********************************************* PALEOCLIMATE MODELING INTERCOMPARISON PROJECT ********************************************* Newsletter N. 6 ----------------- 28 July 1995 Dear participant, Now that most modeling groups have completed some of their PMIP simulations, it is time to prepare the output for immediate use in model-model comparisons and for later use in model-data comparisons. We hope this will lead to a better understanding of our models' ability to simulate PMIP paleoclimatic periods and will indicate how sensitive the results are to various model formulations. As you may already know, we have planned a first PMIP workshop devoted to model-model comparisons for the week of October 1 - 6, 1995 in Collonges-la-Rouge (France). Representatives of each modeling group (see table in newsletter 5) along with members of our scientific advisory committee and data-model subcommittee (see newsletter 4) have been invited to this workshop, making up a group of about 45 scientists. The workshop is sponsored both by the European Union and the NOAA Office of Global Programs (Paleoclimatology Program). The purpose of the workshop is for the participating modeling groups to share results from the 21 ka and 6 ka PMIP simulations and to begin exploring differences. We envision that special common interests will draw together groups of participants into sub-projects which will focus on various specific aspects of the simulations. We shall also initiate planning for systematic comparisons with paleoclimate data. In preparation for the comparisons, we are now requesting that participating groups fulfill their obligation to contribute model output to an archive that will be accessible by the other PMIP participants (initially only by those who have contributed data to the archive). The data will be stored at PCMDI (Program for Climate Model Diagnosis and Intercomparison), and the following specifications for preparing data for transfer to PCMDI are based largely on their experience in collecting and archiving data for AMIP (Atmospheric Model Intercomparison Project). Although this newsletter necessarily contains a rather detailed description of the procedure for preparing your output for the PMIP archive, it is really quite straight-forward. You will need to: *** compute climatological means of monthly, seasonal and annual data from your model, and compute the variance of the individual samples comprising the climatologies (i.e., interannual variances of the monthly, seasonal, and annual data). These means and variances will need to be calculated for each of the variables comprising the PMIP standard output (listed in Appendix I of this newsletter) and for each of the model simulations you perform (6k, 21k, and controls). *** convert your data (if necessary) to the PMIP standard units (given in Appendix I) and order your data as specified in section -A- of this newsletter. *** output the data in the PMIP standard ascii format by calling a fortran subroutine described in section -B- and obtainable from Karl Taylor. *** prepare short README files (one for each simulation), according to directions given in section -C-, and ftp them to PCMDI. *** collect your output files into a few "tar" files (constructed according to the directions given in section -D-), and then compress and ftp them to the PMIP archive at PCMDI. As soon as you contribute your output to the archive, you will be given access to the output from the other models. In the present newsletter we therefore provide information you will need to archive your model output. Specifically, we provide instructions for: -A- Computing and Structuring Standard Output -B- Writing PMIP Output in the Standard ascii Format -C- Creating Standard README Files -D- Transferring Data to the PCMDI Archive Appendix I List of Variables Requested for Archival Storage Appendix II Standard File Names Constructed by subroutine wrtascii Appendix III Sample Output File Produced by subroutine wrtascii *********************************************** -A- Computing and Structuring Standard Output *********************************************** Data should be sent to PCMDI by ftp or, if that is impossible, by exabyte tape (details to follow in -D- below). (Note that your data will only be made available to other contributing PMIP members.) Ascii files in the PMIP standard format are required. It will be easy to put your data in this format because a fortran subroutine (called wrtascii) has been written for this purpose and is available from anonymous ftp at PCMDI. Documentation of this subroutine and directions on how to obtain a copy of it are given in -B- below. The PMIP standard ascii files are written consistent with CDL (network Common Data form Language), which makes it trivial to translate the data into netCDF format (which like DRS and grads files can be ingested by the PCMDI visualization package called VCS). Through this established PMIP standard file format, the tasks of archiving the data and performing quality control checks can be simplified and automated. Furthermore you will be able to obtain the data in this format from all the PMIP models along with programs for reading it or translating it into netCDF or DRS format. The standard PMIP fortran program for writing model output (subroutine wrtascii) will correctly format your data if you prepare it as described here: *1* For each simulation and each variable (listed in Appendix I) you should produce the following climatological statistics which will reside in separate files: a) climatological annual mean data (i.e., averages over at least 10 years) b) interannual variance of the annual mean data (see 2g below) c) climatological seasonal mean data (i.e., averages over several winters, several springs, etc. as defined in 2b and 2e below) d) interannual variance of the seasonal mean data e) climatological monthly mean data (i.e., averages over several Januarys, several Februarys, etc.) f) interannual variance of the monthly mean data Thus for each simulation and each variable you will call subroutine wrtascii 6 times (with different arguments) to produce the above 6 files. *2* In preparing to call subroutine wrtascii note the following: a) Remember that means for the months and seasons should be defined in the conventional way with the calendar defined in all experiments such that vernal equinox coincides with noon of March 21, and months kept the same as for the control run (See Newsletters 1, 2 and 3 for further discussion.) b) Seasons are defined as: N.H. winter = DJF (December, January, February); N.H. spring = MAM (March, April, May); N.H. summer = JJA (June, July, August); N.H. autumn = SON (September, October, November) c) Although not a requirement of PMIP, some of you may want to ensure exact correspondence between the data used to define your climatological monthly, seasonal and annual means. If you do, we recommend that you define your climatological year as beginning in December and ending in November so that the same months that contribute to the climatological monthly mean data also contribute to the climatological seasonal and annual mean data. d) Although the climatological seasonal mean data and the climatological annual mean data can obviously be calculated from the climatological monthly mean data, the interannual variance of the seasonal and annual data cannot be calculated from the interannual variance of the monthly mean data. For completeness and consistency, therefore, please contribute all 6 files listed in *1* above to the archive. e) When computing seasonal means, weight the monthly data by the number of days in that month. For example, if your model has a 365 day year, then the mean autumn = (SEP*30 + OCT*31 + NOV*30)/91, where SEP, OCT, and NOV represent the monthly mean data for those months. f) When computing the annual mean, weight the monthly data by the number of days in that month. (Alternatively, compute the annual means from the seasonal means, weighting the seasonal data by the number of days in the season.) g) An unbiased estimate of the interannual variance should be computed according to the conventional formula: sum over N years {(x(n) - x-bar)**2/(N-1)} where x-bar is the climatological mean and n is the year index. h) Be sure to carry out sufficiently long simulations such that the climatologies you calculate are good estimates of the true model climatology. The spin up time for fixed SST experiments will normally be less than a year (unless your soil moisture and snow cover require longer adjustment times), but for computed SST experiments with ocean mixed layers, the spin-up period may vary from one to several decades, depending on the model. Normally you should average over at least 10 years of data, but for computed SST experiments, you may need to average over longer periods to account for variability on longer time-scales. *3* In preparing to call wrtascii be sure your array of data is filled in the following order: a) For each longitude - latitude field, the data should be stored with the longitude index varying most rapidly, latitude next, and finally time (i.e., in FORTRAN: A(i,j,m) where i is the longitude index, j is the latitude index, and m is the month or season index). For annual average data there will be no time index. b) For each longitude - latitude - pressure level field, the data should be stored with the longitude index varying most rapidly, latitude next, pressure level next, and finally time (i.e., in FORTRAN: A(i,j,k,m) where i is the longitude index, j is the latitude index, k is the pressure level index and m is the month or season index). For annual average data there will be no time index. c) For each latitude - pressure zonal cross-section, the data should be stored with the latitude index varying most rapidly, pressure next, and finally time (i.e., in FORTRAN: A(j,k,m) where j is the latitude index, k is the pressure index, and m is the month or season index). For annual average data there will be no time index. d) Longitudes should be stored west to east beginning at 0 degrees or if your model does not have a grid point at 0 degrees, then beginning at the first longitude to the east of 0 degrees. Do not include a "wrap-around" point (i.e., do not duplicate your first grid point as your last grid point). e) Latitudes should be stored north to south, beginning with the northern most grid cell. f) Pressure levels should be stored in order of increasing pressure (i.e., top of atmosphere to surface) g) Seasons should be stored in the order: N.H. winter (DJF), N.H. spring (MAM), N.H. summer (JJA), N.H. autumn (SON). h) Months should be stored in calendar order beginning with January and ending with December. *4* Please note: a) All variables from your model must be reported on the same grid. If you have a staggered grid (e.g., some variables calculated at the center of a grid cell and others calculated at the edges or corners) or if your model uses a grid with the number of longitude grid-cells varying with latitude (e.g., the new ARPEGE/IFS model with a grid that "thins" at higher latitudes), then you should interpolate to a common rectangular longitude-latitude grid. Note that it is not necessary that the grid spacing be uniform (Gaussian grids, for example, used in spectral models are perfectly acceptable.) b) Interpolate, when necessary, to standard pressure levels (see Appendix I for the list of standard levels). Ideally this should be done before computing averages. ***************************************************** -B- Writing PMIP Output in the Standard ascii Format ***************************************************** You can use anonymous ftp (to PCMDI) to obtain subroutine wrtascii, which will write data in the PMIP standard ascii format, as follows: enter: ftp sprite.llnl.gov Respond to login prompt: anonymous we ask you to enter your e-mail address as the password, for example: myemail@myhost.mydomain change directories: cd /pub/ktaylor/pmip transfer files, for example: get wrtascii.f quit For each variable in the list given in Appendix I and for each simulation you will need to read your data, calculate appropriate time means and interannual variances, possibly reorder the data (to be consistent with the specifications given in -A- above), and call subroutine wrtascii 6 times (once for each file written). The arguments for the subroutine are documented here: subroutine wrtascii(modelid, modelnm, run, varname, stattype, & nlon, nlat, nlev, ntim, lensim, undef, tol, alon, alat, & alev, field, ierr) c ---------------------------------------------------------------------- c c This subroutine creates a file and stores a single output field c in it. The ascii output written by this subroutine is in a form c called CDL (network Common Data form Language) that can be easily c translated into a binary netCDF file or a DRS file. The output c is the standard form required by the PMIP project (ca. 1995). c c Author: Karl E. Taylor c c Date last revised: 27 July 1995 c c c INPUT: c c character strings: c c modelid = ascii string (<= 8 characters) containing c abbreviated model i.d. for use in file names (e.g., c gen, ccm2, ech2, ccm-llnl, etc.) No blanks or upper c case characters are permitted. c c You should be sure to plan ahead in choosing your c model i.d.; if you plan on running your model again c at a different resolution or after modifying it, for c example, then leave room for some suffix modifier, c such as ccm2_2a c c modelnm = ascii string (<= 48 characters) containing full c model name and version along with resolution c information (e.g. "Genesis 2.3 T42L18" or c "LMCE - modele du LMD version 4ter - 5x7L11"). c Blanks, lower and upper case characters are o.k. c c Although the model name and version are up to you, you c must include the resolution information (e.g., T42L18 c or 5x7L11) consistent with the following rules: c For spectral models use the common spectral truncation c abbreviations (e.g., T42, R30). For finite-difference c models use average grid cell size (i.e., 180/nlat by c 360/nlon) and write resolution similar to 5x7 (for a c 5 degree latitude by 7 degree longitude grid (round c to nearest degree). Indicate the number of levels c with an L prefix (e.g., L18 or L11) c c run = ascii string (<= 5 characters) c = '6fix' (for 6 ka BP simulation with fixed SSTs) c = '21cal' (for 21 ka BP simulation with computed SSTs) c = '21fix' (for 21 ka BP simulation with fixed SSTs) c = '0fix' (for fixed SST control run) c = '0cal' (for calculated SST control run) c c varname = ascii string (<= 8 characters) containing variable c name (exactly as specified in standard list given in c Appendix I) c c stattype = ascii string (<= 8 characters) c = 'mean' (for time-mean field) c = 'variance' (for interannual variance data) c c integers: c c nlon = longitude dimension (i.e., number of longitudes). c nlon should be set to 0 for fields that are c independent of longitude (i.e., for zonal mean cross c sections) c c nlat = latitude dimension (i.e., number of latitudes) c c nlev = pressure level dimension (i.e., number of pressure c levels). nlev should be set to 0 for fields that c are independent of pressure level (i.e., for data c that represent a surface field (e.g., pmsl), c or a field at the top of the atmosphere (e.g., rsdt), c or a column integrated total (e.g., clt), then nlev c should be set to 0). Allowed values of nlev are: 0, c 2, 3, 15. c c ntim = time dimension c = 0 for annual mean data c = 4 for seasonal data c = 12 for monthly data c c lensim = number of years of data from which the c climatological statistics were calculated (i.e., c the averaging period). c c real: c c undef = the value you use to identify missing or undefined c data. (This value will be replaced with the PCMDI c standard missing value identifier, 1.0e20, before c the ascii output file is written by this c subroutine.) c c tol = tolerance for determining how close data must be to c to your "undef" value before it will be considered c missing. if (abs(field-undef) .le. tol) then field c will be reset to 1.0e20, which identifies it as c "missing" or "undefined". c c alon = a vector containing the longitude coordinate values c (e.g., 0., 5., 10., ... 355.) The values stored in c alon are immaterial (i.e., they are ignored) if you c are writing a zonal mean cross-section. c c alat = a vector containing the latitude coordinate values c (e.g., 88. 84., 80, ... -88.) c c alev = a vector containing the pressure-level coordinate c values (e.g., 200., 850.). The values stored in alev c are immaterial (i.e., they are ignored) if the data c set you are writing represents a surface field c (e.g., pmsl), or a field at the top of the atmosphere c (e.g., rsdt), or a column integrated total (e.g., c clt). c c field = array containing your data stored in the order c described in -A- above. c = field(nlon,nlat,nlev,ntim) for monthly or seasonal c data that are functions of longitude, latitude c and pressure level. c = field(nlon,nlat,ntim) for monthly or seasonal data c that are functions of longitude and latitude. c = field(nlat,nlev,ntim) for monthly or seasonal zonal c mean cross-sections that are functions of c latitude and pressure level. c = field(nlon,nlat,nlev) for annual data that are c functions of longitude, latitude and pressure c level. c = field(nlon,nlat) for annual data that are functions c of longitude and latitude. c = field(nlat,nlev) for annual mean cross-sections that c are functions of latitude and pressure level. c c OUTPUT: c integer: c c ierr = error flag c = 0 if no error detected by subroutine wrtascii c > 0 if fatal error detected by subroutine wrtascii c < 0 if warning message has been written by subroutine c wrtascii c c ---------------------------------------------------------------------- ************************************ -C- Creating Standard README Files ************************************ For each of your simulations, please provide a README file which will contain some important information about your model and simulation. The README file names should be of the form: modelid_run_README where modelid and run are defined in -B- above. Example: ccm-llnl_21cal_README To create a standard format README file, you will probably find it easiest to simply edit the following example: ---------------------------------------------------------------------- # Comments or additional information should be given on lines with # a "#" in the first column. # Blank lines are allowed. # KEYWORD YOUR INPUT MODEL_ID ccm-llnl MODEL_NAME CCM1/LLNL Version 1 R15/L12 PMIP_RUN 21cal YEARS_MEAN 20 YEARS_SPUNUP 25 NB_SAMPLES 24 DIM_LONGITUDE 48 DIM_LATITUDE 40 DIM_LEVELS 12 MODEL_MONTHS 31 28 31 30 31 30 31 31 30 31 30 31 PMIP_CONTACT Karl E. Taylor ADDRESS P.O. Box 808 L-264 ADDRESS Lawrence Livermore National Lab ADDRESS Livermore, CA 94550 ADDRESS USA PHONE 510 423 3623 FAX 510 422 7675 EMAIL taylor13 # MODEL_DOCUMENTATION: # # Williamson, D.L., J.T. Kiehl, V. Ramanathan, R.E. Dickinson, J.J. # Hack, 1987: Description of NCAR Community Climate Model (CCM1). # NCAR Technical Note, NCAR/TN-285+STR, 112 pp. # # Taylor, K.E., and S.J. Ghan, 1992: An analysis of cloud liquid # water feedback and global climate sensitivity in a general # circulation model, J. Climate, 5, 907-919. # COMMENTS: # # The model was started from an equilibrium state of a 2XCO2 # experiment. # # The model was run on a CRAY 2 under the UNICOS operating system. # # The model uses a sigma-coordinate in the vertical with the following # levels: 0.009, 0.025, 0.060, 0.110, 0.165, 0.245, 0.355, 0.500 # 0.664, 0.811, 0.926, 0.991 # # [Please be sure to include any special problems encountered # that might affect interpretation of the results or any known biases # in your model's climate.] # ---------------------------------------------------------------------- Note that in the above README file: MODEL_ID should be identical to the "modelid" supplied to subroutine wrtascii and described in -B- above. MODEL_NAME should be identical to the "modelnm" supplied to subroutine wrtascii and described in -B- above. PMIP_RUN should be identical to the "run" supplied to subroutine wrtascii and described in -B- above. YEARS_MEAN should be the number of years contributing to climatological statistics (i.e. averaging period for data) and should be identical to "lensim" supplied to subroutine wrtascii (described in -B- above). YEARS_SPUNUP should be the number of years the model was "spun up" prior to the climatological averaging period (i.e., the spin-up time). NB_SAMPLES is the number of samples per day contributing to your time-means. If you accumulate over each time-step, then the number of samples should equal the number of time-steps per day even though you may only save the accumulations a few times per day. If different variables are sampled at different intervals, enter the fewest samples per day that contribute to any of your time means. You might want to list the sampling interval for each variable in the comments section of this file. DIM_LONGITUDE is your model's longitude dimension (i.e. the number of longitude grid cells in your model) DIM_LATITUDE is your model's latitude dimension (i.e. the number of latitude grid cells in your model) DIM_LEVELS is your model's vertical coordinate dimension (i.e. the number of levels in your model) MODEL_MONTHS is a list of the number of days in each month, according to your model. You should enter these as a list of blank-separated integers. PMIP_CONTACT is the person to contact concerning this simulation. ADDRESS is the contact's address. PHONE is the contact's phone number. FAX is the contact's fax number. EMAIL is the contact's e-mail address. NOTE: The following final two items must appear on comment lines (i.e., lines starting with '#') MODEL_DOCUMENTATION is followed by a list of references that document your model. COMMENTS is followed by any further information of interest concerning your model and the simulation (see example above). *********************************************** -D- Transferring Data to the PCMDI Archive *********************************************** When you are ready to archive your data at PCMDI, you should contact: Karl E. Taylor email: taylor13 (NEW! Please check the PMIP 'Contacts' web page) Karl will provide instructions on where to place the data using ftp. (Note again, that your data will be made available only to other contributing PMIP participants.) You may prefer to send output from each of your simulations as they are completed. This is encouraged. The output files created for each simulation should be put into 6 tar files. You should also ftp the README file (named as specified in -C- above). The directory structure for the 6 tar files should be as follows: For each simulation and each statistic (e.g., climatological annual mean, interannual monthly variance) place all files in a single compressed tar file named modelid_run_stat.tar.Z where modelid should be identical to the "modelid" supplied to subroutine wrtascii, described in -B- above. run should be identical to the "run" supplied to subroutine wrtascii, described in -B- above. (Choose from: 6fix, 21fix, 21cal, 0fix, 0cal.) stat should be identical to the "stat" component of the file name (see Appendix II). Stat should be one of the following: cm, vm, cs, vs, ca, va. Do not confuse "stat" with " stattype". Each tar file should contain all the variable fields for a single statistic and simulation, and should be structured with directories and subdirectories as follows: modelid/run/stat/modelid_run_*_stat_* where modelid, run, and stat were defined just above and modelid_run_*_stat_* represents the set of files you created by calling wrtascii (for each variable) which will be stored in the directory, modelid/run/stat (see Appendix 2 for interpretation of file names). The first * represents one of the standard PMIP variable names and the second * represents one of the three "views" (map, 3d, or xsec). As an example, suppose you have completed the 6 ka BP simulation. You will create 6 tar files plus a README file: ccm2_6fix_cm.tar.Z ccm2_6fix_vm.tar.Z ccm2_6fix_cs.tar.Z ccm2_6fix_vs.tar.Z ccm2_6fix_ca.tar.Z ccm2_6fix_va.tar.Z ccm2_6fix_README Each compressed tar file will comprise the whole suite of variable files for the given simulation and given statistic. For example, ccm2_6fix_cm.tar.Z will contain, in the modelid/6fix/cm directory ccm2_6fix_orog_cm_map ccm2_6fix_sftland_cm_map ccm2_6fix_snow_cm_map ccm2_6fix_sim_cm_map . . . ccm2_6fix_zg_cm_3d ccm2_6fix_ta_cm_3d ccm2_6fix_u_cm_3d ccm2_6fix_v_cm_3d . . . ccm2_6fix_hur_cm_xsec ccm2_6fix_clf_cm_xsec ccm2_6fix_stfmm_cm_xsec We look forward to receiving your output. Remember, you will be given access to the output from the other models only after you contribute your own output to the archive. Please let us know if you have questions. Yours sincerely, Sylvie Joussaume (LMCE) Karl Taylor (PCMDI) --------------------------------------------------------------------- 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) --------------------------------------------------------------------- ************************************************************ Appendix I: List of Variables Requested for Archival Storage ************************************************************* Please send us the following fields for inclusion in the PMIP archive (listed here with standard names that you are required to use when calling subroutine wrtascii). Also listed are the units required of your data and titles and notes for your information and the sign convention for fluxes is given in the NOTES column (either positive up or positive down): GROUP 1: The following are functions of longitude, latitude, time (in the case of seasonal and monthly data), and for zg, ta, u, v, hus, hur, stfuv, and potuv they are also a function of pressure level (as indicated in the NOTES column). NAME UNITS TITLE NOTES orog m Surface Elevation sftland % Percentage Land % of grid cell that is land (# see footnote) seaice % Percentage of Area Covered by Sea Ice % of area of grid cell that is sea-ice covered; i.e., 100 times the product of (fraction of time sea ice is present in cell) and (fraction of cell covered) (## see footnote) snow % Percentage of Area Covered by Snow % of area of grid cell that is snow covered; i.e., 100 times the product of (fraction of time snow is present in cell) and (fraction of cell covered) (## see footnote) sim kg/(m^2) Sea Ice Mass per Unit Area rsdt W/(m^2) TOA Incident Shortwave Radiation insolation (+ down) rsut W/(m^2) TOA Reflected Shortwave Radiation + up rlt W/(m^2) TOA Outgoing Longwave Radiation + up rsds W/(m^2) Surface Incident Shortwave Radiation + down rsus W/(m^2) Surface Reflected Shortwave Radiation + up rls W/(m^2) Surface Net Longwave Radiation + down hfss W/(m^2) Surface Sensible Heat Flux + down hfls W/(m^2) Surface Latent Heat Flux + down mrso kg/(m^2) Soil Moisture AMIP units were cm snm kg/(m^2) Snow Mass per Unit Area mass of snow on ground per unit area mrro mm/day Runoff pr mm/day Total Precipitation include liquid & solid phase, convective & large scale precip. prc mm/day Convective Precipitation if available; (include both liquid & solid phase precip.) prsnwe mm/day Snowfall (Liquid Water Equivalent) include both convective & large scale evs mm/day Surface Evaporation include both evaporation & sublimation prw kg/(m^2) Total Precipitable Water AMIP units were mm tas C Surface Air Temperature 2-meter screen temp. tg C Ground Temperature skin temperature pmsl hPa Mean Sea Level Pressure usfc m/s Eastward Surface Wind Speed at 10 m vsfc m/s Northward Surface Wind Speed at 10 m tauu N/(m^2) Eastward Surface Wind Stress at surface tauv N/(m^2) Northward Surface Wind Stress at surface hursfc % Surface Relative Humidity at 2 m clt % Total Cloudiness % of atmospheric column covered by clouds clw g/(m^2) Cloud Water Content if available rltcs W/(m^2) TOA Outgoing Longwave Clear Sky Method II **; + up Radiation rstcs W/(m^2) TOA Net Shortwave Clear Sky Method II **; Radiation (net=incident-reflected) + down ** rlscs W/(m^2) Surface Net Longwave Clear Sky Method II **; + down Radiation rsdscs W/(m^2) Surface Incident Shortwave Clear Method II **; + down Sky Radiation rsuscs W/(m^2) Surface Reflected Shortwave Clear Method II **; + up Sky Radiation # In calculating the "Percentage Land" for a grid cell, include sea ice as "ocean" and glacial ice as "land". For models with non-fractional coverage of land or ocean within a grid cell, the "Percentage Land" should either be 0.0 or 100.0. ## In calculating "Percentage Sea Ice" for a grid cell, compute (100/n) * [sum-over-n f-sub-n] where n is the time-index and f-sub-n is the fraction of sea ice in the grid cell at time n. For models that allow a grid cell to be partially ocean and partially land, the fraction of sea-ice in the grid cell should be calculated relative to the total area of the cell, not just the ocean area. ** Method II for calculating clear-sky fluxes is the normal procedure used in model-model intercomparisons (see, for example, Cess, R.D. and G.L. Potter, 1987: Exploratory studies of cloud radiative forcing with a general circulation model. Tellus, 39A, 460-473) The clear-sky fluxes are computed during each sampling interval and then time-averaged (without weighting by cloud fraction). GROUP 2: The following are functions of longitude, latitude, pressure level (as indicated in the NOTES column), and time (in the case of seasonal and monthly data). NAME UNITS TITLE NOTES zg m Geopotential Height at 200, 500 and 850 hpa ta C Temperature at 200 and 850 hpa u m/s Eastward Wind Speed at 200 and 850 hpa v m/s Northward Wind Speed at 200 and 850 hpa hus g/kg Specific Humidity at 200 and 850 hpa hur % Relative Humidity at 200 and 850 hpa stfuv m^2/s Horizontal Wind Velocity Stream Function at 200 and 850 hpa potuv m^2/s Horizontal Wind Velocity Potential at 200 and 850 hpa GROUP 3: The following are functions of latitude, pressure level and time (in the case of seasonal and monthly data). For these cross-sections, output should be reported at the standard pressure levels: 10, 20, 30, 50, 70, 100, 150, 200, 250, 300, 400, 500, 700, 850, 1000 hPa. NAME UNITS TITLE NOTES u m/s Eastward Wind Speed v m/s Northward Wind Speed ta C Temperature hus g/kg Specific Humidity hur % Relative Humidity clt % Percent Clouds % of layer occupied by clouds stfmm m^2/s Mean Meridional Stream Function GROUP 4 [You are welcome to propose that other fields you think might be of general interest (especially for model-data comparisons) be archived, but please contact Karl Taylor before including non-standard variables in your contribution to the PMIP archive.] ******************************************************************** Appendix II: Standard File Names Constructed by subroutine wrtascii ******************************************************************** Subroutine wrtascii described in section -B- will use your input to create and write to a file with a descriptive name of the form: modelid_run_var_stat_view Example: ccm2_21fix_pmsl_cm_map The filename comprises 5 segments: 1. modelid -- abbreviated model i.d. supplied by user and described in the documentation of the subroutine arguments given in section -B- 2. run -- model simulation a) 6fix (for 6 ka BP simulation with fixed SSTs) b) 21fix (for 21 ka BP simulation with fixed SSTs) c) 21cal (for 21 ka BP simulation with computed SSTs) d) 0fix (for fixed SST control run) e) 0cal (for calculated SST control run) 3. variable name (exactly as specified in the PMIP standard list given in Appendix I; for example, pmsl) 4. stat -- statistic a) cm (climatological monthly mean) b) vm (interannual variance of monthly means) c) cs (climatological seasonal mean) d) vs (interannual variance of seasonal means) e) ca (climatological annual mean) f) va (interannual variance of annual means) 5. view a) map (longitude-latitude field) b) 3d (longitude-latitude-pressure field) c) xsec (zonal mean latitude-pressure cross-section) ***************************************************************** Appendix III: Sample Output File Produced by subroutine wrtascii ***************************************************************** Here is a sample, partial listing of the output produced by subroutine wrtascii for the case of a longitude-latitude-season file: netcdf lmcelmd4_6fix_tas_cs_map { dimensions: longitude = 48; latitude = 36; // plev = 0; time = 4; time_3char = 12; variables: float longitude(longitude), latitude(latitude), time(time); longitude:units = "degrees"; longitude:title = "Longitude"; longitude:long_name= "LONGITUDE"; latitude:units = "degrees"; latitude:title = "Latitude"; latitude:long_name = "LATITUDE"; // no plev units // no plev title // no plev long_name time:units = "season"; time:title = "Season"; time:long_name = "SEASON"; char timelabel(time_3char); float tas(time,latitude,longitude); tas:units = "C"; tas:stat_name = "Climatological Seasonal Mean"; tas:missing_value = 1.00000E+20; tas:valid_min = -4.87557E+01; tas:valid_max = 4.29864E+01; tas:mean_val = 1.47086E+01; tas:mean_absval = 1.89489E+01; tas:warning_flag = 0; tas:missing_count = 0; tas:title = "Surface Air Temperature"; tas:source = "PMIP 6 ka BP 15-Year Simulation with Prescribed SST's by LMCE - modele du LMD version 4ter - 5x7L11"; tas:notes = "2-meter screen temperature"; // global attributes :file_name = "lmcelmd4_6fix_tas_cs_map"; :model_name = "LMCE - modele du LMD version 4ter - 5x7L11"; :experiment = "6 ka BP Simulation with Prescribed SST's"; data: longitude = 3.7500, 11.2500, 18.7500, 26.2500, 33.7500, 41.2500, 48.7500, 56.2500, 63.7500, 71.2500, 78.7500, 86.2500, 93.7500, 101.2500, 108.7500, 116.2500, 123.7500, 131.2500, 138.7500, 146.2500, 153.7500, 161.2500, 168.7500, 176.2500, 183.7500, 191.2500, 198.7500, 206.2500, 213.7500, 221.2500, 228.7500, 236.2500, 243.7500, 251.2500, 258.7500, 266.2500, 273.7500, 281.2500, 288.7500, 296.2500, 303.7500, 311.2500, 318.7500, 326.2500, 333.7500, 341.2500, 348.7500, 356.2500; latitude = 76.4638, 66.4435, 59.4416, 53.6639, 48.5904, 43.9830, 39.7090, 35.6853, 31.8554, 28.1786, 24.6243, 21.1684, 17.7916, 14.4775, 11.2123, 7.9836, 4.7802, 1.5918, -1.5918, -4.7802, -7.9836, -11.2123, -14.4775, -17.7916, -21.1684, -24.6243, -28.1786, -31.8554, -35.6853, -39.7090, -43.9830, -48.5904, -53.6639, -59.4416, -66.4435, -76.4638; time = 1.0, 2.0, 3.0, 4.0; timelabel = "DJF", "MAM", "JJA", "SON"; tas =-2.77484E+01,-3.01865E+01,-3.42403E+01,-3.70435E+01,-3.81412E+01, -3.86561E+01,-3.87973E+01,-3.87936E+01,-3.87416E+01,-3.86108E+01, -3.84682E+01,-3.85719E+01,-3.98430E+01,-3.95486E+01,-3.76386E+01, -3.66105E+01,-3.54578E+01,-3.38125E+01,-3.21477E+01,-3.00725E+01, -2.86386E+01,-2.73882E+01,-2.64791E+01,-2.60693E+01,-2.63416E+01, ..., ...,