Ecology

Biology

Watershed

Computer science

Mangrove ecosystem

Geology

Oceanography

Tree canopy

Aerial survey

Geography

Biomass (ecology)

Environmental science

Artificial intelligence

Unmanned ground vehicle

Remote sensing

Mangrove

Canopy

Machine learning

Ecosystem

Remote Sensing of Environment

Mangrove ecosystems are targeted for many conservation and rehabilitation efforts due to their ability to store large amounts of carbon in their living biomass and soil. Traditional methods to monitor above-ground biomass (AGB) rely on on-ground measurements, which are expensive, labour intensive and cover small spatial scales. Structure from Motion and Multi-View Stereo reconstructions from Unmanned Aerial Vehicles imagery (UAV-SfM) have the potential to increase fieldwork efficiency by providing a greater amount of spatial information in less time. However, there is still a need to assess the ability of UAV-SfM to retrieve structural information of mangrove forests, which could pose challenges in areas of high forest complexity and density. In this study we successfully used UAV-SfM data to estimate height, canopy diameter and AGB of natural and rehabilitated mangrove forests across two regions of the southeastern coast of Australia. We used a variable window filter algorithm to detect trees with an 80% detection rate when considering the top canopy. Individual tree canopy segmentation was performed using a marker-controlled watershed segmentation with two sets of constraining markers: treetops and a minimum height below which a pixel is not considered part of a tree. Direct comparison with on-ground measurements at the regional level showed no significant difference in tree height and AGB medians when only top canopy was considered. Similarly, median canopy diameters were not significantly different in natural areas of both regions, but significant differences were found in rehabilitated areas. UAV-SfM estimates of AGB were on average 15% lower in natural areas and 10% higher in rehabilitated areas when compared to on-ground measurements and followed a strong linear relationship close to the ideal one-to-one relationship. Additionally, we performed a cost-benefit analysis of the two methodologies. UAV-SfM methods can save almost AU$ 50,000 per ha when compared to on-ground measurements and become cost-effective (based on total costs) after just 15 days of surveys. The methods described in this study open the possibility for easily repeatable, low-cost UAV-SfM surveys for local managers by providing a faster, more cost-effective approach for monitoring mangrove forests over larger areas than traditional on-ground surveys while maintaining forest inventory data accuracy in both natural and rehabilitated mangrove forests.

The application of Unmanned Aerial Vehicles (UAVs) to estimate above-ground biomass of mangrove ecosystems