 |
| Titel |
A global neural network-based parameterization of biogeochemical water mass properties and processes based on GLODAP data |
| VerfasserIn |
Henry C. Bittig, Raphaëlle Sauzède, Hervé Claustre, Orens Pasqueron de Fommervault, Jean-Pierre Gattuso, Louis Legendre, Ken Johnson |
| Konferenz |
EGU General Assembly 2017
|
| Medientyp |
Artikel
|
| Sprache |
en
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 19 (2017) |
| Datensatznummer |
250145443
|
| Publikation (Nr.) |
 EGU/EGU2017-9385.pdf |
|
|
|
|
|
| Zusammenfassung |
| Global data collections like GLODAP are an extensive source of biogeochemical and
hydrological data. However, data are irregularly distributed in space and time with varying
parameter-coverage. This poses a challenge to data analysis of, e.g., the global
distribution of stoichiometric ratios or temporal trends. Here we utilize a neural
network-based approach called CANYON to estimate carbonate system parameters
(CT, AT, pH, and pCO2) and nitrate, phosphate, and silicate concentrations from
commonly measured quantities (P, T, S, O2, location, and date). CANYON was
derived using GLODAPv2 data but can be applied to any set of input quantities (e.g.,
observations from autonomous platforms like Biogeochemical-Argo floats with
accurate O2 measurements). In essence, CANYON provides a mapping of water
mass properties and biogeochemical relations for those parameters based on the
multidecadal, global observations collected in GLODAPv2. It can thus provide
biogeochemical context and fill observational gaps, e.g., where nutrient or carbonate
system measurements are unavailable. As an example, float-based surface CTD-O2
observations together with the CANYON parameterization are used to obtain surface
pCO2 estimates in the Southern Ocean, complementing sparse surface underway
pCO2 data collected in SOCAT. Moreover, it can shed light on global variations of,
e.g., Redfield ratios of nitrate, phosphate, oxygen, and carbon. We believe that this
parametrization provides a useful alternative to scattered data points or a mapped climatology
to facilitate utilization and exploitation of the unique GLODAP data collection. |
|
|
|
|
|
|