Speaker
Description
- Introduction
Cover crops are an effective tool for decreasing nitrogen leaching in agrosystems, and their biodegradation after destruction provides nitrogen for the following crop, but in highly variable amounts depending on their biomass and the amount of nitrogen that they took up. Remote sensing data from Sentinel-2 satellites show great potential for addressing this variability through the development of predictive models based on vegetation indices and parameterised using field measurements of variables of agronomic interest (Dusseux et al., 2022). These perspectives led the mixed research unit SAS of INRAE (France) to initiate a project, in partnership with the Chambers of Agriculture of Brittany, to develop a web service (WS-CI) designed for farmers and agricultural advisors to characterise the development dynamics of cover crops and help predict their biomass and nitrogen (N) uptake. Developing this tool required setting up an experiment on a network of plots in which the biomass and N of the canopies were measured in elementary sampling units (ESUs), which consisted of homogeneous 20 m × 20 m areas (Verrelst et al., 2015). Here, we present the results obtained from this network and the design and features of the WS-CI web service.
2 Materials and methods
2.1 Data used
We took field measurements from 277 ESUs in Brittany during the autumn/winter of 2022-2023 and 2023-2024. In each ESU, we collected five georeferenced samples of above-ground biomass in a 1 m2 quadrat from each vertices and centre zones of the ESU. A composite sample from each ESU was sent to the laboratory for analysis of dry matter and N content. The data were carefully characterized based on several criteria.
Images from Sentinel-2 satellites were used to calculate vegetation indices. The satellite data for each sample were determined by intersecting the sample’s coordinates with pixels in the image and then averaging the values of each pixel to obtain one value per band for index calculation.
2.2 Selection of the best indices and modelling approach
We calculated 152 vegetation indices selected from the literature. The 20 best spectral bands or indices were selected based on a correlation test with field data, and then all possible pairs of the 20 bands or indices with different combinations were assessed using several modelling approaches (i.e. statistical and machine learning) (Verrelst et al., 2015) to determine which approach was most appropriate for the web service.
2.3 Web service
The WS-CI web service was designed using interoperable standards and free and open-source technology (Bera et al., 2015) (Fig. 1).
Figure 1. Architecture and technology of the web service
3. Results and discussion
We selected parametric regression to parameterise the models because it is highly accurate and simple to implement. We integrated into the web service the most accurate model, which combined GEMI (Pinty et al., 1991) and NIRV (Grayson et al., 2017) (RMSE = 6.53 kg N/ha) and the vegetation index that had the strongest correlation with the amount of fresh biomass (MCARIOSAVI705, Chaoyang et al., 2008) to visualize the development of canopies on a map.
The web service, based on open-source technology, enables users to easily view the development of cover crops over time and the maximum amount of N that they take up at the plot scale. It is now in production and already helping to improve advice on N fertilisation.
References
Dusseux P. et al, 2022. International Journal of Applied Earth Observation and Geoinformation, Volume 111, https://doi.org/10.1016/j.jag.2022.102843
Verrelst J. et al, 2015. ISPRS Journal of Photogrammetry and Remote Sensing, 108, 260-272, https://doi.org/10.1016/j.isprsjprs.2015.05.005
Bera R. et al, 2015. Proc. IAHS, 368, 9–14, https://doi.org/10.5194/piahs-368-9-2015
Pinty, B. et al, 1992. Vegetatio 101, 15–20, https://doi.org/10.1007/BF00031911
Grayson B. et al, 2017. Sci. Adv.3,e1602244, https://doi.org/10.1126/sciadv.1602244
Chaoyang W. et al, 2008. Agricultural and Forest Meteorology, Volume 148, https://doi.org/10.1016/j.agrformet.2008.03.005
| Keywords | cover crops, biomass; nitrogen uptake; web portal; interoperability; WMS and WFS protocols |
|---|
