Descriptive Experimental Protocol for CCSM3

Created April 2006
Last Modified February 2008

1. Experiment 1: Forcing with observed NCEP/NCAR reanalysis (CCSM3 I-configuration)

1.1. Experiment 1.1: Spin-up

1.1.1. CO₂ constant at 283.1878 ppm. This is the value from the standard C⁴MIP forcing file for year 1800 (the first year of that dataset).

1.1.2. N deposition climatology constant at pre-industrial level (circa 1890) for CLM3-CN run. This climatology is provided by J.-F. Lamarque from coupled CAM-Chemistry simulations. The T42 version of this file is regrid_ndep_clm_128x064.nc.

1.1.3. Surface data includes the plant functional type area fractions from Johannes Feddema's dataset for the year 1798, and the standard CLM input datasets for special landcover types (glacier, lake, wetland). This dataset is T42.clamp.surface-data.060412.nc.

1.1.3.1. Note: The surface dataset specification is relevant only to models running with CLM as the base land component (e.g., CN and CASA'). Other models using a different base land component (e.g., IBIS running on LSX) will require appropriate model-specific surface datasets.

1.1.4. River routing module (RTM) turned on.

1.1.5. Memory allocated for all plant functional types on each gridcell in CLM-based models (MAXPATCH_PFT = numpft+1).

1.1.6. Climate drivers: Using the CCSM "data atmosphere" component datm7 to cycle through 25 years (1948-1972) of NCEP/NCAR reanalysis data (from A. Dai and T. Qian, 2006). The originating dataset is currently stored on the NCAR Mass Storage System at mss:/TQIAN/LSMF/NCEPDATA as yyyy-mm.nc (1948-01.nc to 2004-12.nc). New datasets have been created for use with datm7 from these originating datasets. The datm7 datasets are named clmncep.T62.1948-01.nc through clmncep.T62.2004-12.nc.

1.1.7. Run with cyclic surface weather forcing until model reaches steady-state, as defined below.

1.1.8. Spin-up acceleration: The models may employ model-specific mechanisms to accelerate the approach to steady-state conditions. The only criterion is that all models use the same external forcing data, as described above, throughout the spin-up period.

1.1.8.1. Spin-up details for CN model: Uses the accelerated decomposition (AD) mechanism introduced by Thornton and Rosenbloom (2005) to reduce the number of model years required to reach steady state. First step consists of a run initialized with very small leaf carbon and nitrogen pools, and all other C and N pools set to zero. This step is run for 500 years with the base rate constants for litter and soil organic matter (SOM) decomposition multiplied by a factor of 10 (accomplished by setting AD_SPINUP flag as CPP variable in clm.buildexe). Second step consists of a 25-year run (one cycle of forcing) executed as a continuation from previous step, but with AD_SPINUP turned off and EXIT_SPINUP turned on in clm.buildexe. This switches the litter and SOM decomposition base rates back to their normal values, and increases the litter and SOM pool sizes by a factor of 10 on the first time step of the run. Third step consists of a run with both AD_SPINUP and EXIT_SPINUP turned off and continuing to cycle the 25 years of forcing data until the model reaches steady-state, as defined below.

1.1.9. Criteria for definition of steady-state (must meet all of the following criteria):

1.1.9.1. Global land NEE < 0.05 PgC/yr when taken as an average over the 25-year repeat cycle (1948-1972) in the NCEP/NCAR surface weather drivers.

1.1.9.2. For every model gridcell, absolute value of NEE < 1.0 gC/m²/y, taken as an average over the same 25-year repeat cycle (guards against spatially offsetting sources and sinks).

Because experience has shown that thousands of years of model spin-up may be required to achieve this second criteria, the requirement that the absolute value of NEE be less than 1.0 gC/m2/y has been eliminated.

1.2. Experiment 1.2: Control

1.2.1. Start from the end-point of the Experiment 1.1 (spin-up), reset the simulation year to 1798 and run forward with 25-year cycle (1948-1972) of NCEP/NCAR forcing for 207 years, ending in simulation year 2004.

1.2.2. CO₂, N deposition, and landcover held constant, as in spin-up (Experiment 1.1).

1.2.3. The purpose of this experiment is to isolate any remaining drifts in the model state variables, so that they can be removed from forcing experiments by differencing.

1.3. Experiment 1.3: Varying Climate

1.3.1. Branch from Experiment 1.2 at 1948-01-01 and run 57 years, through 2004, using the full NCEP/NCAR data record (1948-2004).

1.3.2. CO₂, N deposition, and landcover are held constant, as in spin-up (Experiment 1.1).

1.3.3. The purpose of this experiment is to isolate the influence of recent historical climate changes on terrestrial carbon cycle.

1.4. Experiment 1.4: Varying Climate, CO₂, and N Deposition

1.4.1. Weather drivers: Start from the end-point of the Experiment 1.1, reset the simulation year to 1798 and run forward with 25-year cycle (1948-1972) of NCEP/NCAR forcing for 150 years (1798-1947), then switch off NCEP/NCAR data cycling and run for another 57 years (1948-2004) using the full NCEP/NCAR data record (1948-2004).

1.4.2. CO₂ follows the C⁴MIP forcing file, which has values from 1800-1999. For 1798 and 1799, the current dynamic CO₂ mechanism in CLM will use the value from 1800, moving into the transient at the start of the C⁴MIP dataset (in 1800). The C⁴MIP dataset was extended out to 2005 by Natalie Mahowald, and the dataset was constructed by Forrest Hoffman (with a year 2006 that contains the value for year 1800 to avoid cyclic interpolation between the two end-points). The dataset is surf_co2.clamp.T42.20070208.nc and the stream file is surf_co2.clamp.T42.stream.20070208.txt.

1.4.3. For CLM3-CN only: This run starts with the N deposition climatology (see 1.1.2, above) until year 1890. At 1890, the CLM3-CN run branches in order to use the T42 N deposition time series, created by Jean-Francois Lamarque. This dataset includes NO_y+NH_x deposition with time slices at 1890, 2000, and 2100. Like the other models, CLM3-CN will branch again at 1948 to use the full NCEP/NCAR data record (1948-2004). This run uses the N deposition time series from NCAR (see 1.1.2, above). The time slices at 1850, 2000, and 2100 are interpolated to create an annual time series, which is ingested by the CLM3-CN code. Code replicates data from the first time slice for any simulation years before the start of the N deposition time series, in this case for the period 1798-1850.

1.4.4. Land cover held constant.

1.5. Experiment 1.5: Varying Climate, CO₂, N Deposition, and Land Cover

1.5.1. Everything identical to Experiment 1.4, but adding the forcing of plant functional type areas from the annual time series in Johannes Feddema's new datasets for the period 1798-2004. This dataset is derived from Ramankutty and Foley (1999) and Klein Goldewijk (2001), and made consistent with the CLM framework by Feddema and NCAR staff. This dataset is located at [to be done]. Treatment of land cover fluxes follows the C⁴MIP protocol. [Provide additional details for landcover, specific to Feddema's datasets and the CLM plant functional type list (Thornton). This experiment requires code modifications for all models (Thornton and Hoffman).]

1.6. Experiment 1.6: Free Air CO₂ Experiment (FACE) Control

1.6.1. This experiment branches off of Experiment 1.4 at year 1997, duplicating the last portion of that experiment but continuing out to year 2100. The last 25 years of the NCEP/NCAR forcing (1980-2004) will be cycled (synchronously for years 1997-2004) while CO₂ concentration and N deposition will be held constant at 2005 values once that year has been reached.

1.7. Experiment 1.7: Free Air CO₂ Experiment (FACE) Transient

1.7.1. Like Experiment 1.6, this experiment branches off of Experiment 1.4 at year 1997, duplicating the last portion of that experiment but continuing out to year 2100. The last 25 years of the NCEP/NCAR forcing (1980-2004) will be cycled (synchronously for years 1997-2004) while N deposition will be held constant at 2005 values once that year has been reached. However, the CO₂ concentration will be set to 550 ppmv at the beginning of the experiment and held constant throughout.

2. Experiment 2: Coupled land-atmosphere, forcing from data sea surface temperature and sea ice distribution, prescribed CO₂

2.1. Experiment 2.1: Spin-up

2.1.1. Experiment 2.1.1: Spin-up Step 1 (F-configuration)

2.1.1.1. For each model: preliminary coupled CAM-land (F-configuration) run to generate 25 years of hourly CAM forcings for use in spin-up. This initial run is 50 years, with output saved from the coupler every timestep for the final 25 years. These outputs are concatenated in a post-processing step to generate a coupler driver dataset for use in spin-up step 2.

2.1.1.2. Each land model is initialized from its own Experiment 1.1 end-point.

2.1.1.3. Data SST and sea ice extent for this 50-year run are provided by cycling twice through 25-years (1875-1899) of the HadISST dataset in the docn component of CCSM. The dataset is hadley.sst.1870-2003.1x1.050730.nc. The accompanying docn7 stream file is hadley.sst.1870-2003.1x1.stream.060209.txt and the CSIM stream file is hadley.sst.1870-2003.1x1.stream.txt.

2.1.1.4. CO₂, N deposition, and surface datasets identical to those used in Experiment 1.1.

2.1.2. Experiment 2.1.2: Spin-up Step 2 (I-configuration)

2.1.2.1. Each model uses 25 years of coupler drivers from its own Experiment 2.1.1 as cyclic forcing in data-atmosphere mode. Run until model reaches steady-state.

2.1.2.2. Spin-up acceleration protocols and steady-state convergence criteria as described above in sections 1.1.8 and 1.1.9 respectively.

2.1.2.3. CO₂, N deposition, and surface datasets identical to those used in Experiment 1.1.

2.1.3. Experiment 2.1.3: Spin-up Step 3 (F-configuration)

2.1.3.1. Switch back to active CAM-land (F-configuration), run 50 years to assess the steady-state criteria.Switch back to active CAM-land (F-configuration), run 50 years to assess the steady-state criteria.

2.1.3.2. Use the spun-up land state from the end-point of Experiment 2.1.2 as the initial condition for land.

2.1.3.3. Use the end-point for Experiment 2.1.1 as the initial condition for atmosphere.

2.1.3.4. Data SST and sea ice extent for this 50-year run are provided by cycling twice through 25-years (1875-1899) of the HadISST dataset in the docn component of CCSM (same as in Experiment 2.1.1).

2.1.3.5. CO₂, N deposition, and surface datasets identical to those used in Experiment 1.1.

2.1.3.6. Assess steady-state as defined in section 1.1.9 for final 25 years of this 50-year run. Assess steady-state as defined in section 1.1.9 for final 25 years of this 50-year run.

2.1.3.7. If steady-state achieved, move on to Experiment 2.2. Otherwise, run out Experiment 2.1.3 until convergence criteria are met, or, if simulation is still very far from steady-state, perform another segment in I-configuration with new CAM forcing from final 25 years of Experiment 2.1.3, then another segment in F-configuration and test for steady-state again.

2.2. Experiment 2.2: Control (F-configuration)

2.2.1. Start from the end-point of the Experiment 2.1 (spin-up), reset the simulation year to 1800 and run 205 years (1800-2004).

2.2.2. Data SST and sea ice extent from cycling 25 years (1875-1899) of HadISST forcing.

2.2.3. CO₂, N deposition, and landcover held constant, as in spin-up (Experiment 2.1).

2.2.4. The purpose of this experiment is to isolate any remaining drifts in the model state variables, so that they can be removed from forcing experiments by differencing.

2.2.5. Transport land CO₂ flux in CAM.

2.3. Experiment 2.3: Varying climate (F-configuration)

2.3.1. Start from the end-point of the Experiment 2.1 (spin-up), reset the simulation year to 1800 and run 205 years (1800-2004).

2.3.2. Data SST and sea ice extent for the first 75 years (1800-1874) of this run from three 25-year cycles (1875-1899) of HadISST. Starting in simulation year 1875, switch docn cycling off and use HadISST years 1875-2004 for the final 130 years of the simulation (1875-2004).

2.3.3. Radiative CO₂ follows the C⁴MIP historical transient for 1800-2004 (see section 1.4.2), while biogeochemistry continues to see fixed CO₂ at 283.1878 ppmv (the value circa 1800 from C⁴MIP forcing data).

2.3.4. N deposition landcover are held constant, as in Experiment 2.1.

2.3.5. The purpose of this experiment is to isolate the influence of recent historical climate changes on terrestrial carbon cycle, without the direct biogeochemical effects of increasing CO₂ and increasing N deposition.

2.3.6. Transport land CO₂ flux in CAM

2.4. Experiment 2.4: Varying climate, CO₂, and N deposition (F-configuration)

2.4.1. Start from the end-point of the Experiment 2.1 (spin-up), reset the simulation year to 1800 and run 205 years (1800-2004).

2.4.2. Data SST and sea ice extent for the first 75 years (1800-1874) of this run from three 25-year cycles (1875-1899) of HadISST. Starting in simulation year 1875, switch docn cycling off and use HadISST years 1875-2004 for the final 130 years of the simulation (1875-2004).

2.4.3. Radiative and biogeochemical CO₂ follow the C⁴MIP historical transient for 1800-2004 (see section 1.4.2).

2.4.4. This run uses the N deposition time series from NCAR for the period 1800-2004, as described in section 1.4.3.

2.4.5. Landcover held constant, as in Experiment 2.1.

2.4.6. Turn on data ocean CO₂ and data fossil fuel CO₂ fluxes. The dataset for ocean fluxes currently contains data for years 1850-2000. The dataset for fossil fuel fluxes currently contains data for years 1900-1990. [Obtain improved fossil fuel dataset (Mahowald), and examine the use of data from Doney to extend OCMIP CO2 fluxes out to 2004 (Mahowald, Doney)]

2.5. Experiment 2.5: Varying climate, CO₂, N deposition, and Land Cover (F-configuration)

2.5.1. Everything identical to Experiment 2.4, but adding the forcing of plant functional type areas from the annual time series in Johannes Feddema's new datasets for the period 1800-2004 (described in more detail in Experiment 1.5). Treatment of land cover fluxes follows the C⁴MIP protocol.

2.6. Experiment 2.6: Varying climate, CO₂, N deposition with Seasonalized Fossil Fuel Emissions

2.6.1. Everything identical to Experiment 2.4, but the prescribed fossil fuel emissions are perturbed providing seasonality to the fluxes. This experiment will offer an alternative view of the seasonal cycle of CO₂ at monitoring sites as compared with Experiment 2.4.