Is the use of a 5 km grid for the Weather Generator justified?

Yes. The Weather Generator works at a 5 km resolution, although the climate change factors that are supplied to it from the projections are applied at a 25 km resolution. The Weather Generator should be seen as a tool for analysing the the 25 km probabilistic climate projections rather than a standalone climate projection. Therefore, there is no climate change information in the Weather Generator outputs beyond that within the 25 km resolution probabilistic projections.

In detail

The Weather Generator operates with the highest spatial resolution climate projection and observed climate data available. It is of course recognised that there is variability (or noise) in the spatial data sets, arising from various sources, which means that there is a limit to the resolution that is meaningful for such analyses, as it can give illusory detail or variation. However, in the case of the current application, the resolution is justified, as there is an important and significant spatial signal in the long term weather statistics (or climate) which is captured by the 5 km grid.

The 5 km grid provides useful information on how the long term average statistics of weather vary across the country, derived from 30 years of observations from thousands of weather stations. The variation across the country is resolved at 5 km because some weather variables (e.g. rainfall and temperature) depend on the ground surface elevation and geography in a systematic way, and at a similar spatial scale. It is important to include this variation in the Weather Generator so that its outputs can be used for specific locations. For example, the rainfall on a mountainous 5 km grid square will be considerably higher than on a neighbouring low-lying square.

The dependence of rainfall and temperature on the topography is likely to be essentially the same in the future as it has well understood meteorological causes (from the laws of physics, or at least thermodynamics). So this downscaling provides our current best estimate of the spatial variation across the country. In any case, if such a method wasn't followed, users would apply their own, different downscaling procedures, thus causing inconsistency and confusion.