Climate Scenarios and Models Used For Developing the 2008 Pacific Northwest Climate Change Scenarios
As part of the global effort to quantify future changes in climate, the Intergovernmental Panel on Climate Change (IPCC) has developed different scenarios of change in greenhouse gas and sulfate aerosol emissions for use in global climate modeling efforts. These scenarios are grouped in four categories, or storylines, based on different assumptions about demographic, social, economic, technological, and environmental change.
The CIG chose two scenario families for its most recent update of the Pacific Northwest climate change scenarios:
Both scenario families are considered equally probable. For more information on these and other scenario families, please see the IPCC's Special Report on Emissions Scenarios (2000).
The CIG has analyzed simulations of future Pacific Northwest climate using twenty global climate models (GCMs) run with the A1B and B1 emissions scenarios. These simulations were prepared by climate modeling centers worldwide for the Fourth IPCC Assessment (also referred to as AR4). For consistency with prior CIG work, we include the current versions of the ten models used previously: HadCM3, ECHAM5, CGCM_3.1(t47), CCSM3, CNRM_CM3, GISS_ER, MIROC_3.2, IPSL_CM4, CSIRO_MK3, and PCM1 (CCSM3 is the successor to both the NCAR CSM and DOE PCM). To these we added ten additional models to better represent the range of models participating in AR4. These are: GISS_AOM, INMCM3_0, BCCR, GFDL_CM2_1, GFDL_2_0, CGCM3.1_t63, ECHO_g , FGOALS, MIROC_3.2(high resolution), and HADGEM1.
GCMs are designed to represent the physical processes leading to climate change on the global scale. They resolve features on the Earth on the continental scale. The models can show differences in the rate of climate change at different locations, but only on the continental scale. Differences can be seen, for example, in warming and rainfall between the Pacific Northwest and the Southwest. However, GCMs do not contain the smaller mountain ranges and water bodies that are important to regional climate.
There is no information in the GCMs to allow them to distinguish between the Puget Sound lowlands and the Columbia Basin, and thus we assume a uniform rate of climate change for the entire region. The regional trend in temperature and precipitation is sufficient for most climate impacts studies. In the past, changes in temperature and rainfall have been fairly uniform across the region. We can understand regional differences in climate impacts by using emiprical information about the present climate.
In some locations, however, there are important physical processes that can affect the response to climate change. For example, a location that is currently mostly snow covered in the present climate and will be mostly snow-free in the future would see very different climate trends than the region as a whole. These effects can be represented by using regional climate models. A regional climate model is very much like a global climate model except that it uses much finer resolution and covers a limited area. Since it covers a limited area, the regional model needs to receive information about its boundaries from a global climate model, which is used to "force" the regional model. Regional models run at about 20-km or finer resolution are able to represent the geographical and land-use features of the region that determine its various climate zones.
Regional climate modeling is a new tool in climate impacts assessement. We are still learning how these models respond to climate change forcing and how to best represent uncertainty in climate projections.
As part of the 2008 climate change scenarios update, the CIG implemented a new approach to evaluating climate change projections for the Pacific Northwest. Reliability Ensemble Averaging, or REA (Giorgi and Mearns 2002), weights regionally-averaged GCM simulations in accordance with each model's ability to replicate 20th century Pacific Northwest climate. The REA value for each season and decade is calculated by weighting each model’s output by its bias (i.e., the model is generally too cool or warm, or wet or dry relative to observed 20th century climate) and distance from the all-model average.
Multi-model averages in weather forecasting, seasonal forecasting, and climate simulations often come closer to observations than single models. REA should produce better results for the future than an unweighted average. For more details on the REA calculation, see Mote et al. 2008, Appendix B.