 |
Incorporation remote sensing retrievals into biogechemical models will provide more reliable estimates of production and C and N fluxes in crop systems in the perspective of climate mitigation.Traditional remote sensing techniques face different limitations when applied to precision agriculture or conservative agriculture purposes. In this aspect, the spatial heterogenity of field crops, that can be constitued by different fractional covers of green vegeation, non-photosynthetic vegetation and soil make that the observed spectrum is affected by the different covers, making it difficult to identify the individual contribution. Also the traditional interpretation of spectral data is often based on vegetation indices (VIs) that consider only few spectral bands with no physical-based interpretation, preventing direct estimation and quantification of plant traits and surface pure components. This study validates different modeling techniques to overcome this limitation and provide accurate information to biogechemical models: i) a hybrid machine learning-radiative tranfer model pipeline is tested in a field experiment to retrieve key plant traits using satellite imagery and ii) a multiple endmember spectral mixture analysis (MESMA) is tested to monitor farmer strategies at regional sale by mapping the fractional covers in crops fields.
