Drivers of spatio-temporal variability of carbon dioxide and energy fluxes in a Mediterranean savanna ecosystem

To understand what is driving spatial flux variability within a savanna type ecosystem in central Spain, data of three co-located eddy covariance (EC) towers in combination with hyperspectral airborne measurements and footprint analysis were used. The three EC systems show consistent, and unbiased mass and energy fluxes. Nevertheless, instantaneous between-tower flux differences i.e. paired half hourly fluxes, showed large variability. A period of 13 days around an airborne hyperspectral campaign was analyzed and proved that between tower differences can be associated to biophysical properties of the sampled footprint areas. At high photosynthetically active radiation (PAR) net ecosystem exchange (NEE) was mainly controlled by chlorophyll content of the vegetation (estimated through MERIS Terrestrial Chlorophyll Index (MTCI)), while sensible heat flux (H) was driven by surface temperature. The spatial variability of biophysical properties translates into flux variability depending on the location and size of footprints. For H, negative correlations were found with surface temperature for between-tower differences, and for individual towers in time, meaning that higher H was observed at lower surface temperatures. High aerodynamic conductance of tree canopies reduces the canopy surface temperature and the excess energy is relieved as H. Therefore, higher tree canopy fractions yielded to lower surface temperatures and at the same time to higher H. For NEE, flux differences between towers were correlated to differences in MTCI of the respective footprints, showing that higher chlorophyll content of the vegetation translates into more photosynthetic CO2 uptake, which controls NEE variability. Between-tower differences of latent heat fluxes (LE) showed no consistent correlation to any vegetation index (VI), or structural parameter e.g. tree-grass-fraction. This missing correlation is most likely caused by the large contribution of soil evaporation to ecosystem LE, which is not captured by any of the biophysical and structural properties. To analyze if spatial heterogeneity influences the uncertainty of measured fluxes three different measures of uncertainty were compared: the standard deviation of the marginal distribution sampling (MDS), the two-towerapproach (TTA), and the variance of the covariance (RE). All three uncertainty estimates had similar means and distributions at the individual towers while the methods were significantly different to each other. The uncertainty estimates increased from RE over TTA to MDS, indicating that different components like space, time, meteorology, and phenology are factors, which affect the uncertainty estimates. Differences between uncertainty estimates from the RE and TTA indicate that spatial heterogeneity contributes significantly to the ecosystem-flux uncertainty.