Can small clusters of trees add value to rural landscapes?
We are leading a five year programme looking at the benefits of increasing biomass and soil carbon stocks from clusters of trees established in pasture as an alternative to plantation forest.
The Trees in Landscapes programme, led by David Whitehead and Sam McNally, is focusing on enhanced carbon stocks around the edges of tree clusters compared with the same land area of plantation forest with a number of additional co-benefits for landowners:
- Provide shelter and food for animals
- Reduce nitrate leaching
- Reduce erosion by holding hillsides together
- Improve productivity
- Maintain rural livelihoods
- Aligns with te ao Māori world views.
Can you help our researchers?
The programme is looking to recruit 32 farms to obtain baseline data to test the idea that there are carbon and biomass benefits around clusters of trees.
We have strict criteria for site selection:
- Blocks or belts of trees alongside grazed pasture
- Can be fenced or unfenced
- Land cannot be too steep (less than 20 degrees)
- Tree clusters must be at least 20 meters wide
- Trees must be at least 7 years old
- Can be any species.
Once a farm has been selected, our research staff will be on site for 2-3 days taking soil samples and doing lots of measurements around the tree clusters. No destructive work will take place and the site will be left in the same condition as it started in.
From there, four sites will be selected that will have regular seasonal visits over the next four years for more detailed measurements. Any findings from individual farms can be shared back with the landowners. No reference to their location or identity will be made publicly available.
Please contact David Whitehead if you would like more information on the trial.
The future of tree mapping
Trees provide many benefits for our environment:
- Carbon sequestration
- Climate regulation
- Reduce soil erosion
- Habitat for native species.
There is a clear need to develop better tools that map, model and manage individual trees in different environments.
We use various Artificial Intelligence (AI) technologies and apply them to any type of imagery we work with that include hi-res aerial photos, remotely sensed images from a satellite, or raster-based images. The programs we use require accurate training data to develop any algorithms.
Through the MBIE Endeavour-funded Advanced Remote Sensing Programme, an algorithm has been developed to detect and segment individual trees crowns from aerial LiDAR data. LiDAR is a technique that can measure the height of objects by sending out laser pulses from a sensor that is usually flown over regions on a plane. The algorithm identifies heights of trees (red area – taller trees; yellow area – shorter vegetation). It is then able to segment the image into tree crown objects. Simply explained, the algorithm starts by identifying the tree top locations, then grows a crown polygon around these points to generate tree crowns.
Our researchers are starting to explore the use of AI methods to analyse high-resolution RGB aerial imagery which is widely available in New Zealand. This would allow for finer detail of crown polygons and a better detection rate.
Wellington Urban Tree Explorer
In collaboration with Scion, University of Canterbury, Victoria University, the Institute of High Performance Computing Singapore and Nanyang Technological University, we created the Wellington Urban Tree Explorer website that has over 1.8 million trees mapped to support the city council with urban planning.
The tree height values were measured by LiDAR aerial surveys in 2019 and 2020. The data was produced by a deep learning model to identify individual tree objects and output geospatial polygons to represent tree crowns. The interactive map is easy to use and has three layers including tree crowns, height and diameter. From there, you can filter by tree height and diameter.
Tree species mapping in pastoral hill-country
This work was a collaboration with the Smarter Targeting of Erosion Control programme to find out what the effect of individual trees is on landslide risk. By adding individual trees and different species to a trained model, we are able to assess the landslide risk and their effect on the surrounding slope stability.
The project compared two landslide susceptibility models at a scale of individual trees on two farms in the Waiarapa that were prone to landslides. In order to run these models, you need information about the vegetation on the land. Previous methods of obtaining this data would be by using the Land Cover Database but this is on a different scale and doesn’t define individual trees, just groupings of vegetation. Our researchers mapped individual trees and classified tree species. This information was then used in the model to measure the effect of tree species on slope stability. Image g shows the landslide risk with no trees (red & yellow areas). Image h shows how adding trees and species updated landslide risk and their effect on the surrounding slope stability.
For more information on these projects, please contact Jan Schindler.