Leaves were randomly selected and collected from healthy Ceriops australis mangrove trees in the Jack Barnes Bicentennial Mangrove Boardwalk site in Cairns (16° 52.976′S, 145° 45.663′E) (n = 96). We inspected the samples to avoid any obvious damage by insects, sunburn or disease and then placed them into coolers with ice on sealed plastic bags and taken to the remote sensing laboratory at James Cook University - Cairns. We selected this species because it is widespread throughout Australia and the size of their leaves (5.5–10 cm x 2.0–3.4 cm) was suitable for this experiment (Duke et al., 2006; Wightman et al., 2006). The leaves were divided into three groups of 32 leaves, where each group represented a different tidal height scenario. Each group of leaves was then divided into stacks of one, three, five and seven leaves and attached to a wooden platform as shown in Fig. 1. Stacks were arranged perpendicularly to each other to reduce the effects of the position of the sun, shadows or other factors when scanning. Furthermore, to reduce the effects of leaf dehydration and solar angle changes, all scenarios (i.e. 5, 15 and 30 cm of water) were scanned at the same time. Because our aim was to determine if water depth affects our ability to estimate FVC from remotely sensed data, we simulated low, transition and high tide by filling the containers with 5, 15 and 30 cm of water respectively. We decided to use 5, 15 and 30 cm of water for our experiment for several reasons: i) we were able to replicate realistic scenarios where mangrove mud is exposed and fully covered with water; ii) if 15 or 30 cm of water have an effect on the spectral reflectance of mangrove leaves, a more thorough experiment should be proposed; however, if there is no effect at these depths, no effect could be expected from other tidal heights; and iii) the containers used limited our ability to simulate the entire tidal range of the Cairns region (i.e. 3.6 m). We also collected mud from the study site and used it to create a 5 cm coating for the bottom of the containers. We did this for two main reasons: i) to better simulate the mangrove environment in our study site (turbid water) and ii) to prevent the bottom of the container from contributing to the spectral readings from the leaves. With the mud in place, water was poured into the container and left for 24 h to ensure any solids would settle, though water remained turbid throughout the experiment, as it is often the case in our study area. Lastly, our design allowed us to isolate the effects of tidal height on FVC estimation from the interference of stems, branches and dead leaves commonly found in mangrove ecosystems.
Hyperspectral imagery acquisition and pre-processing: We obtained hyperspectral imagery over the leaf arrangement using a Headwall NANO Hyperspectral Scanner (Headwall Photonics Inc.), with a spectral resolution of 270 bands between 400–1000 nm (spectral bandwidth of 1.4–2 nm) and a Field of View equal to 15.3 degrees. The spectral range of this pushbroom scanner incorporates the visible (400–650 nm), red edge (650–750 nm) and NIR regions (750–1000 nm), which are often used to discriminate plant species. The scanner was turned on and left to warm up for 3 min before the first scan. Dark signal measurements were taken before and after the scans by completely covering the sensor and recording the signal in each spectral band. The samples were illuminated by direct sunlight and the scanner was positioned two meters above the samples, centered above the leaf arrangement. A 75% white Spectralon® panel was scanned alongside all images, thereby ensuring that any corrections due to changing illumination could be made during the image pre-processing stages. This set-up allowed us to i) reduce to the minimum any bidirectional reflectance distribution effects, ii) attain imagery of the mangrove leaves such that each pixel represented approximately 1 x 1 mm, and iii) isolate the effects of tidal height on mangrove leaves to assess FVC.