A sensitivity study of the sea ice simulation in the global coupled climate model, HadGEM3 Jamie Rae, Helene Hewitt, Ann Keen, Jeff Ridley, John Edwards, and Chris Harris 5.0 2. Model 3. What parameters can/should we adjust? 1. Motivation Coupler Test sensitivity to downward SW radiation and surface effects by adjusting albedo. αb: Bare ice albedo. αc, αm: Cold snow and melting snow albedos. Parameterisations for: αc to αm transition (function of T). Effect of meltponds on αb (function of T). Scattering effects in zero-layer model (Semtner, 1976). Test sensitivity to ocean-ice heat transfer by adjusting transfer coefficient, cH (McPhee, 1992). Ice salinity, S, which affects brine rejection, ocean stability, and advection of oceanic heat to ice base. Ice dynamics/ridging: test sensitivity to ridging parameter μrdg (Hunke, 2010) Snow and ice thermodynamics: test sensitivity to thermal conductivities of ice (κi) and snow (κs). Also tested sensitivity to resolutions of ocean-ice and atmosphere-surface models. Arctic summer sea ice extent and thickness too low in previous model version. Desire to improve this by adjusting sea ice parameters within the range of observational uncertainty. Desire to inform future model enhancements through studying sensitivity to parameter perturbations. Extent seasonal cycle Volume seasonal cycle CICE The Los Alamos Sea Ice Model Surface Ocean Atmosphere Sea ice HadISST Sept ice concentration Model Sept ice concentration Model Model HadISST ± 20% PIOMAS HadGEM3: Fully-coupled global atmosphere-ocean-ice model. Sea ice component is Los Alamos CICE model. CICE currently run in zero-layer configuration. Atmosphere-surface and ocean-ice components run at different resolutions. (Model output plotted here is for the mean of years 16-30 of a 30-year coupled HadGEM3 run with constant year-2000 greenhouse gas concentrations and aerosol emissions. HadISST and PIOMAS data are means for the period 1995-2004.) 4. Model experiments (all parameters defined in box 3) Parameters perturbed separately and in combination. Selected experiments: Control experiment: αb = 0.61 αc = 0.80 αm = 0.65 : κi = 2.09 Wm-1K-1 κs = 0.31 Wm-1K-1 Ocean-ice res: ORCA1 (~1°) Atmosphere res: N96 (~130 km) αb = 0.58 αc = 0.85; αm = 0.72 cH = 0.003 S = 8 ppt All experiments ran for 30 years with year-2000 greenhouse gas concentrations μrdg = 3 m1/2 κi = 2.63 Wm-1s-1 κi = 2.63 Wm-1s-1; κs = 0.50 Wm-1s-1 cH = 0.006 S = 4 ppt μrdg = 4 m1/2 Colour-highlighted experiments are discussed in box 6. 6. Sensitivity to sea ice parameters 5. Sensitivity to ocean-ice model resolution Arctic extent seasonal cycle Arctic volume seasonal cycle March ice concentration All experiments performed at ORCA1 (~1°) ocean-ice model resolution. Arctic extent seasonal cycle Arctic volume seasonal cycle HadISST ORCA1 (~1°) ORCA0.25 (~0.25°) PIOMAS Control Change αc, αm Change κi only Change κi, κs Control Change αc, αm Change κi only Change κi, κs Sept Arctic ice concentration HadISST Control Change αc, αm Change κi, κs HadISST±20% PIOMAS HadISST±20% ORCA1 (~1°) ORCA0.25 (~0.25°) ORCA1 (~1°) ORCA0.25 (~0.25°) Antarctic extent seasonal cycle Antarctic extent seasonal cycle Antarctic volume seasonal cycle Antarctic extent seasonal cycle Antarctic volume seasonal cycle Increased ocean-ice model resolution from ORCA1 (~1°) to ORCA025 (~0.25°) Leads to disappearance of winter ice in Labrador Sea, and so to closer agreement with HadISST Linked to increased SSTs. Control Change αc, αm Change κi, κs Arctic sea ice found to be most sensitive to snow albedo (αc, αm – see box 3 above) and ice and snow thermal conductivities (κi, κs). Sensitivity to snow albedo is linked to top melt in early summer before snow layer has melted completely. Sensitivity to conductivity is linked to ocean-atmosphere heat flux through ice in autumn, and basal ice growth. Antarctic sea ice more sensitive to changes in atmosphere and ocean, but is sensitive to ice salinity (not shown here). Sept Arctic ice thickness In the Antarctic, increased resolution leads to exacerbation of existing warm bias in ocean. This causes a large negative bias in sea ice extent and volume. Some model development work is focussed on solving this Southern Ocean warm bias. Annual mean SSTs: ORCA025 minus ORCA1. Note increase in Labrador Sea. March Arctic ice thickness metres Last 15 years of 30-year simulation are used in all analysis. 7. Effect of combining the perturbations and sensitivity to atmosphere resolution Ran at two atmosphere model resolutions: N96 (~130 km) N216 (~60 km) Combined-N96, compared to Control-N96: Summer (Sept) Arctic ice concentration has increased, although now larger than HadISST. Arctic ice now thicker in all seasons (but volume now greater than PIOMAS). Antarctic ice extent too small because of Southern Ocean warm bias (see box 5). N216 (~60 km), compared to N96 (~130 km): Arctic ice thinner and less extensive because of increased poleward heat transport. Antarctic ice shows similar bias, because of clouds / SW radiation effects. 8. Summary, conclusions, and future work Performed simulations with a coupled atmosphere-ocean-ice model. Perturbed various sea ice parameters within the range of observational uncertainty. Also studied sensitivity of sea ice to changes in atmosphere and ocean. Arctic sea ice most sensitive to snow albedo, and to ice and snow thermal conductivities. Antarctic sea ice most sensitive to salinity, and to changes in atmosphere and ocean. Both Arctic and Antarctic ice are sensitive to ocean-ice and atmosphere model resolutions. Paper (Rae et al., 2013) submitted to Ocean Modelling. Parameter changes reflect future model enhancements (e.g., improved albedo scheme, impact of meltponds on albedo, multi-layer ice model). Perturbed the following parameters in combination: αc: 0.80  0.85 αm: 0.65  0.72 Scattering fraction: 0.068  0.12 S: 4  8 ppt μrdg: 4  3 m1/2 κi: 2.09  2.63 Wm-1K-1 κs: 0.31  0.50 Wm-1K-1 Changes to atmospheric physics and dynamics, and to ocean model, also included. ORCA-025 (0.25°) ocean-ice resolution. Arctic extent seasonal cycle Arctic extent seasonal cycle Arctic volume seasonal cycle Arctic volume seasonal cycle HadISST±20% PIOMAS Control-N96 Combined-N96 Combined-N216 Control-N96 Combined-N96 Combined-N216 Antarctic extent seasonal cycle Antarctic volume seasonal cycle HadISST Control-N96 Combined-N96 Combined-N216 Control-N96 Combined-N96 Combined-N216 Sept Arctic ice concentration Sept Arctic ice thickness References Hunke, E.C., 2010, Thickness sensitivities in the CICE sea ice model, Ocean Modelling, 34, 137-149. McPhee, M.G., 1992, Turbulent heat flux in the upper ocean under the sea ice, J.Geophys.Res., 97 (C4), 5365-5379. Semtner, A.J., 1976, A model for the thermodynamic growth of sea ice in numerical investigations of climate, J.Phys.Oceanog., 6, 379-389. March Arctic ice concentration March Arctic ice thickness metres Met Office Hadley Centre, FitzRoy Road, Exeter, Devon, EX1 3PB United Kingdom Tel: +44 1392 884442 Fax: +44 1392 885681 Email: jamie.rae@metoffice.gov.uk © Crown copyright 07/0XXX Met Office and the Met Office logo are registered trademarks