Most of the data files provided here are direct copies of the original PMIP3 or CMIP5 datasets. If you use any of these datasets for publication please reference the data as indicated on the PMIP3 or CMIP5 web sites. Files for the RCP extensions, which are specified to year 2500, but only used here to 2300, have been shortened to avoid potential confusion.

Maps of land-use change and emissions of aerosols are specified as RCP forcing for GCMs. Some EMICs can not use spatial maps of vegetation to simulate forcing from land-use change. Most EMICS do not have the appropriate atmospheric chemistry to calculate the direct and indirect effects of aerosols from emissions. Equivalent modelled SO4 concentrations are also specified for the RCPs which can be used with models (including GCMs) that lack the appropriate aerosol chemistry. Emissions and concentration data are available from the RCP download page. This page may require signing in to gain access. If the spatial emissions or concentrations of aerosols or maps of land-use change can not be used, then it is recommended that the globally averaged estimates of radiative forcing from Task Group: RCP Concentrations Calculation and Data be used. For sulphate aerosols use "Global SOX (Fossil & Ind) Radiative Forcing" and for land-use use, "Global LANDUSE Radiative Forcing". Radiative forcing data can be found in the file

Links to modified versions of some of these datasets are available at the bottom of this page and are provided for convenience. See UVic Forcing Data below. These processed files are all in compressed netcdf format and often are considerably smaller than the original datasets. If used, they should be checked against the original datasets. Use with caution. Even if you do not use the processed versions of these datasets, this section may provide hints on how to process your own data.

4. Historical Simulations

Initial Condition
Spin up at 850 boundary conditions or alternatively, spin-up as pre-industrial (PI), where PI is defined as in CMIP5 (~1750 to ~1850). A CO2 concentration of 280 ppmv is preferred but not required. It is recommended that you spin up with the average of the solar irradiance over a solar cycle. The average irradiance from the recommended Delaygue and Bard dataset between 850 and 861 (next solar cycle) is 1365.763 W m-2.

Control Simulation (850 to 2005)
Control simulation with constant 850 forcing (from the forcing specified below) or alternatively, a control simulation with constant PI forcing for models spun up at PI. This is usually just a continuation of the constant forcing simulation used to provide the initial condition.

The simulations over the last millennium will follow the PMIP3/CMIP5 protocol. Further details are available from Schmidt et al. (2010) and: All forcing that is compatible with the model should be used. If a model is not capable of using the prescribed forcing then the globally averaged radiative forcing from Task Group: RCP Concentrations Calculation and Data may be used for forcing beyond 1765. If radiative forcing estimates are required before 1765, it is left to the modelling groups to decide how best to specify the forcing. If any forcing is applied differently than specified here or not applied, please indicate this when submitting any model results.

Forcing (850 to 2005):

Compatibility with RCP forcing:
All forcing from 1850 to 2005 should be the same as for the historical portion (1850-2005) of the RCP scenarios (see below). Any blending of datasets should happen before 1850.

Additional optional forcing:
PMIP3 does not recommend singular volcanic, solar irradiance or land-use datasets. Although it is recommended here to use the datasets indicated above, please consider doing additional simulations using other PMIP3 recommended volcanic (Gao-Robock-Ammann), solar irradiance (from Muscheler, Joos, Beer, Müller, Vonmoos and Snowball: 850-1609, Vieira, Krivova and Solanki: 850-1849 and Steinhilber, Beer, and Frohlich: 850-1849 - see "Solar Irradiance" file above) and land-use (Kaplan et al.) forcing data .

5. Representative Concentration Pathways

Initial Condition
Use the end state (2005) of the "total" forcing simulation from the "historical" last millennium experiment above or alternatively, spin up as for CMIP5 PI and perform the simulation from PI to 2005 using the historical portion of the RCP forcing. It is recommended that you spin up with the average of the solar irradiance over a solar cycle and a CO2 concentration of 280 ppmv is preferred but not required.

Control Simulation (2005 to time of longest experiment)
A continuation of the last millennium "historical" control simulation with constant 850 forcing (see above) or alternatively, a continuation of the spin up at PI with constant PI forcing. This is usually just a continuation of the constant forcing simulation used to provide the initial condition.

Representative Concentration Pathways (RCPs) have been designed to provide radiative forcing data for modelling groups wishing to contribute to the IPCC AR5. Information on the RCPs is available at: The Coupled Modelling Intercomparison Project (CMIP) have taken the RCP data and designed a suite of experiments for the CMIP5 project. See Taylor et al. (2009) and for details.

All forcing that is compatible with the model should be used. If a model is not capable of using the prescribed forcing then the globally averaged radiative forcing from Task Group: RCP Concentrations Calculation and Data should be used.

f Historical RCP Forcing (PI to 2005):
Only required if not starting from the end of the last millennium experiment. RCP 2.6 (3PD) Forcing (2005 to 2300):
As for historical RCP with: RCP 4.5 Forcing (2005 to 2300):
As for historical RCP with: RCP 6 Forcing (2005 to 2300):
As for historical RCP with: RCP 8.5 Forcing (2005 to 2300):
As for historical RCP with:

8. Allowable cumulative emissions (2005 to 2500)

For these simulations, only EMICs coupled to a dynamic carbon cycle component are appropriate. The simulations outlined in this section require code modifications that allow the model to "track" a specified temperature profile. This is accomplished by diagnosing emissions that will produce a level of CO2 that is compatible with the desired temperature.

Generating temperature profiles
Temperature profiles should change smoothly from the historical simulation to the idealized profile in order not induce large emissions near the transition. The slope from the last 10 years of the historical simulation can be used to specify the starting slope for a spline fit to a final temperature. A simple cubic spline interpolation program that generates a temperature profile from the end of a historical simulation is available as a Fortran 90 routine (sat_profile.f90) or as an IDL procedure ( Temperature profiles that change smoothly to 1.5, 2, 3, and 4 C above PI are needed. It may take some experimentation to achieve smooth profiles.

Generating running average temperature
A running annual average of the global average surface air temperature (SAT) needs to be calculated so that it can be compared to the temperature target. This average needs to be calculated whenever emissions are calculated. This can be done by storing global average SAT in a temporal array. As an example, in the UVic model emissions are specified every 5 days (coupling time between model components) so the array holding a years worth of global average SAT has a dimension of 72. If a model has a great deal of interannual variability the average can be taken over several years by increasing the size of the SAT accumulator array. Every time emissions are calculated the appropriate global SAT value is updated in the array and the annual (or multi year) average is calculated by averaging all values in the array. The array index to be updated can be obtained with a simple counter that continuously cycles through the array. The target air temperature is compared to this average and appropriate emissions are derived. If the model is stopped and restarted during a "tracking" simulation then the global average SAT array and index counter must be saved when the model stops and read in again when the model starts. The SAT accumulation array must be full, before emissions can be calculated.

Calculating emissions
Emissions are calculated in order to produce desired changes in SAT. If the global average temperature becomes too high or too low then emissions are reduced or increased until the desired temperature is achieved. This control can be accomplished by using a simple, heavily damped proportional control algorithm. emissions = k*(delta SAT) where delta SAT is the desired change in temperature and k is a constant that relates emissions to temperature change through CO2. This constant is roughly inversely proportional to the model's climate sensitivity per unit of CO2 emission. For levels of CO2 near present day this might be something like 2.13*280/3/timestep for a model with a climate sensitivity of 3C, for a doubling of PI CO2, and a ppmv to Pg emissions conversion of 2.13. This constant is usually reduced considerably from this idealized value to ensure an over damped system.

Timing considerations
The time associated with the annual average temperature profile is usually reduced by half of the averaging time that is being used to calculate the model's running average temperature (see above). For example, if the model is simulating the start of 2011 and the model is calculating a running average over just one year, then the running average represents the temperature for the middle of 2010 and any calculated emissions should be calculated from the difference in temperature (target-model) for the middle of 2010 but applied at the beginning of 2011. In other words there is a delay of 6 months from when emissions are calculated and when they are applied. This does not significantly affect the ability of the model to "track" global annual temperature as long as the SAT profiles are relatively smooth but it is one of the reasons why the minimum averaging time, that removes most of the model's natural interannual variability, should be used. Changing from specified emissions or concentrations, to emissions calculated from lagged temperature change can cause some small discontinuities in emissions. This is why some care should be taken to match the slope at the end of the simulated portion of the temperature profile to the slope at the beginning of the specified portion. Usually these are discontinuities are small but if they are large it may be best to construct a temperature time series that includes both the historical period and the future smooth profile, and perform temperature "tracking" from PI though to 2500.

Other forcing
This experiment calculates the radiative forcing required to follow a specific temperature profile. The radiative forcing is assumed to come from emissions of CO2, however, changes in other forcing would also change the calculated amount of CO2 required to meet a specific temperature target. Here it is assumed that other forcing either remains constant (such as natural forcing from changes in solar irradiance or volcanic activity) or decreases to zero by 2300. Best estimates of other anthropogenic forcing such as sulphates and additional greenhouse gases nearly cancel giving a near net zero contribution to the radiative forcing at present day. An assumption is that this will continue into the future so changes in radiative forcing can be expressed in terms of CO2 alone. Although the radiative forcing from sulphates and other greenhouse gases nearly cancel, these should be linearly reduced from their levels at 2005 to those at PI, by 2300. All other forcing should be held constant at 2005 levels. Beyond 2005, solar irradiance should be held fixed at the average of the last solar cycle (years 1996 to 2008 inclusive).

Code changes
The details of how to implement temperature "tracking" will depend on specifics of the model but the following code details how this is done in the UVic model. Search for variables with "track_sat" in them to find the appropriate code.

Any questions about implementing temperature "tracking" should be addressed to the contact person given on the UVic model page.

UVic Forcing Data

This data has been converted into a form that can be used by the UVic ESCM. Other models may require different conversions. No spatial interpolation has been performed. All data is in netcdf with the time variable specified in years. The time variable is referenced to year 0, which technically does not exist and although some netcdf browsers that interpret calendars may be confused by this, it allows calendar years to be specified which may be less confusing for modellers. The calendar is for a constant 365 day year (no leap years).The data is compressed with the gnu gzip utility. Fortran extraction and conversion routines are available upon request. Any questions or concerns about this data should be addressed to the contact person given on the UVic model page. A single gzipped tar file containing all of the following data can be downloaded here: UVic_forcing_data.tgz (719 MB). The file can be uncompressed into a UVic_forcing_data subdirectory with the command "untar -xvzf UVic_forcing_data.tgz".