This file contains the results from blending land and ocean data from
RCP climate model runs using various blending options to try to reproduce
the data and methods used in the observational datasets.

THE FILE CONTENTS MAY NOT BE INTUITIVE, SO PLEASE READ THIS CAREFULLY.

The zip file contains this file and the following subdirectories:

rcp85-xxx/
rcp85-xax/
rcp85-mxx/
rcp85-max/
rcp85-mxf/
rcp85-maf/
rcp85-xxf/
rcp85-xaf/
rcp85-had4/
rcp45-xxx/
rcp45-xax/
rcp45-mxx/
rcp45-max/
rcp45-had4/

The first part of the directory name is the RCP dataset (45 or 85).
The second part of the directory name is the blending method which
is one of:
 xxx - unmasked, absolute temperatures, variable ice
 mxx - masked, absolute temperatures, variable ice
 xax - unmasked, temperature anomalies, variable ice
 max - masked, temperature anomalies, variable ice
 xxf - unmasked, absolute temperatures, fixed ice
 mxf - masked, absolute temperatures, fixed ice
 xaf - unmasked, temperature anomalies, fixed ice
 maf - masked, temperature anomalies, fixed ice
 had4 - HadCRUT4 emulation

In each subdirectory, there are a list of temperature files (.temp).
The columns in these files are as follows:

1. Date. (As fractional year)
2. Air temperature (tas)
3. Blended air/sea temperature (tas/tos)
4. Difference column 3 - column 2

BE CAREFUL WITH COLUMN 2. It is calculated using the same method as
column 3, but using tas for both fields in the blend calculation. This
is useful for trying to separate out the effect of blending versus the
other factors affecting the results (as far as that is possible), but
that does not mean that column 2 is necessarily meaningful outside this
context.

In practice, column 2 is only meaningfull on its own for the
-xxx and -mxx files: these correspond to the mean model air temperature
either for the whole globe or for the HadCRUT4 coverage-masked model
air temperature.

I recommend against using the BNU-ESM, CMCC-CESM, and bcc-csm1-1-m data
for reasons mentioned in the paper.

