To help make a selection of GCMs to incorporate into an analysis we were running, we wanted to see by just how much the IPCC AR5 climate projections differed, and how they different in geographic space. While I thought this information would be relatively easy to find, it proved elusive (or my Google skills proved inadequate). Either way, I ended up building some geographic summaries of the WorldClim CMIP5 / AR5 climate data. Tabular summaries of this data work are posted on my former supervisor’s website for the ClimateNA software package: a topographic downscaling of climate data. I also did something similar for ClimateWNA (western North America) for the AR4 models.
This post comes out of some data work I did for a research project a while back and I figure it might be useful to someone else too. So I’ll post it here. Full disclaimer: this was data exploration work for our own purposes that hasn’t been peer-reviewed or really thoroughly double-checked. But it’s pretty straightforward, so hopefully I didn’t mess it up. But I might have. Always possible. I have the CMIP5 palaeoclimate data (also supplied by WorldClim) that I might analyse in the same fashion one of these days. Will post that if I get it done.
Incidentally, we went on to generate a huge amount of data in the form of climate variable surfaces at fine resolution (1km) for all of North America. These files are all available on the AdaptWest DataBasin for public download. We have also posted the climate velocity surfaces (for North America at 1km resolution) that came out of a recently published algorithm, where we offer a twist on Loarie’s standard climate velocity that optimises the metric for conservation and management applications.
The data used to create the summaries is supplied by CMIP5 as temperature and precipitation anomalies from the present day baseline. The data is in Lat/Long, and I have run the analysis on the 10 arc minute WorldClim data. Because of convergence, this has the effect of over-weighting the anomalies in the polar regions, as there are simply more data points closer per unit area as you go farther north. I didn’t deal with this when I ran this back-of-the-envelope calculation and won’t address it here. Sorry, the data are north-biased. But they’re consistently north-biased across all models, if that’s any consolation.
Basically, here’s what I did:
- Created a Lat/Long grid coverage for North America of temperature and precipitation anomalies (projection minus baseline) from various GCMs and time periods from the WorldClim CMIP5 RCP4.5 2050s future data.
- Summarised the data by country and state/province in North America and by country globally, based on the grid points falling within country/state polygons (no interpolation applied, just a straight average of points therein).
- Plotted it all out by continent, country, and state/province.
Plots are of mean annual temperature vs. mean annual precipitation anomalies for various GCMs for the 2050s for the AR5 RCP 4.5. Individual GCMs are shown as coloured dots. To show the breadth of the anomalies, all countries or states/provinces are shown on all plots in each series as grey dots.
Here’s the overview GCM distribution plot for the globe (North America is below):
All the individual country plots are available in a single .zip file here.
(Hosted at my old domain, so please let me know if any links are broken.)
Here’s the overview GCM distribution plot for North America:
And here are the plots of the individual regions of North America. I’ve posted thumbnails of summary plots for Canada, and the continental USA, and provide zip files containing all plots for US states and Canadian provinces. Complete tables of anomalies values (for states/provinces) for a selection of GCMs (RCP 4.5 for the 2050s) are posted on the Adaptwest DataBasin.
Download all provincial plots in a .zip file here.
Continental United States
Download all state plots in a .zip file here.