Monte-Carlo Regional Temperatures
Climate change won't impact everywhere equally. Some regions will warm faster than the global mean, while some others may warm slower. One such area which is expected to warm significantly faster than the global mean is the Arctic/Greenland regions.
Global ocean circulation patterns and changes in net incoming solar radiation due to changing sea ice/snow cover, a phenomenon known as Polar amplification, will cause rapid temperature increases in the Northern pole.
There is a lot of focus on quantifying the change in the global mean, but it is also important to know how uncertainty features in these projections. An understanding of how different locations will be impacted under different scenarios is an important tool in evaluating climate risk.
Below we show the results from a large 600,000 member Monte-Carlo analysis for exploring the range of possible temperatures for various regions around the globe in 2100. Regional scaling factors calculated from CMIP6 model output were used in combination with a 600 member MAGICC probabilistic ensemble to explore 3 sources of uncertainty in regional temperatures:
- Natural variability
- Dependence of the scaling factors on global mean temperature
- Structural uncertainty from the CMIP6 models
Update the region and emissions scenario below to see how different regions will warm under different scenarios. All temperatures shown are changes in surface temperature with respect to pre-industrial levels.
Methodology
A large ensemble of CMIP6 annual-mean surface temperatures was used to calculate regional scaling factors. From each ensemble member, annual-mean surface temperatures are calculated for each of the AR6 regions under different SSP scenarios, available from https://cmip6.science.unimelb.edu.au/. These temperatures were normalised over pre-industrial period (1850-1900) and scaling factors are calculated for all available regions, scenarios, models and years.
These scaling factors can be used to obtain a regional annual mean surface temperature given a global mean surface temperature.
In this case a simple climate model, MAGICC7 was used to generate a 600 member probabilistic ensemble of surface temperatures. The use of an emulator better samples the structural uncertainty from the climate system and enables arbitrary emissions scenarios to be used.
For a given year, scenario, annual mean global mean surface temperature values were extracted from the MAGICC ensemble. For each of these global mean temperatures, a subset of scaling factors within were selected for each region, of which 1000 of these scaling factors were drawn. These drawn scaling factors were used to generate 1000 regional temperatures.