Research project S0/19/015 (Research action S0)
Project summary:
Nitrogen fertilization practices in Belgian agriculture are based on parcel nitrogen balances. In this N-balance, one term appears extremely difficult to estimate : the nitrogen provided by the mineralization of the soil organic matter (SOM). Indeed this term can be far from negligible as it can sometimes exceed the amount of nitrogen from the mineral fertilizer input on crops. One of the reasons explaining this difficulty stands in the high inter- and intra- variability of SOM. A better knowledge of the soil organic matter would allow not only to optimize the fertilization but also to reduce the impact of excessive fertilizations on the environment. This knowledge is necessary in precision farming.
The objective of the proposal is the determination of the soil surface organic matter in bare soils by an hyperspectral analysis in a region selected for its large range of organic matter content.
15 experimental fields selected in the study zone in Belgian Lorraine will be submitted to a field campaign in order to measure their water content and their organic matter in 10 points for each experimental fields precisely located with GPS. These field observations will be connected to their hyperspectral signal taken during the APEX 2002 air campaign. A multiple regression will select the spectral bands by stepwise analysis in order to extract the best correlated combination of bands with the observed soil surface organic matter. Next, a principal component analysis based on the selected bands only will serve for the simplification of the spectral signal necessary before the clustering process that will provide classes for maps of soil organic matter content. This analysis will be confronted with ground truth.
The study will also verify the interest of such an hyperspectral approach by a comparison with a more classical multispectral approach.
