Abstract:

We describe an approach for evaluating the representativeness of eddy covariance flux measurements and assessing sensor location bias (SLB) based on footprint modelling and remote sensing. This approach was applied to the 12 main sites of the Fluxnet-Canada Research Network (FCRN)/Canadian Carbon Program (CCP) located along an east-west continental-scale transect, covering grassland, forest, and wetland biomes. For each site, monthly and annual footprint climatologies (i.e. monthly or annual cumulative footprints) were calculated using the Simple Analytical Footprint model on Eulerian coordinates (SAFE). The resulting footprint climatologies were then overlaid on to images of the Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI) derived from LANDSAT Thematic Mapper (TM) imagery, which were used as surrogates of land surface fluxes to estimate SLB. Results indicate that (i) the sizes of annual footprint climatology increased exponentially with increasing cumulative footprint percentages and, for a given percentage of footprint climatology, the footprint areas were significantly different among the sites. Typically, the 90% annual footprint climatology areas varied from 1.1 km² to 5.0 km²; (ii) using either NDVI or EVI as the flux surrogate, the SLB was less than 5% for most sites with respect to the reference area of interest (A_r) at 90% annual footprint climatology (scenario A) and a circular area with radius of 1 km centred at the individual tower (scenario B), with several exceptions; (iii) the SLB decreased with increasing size of footprint climatology for all sites for both scenarios A and B; (iv) out of 12, eight flux towers represented most of the ecosystem surrounding the towers for an area of 0.3 km² up to 10 km² with a satisfactorily low bias of <5%, whereas four towers represented areas ranging from only 0.75 to 4 km²; and (v) the seasonal differences in monthly SLB using NDVI as a flux surrogate were about 1–4% for most sites for both scenarios A and B.