Abstract
Mangroves are tropical and sub-tropical coastal forests which provide numerous ecosystem services related to coastal protection, fisheries, erosion, and carbon uptake, among others. Mangroves are vast and difficult to traverse. Litterfall and Hemispherical Photography (HP) Effective Leaf Area Index (effLAI) are methodologies widely used in forestry as monitoring tools. Although, both methods measure canopy dynamics, they obtain data in different timeframes (average over time (Litterfall) vs instantaneous (effLAI)), measure different characteristics of the canopy (loss vs loss/gain), work under different assumptions and use different calculations (litter fall rate vs index). HP equipment has become easier to use, smaller, more affordable and provides better quality images than in the past. We compared the utility of litterfall and HP LAI in two different mangrove forests in Southwest Florida after hurricane Irma. HP data complemented our litter data, providing additional information about the canopy that litterfall did not provide. The mean daily litterfall rates (g/m²/day) were statistically different between sites (p < 0.01), seasons (p < 0.01), years sampled (p < 0.01), seasons by sites (p < 0.01) and seasons by years (p < 0.01). The mean effLAI demonstrated statistically significant differences between sites but not by the years sampled or the seasons. The dynamic nature of a leaf area changes in the canopy and the relatively quick occurring over smaller timeframes such as seasons were not highlighted by our sampling efforts and periodicity. Litterfall data however takes tissue density into account, thus dense materials (propagules, branches, etc.), which may take longer to produce and replace, will drive or influence seasonal, site and sampling year differences. Leaf area index only considers area and not volume or density, thus losing dense materials (e.g., branches) from the canopy and replacing it with less dense material (e.g., leaves) could result in similar effLAI. Litterfall demonstrated clearer cycles and trends among seasons and years sampled in this study. The more constant sampling (monthly) and slower processes of litter loss may help explain litterfall’s ability to better differentiate between seasons and years. Litterfall did show larger production rates in what would typically be expected to be the less productive system in this study. This pattern appears to be driven by the larger amounts of vegetative material being lost from the canopy as a response to greater damage suffered at the Big Hickory (BH) site. Linear regressions demonstrated that triannual effLAI explained more of the variability in compounded triannual mean daily litterfall for the BH site compared to the Imperial Shores (IS) site (BH (y = 52.50x - 59.11; R²=0.95) and IS (y = 17.60x – 20.53; R²=0.66)). Thus, showing a larger amount of litterfall rates in BH as LAI increased, but also highlights a larger influence of litter fall rates on LAI in BH. In other words, the loss in LAI, which translates into litterfall, is more influential in the linear model explaining this relationship for BH. A possible explanation is the slower recovery process of the canopy and thus the LAI is replaced at a slower pace than for IS. Our results and experiences suggest the HP approach we employed is affordable and provides quality data, but it is not efficient to employ in a mangrove forest. The timing requirements tied to weather and lighting further limits when the LAI data can be collected, coupled with the difficulty in moving through the system makes the methodology used in this study inefficient. Additionally, the combination of LAI and litterfall data provides a unique opportunity, as the datasets allow for the opportunity to better explain the dynamics of a mangrove forest canopy recovering from a hurricane impact. Litterfall, coupled with LAI data, showed the recovery dynamics of two different forest stands subject to different degrees and extents of damage and how these influence the litter and canopy dynamics through time. The site with the largest damage (BH) has undergone a slower recovery process in LAI and litterfall. BH has continued losing heavier/larger amounts of litter, while IS appears to have peaked and reached a more stable litterfall production rate.
Mangroves cover vast areas of tropical and subtropical coastal regions. These ecosystems are important due to the numerous services they provide. Unfortunately, they have undergone a considerable decline, to the degree that they are considered a threatened system due to human activities and climate change. For better management of these systems, technology offers new methods to monitor larger areas within shorter intervals. Remote sensing tools (e.g., airplanes and satellites equipped with cameras) have been used in the past to aid this endeavor, but the cost and lack of temporal detail makes them impractical to adequately answer certain ecological questions. Traditional in situ data collection methods, such as the use of a pole to measure tree heights, may provide great data but are labor intensive, slow and costly. Unmanned Aerial Vehicles (UAVs) and more portable iterations of remote sensing sensors have undergone rapid development during the last several years. Coupled with better computing power and processing software, UAVs have become more accessible, affordable, and easy to use for environmental research and monitoring. Structure from Motion (SfM), a form of photogrammetry, uses images from different perspectives to identify points in common and then renders a 3D representation of an ecosystem. Although forests present a challenge for SfM, several practices and settings can aid in obtaining good data, capable of estimating forest characteristics such as tree height. Tree height can be an important metric for forests, providing information about the tree biodiversity, population dynamics, productivity, biomass, among others. A commercial UAV leveraged with GCPs and a GNSS RTK unit can provide estimates of mangrove tree height, comparable to the traditional pole method. In this study, a total of 22 trees were sampled in a 20 m by 100 m area within a mangrove forest in Estero Bay, FL. The height, tree base geolocation and species were collected. The tree heights obtained in situ with a telescoping pole were compared to the tree height estimates of the same trees estimated from the point cloud (Figure 7.5) rendered using UAV SfM. Based on a linear regression our SfM point cloud tree height explained 89% of the variation in tree heights measured with a pole. Two of the 22 surveyed trees could not be readily identified in the SfM point cloud, possibly due to occlusion, or merging with taller neighboring trees. A matched pairs t-test found no significant differences between the heights measured between the two techniques for A. germinans and R. mangle. This study suggests UAV SfM can help in providing accurate tree height estimates for mangroves, with or without leaves. The ability to obtain accurate tree heights using UAV SfM provides the potential to rapidly estimate other important forest metrics such as stand biomass, population height dynamics, and canopy succession among others over larger spatial scales moving forward.