Reanalysis of Arctic Climate

For years now I’ve been using the convenient tools provided at the NCEP/NCAR reanalysis web site to generate custom maps and time series illustrating the climate of the Arctic. By way of example see last December’s “Post-Truth Global and Arctic Temperatures“:

Prompted in part by the obvious difficulty the different models are currently having in generating accurate short term forecasts for the “New Arctic”, I’ve been recently been comparing assorted reanalysis products. For example the UCAR Climate Data Guide points out that:

NCEP Reanalysis (R2) is better than NCEP-NCAR (R1) but still a first generation reanalysis. It is best to use 3rd generation reanalyses, specifically, ERA-Interim and MERRA.

I recently discovered that Richard James has performed a similar analysis for the Arctic, which can be viewed at:

Arctic Winter Warmth

wherein I mentioned the NOAA ESRL Web-based Reanalysis Intercomparison Tool, which allows you to produce plots and timeseries for arbitrary areas of Planet Earth using NCEP/NCAR, ERA Interim, MERRA-2 and numerous other reanalysis products. Here’s one little example:

NCEPr1-ERA-T2m

which makes it evident that NCEP-NCAR (R1) and ERA Interim have different ideas about surface temperatures in the Arctic. So does MERRA-2!

MERRA2-ERA-T2m

For a graphic example of the differences between the three products here is my version of Richard’s Arctic winter temperature comparison (note that currently ERA data is only available up to January 2017):

NCEPr1-T2-Oct-Jan-2017

MERRA-T2-Oct-Jan-2017

Era-T2-Oct-Jan-2017

Can you spot the difference? In conclusion, here’s the Era Interim version of the High Arctic autumnal 925 hPa temperature trend graph at the top:

ERA-SON-80N-T925

2 thoughts on “Reanalysis of Arctic Climate

  1. Thanks Jim !
    One comment : The “2m air temperature” is kind of a tricky metric to determine from satellite sources, especially during the Arctic winter. You know, “inversion” and things like that. In fact, NCEP/NCAR does not even have a “2m air temperature” tag on their web portal. Only “surface”.

    Maybe a better metric would be “925 mb temperature”.
    That is away from the inversion layer, and thus should be more consistent across different data sets.

    So could you do your same analysis between NCEP/NCAR and MERRA-2 and ERA-Interim for 925 mb for the 2016/2017 winter ?

Leave a Reply

Your email address will not be published. Required fields are marked *