Sea-floor videos and locations of bottom video tracklines collected in Cape Cod Bay, Massachusetts, in September 2019 by the U.S. Geological Survey during field activity 2019-034-FA (MP4 video files and polyline shapefile, GCS WGS 84)

Step 1: Collected data.
A marine geological survey (field activity 2019-034-FA) was conducted in Cape Cod Bay, Massachusetts, in September 2019. The R/V Tioga occupied one of the target sites, and the MiniSEABOSS was deployed off the vessel's A-frame on the stern of the ship. The MiniSEABOSS was equipped with a modified Van Veen grab sampler, a GoPro HERO4 Black digital camera, an oblique downward-looking SeaViewer 6000 HD Sea-Drop video camera with a topside feed, a Paralenz DiveCamera+ recording backup video, and a dive light to illuminate the sea floor for video and photograph collection. The elements of this particular MiniSEABOSS were held within a stainless-steel frame that measured ~1 x 1 meter. The frame had a stabilizer fin that oriented the system as it drifted over the seabed. Two red lasers were set 20 centimeters apart (both as they were mounted on the MiniSEABOSS frame and as seen in photographs and video on the seabed) for scale measurements. The red laser dots can usually be seen in the sea-floor photos and videos depending on the bottom type and distance to the sea floor. The winch operator lowered the MiniSEABOSS until the sea floor was observed in the topside live video feed. Generally, the vessel and MiniSEABOSS drifted with wind and current for up to a few minutes to ensure a decent photo with a clear view of the sea floor was acquired. The GoPro camera was set to take photos every 2 seconds during a MiniSEABOSS deployment. Bottom video was also recorded during the drift from the oblique downward-looking SeaViewer 6000 HD Sea-Drop video camera directly to a solid-state drive using an Odyssey7 video recorder. Then, the winch operator lowered the Van Veen grab sampler until it rested on the sea floor. When the system was raised, the Van Veen grab sampler closed and collected a sample as it was lifted off the sea floor. The sampler was recovered to the deck of the survey vessel where a subsample was taken for grain-size analysis at the sediment laboratory at the USGS Woods Hole Coastal and Marine Science Center. During the survey, DGPS navigation from a Hemisphere R131 DGPS receiver was logged through a DataBridge data logger and QGIS (version 3.0.0) GPS Tools. The DGPS was set to receive fixes at a 1-second interval in geographic coordinates (WGS 84). Dates and times were recorded in Coordinated Universal Time (UTC). Log files were saved for each Julian day in NMEA text format. The GPS data were also overlaid onto the SeaViewer video using a Proteus-V Pro video overlay device. A total of 49 sites were occupied aboard the R/V Tioga with the MiniSEABOSS during field activity 2019-034-FA, and bottom videos were collected at all 49 sites.

Step 2: Processed video files.
A shell script (do_concat.sda) was run on the original video files to concatenate the video clips for each site as needed (the Odyssey7 splits clips into less than 4 GB segments) and convert the video files to MP4 files using FFmpeg. The script also created a text file with the start time and duration of each video recording.

Step 3: Processed navigation.
To process the NMEA navigation data, GPSBabel (version 1.7.0) was used to read the GPRMC navigation sentences from the NMEA log files, translate them to waypoints, and write the waypoints to a GPX XML file (nmea_nav_jd260_jd262.gpx). The GPX file was then converted to a shapefile in QGIS (version 3.10.9). New fields for the latitude and longitude were added to the shapefile and calculated to six decimal places. The shapefile was exported as a Comma Separated Values (CSV) text file, which was then formatted in Microsoft Excel for Mac (version 16.16), creating a processed NMEA navigation CSV file for the survey (nmea_nav_jd260_jd262.csv).
Due to an issue with the data logger, the raw navigation data were recorded to NMEA files for only 22 of the 50 sampler deployments. For the remaining 28 deployments, the latitude and longitude recorded on the video overlay were used for the navigation data. To obtain the navigation data from the overlay, first, a command was run to extract a still-image frame grab every 1 second from the videos using FFmpeg (version 4.3.1). Next, a command was run to crop the latitude from each frame grab. The cropped image of the latitude was then converted to text using the optical character recognition engine Tesseract (version 4.1.1) and saved as a text file. These commands were repeated to crop the longitude from each frame grab and convert it to text, and then again to crop and convert the time to text. All the latitude, longitude, and time text files were concatenated and merged into a single text file. This text file was then imported and formatted in Microsoft Excel for Mac, creating a processed video overlay navigation CSV file for the survey (videocrop_coords_and_times.csv). The video overlay navigation data were checked to see if any times were missing. Occasionally, frame grabs were not extracted for the first or last second of the video. These missing frame grabs were identified, and the navigation data from the missing frame grabs were added to the video overlay navigation CSV file.
Shapefiles of the NMEA and video overlay navigation CSV files were created in ArcGIS (version 10.7.1). The sites with no NMEA navigation were identified, and the video overlay navigation data for those sites were exported as a table. The exported table was added to the NMEA navigation in Microsoft Excel for Mac and saved as a CSV file, creating a final navigation file for the survey (nmea_and_videocrop_nav.csv). This process step and the subsequent process steps were performed by the same person, Emily Huntley.

Step 4: Trimmed and renamed video files.
The videos were prepared for publication. First, videos used for camera calibration or less than 1 second were removed. Next, the bottom videos were trimmed as needed using FFmpeg (version 4.3.1). For the first survey day (Julian day 260), the videos were trimmed to the duration that the camera was within view of the sea floor if more than about 20 seconds of sampler retrieval was originally recorded. For the second survey day (Julian day 262), both the sampler deployment and retrieval were recorded, so all the videos were trimmed to the duration that the camera was within view of the sea floor. One site (site 2019-034-FA-034) had one continuous video of two separate deployments, so the video was split into two videos and trimmed to the duration that the camera was within view of the sea floor. Finally, the videos were renamed using a shell script to include the field activity identifier, camera, and date and start time in the ISO 8601 standard (YYYYMMDD T [time separator] HHMMSS Z [Zulu/UTC time]) in the filename.

Step 5: Created a CSV file of the bottom video trackline points.
The text file with the start time and duration of each video recording and the final navigation data were prepared for the video trackline script. Since the videos were trimmed, the start time and duration of each video was checked and corrected as needed in the start times/durations text file. Then, the fields in the final navigation CSV text file were reordered and formatted to be used with the video trackline script. Finally, a Jupyter Notebook Python script (Video_trackline_prep_WORKING_v2.ipynb) was run to create a CSV file of the bottom video trackline points by extracting the navigation data for each video drift using information from the start times/durations text file. The script reads the video start time and duration from the text file, calculates the video end time, extracts the navigation points that fall within those start and end times, and exports the navigation points to a CSV file.

Step 6: Created the final bottom video tracklines shapefile.
A point shapefile was created using the bottom video trackline points CSV files in Esri ArcGIS Pro (version 2.4.1). The Points to Line tool was then run with the video trackline points as the input features and the video filenames as the line field to create a polyline shapefile of the video tracklines. New fields were added to the polyline shapefile, including an attribute for the site number of the video trackline (FIELD_NO), bottom video filename (LINENAME), start time of the bottom video drift in UTC (STARTTIME), end time of the video drift in UTC (ENDTIME), Julian day of collection (JD), date of collection (DATE), year of collection (YEAR), trackline length in meters (LENGTH_M), camera used (CAMERA), field activity identifier (FA_ID), sampling device used to collect the video (DEVICE_ID), and survey vessel (VEHICLE_ID). The bottom video tracklines shapefile was joined with the trackline points shapefile to add the start and end times of the video drifts (STARTTIME and ENDTIME, respectively), Julian day of collection (JD), and date of collection (DATE). The tracklines shapefile was joined with the survey log to assign the site number (FIELD_NO) of each video trackline. The trackline length (LENGTH_M) was calculated using the Calculate Geometry Attributes tool (Property=Length; Length Unit=Meters; Coordinate System=WGS 1984 UTM Zone 19N).