Since there are no currently active contests, we have switched Climate CoLab to read-only mode.
Learn more at https://climatecolab.org/page/readonly.
Skip navigation

EMF 22 response surfaces


Overview #

Model name

EMF 22 response surfaces

Brief description

Estimates of the mitigation costs associated with varying levels of emission reductions are generated through use of a response surfaces derived from the international scenarios developed in the Stanford Energy Modeling Forum's EMF 22 exercise.

Models used

These response surfaces are based on runs of 6 integrated assessment models that participated in the international scenario exercise:

Some of the models used in the EMF 22 exercise were not used to create mitigation cost response surfaces because they did not report variations in GDP across the scenarios (ETSAP-TIAM, IMAGE, POLES) or because of large differences in the costs associated with their stabilization vs. their overshoot scenarios (MESSAGE).
\ Key publications

Leon Clark, Jae Edmonds, Volker Krey, Richard Richels, Steven Rose and Massimo Tavoni. International climate policy architectures: Overview of the EMF 22 International Scenarios. Energy Economics, 31, Supplement 2, (2009): S64-S81.

Papers based on the results from individual models were published in a 2009 special issue of Energy Economics.

Data sources

Published data from the international scenarios exercise used to generate the response surfaces is downloadable from EMF Briefing on Climate Policy Scenarios: U.S. Domestic and International Policy Architectures, June 4, 2009.

Variables and key assumptions #

Rationale for generating response surfaces

The response surface methodology used is intended to approximate the results from the original EMF 22 exercise.

In some cases the assumptions used in the original models differ from those in C-LEARN runs that generate inputs for the response surface, and this may lead to further inaccuracies.

For example, the allocation of emission cuts across regions could vary between a C-LEARN run and a model run in the EMF 22 exercise.

We believe, however, that the results of these response models are accurate enough to provide a useful estimate of the likely impacts.

Input variables

Global reductions in 3 greenhouse gases (CO2, NH4 and NO2 expressed as CO2 equivalent of CO2e values) at ten-year intervals between 2000 and 2100. (e.g. 2000, 2010, 2020, etc.)

Output variables

Reduction in global gross domestic product (GDP) vs. the reference scenario for the given level of the input variable ten-year intervals between 2000 and 2100 (e.g. 2000, 2010, 2020, etc.).

Key assumptions

The EMF22 international scenarios exercise involved creating a series of ten model runs that simulated potential emission stabilization scenarios.

The scenarios involved stabilizing greenhouse gas concentrations at 3 levels:

They also involved 2 assumptions concerning the timing of participation in mitigation efforts:

And there were two assumptions concerning whether emissions could ever exceed the stabilization target:

The modeling teams were asked to create 10 stabilization scenarios.

Timing 450CO2e 550CO2e 650CO2e
Full Not-to-exceed Not-to-exceed Not-to-exceed
Full Overshoot Overshoot
Delayed Not-to-exceed Not-to-exceed Not-to-exceed
Delayed Overshoot Overshoot

The models model were run for each of these ten scenarios, and key data outputs at 10 year intervals were published in the final report and in the accompanying data tables.

These outputs included emissions (globally and by key region) of CO2 and other greenhouse gases (GHGs); carbon absorption by the ocean and land; atmospheric concentrations of CO2 and other GHGs; radiative forcing caused by accumulation of GHGs in the atmosphere; marginal costs of emission abatement; detailed information about energy production and prices; amount of carbon sequestration by region and sector; and GDP for the world as a whole and for each key region.

Methodology for creation of response surfaces

Response surfaces were initially created from the results of the Full participation scenarios only.

This choice was based on the assumption that the key developing economies in the EMF 22 scenarios, the so called BRIC countries (Brazil, Russia, India, China) would likely anticipate the introduction of emission reduction policies after 2030 in the energy sector investments the make between now and then.

One of the modeling teams examined the likely outcomes of such delayed participation with foresight. It found that the costs of delayed participation with foresight were:

For more, see Valentina Bosettia, Carlo Carrarob, and Massimo Tavoni. Climate change mitigation strategies in fast-growing countries: The benefits of early action. Energy Economics, 31, Supplement 2, (2009): S144-S151.

Given this finding, use of the Full participation scenarios can provide an order-of-magnitude estimate of overall costs of various stabilization policies.

In addition, it is assumed that financial transfers between developed and developing countries, including those that occur under the UN's Clean Development Mechanism can encourage emission reductions in line with the Full participation scenarios.

The CoLab team created a spreadsheet for each model based on two outputs from the EMF 22 data:

The CoLab team plotted all the points for all scenarios for which results were reported by the modeling team.

Each line in the plot corresponded to one of the emissions scenarios (e.g. Reference, 650 ppm CO2e stablization) for each year reported (e.g. 2010, 2020, etc.)

Reductions in GHG emissions vs. 2005 were plotted on the x-axis and reduction in global GDP vs. Reference scenario on the y axis.

The team then derived equations that plotted a linear interpolation between the points.

After deriving these equations, the team created a response surface based on them.

The reduction in GHG emissions vs. 2005 for each year in the time series (e.g. 2020, 2030, etc.) was used as the input, taken from runs of C-LEARN generated by Climate CoLab members in the course of creating proposals.

The response surface equations then derived the output, reduction in global GDP vs. the reference scenario for each year (e.g. 2020, 2030, etc.)

These outputs are plotted, with lines connecting the points, in the "Mitigation costs" section of the Climate CoLab's Actions/Impacts tab.

The low and high values from the EMF 22 data sets are the lower and upper bounds for the response surface.

If the input value of atmospheric concentration of CO2 is below the lower bound or above the upper bound, the response surface will not calculate a value for reduction in global GDP.

In such a case, that portion of the curve in the "Mitigation costs" plot would be left blank, and a message is generated on the chart.

For more on that rationale behind this approach, see Mitigation cost response surface boundaries.

Strengths and weaknesses #

The response surface derived from the EMF 22 exercise provides Climate CoLab users with a quickly calculable estimate of the projected economic impact of emission reduction pathways.

These estimates are approximations, since the C-LEARN runs that generate inputs for the response surface do not correspond in every detail with the runs of the actual models that served as the basis for the response surface.

In relying on these response surfaces, the Climate CoLab team seeks to provide users with a quick-running estimate of the relative environmental and economic trade offs involved in various proposals to address climate change.

To address computational constraints, these response surfaces necessarily sacrifice precision.

The Climate CoLab team hopes to undertake sensitivity analysis in the near future to assess how great a loss of precision results from the use of response surfaces like this one.