favorite climate scientist and several of his colleagues have a new paper out on global land surface temperature trends (Montandon et al. 2011). They perform an interesting analysis in asking the degree to which the spatial distribution of land surface stations is representative of land surface types found on Earth.
They find that the major surface temperature records (i.e., NCDC, GISS, CRU, GHCN) are not spatially representative (see their Figure 2 above).
What happens if you weight the land surface record to account for this bias? Their preliminary result (which they emphasize is preliminary) is that land surface trends would actually increase if properly weighted. If this is the case then it potentially presents a headache for the climate modeling community because it would exacerbate the divergence between land surface and tropospheric trends that we documented in Klotzbach et al. 2009 (see this, this, and this).
Does this study have policy implications? Not for me, but it does show that there remain interesting scientific questions to ask and maybe surprises yet to be found in the greenhouse.