Error in Princeton Data 1901-1947


Posted by Lila Warszawski on Feb. 17, 2016

As we reported to you at the end of last year, the Princeton data set provided to you for use in the ISI-MIP simulations for the period 1901-1947 has been superceded by a corrected version. The problem was a mistake in the processing scripts, not a systematic methodological mistake, so the data were just wrong.  Unfortunately we only became aware of this well after you started with your simulations.

We have now consulted with the data provider and compared the corrected Princeton data to those supplied for use in ISI-MIP and have found that they differ significantly.  In the attached document you will find a set of plots comparing the two versions.  As you will see, the differences in, for example, monthly mean temperature over the time period 1901-1947 are significant, reaching up to 4 degrees.

A comparison between the corrected Princeton data and the WATCH forcing data also supports the conclusion that the corrected data are more reasonable.

We have also used the LPJmL model to examine how the effect of these differences during the spin-up propagates into model results during the reporting period.  The results show that the state of carbon stocks in particular following the spin-up period is significantly different for the two data sets. You can find the related plots in the attached document.

We therefore would like to recommend the following course of action:

  • All models requiring a spin-up period longer than the 22 years between 1948-1970 (the start of the reporting period) should re-run the Princeton data with the corrected data set.

We understand that these recommendations represent a significant investment of time for some modelling teams, but hope that you will agree that under the circumstances, it is the best way forward. 

Since it is central to the ISI-MIP objectives that a common climate-data set has been used to produce the simulations, it is important that we reach a common decision on this. We would like to invite your feedback, concerns and comments until Tuesday, 23 February, after which a final decision will be reached.