Skip to: [content] [navigation]

The purpose of these meetings was to present a preliminary analysis of our current 1860 run, and more clearly define the guiding philosophy for ESM development and implementation through a discussion of the various potential physical initialization strategies, carbon initialization strategies and potential carbon system "show stoppers".

A) Preminary Analysis of 60 year ESM2.1 1860 control run:

Anand Gnanadesikan presented the model results. In the global scale, results were surprisingly good. However, the Amazon is drying out considerably, much more than in CM2.1U. His presentation can be downloaded here

Result - LMDT will take on the issue of the desertification of the Amazon to diagnose where the problem lies, and hopefully how to improve it.

B) EM2.1 Initialization procedure

This run was performed with the radiation configuration of Jan 1, 1860 (CO2 = 285.97865 ppmv), but with the land and ocean biogeochemistry "feeling" 276 ppmv CO2 for equilibration.

Issues:
In this run, we are attempting to accomplish three separate goals:

Spinning up land and ocean carbon cycle
Assessing the fidelity of the model solution by comparing with present-day obs.
Assessing biospheric feedbacks/impacts to the fidelity of the model solution by compare directly to CM2.1U results
Analysis of the 1860 control run is non-ideal because it makes it difficult to do a direct comparison with "present day" data.
Analysis of the 1990 control run is non-ideal because the model and real-world systems are both out of radiative, ocean nutrient, and carbon balance

Lessons from CM2.1 configuration for initialization, development and spinup:

All configurations with first initialized from data, and then spun-up as component models with data boundary conditions (AMIP/CORE) then further spun up in the intermediate coupled model configurations, before the final runs were performed.
Ocean - Levitus temp and salt was run for 1 year with CORE forcing to spin up currents while retaining large scale temp/salt structure.
Ice - climatology
Atmosphere and Land - A cold start (isothermal, stationary atmosphere) was done with the atmosphere AMIP was then run for 1979-1986
Control_1860_D4 - 1860 greenhouse gases i.e. CO2 = 286 ppm) were run for 241 years, then restarted in testing (and final) configuration
Control_1990_E1 - 1990 greenhouse gases (i.e. CO2 = 353 ppm) were run for 241 years, then restarted in testing (and final) configuration
Historical run (1860_2000_H2) - Time-varying greenhouse gases were run from year 41 of Control_1860_D4.
Even in the case of the physics-only CM2.1U, many steps were required for initialization. In particular, efforts were made to assure that each component was both stable and as close to observations as possible in it's interface with other models - e.g. allowing the ocean to have currents but not let it change SST too much before coupling with the atmosphere. Ron Stouffer commented that, in retrospect, the physics spin-up should have been given at least 500 years to decrease drift, but that time-pressure did not allow it.

ESM2.1 greenhouse gas configuration options:

1750 - undisturbed vegetation, CO2 = 278 ppmv - the benefit is that this scenario allows us to cleanly spin up the carbon system before anthropogenic impacts. The liabilities are that there is no data to establish fidelity, we do not know how how adversely the CO2 will impact vegetation, the CO2 level from the middle ages with more like 283 ppm, so 278 may be overextending the industrial change, and there is not currently a CM2.1U run for these exact conditions available for direct comparison with the ESM... perhaps we should run a Control_1750 in order to do these comparisons?
1860 - beginning of the industrial revolution, CO2 286 ppmv - the benefit is in comparability to the CM2.1U control run. The liability is that land use and CO2 have some anthropogenic impacts, such that the total CO2 change from full "preindustrial" to 1990 is diminished by 11%.
1990 - Present day, CO3 = 353 ppmv - the benefit is that these are the current conditions of the Earth system under which most of the observations were taken. The liabilities are that the spin-up result would not be applicable as an initial condition for the pre-industrial era and that, at the present time, the earth system is currently out of balance in terms of radiation, CO2, and land use, all of which are increasing. Since it takes on the order of 100 years for the vegetation and surface ocean biogeochemistry to come into balance, the conditions would be difficult to interpret, since the observations have their own time-scale of response embedded within them.
Fixed to historical concentrations - the benefit of this would be that it would follow the real forcing. The liability is that the system would not be in steady state, and so temporal changes would be impossible to attribute uniquely to model drift versus climate response.
Result - Until we get through the development, initialization and spin-up and phases, all ESM runs will be in the 1860 configuration - this means that BOTH RADIATION AND BIOGEOCHEMICAL CO2 will need to be equilibrated to the Jan 1, 1860 value (285.97865 ppmv) to avoid shock after spin-up. The previous run had used 276 ppmv as the biogeochemical CO2. As we assess the fidelity of this run, we will be comparing results to present day data for land and ocean biogeochemistry with the implicit assumption that large-scale features have not changed.

C) Fidelity Targets

The central question is whether we need to establish strict fidelity criteria that determine when model development is complete...

CM2 analogy - CMDT established the following three criteria that the model development was forced to satisfy before it could be considered viable and complete:

Modeled sea surface temperatures everywhere < 10 deg. C from Levitus on the annual mean
North Atlantic Deep Water formation rates had to be at least 10 Sverdrups.
The model had to exhibit some form of El Nino - where El Nino was somewhat loosely defined as a tropical variability with a period between 1 and 5 years with an intensity of the same order as that observed.

The ESM effort, however, is different from the CM2 effort in at least four important ways:

Most of the physics driving the biogeochemistry has already been optimized for the particular configuration, so simple physical improvements are unlikely

The ESM is more complex than a climate model alone as it includes both physical and biogeochemical components with potentially much more sophisticated representation of the role of and impact on humans. Thus, it behooves us to charge ahead and not get too bogged down in details which may prove irrelevant.

We don't have nearly the kind of experience with the ESM class of models as with do with climate models, so we have much less idea as to what features and defects of the model the results will be most sensitive.

The experience we do have tells us that some of the biogeochemical time scales are much longer because of terrestrial soil processes and ocean interior sources and sinks of tracers.

Result: For now, the only target that we will consider absolutely necessary is that both the land and ocean models spin up to an equilibration criterion wherein the net flux to/from the atmosphere - which we will specify as having to be less than 2.0 Pg C / decade - is dwarfed by the anthropogenic increase which will be applied (2 Pg C / yr).

Other targets we will consider aiming for are:
total land biomass (600PgC)
total land soil carbon (2000 PgC)
average surface nitrate concentration (5 uM)
average 1000m nitrate concentration(34 uM)

This discussion brought up the larger question of whether or not we even attempt to tune in the ESM, or simply incorporate tunings from component models?
Result: Tuning within the ESM should be avoided as much as possible. When problems are suspected in the simulation, they should be worked out in the component models wherever possible.

D) Plans for next two months/continuing action items.

1) Achieve frozen rts checkout based on Lima that is capable of reproducing CM2.1

2) Finalization of ocean carbon system initialization plan.