Biodiversity Restoration using Drones, Mapping and Spraying!

We are excited to share that we have embarked on a sophisticated journey of biodiversity restoration using drone technology and GIS mapping!  Drone mapping is assisting us to more efficiently monitor and target Invasive Alien Plants (IAP) on the reserve using Pilot area 1 as a test site.

On the 11th and 12th April 2023 the GCF deployed two drones to analyse a predetermined target area for IAP analysis, control and testing. An initial scout flight was conducted with DJI Phantom 4 Pro and DJI Phantom 4 Multispectral drones to establish exactly where alien vegetation clusters or individuals are as well as vegetation health within Pilot area 1. With the data captured on this flight we created baseline maps and models of the vegetation and terrain. This will form historical data that will be used in the future for comparison, analysis and monitoring of spray sites.

The mapping technology is so advanced that it is able to extract information such as canopy density, healthy and unhealthy vegetation, the height and size of adult trees, density and mass of the targeted areas and it even goes as far as being able to identify live and dead plant matter.

The GIS mapping technology makes use of Machine-learning algorithms as a powerful tool to classify objects within the drone maps. One of these custom machine-learning models have allowed us to get a good approximation of what is happening in the target area. The model “is trained” to identify specified objects, for example an IAP species versus natural vegetation versus soil. In this way a high level of classification accuracy is achieved because each pixel within the map represents 6.7cm on the ground. This is illustrated for two detections areas of the map, as examples, in the figures below.

During the first flight in April, we mapped a specific flight path at a specific altitude over the target areas and sprayed the API’s. In June we repeated the exact same flight path with the drones at the same altitude to document a comparison between the data collected across the two flights over the same target areas.

The findings measured a success which was hampered by certain factors explained in the discission and conclusion listed below the table summaries.

Table 1: Summary of vegetation cover metrics before and after drone spraying for site 1

Metric	Before Spray	After Spray
Total Site Area (ha)	1.03	–
Total Spray Area (ha)	0.31	–
Total Vegetation Cover Area (ha)	0.20	0.11
Vegetation Density (%)	63.9	34.1
Kill (%)	–	46.7%

Table 2: Summary of vegetation cover metrics before and after drone spraying for site 2

Metric	Before Spray	After Spray
Total Site Area (ha)	0.606	–
Total Spray Area (ha)	0.34	–
Total Vegetation Cover Area (ha)	0.069	0.026
Vegetation Density (%)	20.2	7.6
Kill (%)	–	62.3%

Here is a direct extract from the June report on comparing data after the initial two flights.

“Discussion

Based on the analysis of the vegetation cover before and after the drone spraying and, to some extent, ground-truthing, the following observations can be made: 

• Comparison of the total vegetation area before and after the spray provides an understanding of the extent of effective vegetation die-off. 
• Overall, the results were not entirely satisfactory, particularly with regard to the mature plants. 
• There was some success on the TigerCat Regrowth site, but the results were less than what the drones are capable of achieving. There are a number of parameters that may have led to the ineffectiveness of this spray operation; nozzle (and therefore droplet) size and spray route width. These were part of the test work on these sites. 
• It was noted from ground-truthing that some fresh regrowth was present, but the extent of this cannot be readily quantified against the originally targeted areas. 
• In the case of the mature plants, the chemical concentration was insufficient to effectively kill off the trees. Other factors such as heavy rains post-spraying together with location of the trees within a water course, both of which increase growth vigour, may have also played a roll in the ineffectiveness of this operation. 
• It is also interesting to note that where streaking occurred on all sites, it was directly underneath the drone which indicates that altitude corrections may need to be made. 

Conclusion 

The comparison of vegetation cover before and after drone spraying provides valuable insights into the effectiveness of the spraying operation. By analysing metrics such as total vegetation area, average VARI, and vegetation density, we can assess the impact of the spray on the site. The visual comparison through aerial imagery also allows for a visual understanding of the changes in vegetation cover. These findings can assist in monitoring and optimizing future drone spraying operations for vegetation management.”

We will be flying the drones again in July to spray the same areas again with adjustments based on the data collected previously. We look forward to a follow up audit on this exciting biodiversity restoration project.