This page explains how we post-processed the original observation and model output data.
The R data file contains the location of grid (longitude/latitude)-loc, observations (DJF and JJA averages)-obs.DJF and obs.JJA, and model outputs-mx.DJF and mx.JJA, for surface temperature of 1970 - 1999. Note that x gives the model number and it runs from 2 to 20. These numbers should correspond to the numbers given in Table 1 of the paper.

The model outputs and observations are matrices of size 2448 x 30. The observations are originally given on a 5 degree grid (the location of grid is given in loc), which gives total of 2448 grid points. The model outputs, originally on finer grids (see Table 1 of the paper), are bilinearly interpolated to the observation grid. Each column gives the annual DJF (or JJA) averages for each year - we have total 30 years of data. There is originally no missing data in model outputs. In taking annual average of observations, no imputation is performed and missing months are simply ignored. Note that DJF average for the year 1970 is the average of December 1969, January 1970 and February 1970. Likewise, DJF average for the year 1999 is the average of December 1998, January 1999 and February 1999.