Precise herbage biomass estimation on pastures using a UAV-based model
Photo: BFH-HAFL
A new model for predicting herbage biomass in pastures uses UAV-based image data and machine learning. The method promises greater accuracy and flexibility compared to conventional measurement methods.
More than 70% of Switzerland’s agricultural area is covered by grasslands. A large proportion of this land is used for livestock grazing. Efficient grazing management is challenging because yields are highly dependent on weather conditions, management intensity and sward species composition. The precise estimation of herbage biomass is essential for targeted grazing management. However, the currently used manual methods such as such as Rising Plate Meters or C-Dax Pasture Meters are often time-consuming and only measure a subset of locations.
New method
To better meet the requirements of accurate herbage biomass estimation, the estiGrass3D+ model has been developed. In this model, a digital terrain model (DTM) is derived from the digital surface model (DSM) based on UAV imagery. The difference between DSM and DTM is used to determine the sward height, eliminating the need for absolute georeferencing. In addition, spectral vegetation indices are obtained from the UAV data and processed using a Random Forest algorithm. The method has already demonstrated high accuracy on test data (NRMSE: 20.3%) and provides a comprehensive and reliable data base compared to conventional techniques. With further automation, the model could provide a valuable and efficient tool for long-term pasture management.
The red curve shows a smoothed trend (LOWESS).
Applications and opportunities
Various trials at sites in the Swiss central plateau demonstrate the versatility of the model for both grazed and cut areas. In practice, the model could help farmers to more accurately monitor herbage biomass and to tailor feed planning more closely to management objectives. However, further developments are needed to increase the model’s efficiency and user-friendliness through automated processes before it can be truly integrated into agricultural practice.
Conclusions
- Herbage biomass estimation with the estiGrass3D+ model presents promising approaches in support of pasture usage.
- In the long term, the model could enable farmers to use flexible, data-driven planning that offers both ecological and economic advantages for pasture farming.
- Although the model provides an accurate basis for decision-making, automated software solutions that would facilitate widespread adoption among practitioners are still lacking.
- Future research and development work could further optimise the workflow and add software support for greater efficiency.
Bibliographical reference
Herbage biomass predictions from UAV data using a deriveddigital terrain model and machine learning.
