Dataset: Global depth resolved reconstruction of particle size distribution and carbon export

This work is based on the compilation of over 6800 profiles of particulate matter observations from Underwater Vision Profilers (UVP5) (Rainer, 2021). The biovolume of the particle size distribution is calculated as the equivalent spherical volume of the particle size distribution  (PSD), by summing the product of particle counts time particle volume in each size class. The slope of the PSD is calculated assuming a power law distribution for the particle abundance, by linear least square fit of the log of particle counts vs. the log of particle size. These quantities are estimated for the upper 2000m of the water column (Global_POC_Export_2024.nc). To convert sparse observations to a global climatology, we trained 50 ensembles of regressions trees (Random Forests, RF) to predict biovolume and slope based on their relationship to well-sampled physical and biogeochemical predictors (predictors_3D.mat).

We calculate the particle sinking speed and carbon content by combining PSD reconstruction (biovolume and slope) with an empirical relationship between particle size, carbon content and sinking speed, the parameters of which are optimized to match in situ particle flux observations (Mouw et al. 2016). The flux values are calculated as the sum of the PSD times the sinking carbon parameters, for each grid cell. The error for reconstructed quantities is given by the standard deviation of 50 independent realizations of the RF reconstructions.

Note: The "Deployments" section of this dataset is empty.  See the supplemental file "Cruise List" cruise_list.csv which contains the cruise identifier in EcoPart (EcoTaxa), year (2008-2020), and number of profiles conducted during the cruise.  More information about these cruises can be found at EcoTaxa (http://ecotaxa.obs-vlfr.fr).