Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...


HindcastForecast
Atmosphere initialization
ERA5ECCC GEPS
Atmosphere IC perturbationsRandom isotropic perturbationsEnsemble Kalman Filter

Land Initialization

Off-line ISBA-based Surface Prediction System (SPSv6.2)

forced by 1-hourly lowest-level ERA5 fields

ECCC GDPS output adjusted to hindcast climatology
Land IC perturbationsNoneNone
Soil moisture initialization

Off-line ISBA-based Surface Prediction System (SPSv6.2)

forced by 1-hourly lowest-level ERA5 fields

ECCC GDPS output adjusted to hindcast climatology
Snow initialization

Off-line ISBA-based Surface Prediction System (SPSv6.2)

forced by 1-hourly lowest-level ERA5 fields

ECCC GDPS output with prescribed climatological snow over Tibetan Plateau
Unperturbed control forecast?NoneNone

Data assimilation method for control analysis: 

Horizontal and vertical resolution of perturbations:  

Perturbations in +/- pairs: 

Detailed documentation:

Gauthier, P., et al. 1999: Background-error statistics modelling in a 3D variational data assimilation scheme: Estimation and impact on the analyses. Proc. ECMWF Workshop on Diagnosis of Data Assimilation Systems, Reading, United Kingdom, ECMWF, 131–145, 

...

6. Forecast system and hindcasts

Forecast frequency12-month forecast is produced on the last day of each month
Forecast ensemble size20 (ensemble members 1-10 initialized at 00Z on last day of month, 11-20 at 00Z on 5th to last day of month)
Hindcast years1980-2023
Hindcast ensemble size20 (ensemble members 1-10 initialized at 00Z on first day of month, 11-20 at 00Z on 5th to last day of previous month)
On-the-fly or static hindcast set?static
Calibration (bias correction) period1991-2020

7. Other relevant information

...

Vertical interpolation pressure levels: 

  • For temp/phi near the surface, log-linear interpolation (linear in ln(eta)) with the lapse rate of 6.5e-3 deg/m
  • For remaining variables near the surface and all variables near the top, constant extrapolation, i.e. keep same value from the closest model level

8. Where to find more information

...