Comparing the scatter in PILPS off-line experiments with that in AMIP I coupled experiments

In PILPS off-line experiments, quite large discrepancies have been identified among the land-surface schemes in terms of the partitioning of available energy into sensible and latent heat fluxes and the partitioning of precipitation into evaporation and runoff plus drainage. In order to determine whether the extent of such differences found in off-line experiments is replicated in coupled experiments, a direct comparison between the range of scatter in results from PILPS off-line Ž . experiments and that from Atmospheric Model Intercomparison Project AMIP I coupled experiments is made. The surface response variables are normalized by forcing variables so as to reduce the influences of differences in forcing between the AMIP and off-line simulations. Monthly data of surface variables from 15 AMIP I experiments and from two PILPS off-line Ž . experiments 23 simulations from Cabauw and 21 simulations from Valdai were used for the analysis. The normalized measures of the scatter among land-surface simulations show that the range from the coupled simulations is larger than that from the off-line simulations, on both annual and seasonal time scales. Feedback from the host atmospheric models seems not to have reduced the disagreement among PILPS land-surface parameterization schemes. q 1998 Elsevier Science B.V. All rights reserved.