How Much Spread Do We See In The Models?

All of these analyses use a combination of models that I've referred to as ensembles. Specific analyses vary in which models they use based on what data are available and what data make sense to combine. What sort of spread exists in the temperature models?
This is not a very easy question to answer as it can mean a massive number of things and there's just so much data. The interpretation of the question I'll use is:

How much does the model that predicts the least warming differ from the model that predicts the greatest warming in a specific time period?

Note that throughout I am using RCP8.5 as the assumption. RCP's below that have less variation between models.

A way to test that is to pick a year, get the total warming over the US predicted by each model for that year, and compare the max and min models with each other. I don't know of a great way to present that data so here's a plot of the max and min projections for monthly average high in July of 2050:

Spread in global warming model projections by latitude and longitude
And the average high over the country for the month of July is:

  • Max projection: 90.7 degrees (F)
  • Min projeciton: 86.8 degrees (F)
It's a bit hard from that figure above to see the spread, and that's only July. I could do hundreds of plots to cover all months, max/avg/min, all years, etc., but that's insane and we don't have infinite time and bandwidth.

Another way I thought to present it then is to pick some sample areas and look at their monthly temperature plots. To do this then, I took the models that project the greatest and least warming for this specific location in 2050 and plotted those. The three sample areas are:

  1. Lat/Lon = 33.6/-84.4; close to Atlanta, Georgia
  2. Lat/Lon = 43.1/-89.4; close to Madison, Wisconsin
  3. Lat/Lon = 33.4/-112.1; close to Phoenix, Arizona
The monthly averages for each are presented below:

Atlanta temperature projections

Madison temperature projections

Phoenix temperature projections

You can see in each of these that the spread varies from month to month (i.e., the min model is not just something like 'max model minus 3 degrees') and the spread for them in 2050 is 3-5 degrees. If you'll recall from the methods page, what's presented in the tools and the analysis is generally the median of models. Also...just to reiterate...what's plotted above is not the max and min from all projections at each is the projections from the two models that project the greatest and least average warming across the areas in question over the course of the year 2050.

As a final way to look at it, I took the annual average high for the continental US for each model and have listed the spread below:

Min Projection (F)Max Projection (F)

As you'd expect..the further into the future you get, the larger the higher the absolute temperatures, and the larger the spread between the projections.

Finally...this is not an exhaustive analysis of the spread and uncertainty in the CMIP5 models and you can find much more in depth information if you google it...this was just meant to give a quick overview of how the models vary.

No comments:

Post a Comment