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

Highlights

• UAV imagery can be used to assess AGB of natural and rehabilitated mangrove forests.

• Mangrove tree height and canopy diameter accurately estimated

• UAVs are more cost-effective than on-ground surveys (AU$ 50,000 savings per ha).

Abstract

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.

Introduction

Mangrove ecosystems provide unparalleled economically and ecologically critical services to coastal areas (Alongi, 2008), including: 1) provisioning (Hemminga and Duarte, 2000); 2) coastal protection (Badola and Hussain, 2005; Das and Vincent, 2009; Koch et al., 2009); 3) recreational and aesthetic uses (Bergstrom et al., 1990); and 4) soil formation and carbon sequestration (Atwood et al., 2017; Donato et al., 2011; Mcleod et al., 2011). Despite their importance they are one of the most threatened and vulnerable ecosystems worldwide (Hamilton and Casey, 2016; Thomas et al., 2017).

As they occur across the land-sea interface, mangrove forests are subject to both terrestrial and marine pressures. A large percentage of mangrove ecosystems have been altered, destroyed or degraded worldwide as a result of anthropogenic impacts, particularly near populated areas (Clark and Johnston, 2016; Richards and Friess, 2016). Consequently, mangrove ecosystems are the focus for many conservation and rehabilitation efforts (Duncan et al., 2016; Nam et al., 2016; Ren et al., 2010), which require regular monitoring of the forest structure characteristics for better management strategies. This is particularly relevant for any projects seeking to gain carbon credits for mangrove rehabilitation under the impacts of climate change related issues (sea level rise and changes in precipitation and temperature; Ward et al., 2016).

Mangrove forest monitoring and management has traditionally relied on regular on-ground surveys for collecting forest inventory data (Nam et al., 2016; Ren et al., 2010). This data collection is important because it provides a better understanding of species composition and carbon biomass of the survey areas. However, these surveys usually cover relatively small spatial scales (<0.5 ha), and can often be expensive, labour intensive and time-consuming due to the tides, mud, and general difficulty in accessing these remote coastal ecosystems (Lee and Lunetta, 1995).

Remote Sensing data, on the other hand, provides a fast, cost-effective, and efficient method to estimate the biological, biophysical and biochemical factors that translate into some of the services provided by mangrove ecosystems (Giri, 2016; Pham et al., 2019). The type of remote sensing platform (ground-based, airborne or satellite) and sensor (photographic, LiDAR or radar) used depend on the scale and the goal of the research (Wang et al., 2019). Remotely sensed satellite archive data offer broad-scale (nation-wide) and long-term (up to 40 years) monitoring for detection of change over time (Giri, 2016; Kuenzer et al., 2011; Pham et al., 2019). However, it might not necessarily provide the resolution and precision required for local accounting (Ruwaimana et al., 2018). To increase spatial resolution while still covering large areas, airborne LiDAR systems and ground-based platforms like terrestrial laser scanners have been used in the past for retrieval of forest inventory data of coastal wetlands (Feliciano et al., 2014; Wannasiri et al., 2013). However, they are expensive, not widely available, and data management and processing is often difficult and requires specialized software (Wallace et al., 2016; Yin and Wang, 2019).

Unmanned aerial vehicles (UAVs) paired with Structure from Motion and Multi-View Stereo photogrammetric procedures (from now on UAV–SfM) have the potential to: 1) increase fieldwork efficiency by collecting broader spatial information in less time than traditional ground-based surveys (Dandois and Ellis, 2013; Messinger et al., 2016; Murfitt et al., 2017), 2) increase spatial resolution obtained from satellite data while still covering large areas (Ruwaimana et al., 2018), and 3) provide a more cost-effective approach than other airborne systems like LiDAR or airborne photogrammetry while maintaining accuracy and resolution (Dustin, 2015; Sankey et al., 2017).

In terrestrial forests, the implementation of high-resolution imagery collected from low-cost UAVs is becoming an increasingly valuable tool for mapping above-ground carbon stock (Dandois and Ellis, 2010; Samiappan et al., 2016; Zahawi et al., 2015). Common forest inventory data derived from UAV-SfM includes tree species, height, canopy diameter and above-ground biomass (AGB), which can complement, and eventually replace, traditional forest inventory techniques (Dandois and Ellis, 2013; Messinger et al., 2016; Panagiotidis et al., 2017; Zarco-Tejada et al., 2014). Overall, there is a general consensus that UAVs are the most cost-effective solution for sites with an extent between 10 and 20 ha when compared to aircraft and satellite data (Dustin, 2015; Manfreda et al., 2018; Matese et al., 2015). However, no cost-benefit analysis has compared the benefits of using UAVs against on-ground measurements in mangrove ecosystems to date.

Despite the importance of coastal wetlands for carbon accumulation and other ecosystem services, there are few studies evaluating the use of UAV-SfM approaches for assessing the biophysical and biochemical properties of mangrove forests (Li et al., 2016; Navarro et al., 2019; Otero et al., 2018; Tian et al., 2017; Yaney-Keller et al., 2019). Out of these, only two have focused on retrieving mangrove forest inventory data for estimating mangrove AGB (Navarro et al., 2019; Otero et al., 2018). These studies demonstrated that UAV-SfM derived data has the potential for estimating AGB from mangrove plantations, but were unable to predict AGB from natural forests with densely packed mangrove trees. Otero et al. (2018) retrieved information on height from UAV-SfM data at a plantation and natural site with mixed results (only the height medians from the plantation site were significantly similar). Additionally, validation was achieved through visual interpretation of the orthomosaic, with no direct plot by plot comparison between UAV-SfM and field data. On the other hand, Navarro et al. (2019) managed to perform a tree by tree comparison of UAV-SfM data vs field data for two measurements: height and canopy diameter. However, this study is based on trees from a mangrove plantation project <8 years old with clearly defined boundaries between trees. Furthermore, only canopy diameter was not significantly different from on-ground measurements. Consequently, the challenge to effectively estimate mangrove AGB from UAV-SfM derived forest inventory data still remains.

In this study, we propose an approach that combines tree detection and canopy segmentation algorithms applied to UAV-SfM data for quantifying AGB of mangrove forests within natural and rehabilitated (25+ years) areas of the southeastern coast of Australia, and compare it to on-ground measurements at the plot level. Moreover, we perform the first cost-benefit analysis to date of the two different methods (field based measurements vs UAV-SfM data) in mangrove ecosystems to estimate above-ground biomass. The methods described in this study provide local managers with a cost-effective approach for regular monitoring of mangrove forests over larger areas than traditional on-ground surveys while maintaining forest inventory data accuracy.