********************************************* PALEOCLIMATE MODELING INTERCOMPARISON PROJECT ********************************************* Newsletter N. 3 --------------- 23 March 1994 Dear Participant, The present newsletter updates newsletter 2 by giving the final recommendations for the 21 kyr BP experiments. We remind you that the 21 kyr BP experiment should, if possible, be run with computed SSTs and sea ice, in the same configuration that the models would be run for CO2 experiments. For those who may not have access to a coupled atmosphere-mixed layer ocean model, an optional experiment may be performed for 21 kyr BP with prescribed SSTs and sea ice according to CLIMAP (1981). In the present newsletter we give comments on: -A- the new ice sheet dataset provided by D. Peltier, to be used in all LGM PMIP experiments -B- vegetation, soil type, surface albedo, etc. -C- sea ice for the 21 kyr BP experiment with computed SSTs -D- SSTs and sea ice for the optional 21 kyr BP experiment with prescribed SST -E- choice of a calendar for the analyses of model outputs (more relevant to 6 kyr BP) CAUTION : ******** For the 21 kyr BP experiments, YOU ALSO NEED to READ the NEWSLETTER 2, which contains informations not given again in the present newsletter (such as informations about orbital parameters, greenhouse gaz concentrations). All the previous newsletters are available at NGDC through anonymous ftp (see below for the ice sheets) in /paleo/pmip/newsletters_etc ********************* -A- ICE SHEET DATASET ********************* For both 21 kyr BP experiments, we recommend that the new ice sheet reconstructions provided by Peltier et al. be used. Compared to the previous CLIMAP ('81) reconstruction, the new one exhibits a similar extent but a much lower elevation, by as much as 1000 to 1500 meters. If melted, the new ice sheets would add 105 meters of water to the ocean. The change in sea level is thus reduced compared to CLIMAP and requires to modify the land/sea mask for the 21 kyr BP experiments. In both LGM experiments, we recommend you use: - the CHANGE in surface elevation (over ice sheets and land due to sea level lowering) in order that all models will be perturbed (forced) in the same way. This change in elevation would be added by each modeling group to the topography used in their control run. - the 21 kyr BP land-sea and land-ice masks. You may obtain the ice, land/sea mask, and topography in either of the following ways: - Provide Karl Taylor at PCMDI (NEW! Please check the PMIP 'Contacts' web page) with the coordinates of your model, and he will "interpolate" the Peltier 21 ka BP and present-day data onto your grid (in ascii form) and make it available through anonymous ftp. - Obtain the original Peltier data from NGDC through anonymous ftp and interpolate it yourself. *1* To obtain data on your model's grid: ------------------------------------- - Send Karl the coordinates of your grid-cells. What is needed is the location of the boundaries between grid cells, so if you provide the coordinates of the centers of the grid cells, indicate what kind of grid you use (evenly-spaced, gaussian, etc.) - The Peltier data will be interpolated, following the procedure summarized in *3* and *4* below, and 7 ascii (text) files will be placed on anonymous ftp at PCMDI: lndsea0, lndsea21, ice0, ice21, top0, top21, top21m0. The "0" suffix refers to present day, while "21" refers to the LGM. The land/sea mask, glacial ice mask, and surface elevation data will be stored in lndsea, ice, and top, respectively. The top21m0 will contain the difference in topography between 21 ka BP and present. - Karl will send you an e-mail message when the files are ready (within 2 weeks of your request), indicating how to retrieve them and read them. *2* To obtain Peltier's 1x1 degree data: --------------------------------------- The datasets can be obtained from NGDC, through anonymous ftp (NGDC1.NGDC.NOAA.GOV or 192.149.148.109). Topography and ice cover files are provided on a 1x1 regular grid at 1000 year time intervals since the LGM. ftp ngdc1.ngdc.noaa.gov name : anonymous password : your email address when you get the ftp prompt : cd paleo cd pmip cd peltier_ice In this directory you find : readme.pelice ---> a file which explains how to use and check the data TOP.tar ---> a tar file which contains informations on topography at every 1000 year time interval : get the top.21 and top.0 files ICE.tar ---> a tar file which contains informations on ice cover at every 1000 year time interval : get only the is_ice.21 file TOP.21.ascii ---> is an ascii file which can be used to check interpretation of the binary files *3* Recommended procedure for initial processing of 1x1 degree data: --------------------------------------------------------------- a) read the top.21 and top.0 files following the readme.pelice instructions b) compare data read from the top.21 file to data in the TOP.21.ascii file to check that you read the data correctly. c) obtain the land-glacial ice data (i.e., the ice sheet extent) from the is_ice.21 file. d) create a land/sea mask as follows: If the elevation of a grid cell is less than or equal to 0.0 AND if no ice is present, then consider that grid cell to be ocean. Otherwise, consider the grid cell to be land. e) revise the topography data by setting to 0.0 (sea level) the topography of ocean grid points (see 3d). *4* Recommended procedure for regridding data: ----------------------------------------- This procedure assumes that model grid cells do not have fractional glacial ice coverage. a) land/sea mask: Use a simple area-weighted average of the 1x1 degree land-sea mask data (0=ocean, 1=land). Then for the target grid, consider grid cells with values less than 0.5 to be ocean and consider grid cells with values greater than 0.5 to be land. b) For each target grid cell calculate ICE, a simple area-weighted average of the 1x1 degree ice-cover data (1=ice, 0=no ice), and calculate LAND, a simple area-weighted average of the 1x1 degree land/sea mask data computed in 4a) (0=ocean, 1=land). Grid cells where LAND > 0.5 and ICE/LAND > 0.5 should be considered ice-covered. All other grid cells (including all "ocean" grid cells) should be considered free of glacial ice. c) topography: Use the 1x1 degree topography data created in 3e) above and apply an area-weighted averaging scheme to obtain values for the 21 ka BP elevation and present day elevation on your model grid. Set the topography to 0.0 for any grid cell that is "ocean". d) difference in topography: compute the difference between the 21 ka BP and present day elevation (as obtained in 4c), and then add this difference to the topography of your "control" (present-day) simulation. ********************** -B- SURFACE PROPERTIES ********************** For snow-free and ice-free surfaces the surface properties (e.g., vegetation, soil, surface albedo) should be the same as for the control run (i.e., present day). However, with the change in land/sea distribution, we will need to specify land-surface properties for the continental margins that have emerged at 21 ka BP. In PMIP Newsletter 2 we recommended that you use the zonal mean value, averaged over all the snow and ice-free land-surface grid points located at the same latitude. In some models this procedure would make sense for surface albedo, but if the model requires a specification of vegetation type or soil type, numerical averaging might not be appropriate. An alternative is to prescribe surface properties to be the same as properties of the grid cell's nearest neighbor. Normally the nearest neighbor in the east-west direction would take precedence over the nearest neighbor in the north-south direction. ***************************************** -C- 21 KYR BP EXPERIMENT - COMPUTED SSTs ***************************************** The 21 kyr BP experiments with computed SSTs and sea ice should be performed in the same way as the CO2 experiments. Typically this means that a coupled atmosphere-mixed layer ocean model will be used with prescribed present day ocean heat flux. However, simulations of the last glacial maximum raise some specific questions: - how to account for the change of land/sea mask resulting from sea level lowering - how to treat sea ice, in order to allow the model to produce an advance of sea ice at the LGM even in areas with a strong present day convergence of ocean heat flux These questions have been discussed by several participants. - In newsletter 2, Tony Broccoli suggested adjusting the ocean heat flux to account for the change of land/sea mask such that in the zonal average the ocean heat flux would remain the same as today. - However, for J. Kutzbach and B. Gallimore, this procedure enhances the ocean heat flux under sea ice at high latitudes and may inhibit sea ice extent at LGM. For some models the problem of sea ice is even more general: even without changes in the extent of the contents, the ocean heat flux for present day ice-free grid points might be too strong and inhibit sea ice formation at the LGM, thus biasing the simulation of the LGM and introducing strong differences between models. - John Mitchell advocates no change in the continental borders and no changes in ocean heat flux (i.e., keep it "simple"). - Dave Pollard explained how this problem is dealt with in the GENESIS paleoclimate simulations: they assume no change in the heat transport per unit length of latitude circle (effectively increasing the total ocean transport if a basin widens); they also modify interactively the ocean heat flux under sea ice allowing sea ice to advance; in both cases they add or subtract a constant to restore the global integral of heat transport to zero. A copy of the above letters will be put at NGDC (available by ftp) by Robin Webb. As a conclusion for all these discussions, we recognize that: - the treatment of ocean heat flux under sea ice is an integral part of each model's mixed layer ocean and has to be considered in that context. What is appropriate for model may not be appropriate for another. Therefore we do not recommend any specific way to adjust the imposed ocean heat fluxes (there is certainly no unique solution). The important thing is that THE PROCEDURE USED FOR THE LGM SHOULD BE SIMILAR TO THE ONE USED FOR 2xCO2 EXPERIMENTS. - the problem of change in land/sea mask is completely specific to paleoclimate simulations. The first suggestion is to use a procedure consistent with the treatment of heat fluxes in the model. If several choices are possible, then follow the procedure suggested by Tony Broccoli in newsletter 2. ******************************************************* -D- THE OPTIONAL 21 KYR BP EXPERIMENT - PRESCRIBED SSTs ******************************************************* This experiment uses the CLIMAP (1981) dataset for SSTs and sea ice. CLIMAP datasets can be obtained at NGDC in /paleo/climap. Ask Robin Webb if you need more informations about CLIMAP data and how to use them. As mentionned in the newsletter 2, we recommend that: - the CHANGE in SSTs (LGM minus present day) be added to the values of your control run. - the sea ice extent for the LGM be prescribed according to CLIMAP. Differences between PMIP runs may arise from differences in the procedure used to extrapolate from February and August values (given by CLIMAP) to daily values used in the simulations. There is no "best" way to solve this problem. Below we suggest ONE procedure, but : ANY COMMENTS OR SUGGESTION FOR A BETTER METHODOLOGY WOULD BE WELCOME!! *1* About SSTs ---------- - First, interpolate the CLIMAP present day and LGM, February and August, SSTs on your grid, by averaging the CLIMAP SSTS of ocean grid points contained in your ocean grid. - produce daily values for the 4 CLIMAP datasets. The easiest way is to set the mid-month February and August values to the CLIMAP values. Then use a sine function to get daily values, assuming that the February and August values coincide with the extremes. - Finally, add the DAILY CHANGE in SSTs to your daily control SST to get the final daily LGM SSTs. *2* About sea ice ------------- The sea ice seasonal cycle is more difficult to infer from February and August values : - when sea ice occurs for both February and August, then assume a permanent sea ice cover for all months. - when there is sea ice during the winter season but none during summer : we assume a sea ice during N months around the winter season, with N depending on latitude and inferred from the control run. *3* Land-sea mask and CLIMAP SSTs ----------------------------- WARNING : ******** The CLIMAP SST dataset and the land-sea mask inferred from Peltier may be inconsistent. Indeed, when using this new land-sea mask, LGM ocean grid points may exist with no corresponding LGM SST value from the CLIMAP file, since the fall in sea level is reduced in Peltier's reconstruction compared with CLIMAP's. We have not yet checked how many grid points are affected. We hope few are. For such grid points, the simplest suggestion would be to use the same change in SST as that of the grid cell's nearest neighbor. Normally the nearest neighbor in the east-west direction would take precedence over the nearest neighbor in the north-south direction. **************************************************************** -E- INSOLATION AND CALENDAR - For 6 ka and 21 ka BP experiments **************************************************************** At 6 kyr BP, more than at 21 kyr BP, the change in orbital parameters induces a change in the length of the seasons. The problem of the choice of a calendar for past periods was already discussed in newsletter 1. We remind you that for the simulations we recommend you use noon on the 21.00 of March as a reference date for the vernal equinox. This reference ensures that we are all in phase for one specific date or month in the past and will then be able to perform model-model comparisons. However, we can still discuss how to define best the months for past periods to perform model analyses. After several discussions we recommend, as a first step, that monthly average data be kept based on the usual "months" as they are defined today (i.e., using the 21.00 of March as a reference and no attempt to change the length of the months to more nearly reflect the astronomical seasons as defined by the orbital changes). This way has the advantage to provide model analyses that are consistent with the calendar used for surface boundary conditions (SSTs' seasonal cycle). However, for further analyses, it will be interesting to test the sensitivity of model analyses to the choice of calendar, for example by changing the lengths of months or seasons to account for the change of length between solstices and equinoxes. To be able to perform some tests in the future, we SUGGEST that you keep whenever possible your DAILY outputs, at least for the basic variables listed in newsletter 1. The daily outputs will also be interesting for some specific diagnostics, such as transients or monsoon statistics. Sincerely yours, Sylvie Joussaume (LMCE, France) & Karl Taylor (LLNL, USA) & Robin Webb (NGDC, USA) ---------------------------------------------------------------------------- Contact Address: ################ Laboratoire de Modelisation du Climat et de l´Environnement D.S.M. / Orme des Merisiers / Bat. 709 C.E. Saclay 9119 Gif-sur-Yvette cedex FRANCE Tel.: (33) 1 69.08.77.11 Fax.: (33) 1 69.08.77.16 email: paleo (NEW! Please check the PMIP 'Contacts' web page) ----------------------------------------------------------------------------