B.PAS.0510 - Spatio-temporal prediction of pasture dieback using UAVs and remote sensing
Remotely sensed imagery was combined with soil and climate data for cost-effective identification, mapping and monitoring of pasture dieback infestation over time at scale.
| Project start date: | 31 August 2020 |
| Project end date: | 30 January 2022 |
| Publication date: | 17 April 2024 |
| Project status: | Completed |
| Livestock species: | Grass-fed Cattle, Sheep, Lamb |
| Relevant regions: | NSW, Queensland |
|
Download Report
(5.4 MB)
|
|
Summary
Pasture dieback (PD) is causing widespread damages to pastures and beef production in Queensland. However, PD causal agents are still poorly understood. This project integrates remotely sensed imagery and publicly available climatic data for cost-effective identification, mapping and monitoring of PD infestation over time at scale.
Objectives
The main objectives of the project are to:
1. Characterise PD spectral signature using high resolution hyperspectral camera.
2. Develop a predictive model for PD proliferation based on UAV, satellite imagery, and environmental variables.
3. Engage and communicate across stakeholder groups to discuss digital delivery platform (s) suitable to meet industry needs.
Key findings
• Three difference models were built to (1) predict the chances of PD occurrence relative to climatic conditions, (2) classify unhealthy grass, and (3) identify if the unhealthy grass was due to unfavourable weather conditions or other disturbances such as changes in land management or pest infestation.
• The models performed well based on the data they were trained on. However, more training data is required for model applications across multiple climatic regions and pasture species.
• Engagement with stakeholders shows interests for model deployment on an on-demand and easy-to-setup platforms or web applications.
Benefits to industry
This research showcases the usefulness of satellite and gridded climate data in monitoring and predicting PD occurrence. Our models could help producers plan to mitigate the likely impacts of PD on livestock management and grazing.
MLA action
Further research to develop a remote sensing tool for producers will require significant investment. Progress will require a collaboration between CiboLabs (AFM) and QUT to deliver an effective and seamless outcome.
Future research
Future research needs to focus on acquiring more data to improve the model performance as well as building platforms for deployment of the resulting models at scale.
More information
| Project manager: | Felice Driver |
| Contact email: | reports@mla.com.au |
| Primary researcher: | Queensland University of Technology |
