EROCROP - Analysis of the crop productivity - soil erosion relationship using hyperspectral data

Context and objectives

The project aims to identify a relationship between spatial variation of crop response and variation in soil erosion and topographic variables for an area with complex relief in the Belgian Loess Belt, using hyperspectral data. Hyperspectral data are progressively used to detect stress in crops, to allow more efficient application of water and nutrients. This kind of information on crop stress could also be valuable for unravelling the soil erosion – crop response relationship. In the presented research we have further explored these possibilities. We aimed to answer the following questions:
• Can we determine a relationship between soil erosion and crop response for the Belgian Loss Belt using hyperspectral data?
• Can hyperspectral data provide more clues on which, erosion related, factors affect crop growth?
• Is this information available at a regional scale?

Project outcome

Expected scientific results

Despite the promising possibilities of hyperspectral data, we have so far not succeeded to use it for clarification of the soil erosion – crop response relationship in the Hageland area. Two major types of factors may be responsible for this finding: (i) factors relating to technical difficulties while applying the technique and (ii) factors relating to the nature of the soil erosion-crop response relationship under intensive arable cultivation.

We did find significant relationships between field variables and bands or indices, but the interpretation of these results should be done cautiously. Between-field variability may affect the correlations. Small sample numbers may also cause unreliable results. The relationships we found were not comparable to those described in literature.

We did find relatively good relationships between crop response and hyperspectral response in some cases (e.g. a good relationship between crop height and response in some bands). However, neither plant response, nor hyperspectral response could be linked to erosion intensity. Recent research (Bakker et al., 2005) has shown that there indeed exists a relationship between crop response and erosion, but that this relationship is relatively weak, with a 4 % decline in yield for every 0.1 m of topsoil lost. It may therefore well be that we cannot detect this response using hyperspectral remote sensing.