Can I average the probabilistic projections for different grid squares to produce a single probabilistic projection for a user-defined area?

The grid squares from the UKCP09 projections cannot be averaged to produce a probabilistic projection for a user-defined area.

In detail

UKCP09 maps do not present a snapshot of the spatial distribution of the projected climate at a particular 30-year time period with all grid square values spatially coherent (i.e. they do not provide a spatially coherent image of the projected change). When considering how to interpret the maps, users should be aware that if the projected change of a variable at a particular location actually turned out to be the value at the 90% probability level, then this does not mean that the projected change for other variables at that location will also be that projected at the 90% probability level. Nor does it mean that the change that occurs at other locations for that variable will also be at the value associated with the 90% probability level. 

The values from the 11-member RCM are spatially consistent, corresponding values from multiple adjacent grid squares within a single ensemble member can be averaged to produce a single projection for a user-defined multi-grid square area. Users should be aware that the RCM does not incorporate as wide a range of uncertainties as the UKCP09 probabilistic projections. As the 11-member RCM data is not a UKCP09 product, users will have access to the data through the LINK project website. There is little guidance to using this information beyond a limited reference within this user guidance, that in Chapter 5 of the Climate change Projections report and that on the LINK website.