Presentation Link
Transcription
Presentation Link
Evaluation of a new hydroacoustic substrate classification system for oyster reef mapping in Galveston Bay, Texas 2003-2014 Introduction Importance of Oyster Reef Resources • • • • • Vital in maintaining the Galveston Bay ecosystem. Provide habitat for bottom-dwelling fish and invertebrates which attract larger game fish. Stabilize the bay bottom and break wave energy, preventing shoreline erosion. Oysters act as a natural filtration system; they filter silt and contaminants from the water. Support an important commercial fishery. Photo – Jennifer Reynolds, The Daily News 2003-2014 Introduction Researchers at Texas Parks and Wildlife Department (TPWD) have worked for years to quantify and map oyster reefs. Utilize a variety of technology including side-scan sonar & scientific echosounder and various data processing methods. Photo – TPWD Vital in guiding restoration and management efforts. Important in establishing a baseline for future surveys and developing an index for monitoring temporal changes. 2003-2014 Introduction TPWD evaluated a new hydroacoustic processing software (BioSonics Visual Habitat) for the rapid assessment and mapping of various substrate types (sand, cobble, mud, etc.) 2003-2014 Methods – Data Collection Echosounder Configuration • 120 kHz, 7.6 0 single beam transducer • Ping rate: 5 pps • Pulse duration: 0.4 ms • Calibrated, scientifically defensible • Quantitative assessment • All data stored in DT4 file format 2003-2014 Methods – Study Area Hanna’s Reef, Galveston Bay, TX 5.7 sq km rectangular grid of 10 transects spaced 200 m apart. 2003-2014 Methods – Study Area Hanna’s Reef, Galveston Bay, TX 5.7 sq km rectangular grid of 10 transects spaced 200 m apart. 2003-2014 Methods – Data Processing All data were derived from the same DT4 files and processed to generate 5 substrate classes. Hydroacoustic data were processed using three different software programs/methods: 1. (new program) 2. QTC Echo Impact 3. TPWD proprietary SAS method Results plotted over side-scan sonar imagery and survey data in ArcGIS (V. 10.2) 2003-2014 Different software methods resulted in similar substrate classes but assigned to different class/cluster numbers. For ease of interpretation, similar substrates were given matching colors among methods during mapping in ArcGIS. Method: Visual Habitat (Color Code / Substrate) QTC Impact (Color Code/Substrate) TPWD-SAS (Color Code/Substrate) 1 Orange/Mud1 Light Green/Shell Dark Green/Mud Shell 2 Red/Mud 2 Dark Green/Mud Shell Light Green/Shell 3 Light Green/Shell Orange/Mud1 Yellow/Reef 4 Dark Green/Mud Shell Red/Mud 2 Orange/Mud1 5 Yellow/Reef Yellow/Reef Red/Mud 2 Acoustic Habitat Class/Cluster No. 2003-2014 Methods – Data Processing 1. - principal components analysis 2. QTC Echo Impact – PCA, 166 variables 3. TPWD proprietary statistical analysis software (SAS) method – combines output from VBT2 (Beta) with factor analysis (PROC Factor) and K means cluster analysis (PROC Fastclus) in SAS. 2003-2014 Methods – Data Processing • Evolution of BioSonics VBT/VBT2 software based on work of M. Moszynski of Gadansk Univ. of Tech. and J. Tegowski of the Univ. of Gadansk. • Unsupervised classification approach • Generates 14 “features” (variables) • Principal Components Analysis (PCA) of features • Results output to ArcGIS for mapping over side scan imagery and historical survey data, or export direct to KML format for viewing sharing via Google Earth 2003-2014 Methods – Data Processing • PCA result displayed as scatterplot • Users choose the number of clusters (classes) to form 2003-2014 Methods – Data Processing QTC Impact: • Unsupervised classification approach • Generates 166 “features” (variables) • PCA of Features • Required adjustment of “BioSonics Base Gain” value in program’s configuration file (impact.cfg) prior to analysis (which was very time consuming). • No longer commercially available! • Results output to ArcGIS for mapping over side scan imagery and historical survey data. 2003-2014 Methods – Data Processing TPWD Proprietary Statistical Analysis Software (SAS) method: • Supervised Classification Approach • Used 51 Variables generated by BioSonics VBT2 (Beta). • In SAS, factor analysis (PROC Factor) used to assess the number of classes (e.g., clusters) needed to account for ≥75% of variance. • K Means Cluster Analysis (PROC Fastclus) used to bin pings into clusters. • Results output to ArcGIS for mapping over side scan imagery and historical survey data. 2003-2014 Side Scan Sonar A Portion of Hanna’s Reef and Vicinity Side Scan Mosaic • • Benthos C3D bathymetric side scan sonar Oyster habitat (and other hard substrates) indicated by darker areas Oyster Reef 2003-2014 Prior Survey Data Powell et al., (1993) Acoustic Survey • Dual frequency Datasonics DFT-210 echosounder • Towed transducers operating at 27 and 200 kHz • Paper chart recorder • Acoustic technique described by Simmons et al. (1992) • Three bottom types were defined/identified via groundtruthing • ARC/Info used to map reef areas 2003-2014 Methods – Defining Substrate Types Similar results from all 3 methods: Legend: Two classes clearly associated with oyster habitat (Reef and Shell). Reef habitats are best left alone. Shell habitats good candidates for restoration. Class 1. Shell 2. Reef/shell Remaining 3 classes are sedimentary. 3. Mud - dredge dug One sedimentary class (Mud Shell) appears to be a mud/shell mix or transitional class. This is the least abundant of the 5 classes and thus is the most difficult to sample precisely. 2003-2014 4. Mud 5. Reef Merging Data Layers A Portion of Hanna’s Reef and Vicinity Hydroacoustic Data Processed using BioSonics Visual Habitat Legend: Visual Habitat 1. Shell 2. Reef/shell 3. Mud dredge dug 4. Mud 5. Reef Powell et Al 1993 Mud Reef Shell 2003-2014 Merging Data Layers A Portion of Hanna’s Reef and Vicinity Hydroacoustic Data Processed using QTC Impact Legend: QTC Impact 1. Shell 2. Reef/shell 3. Mud dredge dug 4. Mud 5. Reef Powell et Al 1993 Mud Reef Shell 2003-2014 Merging Data Layers A Portion of Hanna’s Reef and Vicinity Hydroacoustic Data Processed using TPWD Proprietary Method Legend: TPWD 1. Shell 2. Reef/shell 3. Mud dredge dug 4. Mud 5. Reef Powell et Al 1993 Mud Reef Shell 2003-2014 Merging Data Layers Visual Habitat QTC impact TPWD 2003-2014 Results Results with Visual Habitat very favorable compared to both TPWD’s method and QTC Impact (the "gold standard" IMHO) for classifying reef and shell habitats. In non-reef areas, QTC Impact results tended to be more homogenous. Visual Habitat and TPWD’s method gave more heterogeneous results in nonreef areas. Is this a more accurate representation of small scale habitat patchiness? More precise ground truthing needed. If identifying oyster habitat is the only concern, fewer classes might be better. Visual Habitat performed quite well using a 4-class analysis, especially over the shell habitats. While all 3 methods performed well at classifying oyster habitat, BioSonics Visual Habitat was the least time consuming method and significantly reduced level of effort as compared to QTC and TPWD SAS method. 2003-2014 Results - 4 Classes w/Visual Habitat A Portion of Hanna’s Reef and Vicinity Hydroacoustic Data Processed using BioSonics Visual Habitat Legend: Visual Habitat 1. Shell 2. Mud dredge dug Visual Habitat performed quite well using a 4-class analysis, especially over the shell habitats. 3. Mud 4. Reef Powell et Al 1993 Mud Reef Shell 2003-2014 Results In Galveston Bay habitat types are patchily distributed at small spatial scales, especially in non-reef areas. Visual Habitat showed a high agreement with prior results and resulted in a significant reduction in effort as compared to processing with other software. In addition to the significant reduction in effort, Visual Habitat provided data visualization tools in the form of a transect map view displaying user-defined color gradients for each data layer. Visual Habitat also provides the ability to export results in KML file format for rapid interpretation via Google Earth as well as the ability to export as csv for easy import into other applications (e.g., AutoCAD and ArcGIS). 2003-2014 Evaluation of a new hydroacoustic substrate classification system for oyster reef mapping in Galveston Bay, Texas Thank you for your time. Questions: Bill Rodney bill.rodney@tpwd.texas.gov 281-534-0127 Eric Munday eric@subsea2020.com 855-SUBSEA1 2003-2014