A recent study published in Global and Planetary Change demonstrates cost-saving opportunities in the otherwise computationally expensive process of high-resolution climate modeling. General Circulation Models (GCMs, occasionally also referred to as ‘global climate models’) are extremely useful tools used to understand how our climate system works. The models use mathematical equations to describe the physics of the oceans and atmosphere. These equations quantify, for example, how hot air tends to rise (the reason hot air balloons fly) and how the wind flows from high-pressure areas to low-pressure areas (the reason it’s often windy during storms). The equations in each box of a three-dimensional grid representing our planet provide us with estimates of climatic variables, such as temperature and precipitation, in each place and time. However, due to their computational complexity and other factors, the spatial resolution (size of the grid boxes) of most modern GCMs is too coarse to understand how climate change will impact a particular place. For example, most modern GCMs have boxes between 250-600km (150-300 miles) on a side, which is greater than the distance between Boston and New York. So GCMs cannot provide great detail about what will happen to specific communities or ecosystems in the future, limiting our understanding of the local impacts of climate change. To overcome this limitation, several methods are used to ‘downscale’ or increase the spatial detail of the model output. The two primary methods include ‘dynamical downscaling,’ which uses a nested Regional Climate Model (RCM) to estimate the fine-grain variability (like a model within a model), and ‘statistical downscaling,’ which uses observations (typically from weather stations) to statistically adjust and account for any biases in the GCM output. But many questions remain about which method (or combination of methods) results in the most reliable projections of future climate change.
In the study co-authored by YCEI post-doc Dr. Adam Wilson, the outputs from six GCMs and four RCMs representing past (1961–1999) and future (2046–2065) climate were downscaled to about 12km resolution (~8 miles) for the Northeastern U.S. to facilitate regional impact assessments. The study demonstrated that statistical bias correction is vital for both GCM and RCM outputs and that using bias correction for projected (future) data resulted in a greater agreement among GCMs in capturing the spatial pattern of extreme climate indices. The study also concluded that statistical downscaling alone (without using an RCM) produced similar results, suggesting that the computationally expensive RCMs may not be necessary if the output will also be statistically downscaled. The multi-model analysis projects a significant decrease (27-36 days) in the number of frost days for most of the Northeastern US. Growing season length is also likely to increase from 15-25 days and there are likely to be more frequent heavy precipitation events occurring in the coastal area. Nearly the entire region is projected to experience a small increase in the number of days per year with precipitation greater than 10 mm (0.4 inches).
Ahmed, KF, GL Wang, John. A Silander, Adam M. Wilson, and Jenica M Allen. 2013. “Bias Correction and Downscaling of Climate Model Outputs for Climate Change Impact Assessments in the U.S. Northeast.” Global and Planetary Change 100 (2013): 320–32. doi:http://dx.doi.org/10.1016/j.gloplacha.2012.11.003.