Grid box or grid point: What type of data do GCMs deliver to climate impacts researchers?

Simulated data from General Circulation Models (GCM) are an important source of climate data and appropriate interpolation and subsequent interpretation of these data are required when used at the subgrid scales demanded in the study of climate change impacts. Although the techniques used to interpolate observed climate data can also be used to interpolate GCM output, there has been some hesitancy in interpreting subgrid-scale GCM data. We believe that a lack of understanding of this data source contributes to this hesitancy, in particular the spatial nature of GCM data is unclear. Are GCM data point or areal quantities; how does this affect spatial interpolation; and, how should climate impacts researchers use and interpret subgrid scale GCM data? Above all, it bears remembering that interpolation does not create new information, it only increases the spatial precision of the existing data and it does this at the cost of increasing uncertainty. We believe that one can adopt either an areal or point approach, at least until there is some definitive evidence to support one approach over the other, but users of this data need to interpret simulated climate surfaces in light of whether they have adopted a grid-box or a grid-point approach to GCM data. In particular, the larger the area represented by a GCM value the smoother the resulting climate surface is likely to be and the more different from a similarly constructed observed climate surface. Although the grid-box approach seems more intuitive, at least to us, researchers will not find many commercially available interpolation packages for this type of data.