User Tools

Site Tools


Export page to Open Document format

  Go to ⇒ [ PI ] - [ 6ka ] - [ 21ka ] - [ LM ]- [ 8.2ka ] - [ LI ] - [ Plio ]

8.2ka Experimental Design

Page in progress: this is not the final version!

Use the discussion panel at the end of the page for comments
Information about metadata and variables to be contributed can be found here: 8k-variables.xlsx

Boundary conditions

The following are ideal boundary conditions, but experiments that deviate from these boundary conditions may also be contributed if complete metadata are provided.

Summary of 8.2ka boundary conditions

Boundary conditions below are for 8.5 ka (i.e. before lake drainage)

PMIP3/CMIP5 Minimum solution
Orbital parameters [ ecc = 0.19199 ] - [ obl = 24.222° ] - [ peri-180° = 319.495° ]
Date of vernal equinox March 21 at noon
Trace gases [ CO2 = 260 ppm ] - [ CH4 = 660 ppb ] - [ N2O = 260 ppb ] - [ CFC = xxx ] - [ O3 = same as PI ]
Aerosols Same as PI
Solar constant As in PI
Vegetation prescribed or interactive as in CMIP5 PI
Ice sheets Remnant Laurentide ice-sheet (provided*) Same as PI
Topography and coastlines Modification of Hudson Bay coastlines (provided*) Same as PI

*For the ice-sheet reconstruction, see discussion on LGM wiki page

Please use the discussion panel to comment this table!

Freshwater forcing

Two different types of freshwater forcing: the background melt-flux from the Laurentide Icesheet, and the Lake Agassiz drainage. The former supposedly prevented Labrador Sea deep convection (see Hillaire-Marcel et al. 2006 GRL)

Proposed Background Early Holocene melt-water forcing (mandatory)

Flux 0.05 Sv over 500 years
Location Either part or all of the Labrador Sea (up to the individual modeling groups, as long as it is not added outside the Labrador Sea)

Proposed Lake Agassiz drainage (mandatory):

Flux 2.5 Sv over 1 year
Location Either part or all of the Labrador Sea (up to the individual modeling groups, as long as it is not added outside the Labrador Sea)
Ensemble this experiment should be completed at least three times, beginning at different years in the control simulation
Duration experiment should be integrated until the AMOC recovers, except in the case of unrealistically long (i.e., multicentury) recoveries


The vegetation should be treated as in the CMIP5 PI experiment. The reason is that in CMIP5 we test the version of the model used for future climate projections. Since OA and ESM models will be considered, depending on the model used the vegetation will be

  • prescribed to PI (which means both vegetation types and LAI are prescribed )
  • Prescribed to PI with interactive LAI (models with interactive carbone cycle, but no vegetation dynamics)
  • Computed by the model (models with dynamic vegetation)

For Earth System Models with interactive carbon cycle

The simulations are forced by CO2 concentrations. Please use the same protocol as in CMIP5 to store the carbon fluxes and the variables needed for PCMIP (see list here)


Note that insolation should follows PMIP requirements. Please check it carefully using the following tables (6ka BP insolation tables)

Initial conditions

PMIP3 Alternative solution
Initial conditions Branch off PI after adjustement Same as in PI
Model spin up Same as in PI

[ PMIP3 Wiki Home ] - [ Help! ] - [ Wiki syntax ]   - [ Top ]


Anders Carlson, UW-Madison, Dept. of Geoscience & CCR, 2010/04/12 13:52

Overall, I think the outlined plan looks quite good; a few comments 1) I would up the pre-flood flux to 0.1 Sv reflecting the decay of the Laurentide up to the opening of Hudson Bay. The majority of this freshwater would be routed out the St. Lawrence with the remainder predominately discharged through Hudson Bay (only a little to the Arctic) 2) Which ice sheet configurations are going to be tested? I haven't seen the deglacial simulations of 6G, but 5G had too much ice in Keewatin and too little over Quebec up to and after the opening of Hudson Bay. Quebec deglaciated later than Keewatin and the 5G configuration would predict the opposite. 3) Are drainage patterns in the model going to be changed to reflect ice configurations and thus P-E's routing predicted by the model? 4) Are sensitivity experiments going to be performed to test the effects of the flood, vs. the routing of P-E vs. changes in Laurentide topography? 5) Is the attendant routing of Laurentide melt going to be included as well? 6) I think an important thing to get right is the duration of the event in addition to the magnitude

Andreas Born, Climate and Environmental Physics, U of Bern, 2010/12/17 12:11

I like the set-up. 1 question, 1 comment:

1) What is the flux estimate (2.5 Sv over 1 year) based on?

2) I think the location of the freshwater flux is critical. If the freshwater is added to just part of the Labrador Sea, e.g. at the coast which would arguably be the most realistic choice, it will stay at the surface and be exported out of the basin by strong boundary currents. In this case, the freshwater does not reach the convection region (directly). This is obviously different if the freshwater is spread out over the entire basin and the effect on deep convection will radically differ. Actually, the freshwater did very likely not reach the central Labrador Sea (Keigwin et al, 2005, Hillaire-Marcel et al, 2007) and Labrador Sea convection even intensified shortly after the lake drainage according to proxy data.

Anders Carlson, UW-Madison, 2010/12/18 01:57

I think the 2.5 Sv flux is from Teller et al (2002) and Clarke et al. (2004). As for the water pathway, we published records from Hudson Strait last year in GRL that showed the accompanying routing event during/following the flood (Carlson et al., 2009, GRL). We have new records from Cartwright and Eirik Drift that show light d18Osw once the temp effect on d18Ocalcite is removed by Mg/Ca temp estimates. Combining these with Keigwin et al.'s record from the St. Lawrence (a reduction in runoff when we show an increase out the Hudson Strait) and the Came et al./Thornalley et al. records in the NE Atlantic, you can trace the freshwater signal from Hudson Strait down the west side of the Lab Sea, with it then transported in the North Atlantic current over to Iceland. Some of this water peals off and recirculates in the Lab Sea to explain the light d18Osw at Eirik Drift. So I think there's pretty good coverage now on where the water went

Andreas Born, Climate and Environmental Physics, U of Bern, 2010/12/19 15:10

This is what I mean, thank you for adding the references! We know from proxy data which path the water took and it did not spread out across the entire Labrador Sea. If you do spread it out and put it on top of a deep convection region, some models will remove some of the buoyant freshwater anomaly from the surface into the deep ocean and the result on dynamics will be very different.

Avoiding adding freshwater to the convection region, we found that up to a quarter of its volume is advected into the Nordic Seas where it impacts deep convection the most (Born and Levermann, 2010, G-Cubed). This perturbation is then communicated back to the North Atlantic through the Greenland Scotland ridge overflows and eventually strengthens Labrador Sea convection–quite contrary to what happens if the water is added directly to the Labrador Sea convection site.

Generally, I think the experimental design should let the models the freedom to simulate the advection of the freshwater rather than to prescribe. Most if not all models involved in PMIP3 include 3D OGCMs that should be able to reproduce the advection path reasonably well.

pmip3/design/8k2/index.txt · Last modified: 2012/03/07 21:59 by morrill