Drone Mapping UK Cornwall On The Way To Africa

A blog by Daniel Ronan, a Frontier Tech Hub Implementing Partner.

Explore key learnings from our pilot, Project Sapling.

Sierra Leone has suffered decades of deforestation. In 2020 alone, 155kha of tree cover (79.9Mt of CO₂) was lost. This has had a direct negative impact on the environment, biodiversity, ecosystems, local communities and exacerbated the impact of climate change on the country.

A UKAid Frontier Technologies Livestreaming (FTL) project was subsequently established in late 2022 to investigate and trial specific innovations to support reforestation in Sierra Leone, with the view to developing a methodology which, where applicable, could be scaled across Africa. This innovation centred around a reforestation process with local communities planting and becoming custodians of new forests, with robust data collection and remote monitoring of the trees for verification and ability to raise carbon funding for sustainable reinvestment. One of the essential components of this innovation was the ability to map rural areas quickly and at an image quality sharper and with more detail than achievable by commercial satellite. Bring in the drones!!

Drone specialist UAVaid is the technical lead for this FTL project, partnering with Crown Agents and Tacugama Chimpanzee Sanctuary. Press release available HERE.

The use of drones for aerial mapping has significantly moved towards the mainstream in the last few years, with their application becoming increasingly common worldwide for land surveys and agriculture. This has been brought about by technological advances in sensor (i.e. camera) miniaturisation and micro electronics, as well as evolving regulatory frameworks. Through their ability to fly over land, free from constraints of the terrain, increasingly capable drones with payloads of high definition cameras are creating the prospect of rapid aerial map and survey generation of very high image standard.

Taking a structured approach to how drone mapping would be applied to this particular FTL reforestation innovation, a micro project was run in England to test the ability of drone tech to map at high detail and to establish optimal operating parameters. This involved the mapping of a section of rural land by drone, and testing the various parameters for best results.

Selecting the Test Site

Cornwall, a region in southwest England, was selected for its proximity to the drone team and rural landscape. Using local knowledge and contacts, an area of private farmland was identified as of similar size and potential land features to the 6 hectare project test sites in Sierra Leone. Exclusive use permission was sought, and secured, from the land owners. This approach of using private farmland gave key risk-reduction benefits (1) safety and security of people and equipment on the site (2) no distractions or interference from/with people not associated with the test (3) range of (agri) land features. The site was visited in advance to ensure suitability and confirm exact site orientation to take advantage of the land features: water (large pond), forested area, line of new trees, open grassland, hedges and ploughed land, flat and sloped terrain.

Regulatory compliance was confirmed, equipment and logistics booked.

The Day of Testing

As not unusual for early February in Cornwall, the day started off frosty.

[Unanticipated bonus learning point: frost changes the reflectance of the land, impacting on multi-spectral imaging.]

Although a public RTK (think of it as technology to ‘superboost GPS positioning accuracy’ to centimetre level) service was available in the area, to replicate the ‘don’t rely on anything you are not bringing with you’ principle when deploying in remote and challenging contexts, a portable RTK base station was setup at a high point overlooking the test site.

What followed throughout the day was a series of drone flights, mapping the area with different combinations of settings and parameters. Tested were different altitudes, RGB (optical camera) and RGB+MS (multi spectral) imaging, with / without RTK, with / without Ground Control Points (GCP’s) and others. An unexpected variable was the morning ground frost, which fortunately melted away as the day progressed into glorious sunshine. An attempt was also made to capture video footage of the mapping drone flying, by another drone (spoiler alert: very difficult to get good air-to-air footage manually due to need to maintain safe distance between aircraft).

For flight operations, the drone routes and flight parameters were pre-programmed and uploaded onto the drone and ‘executed’, with line of sight maintained throughout the flight.

Image Processing

Following the data collection, the imagery was orthomosiac and NDVI processed using industry standard software, Pix4D. This provided image files in formats suitable for display on GiS systems and for the data fusion component of this FTL pilot.

Results

The drone-generated imagery was SIGNIFICANTLY sharper and clearer than the benchmark satellite derived maps. The tree canopy, water, grass and even tractor tire marks were all clear. The drone image acquisition was able to handle the changes in light, as the day progressed, well. Although the aerial imagery was able to clearly distinguish features, the small newly planted trees were only indirectly observable by their shadow, when viewed directly from above.

*Orthomosiac map of the UK test site in Cornwall*

Top 5 Learning Points

Drone mapping imagery: The programme demonstrated and confirmed that drones are capable of rapidly mapping areas of landscape and ground features in higher clarity and resolution than commercial satellites. Stunning image clarity was achieved which showed off the different ground features. With RGB-only data acquisition, a GSD of almost 2cm was achieved, compared to 30cm native data from commercial satellites.

Imagery Limitations: the observable area of a baby tree (sapling) when viewed directly from above is very small. Aerial imagery, however derived, will be of limited value for individual plant inspection until the plant reaches a size large enough that, when viewed from above, it is able to be distinguished from its surroundings. This exact size will require futher study.

Drone mapping variable terrain: utilising the ‘terrain following’ capabilities of the tested drone system provided additional collision safety and successfully demonstrated the capability of mapping landscape elevations (hillsides) while maintaining constant GSD. However, there were constraints on the range of height Above Ground Level (AGL), limiting the mapping GSD flexibility.

Data: Not unexpectedly, the acquisition of combined RGB and MS imagery increased the data storage and processing requirements by over 400% when compared to RGB alone. Further, orthomosiac and NDVI image processing increased the data storage requirements significantly, in some cases by over 300%.

Bring what you need: to reduce risk, do not rely on anything outside of your direct control when operating in new or potentially complex settings.

So what?

With the techniques and technology tested, the next steps will include preparing the data fusion component of the project and baseline mapping the test site(s) in Sierra Leone. This mapping of the test site(s) will apply the learning from this sprint and the team will be making sure of local community engagement for drone flights and taking all equipment and parts with them to reduce reliance on externalities and therefore reduce risk.

…and a little extra

So successful was the mapping at the test site, the team secured permission to map the local St. Austell golf course linked here.

If you’d like to dig in further…

📚 Learn about the pilot’s efforts in determining data literacy —“Determining Digital Literacy for Reforestation in the Mansonia Community, Sierra Leone”

📚 Explore learnings from the pilot’s field trial — “Investing In New Forests ? — How to prove trees have actually been planted.”

📚 Learn about the pilot’s recent planting event — “Transforming Sierra Leone’s Landscape: A Community Tree Planting Event Powered by Mobile App Geotagging”