Atmospheric Chemistry and Climate Model Intercomparison Project (ACCMIP)

Experiment Specifications

Emissions

Consistent gridded emissions dataset from 1850 to 2100 for modeling studies in support of IPCC AR5 have been created by an international team recently (Lamarque et al., 2010). Emissions of gaseous and particulate species (i.e. aerosols, ozone and aerosol precursors) from anthropogenic activities and biomass burning have been estimated over the full period, using the 2000 dataset for harmonization of the past and current 1850-2000 emissions with the future emissions determined by the Integrated Assessment Models (IAMs) for the four IPCC Representative Concentration Pathways (RCPs). These emissions will be used as boundary conditions for chemistry/aerosol model simulations in ACC-MIP.

ACCMIP_1: Timeslice runs complementing CMIP5

Timeslice runs including detailed chemistry diagnostics to provide information on the forcings of historical and future climate change in the CMIP5 simulations. Each run 4-10 years with prescribed SSTs taken from AR5 runs (SSTs should ideally be decadal means around given years), 2-month initialization suggested. Core runs will ideally be 10-years, with future analyses to test whether 4 years of simulation is adequate (or which quantities it is adequate for).

Additional runs for 2000 with 1850 climate and for 2030 and 2100 (RCP 8.5) with 2000 emissions are designed to separate the effects of climate change on constituents and for isolating aerosol indirect effects more cleanly using the clear-sky/all-sky flux diagnostics. For other time periods and RCPs, these diagnostics will again be used to diagnose the AIE, but removal of climate-induced cloud feedbacks will be required based on the transient 1% per year CO2 CMIP5 runs. Cloud forcing diagnostics need to calculate fluxes with and without clouds, but to isolate cloud forcing only need to use identical 1850 radiatively active constituent fields in all simulations (see output specifications).

Historical Simulations
Emissions/Configuration 1850 1890 1910 1930 1950 1970 1980 1990 2000
Emissions and SSTs/GHGs for given year C 1 1 C 1 1 C 1 C
Given year emissions/1850 SSTs & GHGs 1 1 C
Future Simulations
Emissions/Configuration 2010 2030 2050 2100
RCP 2.6 C 1 C
RCP 4.5 1 1 1 1
RCP 6.0 C C 1 C
RCP 8.5 C 1 C
Year 2000 emissions/RCP 8.5 SSTs & GHGs C C

C = core, 1 = Tier 1, blank = not requested

For the runs with different emission and 'climate' years, e.g. Em2000Cl1850, emissions of aerosol and ozone precursors are set to 2000, methane amounts for chemistry are set to 2000, but ozone and methane at 2000 do not affect the radiation (i.e. radiation sees 1850 'climate' conditions for everything but aerosols). The result is that the aerosols are the only thing affecting radiative fluxes, including the changes they induce in clouds, etc. In essence then, 'chemistry' sees the emissions year, while 'radiation' sees the climate year. For aerosols in a model that includes aerosol indirect effects, the cloud responses could arguably be put in either category. Purely for practical reasons, aerosol direct and indirect effects influence radiation, but not ozone. This is arbitrary and only to allow fewer simulations - if we turned off the AIE we'd have to have additional control runs. This setup allows diagnostics of total aerosol flux perturbations (direct + indirect + rapid responses), and from the same simulations the influence of emissions and 'climate' on pollutants can be separated. The latter is perhaps imperfect, however, as the difference between the standard simulation and these setups for pollutants is due to the climate response to LLGHGs plus ozone. If the ozone changes were to influence the radiation in these setups, then we'd have a cleaner separation of the climate influence of LLGHGs, but we'd compromise our ability to diagnose AIE (and the separation of 'climate' is somewhat imperfect anyway, since the SST changes include the influence of aerosols and ozone, so it's not just LLGHGs). Hence the setup described here.

Total of 14 core simulations, so ~76 model years [leaving out Tier 1 (secondary priority) simulations].

ACCMIP_2: Emission sensitivity studies

Simulations to determine sensitivity to fully or partially natural emissions that will vary between models.

Run at year 2030 (SSTs as in #1), 4 year runs (+2-month initialization), model's own distribution of given emission scaled uniformly:
2.1: +100 Tg/yr isoprene (scale existing source to add 100 Tg/yr).
2.2 +20% biomass burning (all species).
2.3a: +8 ppb (mol/mol) or +50 Tg/yr methane (depending on if running with prescribed concentrations or emissions). This run should be 8 years instead of 4, which will still not necessarily reach equilibrium but should characeterize the exponential approach to equilibrium adequately.
2.3b: 2100 RCP8.5 methane concentration (for chemistry only, not radiation. All else at 2000).
2.4: +2 Tg N/yr lightning NOx (scale existing lightning source to add 2 Tg/yr N).

We assume the effects of varying dust, sea-salt and DMS emissions can be adequately accounted for with emission and removal diagnostics and fully endorse the AeroCom simulations with prescribed optical properties. [total model years requested is 16.5]

ACCMIP_3: Testing the variation in socio-economic modeling of emissions

This set of runs requests the same 4 year timeslice runs as in ACCMIP_1 for 2050 and 2100 but using emissions for the 2.6 and 4.5 RCPs from the other available Integrated Assessment models (IAMs). Runs without AIE only (as these are not climate runs) [8 runs, 32 model years if 2 additional IAM datasets used]

ACCMIP_4: Spread in models using standardized composition

Rerun of 1850, 1930, 1970 and 2000 4 year timeslices with standard 3D constituent fields from ACC Activity 4 Phase 1 climatology. Only applicable to models that did not use climatology in ACCMIP_1 runs. [~17 model years] (Motivation: Variation in the climate response across models will be a function of (a) different climate sensitivity in the GCMs, (b) different impact of aerosols on climate (due to location with respect to clouds, water uptake, natural aerosols, mixing, etc), and (c) different 3D constituent fields from the composition models. ACCMIP_4 will allow us to separate the effect of step 'c' from steps 'a' and 'b'.)

ACCMIP_5: (Proposed) Climate forcing by emission sector

Simulations setting one sector's emissions of all compounds at a time to 1850 while all others remain at 2000.

Output Specifications

Minimal data will be archived from most CMIP5 transient runs. This will include concentrations only. Additional output from the ACCMIP runs will include concentration/mass of radiatively active species, aerosol optical properties, and radiative forcings (clear and all sky) as well as important parameters that do not directly influence climate such as hydroxyl, chemical reaction rates, deposition rates, emission rates, surface pollutants and diagnostics of tracer transport.

CMOR tables have been created, largely based on fields archived for HTAP, AeroCom, and/or CCMVal. All data follow standardized formats and use CF-compliant names whenever available. The tables include several newly created diagnostics for ACCMIP. These include a stratospheric ozone tracer. The stratospheric ozone tracer is defined as equal to ozone in the stratosphere, and including standard ozone removal (but not production) in the troposphere, where the tropopause is the WMO meteorological tropopause. We also include a passive tracer of transport within the troposphere as defined in the HTAP project. Fields requiring high temporal resolution (surface pollutants) have been designated secondary priority. Submission of these fields is encouraged, but we recognize that these can be quite large files and many global models are not ideally suited to air quality studies. Submission of all other fields are strongly encouraged (those that are available, e.g. models without stratospheric chemistry do not need to submit the stratospheric diagnostics).

Isolating aerosol indirect effects (AIE) on clouds requires great care in the calculation of cloud forcing diagnostics. Models participating in the cloud-forcing MIP (CF-MIP) should already have the capability to separately call their radiation code with their GCM's cloud field and with clouds removed. In order to use these cloud forcing calls to diagnose the AIE, it is crucial that both radiation calls have common reference conditions for all radiatively active gases and aerosols. Otherwise, changes in the direct effect of aerosols, for example, would be incorrectly aliased onto cloud forcing. Consider the comparison between the two core runs with 1850 SSTs/GHGs, one with 1850 ozone and aerosols and one with 2000 ozone and aerosols. Changes in clouds between these two runs should be driven by the aerosols (with a minor contribution from ozone via a semi-direct local heating effect). However, were the clear-sky calculations to use the aerosol fields for these two time periods, then the difference in cloud forcing, All-sky(2000)-Clear-sky(2000) vs All-sky(1850)-Clear-sky(1850), would include the influence of aerosol changes on the clear-sky (cloud free) fluxes. Hence all four calls to the radiation must use the same constituent fields. We recommend that these be constant reference preindustrial (1850) fields for aerosols, ozone, and greenhouse gases. We realize that the clear-sky fluxes with reference constituents will be very similar in all runs, but as land temperatures can adjust to constituent changes there may be some small differences so that these diagnostics will still be useful to ensure that the AIE is as clearly isolated as possible. These diagnostics are included in the CMOR tables.

Data is being archived at the British Atmospheric Data Center, with a data access policy providing one year of access to participating groups only followed by general public access.

Return to ACCMIP Homepage