial resolutions of 0.05&#176;, 0.25&#176; and 0.5&#176;. Results showed that RF could well capture the spatial and temporal variability of NPP with the model efficiencies (<em>R²</em>) of 0.55, 0.52 and 0.53 for at the resolution of 0.05&#176;, 0.25&#176; and 0.5&#176;, respectively. Similar spatial patterns were also found for NPP at different spatial resolutions and NPP decreased with increased latitude where the highest NPP was found in the tropical regions and the lowest NPP were distributed in high latitude areas, e.g. alpine tundra. However, a great difference was found for the magnitude of NPP resulting a great difference in total global NPP of 71.5, 78.6, 87.7 Pg C year^-1 from 1981 to 2016 for the resolutions of 0.05&#176;, 0.25&#176; and 0.5&#176;, respectively. These findings suggested the challenges to improve modelling accuracies of the global carbon fluxes used appropriate resolutions.</p><p><em>Keywords:</em> Net primary productivity; Different resolutions; Random Forest; Spatial pattern; Appropriate resolution;</p><p>Acknowledgment: the study was supported by the National Science Foundation of China (31800365).</p>