Near Real Time Validation

Last validation statistics update:

Surface Chlorophyll

Multi satellite observations (MODIS and Sentinel-3) versus (minus) MedBFM model simulation. Statistics of bias and root mean square difference (RMS) are computed daily for the first 3 days of forecast and compared with the Essential Accuracy Number (EAN; QuiD).

BGC-Argo float: profiles

Last available BGC-Argo floats (blue) and model forecasts (1st day of forecast, green) and analysis (after assimilation, red) at the observations location and time. Availability of sensors can vary from float to float.

Observations for biomass of phytoplankton (PhytoC) and particulate organic carbon (POC) are retrieved from Bbp700 data using Bellacicco et al. (2019) relationship. Photosynthetic Active Radiation (PAR) is the 400-700nm downward radiation.

Dashboard Argo floats in near real-time

  • from Med-Argo (float position, type, sensor, transmission, cycle, mission days)
  • from Euro-Argo (float full technical properties and plots, data charts, metadata)

BGC-Argo float: skill metrics

Statistics of bias and root mean square difference (rmse) are computed weekly for the first day of forecast in selected large sub-basins and layers. Number of available floats in grey.

Time evolution of BGC-Argo float - 

Hovmoller diagrams of chlorophyll concentration (mg/m3) from float data (2nd row) and model outputs (3rd row) matched-up with float position (1st row) in the previous 24 months. From 4th to 7th rows, computation of selected skill indexes for model (solid line) and float data (dots). The skill indexes are respectively: surface (SURF) and 0-200m vertically averaged (INTG) chlorophyll, correlation (CORR), depth of the deep chlorophyll maximum (DCM, blue) and depth of the mixed layer bloom in winter (MWB, red). Trajectory of the BGC-Argo floats is reported in the upper panel, with deployment position (blue cross).

Hovmoller diagrams of nitrate concentration (mmol/m3) from float data (2nd row) and model outputs (3rd row) matched-up with float position (1st row) in the previous 24 months. From 4th to 7th rows, computation of selected skill indexes for model (solid line) and float data (dots). The skill indexes are respectively: nitrate concentration at surface (SURF) and 0-200 m vertically averaged concentration (INTG), correlation between profiles (CORR), depth of the nitracline computed as NITRCL1 and NITRCL2. Trajectory of the BGC-Argo floats is reported in the upper panel, with deployment position (blue cross).

Hovmoller diagrams of oxygen concentration from float data (2nd row) and model outputs (3rd row) matched-up with float position (1st row) in the previous 24 months. From 4th to 7th rows, computation of selected skill indexes for model (solid line) and float data (dots). The skill indexes are respectively: oxygen concentration at surface (SURF) and oxygen at saturation, 0-200 m vertically averaged concentration (INTG), correlation between profiles (CORR), depth of oxygen minimum zone (between 200-1000m) and depth of maximum “surface” oxygen (between 0-200m). BGC-Argo floats is reported in the upper panel, with deployment position (blue cross).

DA misfit satellite

Time series of: weekly misfit RMS with respect to satellite observations (top), weekly coverage observations (middle), daily number of available observations and of excluded ones (bottom).

DA misfit float

Time series of:

  • daily misfit RMS with respect to float observations at two/three layers (top), daily percentage of used observations (middle), number of assimilated profiles and of excluded ones (bottom)
  • daily misfit RMS/bias at different layers compared with QuID (CMEMS quality information document) estimated accuracy numbers (EAN; QuiD)
  • monthly misfit RMS/bias at different layers compared with QuID (CMEMS quality information document) estimated accuracy numbers (EAN; QuiD)

The Near Real Time skill performance page aims at delivering sustained on-line information of the quality of biogeochemical forecast products (i.e. to identify main biases and suspicious trends in the time series, and to establish accuracy values for simulated variables). Validation task is performed using both independent and semi-independent data (i.e., data lately assimilated).

Statistics are computed over the following subdivision of the Mediterranean Sea

 

 

Additional detail about CMEMS validation activities at the following link 

Near Real Time on line validation of physical variables at the following link