Diagnostic subproject 12 is analysing models that use, or have as an alternative, a landsurface scheme that is participating in the GEWEX Project for Intercomparison of Landsurface Parameterization Schemes (PILPS). Completed large scale analysis has identified intermodel consistencies, differences and trends, and areas of interest for regional investigation. It is found that no one model is representative of all of the models over all fields and no one model is an outlier over all fields. Seasonal and zonal behaviours are similar but two models show significant spin-up over the first two years. The forcing from the sea surface temperatures is apparent in the continental climates and two maxima corresponding to the 1982-83 and 1987-88 El Niño-Southern Oscillation events are particularly distinct. One model has anomalously high SSTs and this propagates to the results over the land. Cloud cover does not show consistent enough behaviour to indicate that it is simply related to surface radiation or precipitation. The Bowen ratio shows intermodel differences that are not simply related to landsurface scheme complexity. Three models have significant globally averaged surface energy residuals. Two have globally and annually averaged values of about 5 to 6 Wpm2 and one a value of about 13 Wpm2. The importance of model initialisation is very clear in some models where ten year trends in soil moisture are closely related to ten year trends in temperature. Regional analysis using areas around the PILPS tropical rainforest and tundra points show ranges of sensible and latent heats similar to those seen in the PILPS Phase 1 off-line experiments. Differences between the sensible/latent heat partitioning found here and in PILPS are shown to be consistent with differences in precipitation. Many models show large surface energy residuals at individual grid points.