# README - Seafloor Dataset for Bahía Almirante, Bocas del Toro, Panama Title: A Seafloor Dataset for Bahía Almirante, Panama: Single-Beam Echosounder Depth, Acoustic Bottom Indices, and Ground-Truthing Authors: Ximena Boza [1, 2], Carolina I. César-Ávila [1], Viviana D. Bravo [1], Semme J. Dijkstra [2], and Rachel Collin [1] Affiliation: [1] Smithsonian Tropical Research Institute, Balboa, Ancón 0843-03092, Panama [2] Center for Coastal and Ocean Mapping/Joint Hydrographic Center, University of New Hampshire, Durham, NH 03824, USA --- ## Overview This dataset provides the first openly accessible sonar and ground-truth seafloor data for Bahía Almirante, Bocas del Toro, Panama. It includes trackline sonar data (water depth and two relative acoustic bottom indices, E1 and E2, associated with seafloor roughness and hardness, respectively) across the bay and ground-truth data collected at 90 sites. --- ## Data Files Upon entering this dataset on the Smithsonian DataONE portal, you will find two folders: - Bahia_Almirante_raw_DAT_files_imagery: contains the large raw sonar recordings and the original and color-corrected photoquadrat imagery. - Bahia_Almirante_Seafloor_Dataset: contains all tabular and GIS data files. This folder contains a Data/ folder organized into three subfolders (1_Sonar_data/, 2_Groundtruth_data/, 3_Data_supplements/) and a Data_Dictionary.csv file containing column-level descriptions for all variables in all files. All files are described below. **Files available in Bahia_Almirante_raw_DAT_files_imagery:** *Sonar data* - [per-track folders] (DAT, SON, IDX) - Original Humminbird echosounder recordings. *Ground-truth data* - [site image folders] (JPEG) - Photoquadrat imagery; four images per site with original and color-corrected versions where applicable. **Files available in Bahia_Almirante_Seafloor_Dataset:** *Trackline sonar data* - sonar_tracks_metadata.csv (CSV) - Track identifier, start coordinates, and timestamp for each survey track. - data_part_1.csv, data_part_2.csv (CSV) - Raw sonar data (depth, E1, E2, PeakSV, coordinates, track ID); split across two files due to Excel row limits. - all_data (SHP, GPX, KMZ) - Raw sonar data in GIS, GPS, and visualization formats. - processed_data_1.csv, processed_data_2.csv (CSV) - Processed sonar data after outlier removal and tidal correction; recommended for analysis. - combined_data (SHP) - Processed sonar data as GIS-ready shapefile. - SBES_bathymetry_50m.tif (GeoTIFF) - 50 m bathymetry grid of the processed trackline sonar data. - processing_log.csv (CSV) - Soundings excluded from the processed data, with the reason for exclusion. *Ground-truth data* - sites_metadata.csv (CSV) - Coordinates, timestamps, sonar averages, and data availability for all 90 sites. - photoquadrat_point_counts.csv (CSV) - Point-level substrate classifications (25 annotation points per image, 58 sites). - photoquadrat_summary_by_vessel_side_and_site.csv, photoquadrat_summary_by_site.csv (CSV) - Substrate percentages summarized by vessel side and at the site level. - sediment_weights.csv (CSV) - Raw sieve weights and loss-on-ignition data (19 sites). - sediment_calculations.csv (CSV) - Derived grain-size fractions (19 sites). - sediment_site_summary.csv (CSV) - Site-level sediment summary (19 sites). - sonar_data_sites.csv (CSV) - Individual sonar pings recorded at each of the 90 ground-truth sites; these are separate recordings from the trackline sonar data and use distinct track identifiers not listed in sonar_tracks_metadata.csv. - sonar_data_sites_averages.csv (CSV) - Mean depth, E1, and E2 per ground-truth site. --- ## Trackline Sonar Data Location: Bocas del Toro Archipelago - Bahía Almirante, Panama (9.34031°N, 82.24204°W) Time: June 2019 - April 2022. Data collection was interrupted for approximately 7 months due to the COVID-19 pandemic. Equipment: Humminbird HELIX 10 CHIRP SI GPS G2N single-beam echosounder (Johnson Outdoors Marine Electronics, Inc., 2021) Serial number: 17082103-0216 Coordinate Reference System (CRS): WGS 1984 (EPSG:4326). The bathymetry grid (SBES_bathymetry_50m.tif) is in WGS 84 / UTM Zone 17N (EPSG:32617). Data Acquisition: A Humminbird HELIX 10 CHIRP echosounder mounted on a 22-ft vessel was used to collect sonar data. The system provides depth measurements and two relative acoustic bottom indices (E1 and E2), associated with seafloor roughness and hardness, respectively, that were not calibrated against direct measurements of seafloor roughness or hardness. E1 is derived from the first echo return immediately after its peak. E2 is derived from the full second echo return, produced when the signal reflects off the seafloor a second time after bouncing back from the surface. Higher values indicate a relatively rougher (E1) and harder (E2) seafloor. The transducer operated at 200 kHz single-beam frequency with a 20° beamwidth. Depths were recorded by the echosounder relative to the transducer face, and no transducer draft correction was applied (the transducer was approximately 10 cm below the water surface). Data were recorded at 1 Hz while the vessel traveled at 2.5-4.5 knots, with an along-track point spacing of approximately 1.3-2.3 m. Survey line spacing was approximately 800 m, densified to 400 m in areas of high bathymetric relief and 200 m in the channels connecting the bay to the open ocean, and additional lines were run along sections of the coastline. Data were collected over a three-year period (2019-2022) and track orientations were selected based on the feasibility of navigating the bay and its channels. Data files: Sonar data are provided in both raw and processed formats. Track-level metadata (track identifier, start-of-track coordinates, and date/time) are provided in sonar_tracks_metadata.csv. Raw files: The Humminbird echosounder records data natively in a proprietary binary format (.DAT, with associated .SON and .IDX files). These recordings were visualized and exported using ReefMaster v2.0. Raw files are provided in multiple formats: the original Humminbird recordings (DAT/SON/IDX) and ReefMaster exports including two CSV files, a GIS-ready point shapefile, a GPX file, and a KMZ file for Google Earth. The CSV files contain coordinates, water depth, two unitless acoustic bottom indices (E1 and E2), peak seafloor acoustic return strength (PeakSV, unitless), timestamp, and track identifier. Processed files: Processed files are the analysis-ready dataset after depth outlier removal and tidal correction, provided as two CSV files and a shapefile. The data are split across two CSV files because the combined dataset exceeds the maximum number of rows supported by Microsoft Excel; each file can be opened in Excel without truncation. The processed CSV files contain the same variables as the raw CSVs plus an additional column with tidally corrected depth in meters (Depth_m_corrected). Soundings excluded during processing are listed in processing_log.csv. Rows with invalid E1, E2, or PeakSV returns were retained rather than removed because they still contain valid depth measurements; invalid values in these variables are recoded as -9999. Users analyzing E1, E2, or PeakSV should exclude rows where those values equal -9999. A 50 m bathymetry grid (SBES_bathymetry_50m.tif) was created from the tidally corrected depths in QPS Qimera v2.8.2. Each cell contains the mean depth (m, positive down) of the soundings within it. Cells without soundings are NoData, and no interpolation was performed between survey lines. These are the recommended files for analysis. Post-processing: Sonar data were processed in ReefMaster v2.0. Only data recorded within the bay were processed. Depth outliers (zero or physically implausible depths, or abrupt depth changes relative to neighboring soundings) were removed by manual inspection (4,084 soundings; 0.3% of those recorded within the bay). Invalid values in E1, E2, and PeakSV were recoded as -9999. Tidal correction consisted of subtracting the water level, referenced to the mean water level, from the measured depth of each sounding. Tidal heights predicted with the Unified Tidal Analysis and Prediction Functions (UTide; Codiga, 2011) were used for surveys conducted in 2019-2020, and observed water levels for surveys conducted in 2022. In both cases, water levels were linearly interpolated in MATLAB to the timestamp of each sounding. Sea-level measurements used for tidal correction were obtained from the tide gauge of the Smithsonian Tropical Research Institute's Physical Monitoring Program, located on the main platform of the STRI Bocas del Toro Research Station (9.3509°N, 82.2577°W; https://striresearch.si.edu/physical-monitoring/rtg-bocas-del-toro). For access to the sea-level dataset, contact the Physical Monitoring Program team directly. --- ## Ground-truth Data Time: October 2021 - July 2023 Ground-truth data: collected at 90 sites distributed across the bay. At each site, a standardized, decision-based field protocol was followed based on water clarity, substrate type, and sampling feasibility. Site-level metadata for all 90 sites are provided in sites_metadata.csv. The following data are available per site: - Photoquadrat imagery: available for 58 of 90 sites (four images per site, one per vessel side: bow, stern, port, starboard). Images were taken with a GoPro Hero 4 camera using the medium field-of-view setting. Site folders are named Site_XX (e.g., Site_02), corresponding to Site_ID SXX (e.g., S02), and image file names begin with the Photo_ID used in photoquadrat_point_counts.csv (e.g., S02_Bow_edited.jpg). Within each site folder, images are organized into two subfolders: `Original/` contains the unprocessed images as captured, and `Color_corrected/` contains versions color-corrected using the Dive+ application to improve seafloor visibility, particularly for red-channel correction at depth. For 8 sites (Site_28, Site_31, Site_34, Site_38, Site_39, Site_43, Site_46, Site_63), only the `Original/` subfolder is present, as color correction was not required for those images. Point-count annotations were performed on the color-corrected images (or on the original images at these 8 sites) and are provided at the image level (photoquadrat_point_counts.csv), summarized by site (photoquadrat_summary_by_site.csv), and summarized by vessel side and site (photoquadrat_summary_by_vessel_side_and_site.csv). Substrate classes are seagrass, hard coral, soft coral, dead coral, sponges, macroalgae, UHB (uncolonized hard bottom), and sediment; "sediment" includes unconsolidated sediment of any grain size. - Sediment grain-size analysis: available for 19 sites with a measurable coarse fraction. Grain-size fractions (gravel, sand, and mud) follow Wentworth (1922), and sediment textural classes follow Folk (1954). Percentages of each fraction are calculated relative to the sum of all sieve fractions (Sum_all_fractions_g). Each site has three replicate samples (R1-R3), and site-level values in sediment_site_summary.csv are the mean of the three replicates. Grain-size analysis was performed on three subsamples of up to 100 mL each, oven-dried at 70 °C without pretreatment, and dry-sieved. Mud-only samples were classified qualitatively and are not included in the grain-size files. Raw sieve weights are provided in sediment_weights.csv, derived fractions in sediment_calculations.csv, and site-level summaries in sediment_site_summary.csv. sediment_weights.csv also includes loss-on-ignition (LOI) measurements (tin tare, pre- and post-combustion weights, and combustion date) for all 19 sites. After sieving, the fractions of each subsample were recombined and combusted at 520 °C for 5 h, following the MarineGEO Sediment Organic Matter Protocol (2021). LOI is provided as an estimate of organic matter content and was not used in the grain-size calculations. At six sites, sediment samples and their qualitative assessment were collected during a later visit than the one recorded in sites_metadata.csv; the collection date is given in Date_collection (sediment_weights.csv). - Qualitative substrate assessment: recorded at all 90 sites, based on visual observation and tactile hand-texture assessment. - Site-level sonar data: collected at all 90 sites and processed in the same way as the trackline sonar data, except that no tidal correction was applied. Individual pings are provided in sonar_data_sites.csv; site-level averages of depth, E1, and E2 are provided in sonar_data_sites_averages.csv and also in sites_metadata.csv. --- ## Technical Notes Linking files: - Site_ID (format: S01-S90) is the primary key connecting all ground-truth files: sites_metadata.csv, photoquadrat_point_counts.csv, photoquadrat_summary_by_site.csv, photoquadrat_summary_by_vessel_side_and_site.csv, sediment_weights.csv, sediment_calculations.csv, sediment_site_summary.csv, sonar_data_sites.csv, and sonar_data_sites_averages.csv. - Sonar_track_ID (format: R00002) connects sonar_tracks_metadata.csv to the raw trackline CSV files (data_part_1.csv and data_part_2.csv), where it appears as Track_ID, and to the processed trackline CSV files (processed_data_1.csv and processed_data_2.csv) and to processing_log.csv. Note that Sonar_track_ID also appears in sites_metadata.csv and sonar_data_sites.csv, where it identifies ground-truth site recordings; these track IDs are separate from the trackline sonar data and are not listed in sonar_tracks_metadata.csv. - Within sediment_weights.csv and sediment_calculations.csv, Sample_ID (format: S02-R1) links rows across the two files. Track identifier names: The track identifier column is named Track_ID in the raw CSV files and in the processed shapefile (combined_data.shp), TrackName in the raw shapefile (all_data.shp), and Sonar_track_ID in all other files. All refer to the same identifier. Raw vs. processed track coverage: The raw CSV files (data_part_1.csv, data_part_2.csv) contain 215 survey tracks (1,508,128 soundings), while the processed CSV files (processed_data_1.csv, processed_data_2.csv) contain 199 tracks (1,382,019 soundings). Tracks recorded outside the bay were not processed. All 126,109 soundings present in the raw CSV files but not in the processed files are listed in processing_log.csv, with the reason for exclusion (outside_bay or outlier). Every raw sounding therefore appears in either the processed files or the processing log, and can be matched using Sonar_track_ID (Track_ID in the raw files) and DateTime_UTC. Missing values: Invalid or absent sonar returns in E1, E2, and PeakSV are flagged as -1 in the raw CSV files and processing_log.csv, and as -9999 in the processed CSV files. Users should filter these values out before analyzing E1, E2, or PeakSV. Limitations: - The 50 m bathymetry grid (SBES_bathymetry_50m.tif, also shown in Bocas_sites_tracks_labels_map.png) does not represent the resolution of the survey. - The absolute accuracy of the depth data has not been independently established; the data are not intended for navigation or engineering applications. - E1 and E2 are relative indices that depend on water depth and acquisition settings. - The correspondence between acoustic indices and substrate type is exploratory. Column-level descriptions for all variables in all files are provided in Data_Dictionary.csv. --- ## Data Supplements The 3_Data_supplements folder contains supplementary figures and materials intended to support data exploration and interpretation. - **Bocas_map.png:** Map of Bahía Almirante, Bocas del Toro, Panama. - **Bocas_sites_tracks_labels_map.png:** Map of Bahía Almirante, Bocas del Toro, Panama. Colored lines show depth from the processed trackline sonar data, gridded at 50 m cell size in QPS Qimera v2.8.2 for visualization only (only cells containing soundings are shown, and no interpolation was performed between survey lines). Black-outlined circles show the 90 ground-truth sites, labeled with their site identifiers and colored by mean depth from the site-level sonar data using the same depth scale. - **groundtruth_field_protocol.png:** Standardized, decision-based field protocol applied at each ground-truth site. The workflow determines which data types are collected based on seafloor visibility, substrate type, and sampling feasibility. - **site_data_availability.html:** Interactive HTML table showing, for each of the 90 sites, which data types are available (photoquadrats, grain-size analysis, substrate assessment) and summary values. Open in any web browser. - **groundtruth_composite_figure.png:** Summary of ground-truth data availability and acoustic-substrate correspondence across all 90 sites, ordered by increasing E2. Top panel: acoustic indices E1 (blue) and E2 (red). Upper-middle panel: qualitative substrate assessment; sites identified as mud, hardbottom, or mangrove peat are shown (subset of 90 sites). Bars indicate presence only and do not represent a measured quantity. Lower-middle panel: grain-size composition as percent gravel, sand, and mud, available for 19 sites. Bottom panel: substrate cover from photoquadrat point counts, available for 58 sites. In this panel, "vegetation" combines seagrass and macroalgae, and "coral" combines hard coral and soft coral. - **sonar_photoquadrat_scatter_matrix.png:** Scatter matrix showing pairwise relationships among mean depth (m), mean E1, and mean E2 at the 58 photoquadrat ground-truth sites. Points are colored by dominant substrate class from photoquadrat point-count analysis. "Vegetation" combines seagrass and macroalgae, and "coral" combines hard coral and soft coral. - **depth_distribution_kde.png:** Kernel density estimate of the depth distribution of processed trackline sonar data from Bahía Almirante. The y-axis represents depth (m) and the x-axis represents the number of sonar points (×10³). The estimate was computed using a Gaussian kernel with a bandwidth of 0.16 m. The maximum recorded depth is 34 m. --- ## References Codiga, D.L., 2011. Unified Tidal Analysis and Prediction Using the UTide Matlab Functions. Technical Report 2011-01. Graduate School of Oceanography, University of Rhode Island, Narragansett, RI. 59 pp. Johnson Outdoors Marine Electronics, Inc., 2021. HELIX Series Operations Manual (532509-5EN_B). Johnson Outdoors Marine Electronics, Inc. Available at: https://humminbird.johnsonoutdoors.com/us/support/manuals MarineGEO Sediment Organic Matter Protocol, 2021. Tennenbaum Marine Observatories Network, MarineGEO, Smithsonian Institution. doi:10.25573/serc.14925111.v1. Folk, R.L., 1954. The Distinction between Grain Size and Mineral Composition in Sedimentary-Rock Nomenclature. The Journal of Geology, 62, 344-359. doi:10.1086/626171. Wentworth, C.K., 1922. A Scale of Grade and Class Terms for Clastic Sediments. The Journal of Geology, 30, 377-392. doi:10.1086/622910. --- For full methodological details, refer to the associated publication.