CanSIPS

1. Overview of CanSIPS v3.0

• Implemented on: June 11, 2024

• Developed by: Canadian Meteorological and Environmental Prediction Centre (CMEPC)

• Forecast Type: Seasonal to interannual (12-month) probabilistic forecasts

• Ensemble Size: 20 members (10 official-day members + 10 lagged members)

• Processing Time: ~2 hours 20 minutes per 12-month forecast

• Resolution: ~1° x 1° horizontal resolution

• Vertical Levels:

  • Atmosphere: 85 layers (GEM5.2-NEMO)

  • Ocean: 50 levels (NEMO 3.6)

• Model Components:

  • GEM5.2-NEMO (Global Environmental Multiscale Model coupled with NEMO ocean model)

  • CanESM5 (Canadian Earth System Model v5)

Key Updates in Version 3.0:

• CanCM4i replaced with CanESM5

• New ocean-sea ice model: NEMO 3.6 & CICE 6.0

• Refined boundary layer turbulence & gravity wave drag

• Upgraded stochastic physics (SPP) for uncertainty representation

• Higher vertical resolution for improved stratospheric representation

• Bias-corrected long-range forecasting framework

---

2. GEM5.2-NEMO Component (Atmosphere-Ocean Model)

The GEM5.2-NEMO component provides the core physics, atmospheric dynamics, and ocean coupling for CanSIPS.

2.1 Atmospheric Model (GEM5.2)

• Core Model: Global Environmental Multiscale Model (GEM)

• Equations: Hydrostatic primitive equations

• Numerical Scheme:

  • Finite difference discretization (Arakawa C-grid)

  • Semi-Lagrangian time stepping

  • Iterative solver for efficiency (FGMRES)

• Parameterization Upgrades:

  • Revised deep convection scheme (Kain & Fritsch)

  • Improved cloud microphysics for precipitation

  • Updated planetary boundary layer (TKE-based mixing)

• Grid Configuration: Yin-Yang Grid (~1° resolution)

• Vertical Levels: 85 hybrid sigma-pressure levels, top at 0.1 hPa

2.2 Ocean Model (NEMO 3.6)

• Ocean Dynamics:

  • Explicit leapfrog time integration (30-minute timestep)

  • 50 vertical layers, refined near the surface

• Sea Ice Model:

  • CICE 6.0 with Delta-Eddington radiation scheme

  • New roughness parameterization for ice-atmosphere fluxes

• Initial Conditions:

  • Ocean: GIOPS reanalysis

  • Sea Ice: Had2CIS dataset (blended Canadian Ice Service & HadISST)

2.3 Ensemble Configuration

• Initial Conditions:

  • Generated using the Global Ensemble Prediction System (GEPS)

  • 10 members from the main initialization date

  • 10 additional members from lagged initialization

• Uncertainty Representation:

  • SPP (Stochastic Parameter Perturbation) applied to physical processes

  • Ensemble spread enhanced by perturbing deep convection & radiation schemes

---

3. CanESM5 Component (Earth System Model)

CanSIPS v3.0 replaces CanCM4i with the more advanced CanESM5, improving long-range climate variability prediction.

3.1 Atmospheric Model

• Core Model: Canadian Atmospheric Model (CanAM5.1p1)

• Numerical Scheme: Spectral core with hybrid sigma-pressure vertical coordinate

• Resolution:

  • Horizontal: T63 truncation (~1.8°)

  • Vertical: 49 levels (top at 1 hPa)

• New Features:

  • Updated cloud microphysics & aerosol interactions

  • Improved land surface representation (CLASS & CTEM)

  • Prognostic bulk aerosol scheme with full sulfur cycle

  • Bias-corrected radiation & convective schemes

3.2 Ocean Model

• Core Model: CanNEMO (based on NEMO 3.4.1)

• Resolution: 1° x 1° horizontal grid, 45 vertical levels

• Ice Model: LIM2 (Louvain-la-Neuve Sea Ice Model)

• Initialization: Ocean and sea ice fields nudged to GIOPS

3.3 Land Surface Model

• Core Model: Canadian Land Surface Scheme (CLASS 3.6.2)

• Soil Representation: Three-layer soil profile (0.1m, 0.25m, 3.75m)

• Carbon Cycle: Dynamically coupled terrestrial ecosystem model (CTEM)

3.4 Coupling Strategy

• CanCPL coupler used for air-sea-land interactions

• Run-time bias correction for atmosphere-ocean interactions

---

4. Hindcast & Calibration System

4.1 Hindcast Dataset

To improve seasonal forecasting skill, CanSIPS v3.0 includes a 41-year hindcast (1980-2020).

• Reforecasting: 20-member ensemble run for each month

• Initial Conditions:

  • Atmosphere: ERA5 reanalysis

  • Ocean: ORAS5 reanalysis

  • Sea Ice: Had2CIS dataset

• Bias Correction:

  • Systematic errors removed via statistical correction

  • Historical calibration ensures improved seasonal forecasts

4.2 Forecast Post-Processing

• Calibrated ensemble mean used for anomaly predictions

• Probability-based seasonal forecast products generated

• Bias-adjusted predictions for better climate signal detection

---

5. Applications & Use Cases

CanSIPS v3.0 is designed for seasonal and interannual climate forecasting, supporting multiple climate-sensitive sectors.

• Drought Prediction & Water Resource Management

• Winter Severity Forecasting (Snow & Ice)

• Agricultural Risk Assessments (Growing Season Length, Frost Risk)

• Energy Demand Forecasting (Heating & Cooling Requirements)

• Wildfire Risk & Air Quality Predictions

• El Niño/La Niña & Tropical Cyclone Seasonal Forecasts

---

6. Summary of Key Advancements

Feature	CanSIPS v3.0 Update
Core Model	Replaced CanCM4i with CanESM5
Atmosphere	Higher vertical resolution (85 layers)
Ocean	Upgraded to NEMO 3.6 with improved flux exchanges
Sea Ice	New physics (CICE 6.0) for better ice representation
Land Surface	New tiling approach (land, ocean, sea ice, lakes)
Bias Correction	Run-time adjustment for atmosphere & ocean fields
Ensemble Size	Increased to 20 members for better uncertainty quantification
Hindcast Dataset	41-year reanalysis dataset for seasonal calibration

---

7. References & Further Reading

• CanSIPS v3.0 Technical Note: Technical Note

• CMC Operational System Changes: Changelog

• Ocean Model (NEMO) Overview: NEMO Ocean

---

Conclusion

CanSIPS v3.0 represents a major advancement in seasonal and interannual climate prediction. By enhancing model physics, improving bias correction, and utilizing state-of-the-art ensemble forecasting, it provides more accurate long-range forecasts, supporting disaster resilience, climate adaptation, and decision-making for government agencies, industries, and researchers.