DAE diagnostics

DAE Diagnostics

DAE (Data assimilation and Ensembles) diagnostics provide a complementary data quality assessment to real time monitoring of satellite observations. These diagnostics provide information on the capacity of observing systems to reduce analysis and forecast errors (so-called ‘forecast sensitivity to observations’ (FSO) diagnostics) and allow users to compare observing systems and identify data that could be detrimental to the quality of NWP forecasts.

The NWP SAF offers web-based DFS (Degrees of Freedom of Signal) diagnostics which measure the relative impacts of different observation types, and the FSOi disanostic.

The FSOI (Forecast Sensitivity-based Observation Impact) diagnostic allows, thanks to the adjoint model technology, to estimate the contribution of the observations to the quality of the forecasts for which they have participated in the production of the initial conditions (via data assimilation). The diagnosis is calculated in the Météo-France global ARPEGE model by focusing on the 24-hour forecast. Here, the FSOi is defined according to several formulations of the adjoint model and of the function which is used to quantify the quality of the forecast (taking into account the humidity or not). In reality, this function is the amplitude of the forecast error and, as a result, the FSOi is a negatively oriented diagnosis: the more negative the contribution of an observation group, the more it contributes to reducing the forecast error and, therefore, to its quality. The FSOi diagnosis is a linear estimate that needs to be accompanied by additional tests (e.g. observation-denial experiments) for the rigorous assessement of the value of these observation groups.

Non-NWP SAF DAE diagnostics

JCSDA FSOI monitoring page contains:

  • 2015 Experiment
  • Data are available for EMC, GMAO, NRL, JMA, and Met Office and all cycles for the dry norm only, from 2014-Dec-01 through 2015-Feb-28.

  • Real-time Data
  • NRL data are available daily starting from 2018-Oct-09 for the dry norm and 00Z cycle only. There is a 4-day lag for the data to become available (e.g., data for Jan 1 will not be available until Jan 5).
    GMAO data are available daily starting from 2019-Feb-20 for the moist norm and 00Z cycle only. There is a 3-day lag for the data to become available (e.g., data for Jan 1 will not be available until Jan 4).
    To see the comparison plots with both GMAO and NRL, the “both” norm must be selected.
    Met Office will be available soon. Data are currently available for 2020 April, moist norm only, and cycle hours 00Z, 06Z, 12Z, and 18Z

  • MERRA2 Re-analysis
  • The MERRA2 Re-analysis data are available from 1980 to 2019 for the months of January and July, moist norm only, and 00Z cycle only.