This dataset consists of ZooProcess and Ecotaxa outputs from ZooSCANs of plankton caught in the upper 600m using Multiple Opening-Closing Net and Environmental Sensing System (MOCNESS) tows during day- and night-time. It includes data for this project from Ecotaxa (export v1.0), an online machine-learning platform that assists in identifying organisms and particles. The dataset also includes particle measurements generated by ZooProcess software. These samples were collected and processed over two years, with three cruises a year to capture distinct seasons. The goal of this data was to assess high-resolution vertical distribution of zooplankton in order to distinguish diel vertical migrators from resident populations and to quantify contributions to particulate organic carbon flux via fecal pellet production. Project description: The oceanic biological carbon pump refers to the export of dissolved and particulate organic carbon to the deep ocean, and it is a significant driver of atmospheric carbon uptake by the oceans. Evidence from long-term research carried out at the Bermuda Atlantic Time-series Study (BATS) site suggests that the spectrum of particles collected by gel-traps below the euphotic zone changes drastically below 150 m, which is attributed to resident populations of zooplankton that feed on vertically migrating zooplankton as well as sinking particles. The goals of this study are to investigate the role of different zooplankton taxa on both particle aggregate formation and in particle transformation, and to compare and characterize the particles generated by the zooplankton communities with those collected by particle traps.

Table of Contents

• Coverage
• Dataset Description
  • Methods & Sampling
  • Data Processing Description
  • BCO-DMO Processing Description
  • Problem Description
• Data Files
• Supplemental Files
• Related Publications
• Related Datasets
• Parameters
• Instruments
• Deployments
• Project Information
• Funding

One pair of 1m2 MOCNESS (Multiple Open Closing Nets with Environmental Sensor System) tows were performed during each cruise- one during the day, and one at night (MOCNESS, Wiebe et al, 1985). Nets with 150um mesh were used to better capture the smaller midwater zooplankton community in the region. Eight nets were fired in sequence along the upcast to capture spatially discrete zooplankton samples between 600m and the surface. While nets one, two, and three consistently targeted depths of 600-500m, 500m-400m, and 400-300m, depths for nets four through eight varied based on hydrographic features including the thermocline, deep chlorophyll maximum, and oxygen minimum zone (Maas et al, 2014, Steinberg et al, 2008).  Once onboard, samples were split in two using a Motoda splitter (Motoda, 1959)  with half preserved with sodium tetraborate buffered 4% formalin in seawater to be scanned with a ZooSCAN (Gorsky et al, 2010) and half placed in 95% undenatured ethanol for metabarcoding. 

​A representative subsample of the formalin-preserved zooplankton community from each net were imaged using a ZooSCAN ver. 4 at either 4,800 dpi or 2,400 dpi (following the methods in: Gorsky et al., 2010, Vandromme et al., 2012 as detailed in Maas et al. 2021). The change in resolution partway through the project was a result of recommendations from Hydroptic and loss of software support for 4800dpi imaging. In order to better represent all size classes in the images, the original sample was divided into three size categories. All individuals larger than 2 cm were selected by eye and scanned separately from all the others (fraction "d1"). The remainder of the sample was sieved through a 1-mm mesh sieve, and both size fractions were individually scanned ("d2" >1000um, "d3" 153-1000um). From these smaller size fractions, at least 1500 particles were scanned after subsampling using a Motoda splitter (Motoda, 1959), requiring generation of two separate scans for both size classes. This resulted in a total of five images per net.

ZooSCAN Image names:

Image names include: cruise#_mocnessID_net#_sizefraction_ and _a|b if a replicate and end in _raw_1.tif

Multiple images of the same size fraction were sometimes taken to obtain a sufficient number of particles. These replicates are named a or b. If there is no replicate they don’t have a letter in the image name. An a and b scan were always done for size classes d2 and d3.  This was important because the split size is for the sum of a+b (e.g. if a is ¼ and b is ¼, the acq_sub_part will be 0.5).

Example of image names:

ae2112_m22_n4_d3_a_raw_1.tif  [a replicate]
ae2112_m22_n4_d3_b_raw_1.tif  [b replicate]
ae2204_m27_n5_d1_raw_1.tif      [no replicate]

This dataset contains the "object_id" (the particle identifier) which is constructed the same way as the image name except it as an additional _# at the end.  This additional number in the object_id is added by the ZooProcess software (Hydroptic, 2016).
e.g.
object_id:       ae1614_m3_n1_d2_a_1_100
image_name: ae1614_m3_n1_d2_a_1.tif

Particle names:

Names for particles follow the pattern "CruiseID_MocnessID_NetNumber_ScanFraction" followed by "_1_XXX", with "1" being automatically added by the software to indicate no duplicates of that scan and "XXX" being the unique particle number within that scan. 

Parameter (column name) nomenclature and data origin: 
* (see the "Parameters" section which contains all column information for the ecotaxa output table)

Parameters (column names) beginning with "object" include basic identifying metadata input by the user as well as all particle measurement data generated by ZooProcess. Any parameters beginning with "object_annotation" parameters are added by Ecotaxa. Parameters that begin with "sample" are sampling metadata input by the user during the scanning process.  "Process" parameters describe the software and assumptions or corrections input during the data processing. "Acq" describes the portion of the sample scanned (input by user) and provides some summary data about the scanned image. 

Interpreting acq_sub_part

For this project, the acq sub part is the fraction of the individual scan only. "A" and "B" scans of the same fraction can be statistically combined for analysis (e.g. d2_a and d2_b from the same net can be combined by adding the sub parts to create just a "d2" group). 

Instruments:

The Multiple Opening/Closing Net and Environmental Sensing System or MOCNESS is a family of net systems based on the Tucker Trawl principle. There are currently 8 different sizes of MOCNESS in existence which are designed for capture of different size ranges of zooplankton and micro-nekton. Each system is designated according to the size of the net mouth opening and in two cases, the number of nets it carries. The original MOCNESS (Wiebe et al, 1976) was a redesigned and improved version of a system described by Frost and McCrone (1974)(from MOCNESS manual). The MOCNESS used in this experiment is a 1m2 (mouth size) rigged with nine 150um mesh nets. One is flown open on the downcast to balance the net (Net 0- contents preserved but not analyzed), and the other eight (Net 1-8) are triggered on the upcast at desired depths. This particular MOCNESS was originally manufactured by Biological Environmental Sensor Systems (BESS), but was refit with new electronics from SIO/STS in 2017 (Net Interface Unit, Net Angle Sensor) to allow it to interface with Seabird instruments (SBE9Plus CTD, SBE3S Temperature, SBE4C Conductivity, SBE11 Deck Box). 

The ZooSCAN (CNRS patent) system makes use of scanner technology with custom lighting and a watertight scanning chamber into which liquid zooplankton samples can be placed. The scanner recovers a high-resolution, digital image and the sample can be recovered without damage.  These digital images can then be investigated by computer processing. While the resolution of the digitized zooplankton images is lower than the image obtained using a binocular microscope, this technique has proven to be more than adequate for large sample sets. Identification of species is done by automatic comparison of the image (vignette) of each individual animal in the scanned image with a library data set which may be built by the investigator for each individual survey or imported from a previous survey. The latest machine learning algorithm allows high recognition levels even if we recommend complementary manual sorting to achieve a high number of taxonomic groups. Scans for this dataset performed with a ZooSCAN (Hydroptic, HYDROPTIC_V4) running with Vuescan (version 9.5.24) and ZooProcess (version 8.22, ImageJ macro suite).

Scans were processed using ZooProcess (version 8.22, ImageJ macro suite). The "Convert and process from RAW" function was used to separate particles into individual vignettes and generate a suite of measurements for each particle. "Doubles" (vignettes containing more than one particle) were manually separated in the software and reprocessed. 

Example vignettes (see Supplemental files for example image download):
  

Processed scans and their corresponding metadata were then uploaded to Ecotaxa (Picheral et al, https://ecotaxa.obs-vlfr.fr/), where a training set was created using manually classified images from this project as well as existing validated images from other projects in the Sargasso Sea. Classification categories were chosen based on taxon of interest, identification level in previous projects, and known limitations of the software. Generally, broader level taxonomic groups are used. Identification of all particles was predicted, then manually validated. 

Data versioning

The ecotaxa data included in this dataset version are "Ecotaxa Export version 1.0."

The ecotaxa export file will be updated in a versioned manner as validation of identification is completed.  Ecotaxa output versions are as follows: 

Version 1: No identifications, predicted or validated
Version 2: All identifications predicted
Version 3: All identifications validated

Version 1:
* Data from source file ecotaxa_export_5446_20240606_1831 v1.0.tsv were imported into the BCO-DMO data system as the primary table for this dataset with "nan" values interpreted as missing data identifiers.
* After import, select columns designated to have the missing data identifier 99999 had a find and replace operation performed to turn 99999 values into system missing data identifiers. This was not done on import since only some columns were described as using this identifier by the data providers. 99999 is within the valid numeric range for columns such as object_intden and object_area_exec so it is possible that if there were any real 99999 values, they were instead interpreted as missing data.

** In the BCO-DMO data system missing data identifiers are displayed according to the format of data you access. For example, in csv files it will be blank (null) values. In Matlab .mat files it will be NaN values. When viewing data online at BCO-DMO, the missing value will be shown as blank (null) values.

* Parameters (column names) renamed to comply with BCO-DMO naming conventions. See https://www.bco-dmo.org/page/bco-dmo-data-processing-conventions

* object_ISO_DateTime_UTC added from local date and time columns.

* extra trailing slash removed in object_link (all values were "http://www.zooscan.obs-vlfr.fr//"). This link may not be persistent for long term curation, a citation for the site was added as a related publication along with the date it was accessed.

* lat lon column names changed for consistency with another dataset of the same data type https://www.bco-dmo.org/dataset/932252. object_lat -> object_lat_start and object_lon -> object_lon_start to correspond to _end columns.
* lat and lons rounded to 5 decimal places.

* References in metadata to Ecotaxa "annotation" columns were removed as those columns were not provided with this version of the data table. They will be included in future updates to this dataset (see Methodolgy and Data Procesing sections).

* format in acq_scan_time updated to newer ecotaxa convention (e.g. times 112023 -> 12:20:23).

Any data showing as a blank denotes an error in data generation or metadata retention in the zooprocess software pipeline. Critical data and metadata needed for processing have been checked and corrected (if needed) in the QC process.

In future updates, this dataset will include annotation columns. Blanks in the annotation columns are due to no identification having been assigned yet.

The process_time and acq_scan_time columns contain variable formats within the column. This will be resolved to have a consistent format with the columns in the next dataset update.

Coverage: Sargasso Sea/BATS area

NSF Award Abstract:
The purpose of this collaborative project is to advance understanding of the role of marine planktonic animals (or zooplankton) in the biological pump, or transport of carbon from surface to deeper ocean waters. This movement of carbon from surface to deep ocean water can ultimately affect carbon dioxide in the atmosphere, with implications for global climate. Many marine zooplankton, including species of copepods and krill, play a direct role in the biological pump both because they are abundant and because they can migrate from surface waters at night, where they feed, to depths of more than 500 m at night. At the same time, some organisms called flux feeders will remain at depth and do not migrate. Instead, they rely on particles produced by other zooplankton feeding in surface waters. In this project, the investigators are focusing on populations of flux feeders in the deeper ocean waters of the Sargasso Sea. They are leveraging an ongoing long-term research program, conducting field collections using specialized nets and particle traps, as well lab experiments, as a way to understand how these organisms modify the particles around them. This project is supporting a postdoctoral scientist and providing research experiences for undergraduates at two institutions. An education specialist is creating lesson plans for an award-winning Ask-A-Biologist website, designed for public and K-12 audiences. Images of zooplankton will be disseminated to the public and scientific community via EcoTaxa (a web platform devoted to plankton biodiversity, with images and taxonomic annotation) and physical samples will be archived as part of a teaching library.

The oceanic biological carbon pump refers to the export of dissolved and particulate organic carbon to the deep ocean, and it is a significant driver of atmospheric carbon uptake by the oceans. Evidence from long-term research carried out at the Bermuda Atlantic Time-series Study (BATS) site suggests that the spectrum of particles collected by gel-traps below the euphotic zone changes drastically below 150 m, which is attributed to resident populations of zooplankton that feed on vertically migrating zooplankton as well as sinking particles. The goals of this study are to investigate the role of different zooplankton taxa on both particle aggregate formation and in particle transformation, and to compare and characterize the particles generated by the zooplankton communities with those collected by particle traps. The investigators are combining field collections with experiments onboard ship and in environmental chambers. They are collecting samples over two years, with three cruises a year to capture distinct seasons. They are assessing high-resolution vertical distribution of zooplankton in the upper 600 m using Multiple Opening-Closing Net and Environmental Sensing System (MOCNESS) tows during day- and night-time, to distinguish diel vertical migrators from resident populations and to quantify contributions to particulate organic carbon flux via fecal pellet production. On each cruise, sinking particles are being collected using gel trap tubes attached to the particle traps deployed monthly at BATS. In addition, roller tank experiments are determining how individual zooplankton mediate aggregate formation. Particle types and fecal pellets are being characterized using image analysis and DNA-based analysis of microbial communities. Finally, ongoing data collection from the long-term BATS program is providing invaluable environmental context and will ensure results from this study contribute to ongoing community efforts to observe and predict the fate of carbon in our global system.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.