Continuous monitoring of forests is possible—up to a point—using remote sensing from satellites, but satellite-based technologies typically cannot pierce the canopy cover to get more details about factors such as soil health and understory complexity. These can only be estimated by on-the-ground measurements. However, such measurements are both expensive and difficult to obtain, especially in tropical rainforests which sequester much of the world’s carbon and have the highest levels of biodiversity. We believe that the availability of drones might change this picture.
Specifically, the goal of this project is to increase the quality and volume of field plot data by using drones equipped with one or more cameras and a hyperspectral sensor to create rich datasets of the forest area. It would then be possible to use machine learning approaches such as partial least squares or RandomForest to analyse the resultant data. For example, it may be possible to work out which particular wavebands have reflectances that predict the underlying soil properties well. Metrics for forest complexity, or even picking up data from camera traps on the forest floor might also be possible.
Useful skills for this project include:
- Machine learning programming and algorithm design
- Data gathering and normalisation and analysis
- Drone assembly, flight and deployment in a control area
- Field work to gather datasets in local forest regions
The successful candidate will join the Application of Artificial Intelligence to the study of Environmental Risks (AI4ER) Centre for Doctoral Training (CDT), based at the University of Cambridge. The AI4ER CDT programme consists of a one-year Master of Research (MRes) course (two terms of formal teaching via lectures, practicals and team challenges plus a three-month research project), followed by a 3 year PhD project. Both the Masters and PhD research projects will be based on the above project description.
For further details on this project and how to apply please visit AI4ER’s applying to us webpages.