Chapter 1 Producing thematic coral reef benthic habitat maps from single-beam acoustic backscatter has been hindered by uncertainties in interpreting the acoustic energy parameters E1 (~roughness) and E2 (~hardness), typically limiting such maps to sediment classification schemes. In this study acoustic interpretation was guided by high-resolution LIDAR (Light Detection And Ranging) bathymetry. Each acoustic record, acquired from a BioSonics DT-X echosounder and multiplexed 38 and 418 kHz transducers, was paired with a spatially-coincident value of a LIDAR-derived proxy for topographic complexity (Reef-Volume) and its membership to one of eight LIDAR-delineated benthic habitat classes. The discriminatory capabilities of the 38 and 418 kHz signals were generally similar. Individually, the E1 and E2 parameters of both frequencies differentiated between levels of LIDAR Reef-Volume and most benthic habitat classes, but could not unambiguously delineate benthic habitats. Plotted in E1:E2 Cartesian space, both frequencies formed two main groupings: uncolonized sand habitats and colonized reefal habitats. E1 and E2 were significantly correlated at both frequencies; positively over the sand habitats and negatively over the reefal habitats, where the scattering influence of epibenthic biota strengthened the E1:E2 interdependence. However, sufficient independence existed between E1 and E2 to clearly delineate habitats using the multi-echo E1/E2 Bottom Ratio method. The point-by-point calibration provided by the LIDAR data was essential for resolving the uncertainties surrounding the factors informing the acoustic parameters in a large, survey-scale dataset. The findings of this study indicate that properly interpreted single-beam acoustic data can be used to thematically categorize coral reef benthic habitats. Chapter 2 A large-scale acoustic survey was conducted in Apr-May 2008, with the objective of quantifying the abundance and distribution of seasonal drift macroalgae (DMA) in the Indian River Lagoon. Indian River was surveyed from the Sebastian Inlet to its northernmost extent in the Titusville area. Banana River was surveyed from its convergence with the Indian River northward to the Federal Manatee Zone near Cape Canaveral. The survey vessel was navigated along pre-planned lines running east-west and spaced 200 m apart. The river edges were surveyed to a minimum depth of approximately 1.3 m. Hydroacoustic data were collected with a BioSonics DT-X echosounder and two multi-plexed digital transducers operating at 38 and 418 kHz. The 38 and 418 kHz hydroacoustic data were processed with BioSonics Visual Bottom Typer (VBT) seabed classification software to obtain values of E1’ (time integral of the squared amplitude of the 1st part of the 1st echo waveform), E1 (2nd part of 1st echo), E2 (complete 2nd echo), and FD (fractal dimension characterizing the shape of the 1st echo). Following quality analysis, a training dataset was compiled from 131 hydroacoustic + video samples collected across the extent of the study area. The 38 and 418 kHz E1’, E1, E2, and FD datasets were merged and submitted to a series of three discriminant analyses (DA) to refine the training samples into three pure end-member categories; bare substrate, short SAV (typically Caluerpa prolifera, ~10cm or less), and DMA. The Fisher’s linear discriminant functions from the third and final descriptive DA were used to classify each of the 480,000+ hydroacoustic survey records as either bare, short SAV, or DMA. The classified survey records were then used to calculate the biomass of DMA as the product of average DMA cover for a block of ten records times the wet weight of DMA. The DMA biomass was found to be 69,859 metric tons (wet weight) within the 293.1 km2 study area. The acoustically-predicted mean percent cover of DMA was (i) significantly greater within the navigation channels (18.3%) than outside (12.2%), and (ii) significantly greater in the Indian River (12.9%) than in the Banana River (9.3%). The overall predictive accuracy of total SAV (i.e. short SAV plus DMA) was 78.9% (n=246) at three levels of cover (0-33, 33-66, and 66-100%). The Tau coefficient, a measure of the improvement of the classification scheme over random assignment, was 0.683 ± 0.076 (95% CI), i.e. the rate of misclassifications was 68.3% less than would be expected from random assignment of hydroacoustic records to total SAV cover. The incorporation of multi-plexed digital transducers in conjunction with new post-processing techniques realized the goal of establishing an accurate, efficient, and temporally consistent method for acoustically mapping DMA biomass. Chapter 3 This chapter presents the results of a large-scale hydroacoustic survey conducted in April-May 2008. The objective of this study was to map the distribution and vertical extent of muck in the Indian River Lagoon, utilizing the data collected during a seasonal drift macroalgae survey. Indian River was surveyed from the Sebastian Inlet to its northernmost extent in the Titusville area. Banana River was surveyed from its convergence with the Indian River northward to the Federal Manatee Zone near Cape Canaveral. The survey vessel was navigated along pre-planned lines running east-west and spaced 200 m apart, except for when muck was indicated by the oscilloscope display, at which point a meandering path was adopted to demarcate the horizontal extent of muck. Hydroacoustic data were collected with a BioSonics DT-X echosounder and two multi-plexed digital transducers operating at 38 and 420 kHz. The vertical extent of muck was derived from the 38 kHz hydroacoustic signal, which was processed with Visual Analyzer, a fish-finding software package produced by BioSonics Inc. The software was adapted to integrate echo energy below the water-sediment interface, and a set of post-processing algorithms were developed to translate the sub-bottom echo energy profile into continuous scale estimates of muck thickness. In this manner 500,000+ 38 kHz pings were translated into 88,927 geo-located estimates of muck layer thickness, down to a minimum bottom depth of 1 m. Ground-truthing was conducted in July 2008 at twenty sites within the Indian River. The predictions of muck layer thickness were found to be accurate over the ground-truthed range of 0-3m (r2 = 0.882, SE=0.52m). The vertical distribution of acoustically-predicted muck demonstrated the tendency for muck to accumulate in deeper areas of the lagoon. For the case of Indian River (excluding navigation channels), muck was not detected in depths shallower than 1.4m and rare in the range of 1.4-2.2 m (only 3.6% of records had a predicted muck thickness greater than 0.5 m). The frequency of muck plateaued between 2.2-3.4 m (9.6%) before making a sharp rise to 82% in the range of 4-5 m. As expected, the mean muck layer thickness was significantly greater within the navigation channels (0.56 m) than outside of them (0.08 m). A significant latitudinal trend of muck thickness was detected within the Indian River navigation channels. The mean muck thickness decreased from 1.38 m at its northernmost origins to 0.83 m in the Titusville area before plateauing at approximately 0.4 m for the remainder of segments. Outside of the main ICW channels, 23 individual muck deposits were identified; 22 in the Indian River and 1 in the Banana River. Factors in descending order of co-occurrence were proximity to causeways or jetties, riverbed depressions, and proximity to shore and drainage channels. In conclusion, this study establishes that a single-beam acoustic survey is a cost-effective and accurate alternative for mapping the distribution and vertical extent of muck deposits in the shallow-water environment of the Indian River Lagoon. Moreover, the temporal consistency afforded by a digital transducer allows for direct and meaningful comparisons between successive surveys. Chapter 4 A thematic map of benthic habitat was produced for a coral reef in the Republic of Palau, utilizing hydroacoustic data acquired with a BioSonics DT-X echosounder and a single-beam 418 kHz digital transducer. This paper describes and assesses a supervised classification scheme that used a series of three discriminant analyses (DA) to refine training samples into end-member structural and biological elements, utilizing E1′ (leading edge of 1st echo), E1 (trailing edge of 1st echo), E2 (complete 2nd echo), fractal dimension (1st echo shape), and depth as predictor variables. Hydroacoustic training samples were assigned to one of six predefined groups based on the plurality of benthic elements (sand, sparse SAV, rubble, pavement, rugose hardbottom, branching coral), visually estimated from spatially co-located ground-truthing videos. Records that classified incorrectly or failed to exceed a minimum probability of group membership were removed from the training dataset until only ‘pure’ end-member records remained. This refinement of ‘mixed’ training samples circumvented the dilemma typically imposed by the benthic heterogeneity of coral reefs, i.e. to either train the acoustic ground discrimination system (AGDS) on homogeneous benthos and leave the heterogeneous benthos un-classified, or attempt to capture the many ‘mixed’ classes and overwhelm the discriminatory capability of the AGDS. This was made possible by a conjunction of narrow beamwidth (6.4o) and shallow depth (1.2 to 17.5 m), which produced a sonar footprint small enough to resolve most of the microscale features used to define benthic groups. Survey data classified from the 3rd-Pass training DA were found to (i) conform to visually-apparent contours of satellite imagery, (ii) agree with the structural and biological delineations of a benthic habitat map created from visual interpretation of 2004 IKONOS imagery, and (iii) yield values of benthic cover that agreed closely with independent, contemporaneous video transects. The methodology was proven on
| Date of Award | Apr 1 2010 |
|---|
| Original language | English |
|---|
| Supervisor | Bernhard Riegl (Supervisor), Richard E. Dodge (Advisor), Samuel J. Purkis (Advisor) & John Brock (Advisor) |
|---|