Deriving Local Climate Information from Global Models

Contributor(s): 
November 27, 2013

A farmer in the central USA and a fisherman in Maine will not experience global warming in the same way.  Dynamic interactions between cities, deserts, and forests mean that a rise in the global mean temperature doesn’t imply uniform warming of the planet.  As climate change advances, wind patterns and ocean currents will shift; the sea may swell in some places, while rivers run dry in others.  Global climate models make predictions on too broad a scale to recognize many of these effects, leaving policy makers unable to prepare their communities.  

As part of YCEI’s Climate Science Speakers Series, Professor Raymond W. Arritt of Iowa State University explained how  the local climate information they need can be derived from global circulation models.  Standard climate models consider the earth in discretized cells 100-400 kilometers wide, and compute the sequential­­­­ effects of weather changes in adjacent boxes.  As they ignore the diverse landscapes that can exist within cells, these large grids overlook small-scale dynamics and lead to inaccurate approximations. 

Lacking the funds and computing power required to refine those grids to the local scale, scientists can use various techniques to infer local climate changes from existing models, such as:

·       The “delta” method, a form of statistical post-processing which subtracts the outcome for a simulation of current conditions from the outcome for a future climate simulation. Once this change is computed it can be added as the predicted change to the observed conditions in the locale of interest.  This method is easy, but fairly crude, resting on assumptions of a uniform rate and trajectory of change across a grid box.

·       Dynamical regional climate modeling, a way to predict small-scale climate changes using finer grid spacing focused on a region and time span of interest. By nesting a fine-resolution region within a model with larger grid boxes, scientists strike a balance between computational cost and accuracy in local prediction. Arritt explained that though inconsistent grid box dimensions can lead to a distortion of weather features, the method is still a valuable option.

Dr. Arritt emphasized that collaborative research is key to improving the predictive power of climate models, empowering people to adapt to climate change before feeling the effects firsthand.