Online Resource 2. Uncertainty in Natural
Variability within the IGSM-CAM Framework and
Alternative Initializing Conditions
Climatic Change Article: Quantifying and Monetizing Potential Climate Change
Policy Impacts on Terrestrial Ecosystem Carbon Storage and Wildfires in the
United States
Authors: David Mills, Russell Jones, Karen Carney, Alexis St. Juliana, Richard
Ready, Allison Crimmins, Jeremy Martinich, Kate Shouse, Benjamin DeAngelo, and
Erwan Monier
Corresponding author: David Mills, Stratus Consulting Inc.,
DMills@stratusconsulting.com
The Integrated Global Systems Model (IGSM) Community Atmospheric Model
(CAM) framework (Monier et al. 2013a) links the Massachusetts Institute of
Technology (MIT) IGSM version 2.3 (Dutkiewicz et al. 2005; Sokolov et al. 2005),
an integrated assessment model that couples an earth system model of intermediate
complexity (with a two-dimensional, zonal-mean atmosphere) with a human activity
model, to the National Center for Atmospheric Research three-dimensional, CAM
version 3 (Collins et al. 2006). The IGSM simulates, among other variables,
greenhouse gases (GHGs), ozone and aerosol concentrations, as well as sea surface
temperatures and sea ice cover, which are then used to drive the CAM (see Monier
et al. 2013a, for more details).
The IGSM couples a two-dimensional, zonally averaged statistical dynamical
representation of the atmosphere to a three-dimensional, dynamical ocean component
based on the MIT ocean general circulation model (Marshall et al. 1997) with a
thermodynamic, sea ice model and an ocean carbon cycle (Dutkiewicz et al. 2005,
2009). Heat and freshwater fluxes are anomaly coupled in order to simulate a realistic
ocean state. In order to more realistically capture surface wind forcing over the ocean,
a six-hour observational dataset of surface 10-m wind speeds from 1948 to 2007 is
used to formulate wind stress (Monier et al. 2013a). For any given model calendar
year, a random calendar year of wind stress data are applied to the ocean. This
approach ensures that both the short-term weather variability and inter-annual
variability are represented in the ocean’s surface forcing. Different random sampling
can be applied to simulate different natural variability in the same way as perturbation
in initial conditions.
In addition to the random wind sampling, different initial conditions in the
atmospheric and land components of the CAM, all consistent with the IGSM forcing,
are considered. This is achieved by running a 50-year control simulation with the
CAM, forced by the IGSM GHGs, ozone and aerosol concentrations, as well as sea
surface temperatures and sea ice cover fixed at year 1980 (the starting year of the
IGSM-CAM simulations in this particular project). Initial conditions are then sampled
every 10 years from the control simulation. These alternative initial conditions
1
represent equally likely potential future climates. The different initial conditions are
represented by labeling the scenarios with the following labels: WIND-1, WIND-13,
WIND-14, WIND-26, and WIND-28. WIND-1 is the scenario that all sectoral impact
models in the Climate Change Impacts and Risk Analysis (CIRA) project have
considered in their primary runs. All five of the different initial condition simulations
have been applied in this paper using MC1, a dynamic global vegetation model
(DGVM), to develop projections of future terrestrial ecosystem carbon storage and
acreage burned by wildfires.
Small perturbations in initial conditions, along with differences in wind stress, lead to
the onset of anomalies in global and regional climates to occur in different years. For
example, two simulations with different initial conditions/random wind sampling can
simulate a particular year – say 2050 – with opposite phases of the El Niño/Southern
Oscillation: one simulation would experience El Niño conditions while the other
would exhibit La Niña conditions. Such simulations would produce very different
global teleconnections with the result that a specific region could experience very
different climates in different simulations.
The uncertainty in future natural variability, commonly considered through initial
conditions perturbation, not only impacts year-to-year variability but has also been
shown to affect long-term projections of climate change (Deser et al. 2012a, 2012b;
Monier et al. 2014, this issue, 2013b). As a result, each IGSM-CAM simulation with
different initial conditions represents a possible outcome where the United States
(U.S.) climate for particular years can be very different and where the 20-year mean
climate may show distinct differences (Monier et al. 2014, this issue).
Differences between IGSM-CAM and IGSM-pattern Scaling Methods
Because the IGSM has a two-dimensional, zonal-mean atmospheric component, it
cannot be directly used to simulate regional climate change. To simulate climate
change over the U.S., we use two complementary downscaling methods. First, a
dynamical downscaling method relies on the IGSM-CAM framework (Monier et al.
2013a) that links the IGSM version 2.3 to the three-dimensional, atmospheric CAM.
Second, a statistical downscaling is based on a Taylor-expansion pattern scaling
algorithm (Schlosser et al. 2012) that extends the latitudinal projections of the IGSM,
two-dimensional zonal-mean atmosphere by applying longitudinally resolved climate
patterns from observations and from climate model projections from the Coupled
Model Intercomparison Project phase 3 (CMIP3). The two-pronged approach is
described in more detail in Monier et al. (2014, this issue), and results and
comparisons between the two downscaling methods are presented. Both methods can
simulate future regional climate change under different climate system responses
(using different values of climate sensitivity, ocean heat uptake rates, and net aerosol
forcing) and different emission scenarios. Each approach has its strengths and
limitations, which are discussed below.
Generally, the IGSM-CAM simulations display a strong inter-annual variability,
especially for precipitation, which is in very good agreement with the observations
over the historical period. As a result, comparing two particular simulations to
2
identify the impact of, for example, the implementation of a stabilization policy or
different values of climate sensitivity is generally difficult. However, by averaging
the results of impacts developed from multiple simulations with different initial
conditions, the signal can be more easily extracted from the noise (note that initial
averaging of the IGSM-CAM data prior to evaluation would instead mute the desired
variability that the climate model introduces). Meanwhile, simulations with the
IGSM-pattern scaling method show limited inter-annual variability, even less than the
IGSM-CAM ensemble mean simulations. This is because the temporal variability in
the IGSM-pattern scaling method is controlled entirely by the IGSM zonal mean,
which displays a much weaker variability than would any particular grid cell along
the same latitude. For this reason, the IGSM-pattern scaling method underestimates
natural variability and its potential changes, as well as climate and weather extreme
events.1
Finally, a major advantage of the pattern scaling method is that it considers the
regional patterns of change of different models and thus takes into account structural
uncertainty. In this project, the pattern scaling method is applied to three CMIP3
models – the driest (Model for Interdisciplinary Research on Climate, MIROC) and
the wettest (BCCR) over the U.S., and the climate model that shares the same
atmosphere as the IGSM-CAM (Community Climate System Model, CCSM) – and
the multi-model mean. The differences between the driest and the other models are
substantial and, as shown in this paper, result in very different impacts. In contrast,
the IGSM-CAM framework revolves around a single atmospheric model, and thus
likely underestimates structural uncertainty.
In conclusion, the IGSM-CAM and pattern scaling methods show complementary
strengths. The IGSM-CAM method simulates realistic natural variability at the global
and regional levels, as well as future changes in natural variability (changes in
magnitude and frequency). This should be particularly important for sectoral impact
analysis where the impact is largely driven by thresholds and/or non-linear response
functions, while the pattern scaling method allows regional patterns of change from
multiple climate models to be considered.
References
Collins W, Rasch P, Boville B, Hack J, McCaa J, Williamson D, Briegleb B, Bitz C, Lin S, Zhang M
(2006) The formulation and atmospheric simulation of the Community Atmosphere Model version 3
(CAM3). J Climate 19(11):2144–2161. doi: 10.1175/JCLI3760.1
Deser C, Knutti R, Solomon S, Phillips AS (2012a) Communication of the role of natural variability in
future North American climate. Nat Clim Change 2:775-779. doi: 10.1038/nclimate1562
Deser C, Phillips AS, Bourdette V, Teng H (2012b) Uncertainty in climate change projections: The
role of internal variability. Climate Dyn 38:527–546. doi: 10.1007/s00382-010-0977-x
1. In the case of how the IGSM pattern scaled data were applied for the MC1 analysis, delta values
were applied to a historic transient of observed monthly data in an effort to better represent climatic
inter-annual variability. Online Resource 4 details the processing of the climate inputs used for the
IGSM pattern scaled models.
3
Dutkiewicz S, Follows M, Bragg J (2009) Modeling the coupling of ocean ecology and
biogeochemistry. Global Biogeochem Cycles 23(4):GB4017. doi: 10.1029/2008GB003405
Dutkiewicz S, Sokolov A, Scott J, Stone P (2005) A three-dimensional ocean-seaice-carbon cycle
model and its coupling to a two-dimensional atmospheric model: Uses in climate change studies. Tech.
Rep. 122, MIT Joint Program on the Science and Policy of Global Change
http://globalchange.mit.edu/files/document/MITJPSPGC_Rpt122.pdf. Accessed 22 July 2013
Marshall J, Hill C, Perelman L, Adcroft A (1997) Hydrostatic, quasi-hydrostatic, and nonhydrostatic
ocean modeling. J Geophys Res 102:5733–5752. doi:10.1029/96JC02776
Monier E, Gao X, Scott J, Sokolov A, Schlosser A (2014) A framework for modeling uncertainty in
regional climate change, Climatic Change (submitted, this issue)
Monier E, Scott J, Sokolov A, Forest C, Schlosser C (2013a) An integrated assessment modelling
framework for uncertainty studies in global and regional climate change: The MIT IGSM-CAM
(version 1.0). Geosci Model Dev 6:2063-2085. doi:10.5194/gmd-6-2063-2013
Monier E, Sokolov AP, Schlosser CA, Scott JR, Gao X (2013b) Probabilistic projections of 21st
century climate change over Northern Eurasia. Environ Res Lett, 8, 045008. doi:10.1088/17489326/8/4/045008
Schlosser CA, Gao X, Strzepek K, Sokolov A, Forest CE, Awadalla S and Farmer W (2012)
Quantifying the likelihood of regional climate change: A hybridized approach. J Climate 26:3394–
2414
Sokolov A, Schlosser C, Dutkiewicz S, Paltsev S, Kicklighter D, Jacoby H, Prinn R, Forest C, Reilly J,
Wang C, Felzer B., Sarofim MC, Scott J, Stone PH, Melillo JM, Cohen J (2005) MIT integrated global
system model (IGSM) version 2: Model description and baseline evaluation. Tech. Rep. 124, MIT
Joint Program on the Science and Policy of Global Change
4