The UAVsoil project (2017-2019) took place in a context of optimisation of agricultural management at field level through the use of images from sensors embedded on UAV to characterise the heterogeneity of the crop and the soil.
Towards a more sustainable agriculture
Excessive use of fertilizers is now recognised as playing a major role in environmental degradation and is encouraging farmers to improve their management of these fertilizers. To assist them, crop growth maps and soil property maps can be used to better characterise intra-field heterogeneity.
UAV to characterize soil properties
The technological innovations of sensors embedded on UAVs enable ever finer characterisation of crops and soil properties. In particular, it is possible to assess the soil organic content by carrying out UAV flights on bare soil.
The main objective of the project was to test the potential of UAV-based sensors to produce parameters to explain yield variations within a field, with a view to improve yield prediction and optimisation.
Specifics objectives of the project were :
- The production of maps of soil and crop properties at the field level
- Analysis of the intra-plot heterogeneity of the soil and vegetation, and their link
- Evaluation of the possibility of using soil maps in a classic nitrogen recommendation tool using the balance sheet method.
- Evaluation of the potential of UAV-based sensors to produce soil properties maps
The final results of the project mainly concern a CRA-W 17 ha field of winter wheat (2018), resulting from an aggregation of 4 former fields.
Vegetation analysis using a multispectral sensor (RedEdge-MTM from MicaSense®) embedded on a UAV revealed that the Red-Edge NDVI (RENDVI) was the most appropriate vegetation index to characterize the heterogeneity of crop development throughout the growing season. The RENDVI maps were also highly correlated with the yield map produced by a sensor embedded on the harvester.
The analysis of the management history of the former fields as well as the soil data collected according to a regular grid showed the significant impact of management differences and the heterogeneity of the field soil properties. This heterogeneity had a direct impact on the development of vegetation, with soil properties explaining from 87% (mid-March) to 78% (at harvest) of the heterogeneity of crop development. The most impacting properties were pH and potassium. The "Conditionnal Inference forest" algorithm allowed this characterization to be useful for guiding the fertilizer management within the plot.
Tests using a hyperspectral sensor (OCEAN FX SpectrometerTM) to determine soil organic matter were promising, in contrast to the multispectral sensor (RedEdge-MTM from MicaSense®). As the work carried out is still very exploratory, this aspect will be further developed.
- Preparation and implementation of field campaigns
- Supply of historical data and existing soil sampling data
- Processing and analysis of field data and vegetation data from UAVs
- Production of scientific publications and vulgarization documents
- Presentations at several conferences and other events
- Université catholique de Louvain (UCLouvain), Earth and Life Institute (ELI), Earth & Climate (ELIC) : Bas van Wesemael
- Helmholtz-Zentrum Potsdam Deutsches GeoForschungsZentrum (GFZ) : Sabine Chabrillat
Prof. Bas van Wesemael
Université catholique de Louvain (UCLouvain)
Earth and Life Institute (ELI), Earth & Climate (ELIC)
Place Louis Pasteur 3 bte L4.03.08, 1348 Louvain-la-Neuve