Proposed Echosounder Data Processing Levels (DRAFT)#

The decades-long experience from the satellite remote sensing community has shown that a set of robust and well-articulated definitions of “data processing levels” [1]–[5] can lead directly to broad and highly productive use of data. Processing level designations also provide important context for data interpretation [6]. However, no such community agreement exists for active acoustic data. The ambiguity associated with the interoperability and inter-comparability of processed sonar data products has hindered efficient collaboration and integrative use of the rapidly growing data archive across research institutions and agencies.

The echopype team is developing a clearly defined progression of data processing levels for active ocean sonar data. The development leverages the collective experience from remote sensing and large-scale, long-term ocean and ecological observatories [7]–[10]. Data processing functions in echopype are clearly associated with specific Processing Level inputs and outputs, and when appropriate, will generate a processing_level dataset global attribute with entries such as “Level 1A”, “Level 2B”, “Level 4”, etc.

Processing Levels and Sub-levels#

Level 0 (L0)#

Description: Raw data in vendor sensor format.

  • raw binary files. Associated metadata may be found in external files.

Level 1 (L1)#

Description: Raw data packaged with ancillary information and converted and standardized to an open convention and standard data formats. May be distributed in the following two forms:

  • as sets of individual converted files as originally segmented into arbitrary time ranges during sensor file creation, or

  • compiled into larger granules corresponding to logical deployment intervals.

  • L1A: Raw L0 data converted to a standardized, open format with geographic coordinates (latitude & longitude) included. Includes other ancillary information extracted from sensor-generated L0 data or other external sources. May include environmental information such as temperature, salinity and pressure. Use of the SONAR-netcDF4 v1 convention is strongly recommended.

  • L1B: L1A data with quality-control steps applied, such as time-coordinate corrections that enforce strictly increasing, non-duplicate timestamps.

Level 2 (L2)#

Description: Calibrated acoustic quantities at raw data resolution, with spatial coordinates included (latitude, longitude and depth)

  • L2A: Volume backscattering strength (Sv) with interpolated latitude, longitude and depth coordinates. May incorporate addition information, such as split beam angle

  • L2B: Sv L2A data with noise removal or other data filtering applied, including seafloor bottom removal.

Level 3 (L3)#

Description: Calibrated acoustic quantities regridded or aggregated to a common grid across channels. May include noise removal or other filtering.

  • Sv resampled to a common, high-resolution grid across channels

  • Mean Volume Backscattering Strength (MVBS)

  • Masks applied to Sv based on different scattering source classification criteria and methods, such as frequency difference across two channels.

  • L3A: The above variables computed on L2A data

  • L3B: The above variables computed on L2B (filtered) data

Level 4 (L4)#

Description: Acoustically derived biological features, involving further processing of L3 data that may include data reduction or incorporation of external sources of data.

  • Nautical Area Backscattering Coefficients (NASC), potentially partitioned into inferred sources of scattering

  • Summary statistics of echogram features (center_of_mass, dispersion, etc)

  • Taxon or species-level data labels (classification). May originate from a variety of methods, including frequency difference thresholds.

  • Estimated biomass, combined or partitioned into different taxonomic sources.

References#