Hyperspectral data for scaling ecosystem traits from point to landscape

An accurate and spatially explicit estimation of ecosystem traits such as leaf area index and chlorophyll content is of fundamental importance for biogeochemical modeling. In this perspective remotely sensed spectral information has been shown to be a powerful tool for investigating ecosystem traits and their dynamics. Unluckily the availability of remote sensed hyperspectral data is still limited. On the other side hyperspectral imaging technology for near surface sensing has been recently quickly developing. Instruments are becoming much lighter allowing for being transported on unmanned aerial vehicles (UAV). UAV imagery is spatially limited but offers more flexibility compared to satellite based sensors allowing for specific and high-resolution scenes acquisition. In this context the use of spectral information for inverting radiative transfer models is a widespread approach to estimate biophysical parameters from earth observation data and its application to near surface sensing is offering new opportunities for spatial explicit estimates of ecosystem traits.

In the present study we apply an inverted radiation transfer model to two grassland sites with differing management regimes. The objective was to derive ecosystem traits and to provide insights on scaling effects by using multiple spectral measurements techniques addressing different spatial scales. The application is to link point to landscape scale estimates through a better understanding of the scaling effects.

For addressing this objective this we made use of a handheld field spectrometer and an UAV-borne imaging Fabry-Peroth interferometer for characterizing the sampling plots. The PROSAIL model parameters were then calibrated by using a Monte Carlo approach. We applied the same method using modis reflectances for the pixels surrounding the experimental plots. The accuracy and reliability of the estimated ecosystem traits with the different methods were then further compared to direct measurements. This allowed to assign uncertainties to the satellite based estimates.

Our approach showed the importance of taking in account the spatial variability of ecosystem biophysical properties addressing scales ranging from plot to landscape.