Table 8.
Obs_diag_output.nc files contain variables with a dimension (“copy”), which organizes the statistics of the contents of the observation sequence files (obs_seq_final). They are ordered in that file according to the CopyMetaData variable. Bold text highlights statistics which are displayed in the Diags_NTrS_YYYY-MM.tgz files. Further details can be found in the DART software package; /your_DART/assimilation_code/programs/obs_diag/threed_sphere/obs_diag.html.
| Variable | Description | Use(s) |
|---|---|---|
| Nposs | Number of observations available | Helps assess the assimilation health |
| Nused | Number of observations assimilated | Helps assess the assimilation health |
| NbigQC | Number of obs rejected due to flag from observation provider | |
| NbadIZ | Number of excessive innovation Z scores | |
| NbadUV | Number of unmatched U/V wind pairs | Can reveal problems with the input observation files |
| NbadLV | Number of obs excluded above the top, below the bottom | Provides context for Nused and Nposs |
| rmse | Root mean square error | Shows the noisiness of the difference of the ensemble mean estimates from the observations. Provides context for other statistics. |
| bias | Average difference of the mean estimates from the observations | Reveals model and/or observation type biases |
| spread | Combined spreads from ensemble estimates of all observations | Focus on ensembles, or remove from total spread to see observation error |
| totalspread | Combined ensemble spread and observation error of all observations | Provides context for RMSE, bias, and assimilation health |
| NbadDARTQC | Number of observations not assimilated due to DART criteria | Total observations that failed DART’s quality control checks |
| observation | Average value of the actual observations in the region + time span | Compare against the ensemble mean |
| ens_mean | Average value of the ensemble mean estimates of the observations | |
| N_trusted | Number of observations designated as ’trusted’ | These will be assimilated regardless of how different they are from the ensemble estimate. |
| N_DARTqc_0 | Number of observations given DART QC 0 | This number were assimilated |
| N_DARTqc_1 | Number given DART QC 1 | Only evaluated, not assimilated, following namelist instructions |
| N_DARTqc_2 | Number given DART QC 2 | Posterior forward operator failed after prior QC = 1 |
| N_DARTqc_3 | Number given DART QC 3 | Posterior forward operator failed after prior QC = 4 |
| N_DARTqc_4 | Number given DART QC 4 | Model estimates of observations could not be calculated |
| N_DARTqc_5 | Number given DART QC 5 | DART’s namelists prevented assimilation |
| N_DARTqc_6 | Number given DART QC 6 | Prior forward operator failed |
| N_DARTqc_7 | Number given DART QC 7 | Observations were too far from the ensemble means |
| N_DARTqc_8 | Number given DART QC 8 | DART failed to convert the observations’ vertical coordinates |