Accounting for flux footprint to enhance the representativeness between remote sensing and in-situ data

The Brazilian Cerrado ecoregion, or wooded Cerrado, is considered one of the biodiversity hotspots. Despite the region’s importance in terms of supplying the water, food, and energy demand, there have not been enough ground-based studies. Furthermore, the lack of validation due to scale incompatibility and the great site-specific heterogeneity transpires in difficulty in the validation process.

The eddy covariance method has the potential to directly measure water vapor or trace gases on an in situ scale. Their measurement directly reflects the surrounding study site; thus, each time interval has a corresponding footprint. So, each study site's heterogeneity can affect the target vegetation's representativeness.

Here, we aimed to assess how two approaches for integrating remote sensing products and in-situ data affected representativeness in the wooded Cerrado. We used the Enhanced Vegetation Index (EVI) in both approaches, which are described as follows: (i) a fixed-fetch approach of the surrounding area considering a radius of 2 km and (ii) a lagrangian footprint approach that varied by a 30-minute time interval.  We assessed their performance based on their hourly and seasonal association with canopy conductance, which was carried out using in-situ data.

Compared to the fixed-fetch technique, the EVI footprint-integrated approach has a smaller range between the lower and upper quantiles, which is indicative of better targeting of the vegetation. Furthermore, we discovered that the integrated footprint technique produced a stronger association between EVI and canopy conductance than the fixed-fetch approach throughout most seasons and examined hours. The difference is most pronounced in the winter season, reaching a gain in the correlation of almost 100%, and for the autumn and spring with consistent gains of about 30%. Our findings highlight that integrating remote sensing products with footprint analysis can significantly improve the analysis's representativeness when targeting a specific land use or land cover, hence improving understanding of complex and heterogeneous areas.