Data Package Metadata   View Summary

Study Sites.— Data for this study was collected at three reservoirs in southwest Ohio, specifically selected to provide imagery of a range of representative habitat features of interest. Acton Lake is a small (≈240 hectare) highly eutrophic shallow reservoir with approximately 14.2 km of shoreline and a maximum depth of approximately 9.1 m. The northernmost portion of Acton Lake is generally non-navigable by boat due to heavy sedimentation from the two primary tributaries to the lake and has a relatively simple shoreline with an overflow dam and rip rap wall at its southern end. Rocky Fork Lake contains approximately 50 km of shoreline and is approximately 791 hectares in size. Rocky Fork Lake has a maximum depth of approximately 12.2 m. Rocky Fork Lake was selected due to the relatively high amount of aquatic vegetation present throughout the lake. Caesar Creek Lake was the largest reservoir in this study (≈1136 hectares) with 58.4 km of shoreline and a maximum depth of 35 m. Caesar Creek Lake is composed of two separate basins and is notable for its diversity of habitat types, including a significant presence of drowned standing timber. All three reservoirs typically have low visibility (≤ 1 m Secchi depth) throughout the year. These reservoirs typically stratify during summer at approximately 5-8 m and have hypoxic conditions in the hypolimnion during the stratified period.

All three study reservoirs are monitored via standardized electrofishing surveys by the Ohio Department of Natural Resources – Division of Wildlife, as part of its Inland Management System (IMS). In brief, the survey protocol divides the navigable portion of these reservoirs’ littoral zones into individual sampling transects (“IMS sampling sites”) that parallel the shoreline and are approximately 500 m in length, with the number of sites dictated by the shoreline of the lake (Acton Lake: 14; Rocky Fork Lake: 49; Caesar Creek Lake: 92). Habitat data in this study was quantified at the scale of an individual IMS sampling site.

Equipment.— The survey-grade sonar system that was used for this project was an EdgeTech 6205 dual frequency (550 kHz and 1600 kHz) side-scan sonar and bathymetry system (EdgeTech, West Wareham, Massachusetts; hereafter, “EdgeTech”). This sonar system simultaneously collects side-scan sonar imagery and multi-beam, point-cloud swath bathymetry data. The physical components of this sonar system include the submersible sonar transducer (759 x 208 x 244 mm, L x W x H), the 6205-R rack mount interface box, and a 20 m deck cable used to connect the sonar head to the interface box. The sonar transducer was mounted to the starboard side of the survey boat using a rigid aluminum mast (Over-The-Side Mini Mount; Marine Survey Fabrication, Coarsegold, CA) that allowed the transducer to be tilted fore and aft, as well as raised from and lowered into the water. The transducer was submerged approximately 0.3 m below the surface of the water. The EdgeTech system was paired with the Applanix POS MV SurfMaster system (Applanix, Richmond Hill, Ontario, Canada) which obtains precise position, heading, attitude, heave, and velocity data. The Applanix POS MV SurfMaster system consists of two Global Navigation Satellite System (GNSS) antennas, an inertial measurement unit (IMU), and PCS interface box. The two GNSS antennas and the IMU were mounted in fixed positions on the boat and were connected to the PCS interface box. An onboard dual-monitor computer setup was used to run the sonar acquisition software programs (Hypack 2020, Xylem, Middletown, CT and EdgeTech DISCOVER BATHYMETRIC acquisition software). Exact positions and offsets relative to the boat’s center of gravity were recorded for the sonar head, IMU, and GNSS antennas. During data collection, all equipment was powered using a 2000W onboard gas-powered generator. Approximately 30 minutes of equipment setup and startup time was required prior to the start of each day of EdgeTech data collection.

The recreational side-scan sonar system used in this study was a Lowrance HDS 12 Carbon fish finder/chartplotter, Active Imaging 3-in-1 sonar transducer (257 x 67 x 35 mm, L x W x H; 800 khz and 455 khz), and Point-1 GPS antenna and heading sensor (Navico, Tulsa, Oklahoma; hereafter, “Lowrance”). The fish finder unit powers the sonar transducer, records data on a removable memory chip, and serves as a means of visualizing the side-scan imagery while on the boat. The Lowrance transducer was mounted on a moveable pole mounted in front of the bow of the survey boat, and submerged approximately 0.2 m below the surface of the water. The transducer was mounted off the bow rather than the stern (typical recreational boat install location) to minimize the impact of turbulence and bubbles from the outboard motor which would reduce image quality (Kaeser and Litts 2008). Less than 5 minutes of equipment setup and startup time was required prior to the start of each day of Lowrance data collection.

Sonar Data Collection.— Sonar data collection was performed in the summer of 2021 and in the spring and summer of 2022. Side-scan sonar imagery of the littoral zone was collected with both types of sonar systems by driving the boat approximately 30-40 m off shore and parallel to the shoreline at speeds of approximately 4.5-6.5 km/hr (Richter et al. 2016). All accessible areas of shoreline with IMS sampling sites (i.e., areas with sufficient depth for boat access and no navigational restrictions) were scanned. Although the EdgeTech is capable of recording at both frequencies, due to swath-width limitations at the 1600 kHz frequency, we only used the 550 kHz imagery. Lowrance side-scan imagery was collected at a frequency of 455 kHz. The total swath width for both systems was set at 100 m (i.e., 50 m on either side of the boat). This range setting was a compromise between achieving the recommended depth-to-range ratio of 0.1 – 0.2 for high quality sonar imagery (Kaeser and Litts 2010) while minimizing the potential for crashing the EdgeTech transducer into submerged obstacles (actual depth-to-range; Acton: 0.06-0.07, Caesar Creek: 0.1, Rocky Fork: 0.6 – 0.7). During data collection, real-time kinematic (RTK) GPS corrections were applied to the EdgeTech data using the Ohio Real Time Network. This resulted in positional accuracy of < 3 cm for the EdgeTech side-scan imagery (as reported in real time by the data logging software). Data collected with the Lowrance system had positional accuracy ≈3 m when paired with the Point-1 GPS antenna (Halmai et al. 2020).

EdgeTech Sonar Data Processing.— Side-scan sonar files collected with the EdgeTech 6205 system were processed and mosaicked using the SonarWiz 7 software program (Chesapeake Technology Inc, Los Altos, California). The bottom tracking and slant range correction functions were used to remove the water column and georectify the imagery and empirical gain normalization (EGN) and the nadir filter were applied to visually optimize the imagery. The side-scan mosaic of the entire lake was exported as tiled GeoTIFF images.

Lowrance Side-Scan Data Processing.— Side-scan sonar files collected with the Lowrance were processed and mosaicked with the ReefMaster 2.0 software program (ReefMaster Software Ltd, West Sussex, UK). After importing the files to the program, the water column was subsequently removed from the imagery using the “water column offset” feature. The files were then added to a mosaic where the sharpness, gain, brightness, and contrast of the image were adjusted for optimum visibility. Lastly, the mosaics were exported as MBTiles files which were then imported into QGIS for habitat mapping and quantifying.

Manual Habitat Quantification.— Sonar imagery was imported into QGIS (3.20.1, QGIS.org) for habitat analyses. A single researcher (T. Fletcher) delineated the patches of littoral zone woody debris, standing timber, aquatic vegetation, and benthic substrate visible in each set of sonar imagery (Figure 1). We defined the quantification area of the littoral zone as the region between the lake shoreline (using existing GIS layers for lake shorelines) and 30 m out into the water using an internal buffer. We quantified areal coverage of submerged woody debris and aquatic vegetation by delineating all visible pieces of wood and patches of vegetation as polygon features (N = 56 IMS sampling sites). If multiple pieces of woody debris overlapped with one another, they were grouped together in a single polygon to avoid double counting the area of overlap. Standing timber within the littoral zone was quantified by marking each individual standing tree or stump as a point feature (N = 42 IMS sampling sites). Areal estimates of each habitat patch type (e.g., wood or vegetation) were summed within each IMS sampling site. Since IMS sites varied to some degree in their total area due to shoreline complexity, areas were converted to percent cover and points to count per 1000 m² by dividing by the quantification area of the IMS site.

Benthic substrate was quantified by dividing a polygon of the littoral area of an individual IMS site into sections based on discernible substrate type (N = 38 IMS sampling sites). These polygon sections were given a specific classification as either “boulder”, “cobble”, “gravel”, “sand”, “fine sediment”, or “other” (artificial structures, e.g., concrete boat ramp, bridge pilings, etc.). Each polygon section was also given a broad classification as either “coarse”, “fine”, or “other”. The “coarse” category included areas classified as “boulder” and “cobble”, whereas the “fine” category included areas classified as “gravel”, “sand”, or “fine sediment”. The classification of “other” was the same for both the specific and broad categories. Areas of blank imagery adjacent to the shore (i.e., areas where the sonar beam failed to reach the shoreline due to shallow depth, dense vegetation, etc.) were labeled as “unclassified” (Supplementary Figure 1). Classification was performed using a training set of reference images of the different substrate types that were taken from areas of known substrate composition. The minimum size for all classifiable sections (i.e., sections not designated as “unclassified”) was 3m x 3m. Within each IMS site, areas of each substrate type were summed and divided by the total area of the IMS site to obtain the percent areal coverage of both the specific and broad substrate categories. GIS mapping times for each habitat feature were also recorded at a subset of IMS sampling sites from each lake to assess how image quality from each sonar system affected processing efficiency.