Author
Paul G Peterson
petersonp@landcareresearch.co.nz
Bioeconomy Science Institute – Manaaki Whenua – Landcare Research Group
Palmerston North, New Zealand
Coauthors
Andrew McMillan, Bioeconomy Science Institute – Manaaki Whenua – Landcare Research Group, Palmerston North, New Zealand
Ben Jolly, Bioeconomy Science Institute – Manaaki Whenua – Landcare Research Group, Palmerston North, New Zealand
Harley Betts, Bioeconomy Science Institute – Manaaki Whenua – Landcare Research Group, Palmerston North, New Zealand
Jan Schindler, Bioeconomy Science Institute – Manaaki Whenua – Landcare Research Group, Wellington, New Zealand
Elizabeth Munro, National Environmental Service, Avarua, Cook Islands
Jessie Nicholson, National Environmental Service, Avarua, Cook Islands
Abstract
We present a case study to map invasive weeds and assess impacts of biocontrol agents being introduced into Pacific Island countries and territories (PICTs). In Rarotonga, land managers did not have the tools to map African tulip tree (ATT) (Spathodea campanulata) distribution and evaluate impacts of biocontrol agents that have recently been released. ATT is considered one of the world’s 100 worst invasive species and is threatening Rarotonga’s natural biodiversity, and ecosystem and community resilience. Climate change is expected to increase the invasiveness of ATT. A multi-scale approach using drone, aeroplane and satellite imagery, coupled with AI (artificial intelligence), was trialled to map ATT and another potential biocontrol target, Tamaligi or Albizia (Falcataria moluccana). This case study also aimed to develop capability in the PICTs to monitor and assess biocontrol programmes locally. An AI mapping algorithm allowed us to estimate the area of Rarotonga infested with ATT as 112 ha, representing 1.6% of the island. We could also assess the density of galls in ATT canopies of the biocontrol agent with high-resolution drone imagery. A combination of aeroplane and drone imagery has provided Rarotongan land managers with a cost-effective weed mapping and biocontrol assessment solution that can be used in future and adapted to weed biocontrol programmes in other PICTs.
Keywords
multi-scale remote sensing
invasive weed management
weed biocontrol assessment
African Tulip Tree
deep learning algorithm
Highlights
A multi-scale remote sensing method was developed to map invasive weeds on Pacific Islands.
A remote sensing method was developed to assess weed biocontrol establishment and impacts.
Capability was developed in the Pacific to assess weed biocontrol programmes with remote sensing.