GEPS

1. Overview of GEPS v8.0.0

  • Implemented on: June 11, 2024 (1200 UTC)

  • Developed by: Meteorological Service of Canada (MSC)

  • Forecast Type: Probabilistic medium-range forecasts (16-day outlook)

  • Grid Resolution: ~25 km Yin-Yang grid

  • Vertical Levels: 84 hybrid levels, model lid at 0.1 hPa

  • Forecast Frequency:

    • Twice daily (00 & 12 UTC) for 16-day forecasts

    • Four times daily (00, 06, 12, 18 UTC) for short-range 72-hour forecasts

  • Key Updates in Version 8.0.0:

    • New data assimilation strategy (LETKF)

    • Improved stochastic parameterization (SPP & SKEB)

    • Updated ocean-atmosphere coupling (NEMO & CICE 6.2.0)

    • Extended reforecast system (39-day monthly forecasts)

2. Data Assimilation Component

2.1 Assimilation Methodology

  • Model Version: Global Environmental Multiscale (GEM) v5.2.1

  • Assimilation Algorithm: Local Ensemble Transform Kalman Filter (LETKF)

    • Uses 256-member ensemble for trial fields.

    • Assimilation window: 6 hours (±3 hours around 00, 06, 12, 18 UTC).

    • Recentered around the 4DEnVar analysis (uses GDPS ensemble mean as first guess).

    • LETKF membership randomized every cycle to improve statistical spread.

2.2 Observational Data Sources

  • Satellite Data:

    • AMSU-A, MHS, ATMS, SSMIS, MWHS-2 (microwave radiance)

    • IASI, CrIS (infrared hyperspectral sounders)

    • GPS Radio Occultation (GPS-RO)

  • Ground-Based Observations:

    • Radiosondes (TEMP, PILOT)

    • Surface Weather Reports (SYNOP, METAR, BUOY, SHIP)

  • Aircraft-Based Observations:

    • AMDAR, TAMDAR

  • Ocean Observations:

    • Scatterometer Winds

    • ATOVS Level 1b Data (RARS & GLOBAL sources)

2.3 Key Improvements in Data Assimilation

  • RTTOV-13 (New radiative transfer model)

  • Correction to water saturation functions

  • Upgraded hyperspectral infrared QC (for albedo variations)

  • Updated sea-ice snow depth & thickness sources

  • Background check & bias correction embedded within GEPS

3. Forecast Component

3.1 Model and Grid Setup

  • Core Model: GEM v5.2.3 (hydrostatic primitive equations)

  • Coupled Ocean-Atmosphere System:

    • GEM coupled with NEMO (v3.6 ocean model)

    • Sea ice model: CICE v6.2.0

  • Numerical Grid:

    • Yin-Yang grid (~25 km uniform resolution)

    • 84 hybrid vertical levels, model lid at 0.1 hPa

  • Time Integration: Semi-Lagrangian (900s timestep)

  • Processing Time:

    • 16-day forecasts computed in ~1 hour 30 minutes (16x80x21 cores)

3.2 Prognostic Variables

  • Core Atmospheric Variables:

    • Temperature, wind components, specific humidity, surface pressure

    • Liquid water content, turbulent kinetic energy (TKE)

  • Derived Forecast Fields:

    • Mean Sea Level Pressure (MSLP)

    • Relative Humidity, Geopotential Height, Dew Point

    • Precipitation Rate, Cloud Amount, Boundary Layer Height

  • Geophysical Variables:

    • Surface & deep soil temperature & moisture

    • Snow depth, snow albedo, sea-ice fraction, sea-ice thickness

    • Sea surface temperature (initially from GIOPS, later from coupled ocean model)

3.3 Key Model Enhancements in Version 8.0.0

  • Delta-Eddington scheme added to CICE 6.2.0 for radiative transfer

  • New GEM-NEMO coupling weights for better ocean-atmosphere interaction

  • Revised turbulence & boundary layer schemes for improved forecast accuracy

4. Ensemble Perturbations & Uncertainty Representation

GEPS uses ensemble spread techniques to quantify forecast uncertainty.

  • Control Member (0): Initialized with ensemble mean LETKF analysis.

  • Ensemble Members (1-20): Initialized from LETKF recentered global analysis.

  • Perturbation Methods:

    • SPP (Stochastic Parameter Perturbation)

    • SKEB (Stochastic Kinetic Energy Backscatter)

  • SPP & SKEB Enhancements in v8.0.0:

    • Better Markovian perturbation fields for turbulence & convection

    • Reduced over-dispersion in tropics by refining SPP element spp_adv_rhsint

    • SKEB backscatter fraction adjusted from 1.0 to 0.7 for physical realism

5. Reforecast System

  • Used for forecast calibration & statistical post-processing

  • Now includes a 5th forecast week (39-day forecast)

  • Runs twice weekly (Monday & Thursday, 00Z)

  • Four-member ensemble initialized from ERA5 reanalysis

  • Ocean & sea-ice fields from ORAS5 & CMC GIOPS analysis

  • Land surface initialization from offline Surface Prediction System (SPS)

6. Applications & Use Cases

GEPS-8.0.0 provides improved probabilistic forecasts for diverse weather-sensitive sectors.

  • Severe Weather Prediction: Hurricanes, storms, heatwaves, floods.

  • Aviation & Transportation: Improved probabilistic flight route forecasts.

  • Energy & Utilities: Supports renewable energy load balancing.

  • Disaster Management: Early warnings for high-impact weather events.

  • Agriculture & Water Resource Management: Drought risk assessment, frost warnings.

7. Summary & Key Advancements in GEPS v8.0.0

  • Higher grid resolution (~25 km) for better spatial detail

  • More accurate initial conditions via improved LETKF assimilation

  • Coupled ocean-atmosphere system (NEMO & CICE 6.2.0)

  • Advanced ensemble perturbation schemes (SPP & SKEB)

  • Extended reforecast system for improved calibration

8. References & Further Reading

Conclusion

The Global Ensemble Prediction System (GEPS) v8.0.0 is a cutting-edge medium- and long-range probabilistic forecasting tool, significantly improving uncertainty quantification for weather predictions worldwide. With enhanced data assimilation, better stochastic physics, and an expanded reforecast dataset, GEPS offers unprecedented skill in ensemble-based forecasting, aiding disaster preparedness, climate resilience, and operational planning.

PreviousNumerical: EnsembleNextREPS

Last updated

Was this helpful?