A Pelagic Size Structure database (PSSdb) to support biogeochemical modeling: update to first release

This dataset is an update to the first release of the Pelagic Size Structure database (PSSdb, https://pssdb.net) scientific project, investigating the global particle size distributions measured from multiple pelagicǂ imaging systems. These devices include the Imaging Flow Cytobot (Olson and Sosik 2007), benchtop scanners like the ZooScan (Gorsky et al. 2010), and the Underwater Vision Profiler (Picheral et al. 2010). The data sources originate from Ecotaxa (https://ecotaxa.obs-vlfr.fr/), Ecopart (https://ecopart.obs-vlfr.fr/), and Imaging FlowCytobot dashboards (https://ifcb.caloos.org/dashboard and https://ifcb-data.whoi.edu/dashboard). Links to the PSSdb code and documentation are available on the PSSdb webpage (https://pssdb.net). 

This updated version includes the following changes:           

• Duplicate data entries and NaN values have been removed.
• Data products now include Normalized Biomass Size Spectra (NBSS), and Particle Size Distribution (PSD), two widely used methods to represent plankton and particles size distribution in marine ecology and biogeochemistry.
• Linear regressions are now performed with log10 transformations of the normalized biovolume/abundance and the size classes.
• Inclusion of UVP6 and other benchtop plankton Scanner datasets from net tows, which expand the temporal and spatial coverage of the data products.
• Unbiased portion of the size spectra is selected by a new thresholding method that accounts for both uncertainties on particle sizes, limited by the camera resolution, and particle count, so that only size classes with less than 20% uncertainty are retained, in addition to gaps in the size spectra.

This PSSdb dataset is composed of two products, specific to each imaging device: 

• Product 1a includes the size distribution , computed from normalized biovolume, for NBSS, and normalized abundance,  for PSD, of plankton and particles within a set of pre-defined size classes (expressed in both biovolume and equivalent circular diameter), averaged by year and month, and in 1-degree longitude/latitude grid cells.
• Product 1b includes the results of NBSS and PSD  regression fit parameters, slopes, intercept, and coefficient of determination (R2), averaged by year and month, and in 1-degree longitude/latitude grid cells. The regression parameters are defined using ordinary least squares linear regressions applied to a log10 transformed normalized biovolume/normalized abundance  and biovolume/ diameter size  class values.

Size spectra parameters were averaged  over a maximum of 16 spatial and temporal subsets (0.5°x0.5°x1 week) to avoid over-representation of repeated sampling events (e.g., time-series datasets) within a grid cell. Linear regressions were performed on the linear portion of the log10-transformed NBSS and PSD estimates, between the size classes with a size measurement or particle count uncertainty greater than 20% (Schartau et al. 2010) , and where the maximum NB/PSD is observed and the largest size class before three empty consecutive size classes.

                                     For additional information, please see the PDF documentation available below ...