Automated grass quality and quantity measurement techniques are becoming increasingly important in the quest to enhance grassland management. Therefore, the focus of this project was to develop and enable intelligent systems that will apply precision management to whole farm grassland and grazing systems. The goal was to optimize grass quality, utilization efficiency, and ultimately profitability, with minimal labour requirement and maximum objectivity. To precisely allocate to the cow herd the absolutely correct area of grass, it is necessary to have an accurate measure of grass quantity as well as quality.
The project team set out to evaluate three different technologies under Swiss conditions. For grass quantity measurements, the Grasshopper® G2 (TrueNorth Technologies, Shannon, Ireland) was included into the investigation. For grass quantity and quality measurements, we focused on multispectral imagery collected by an unmanned aerial vehicle (Parrot Sequoia, Parrot SA, Paris, France; DJI Phantom 4 Pro+ Drone, DJI, Shenzhen, China) and the HarvestLabTM 3000 (Deere & Company, Moline, Illinois, USA). As an add-on, the laboratory NIR System NIRFlex N-500 (Büchi Labortechnik AG, Flawil, Switzerland) was also included in the evaluation. As gold standard wet-chemical analysis was chosen.
The evaluation took place on six different locations across Central Switzerland, simulating grass height for grazing as well as silage cuts at 5 different points in time across the vegetation period. The grass quality measure were defined as % dry matter (DM) % crude protein (CP), % crude fibre, % ADF and NDF as well as % ashes.
In conclusion, none of the smart technologies are currently accurate enough for herbage quantity or quality estimations. However, NIRS as a technology can be very accurate when used under laboratory conditions. Therefore, more research needs to be done how to enable the technology to be used under practical conditions with suitable accuracy. It was also found that dry matter was the parameter easiest to estimate. However, due to the diversity of Swiss grasslands, a more specific approach for algorithms needs to be adapted. Therefore, further work will be conducted at Agroscope to develop the drone approach further for Swiss farmers. Furthermore, as the study identified that the standard algorithms used by the Grasshopper® system is not suitable for multispecies grasslands, we started a project in collaboration with the University of Hohenheim (Stuttgart, Germany) and the LAZBW (Aulendorf, Germany) to develop separate algorithms for Switzerland and South Germany collected during divers field experiments in 2018 and 2019. The developer of the Grasshopper® system is willing to add those algorithms to the Grasshopper App to make it a suitable decision support tool for our regions.
