General circulation models (GCMs or global climate models) have been designed to simulate the planet's future climate. In the past 30 years climate modelers have been improving the GCMs' spatial resolution from the first assessment report (FAR-1995) to the fourth report (AR4-2007) for the IPCC (Intergovernmental Panel for Climate Change) to meet the needs of climate change impacts models simulating finer scale processes. The scale has changed from ~500 km x 500km grid cells to ~110 x 110 km grid cells (see figure to right; source of graphics: IPCC 2007).
Despite these advances, important questions remain about the scale at which spatial and temporal interpolations of GCM projections are reliable. GCM projections cannot be used directly for regional or local impact assessments because the models were designed to simulate the entire planet's climate and their accuracy declines at the local scale due to their inherent coarse spatial resolution. In complex terrain, such as cliffs, deep valleys, rugged mountains, the local climate may decouple from the regional climate and create local processes that are difficult to measure, let alone simulate. Cold air drainage in valleys, temperature inversions, fog banks along large bodies of water, all feed back to the atmosphere but their measurements and representations in models are far from simple.
Downscaling techniques are used to bring GCM results to the scale of interest. Several methods exist, the follwoing are the most commony used:
- Statistical: In the Delta or anomaly method, the difference (or ratio for precipitation) between future and current GCM results, called the delta or anomaly, is calculated and added (or multiplied) to a higher resolution “baseline” or observed long-term average climate (1961-1990 or 1971-2000) in order to generate future climate data. It is important to remember that, despite the fact that the information is now served at fine scale, the original climate change information was generated at coarse scale. The GCM did not take into account feedbacks from local processes and assumed coarse scale homogeneity of land cover.Bias correction: The bias correction minimizes the difference between the GCMs hindcast of historical conditions and the observations. Future projections are corrected with this calculated bias, assuming that the correlation between GCM results and actual climate patterns will not change in the future.
- Dynamic: The most recent downscaling method uses Regional Climate Models (RCMs) that run at finer spatial scale than GCMs, taking into account local climate processes. RCMs incorporate local topography and land-atmosphere feedbacks, and are the most mechanistic way to simulate regional to local climate variables. However RCMs need boundary conditions that are provided by GCMs, so the reliability of their projections is also linked to that of the GCMs. Any bias in GCM results automatically carries to the embedded RCM.
The uncertainty of the original GCM projections result from the imperfect knowledge of:
- initial conditions such as sea surface temperature that is difficult to measure
- levels of future anthropogenic emissions which are unknowable since they are dependent on current and future political decisions and social choices, not on physical laws of nature
- general system behavior (such as clouds, ice sheet melt) that continues to be the subject of basic climate research and constitute the "known unknowns" of the climate system.
- unexpected surprises or "unknown unknowns" like the Larsen B ice shelf rapid collapse, for example.
New findings from on-going changes in climate constantly bring scientists back to the drawing board to improve existing models.
In order to influence decisions with climate information, conservation practitioners need to be asking relevant questions to the providers of climate scenarios. For example, work by Daly et al. (2008) suggests that conservation biologists and climatologists should jointly explore issues relevant to the location of the study area including:
- density of meteorological stations in or close to the area of concern and the length of records from these stations
- topographic complexity that can cause decoupling of local climate from regional trends
- relative proximity to large terrain features that can affect local conditions but not be simulated well by climate models
- proximity to water (lake or ocean) because of its cooling effect and groundwater availability
- influence of human activities (pollution levels and cloud condensation nuclei, fire ignition source, urban heat island effect)
- natural climate variability and records of extreme events that can improve the understanding of ecosystem vulnerability to past climate disturbance
There are several galleries in Data Basin that provide snapshots of GCM projections used by climate change impacts groups.
- The MAPSS team, US Forest Service Pacific Northwest station in Corvallis, uses climate projections to run their climate change impacts models. They provided snapshots of the climate projections they used mostly for the last 30 years of the 21st century.
- Dr. Healy Hamilton and her research team, provided ensemble averages from multiple GCMs and their associated standard deviations to quantify the temporal variability of climate.
Examples of Climate Modeling Teams and GCM acronyms:
- Canadian Centre for Climate Modelling and Analysis, Canada, CGCM3.1 Model
- Meteo-France, Centre National de Recherches Meteorologiques, France, CM3 Model
- CSIRO Atmospheric Research, Australia, Mk3.5 Model
- NOAA Geophysical Fluid Dynamics Laboratory (GFDL), USA, CM2.0 Model
- NASA Goddard Institute for Space Studies (GISS), USA
- National Institute of Geophysics and Volcanology, Italy, ECHAM 4.6 Model
- Center for Climate System Research (The University of Tokyo), National Institute for Environmental Studies, and Japan Agency for Marine-Earth Science and Technology, MIROC model
- Max Planck Institute for Meteorology (MPI), Germany, ECHAM5 / MPI OM
- National Center for Atmospheric Research (NCAR) Community Climate System Model, USA, CCSM 3.0
- Hadley Centre for Climate Prediction, Met Office, UK, HadCM3 Model & HadGEM1 Model
Reference:
Daly, C., Halbleib, M., Smith, J.I., Gibson, W.P., Doggett, M.K., Taylor, G.H., Curtis, J., and Pasteris, P.A. 2008. Physiographically-sensitive mapping of temperature and precipitation across the conterminous United States. International Journal of Climatology 28: 2031-2064.