Reply: (Evaluation of Land-Surface Parameterization Schemes Using Observations)

Robock et al. (1995a) tested a 15-cm bucket model (Manabe 1969) and the Simplified Simple Biosphere (SSiB) model (Xue et al. 1991) by forcing them with observed meteorological and actinometric data and comparing the results with observed soil moisture, albedo, snow depth, and net radiation. Xue et al. (1997) show that it is possible to improve SSiB in response to errors in the simulations by fixing the snowmelt partitioning and adjusting the soil properties to produce a higher effective field capacity. We agree with their conclusion that using real data to test models is a good idea, and this was one of the major conclusions of Robock et al. (1995a). Xue et al. (1997) claim that they demonstrate that the fundamental structure of their model is sound and that it just needs tuning by adjusting parameter values. However, while the bucket model did not have any problem with snowmelt partitioning in the Robock et al. (1995a) experiments, which involved six stations for 6 yr in the former Soviet Union, for some stations the 15-cm field capacity was clearly too low. Schlosser (1995) demonstrated that major improvements are also possible in bucket model simulations for particular stations by taking into account actual conditions at the station. Figure 1, for example, shows a dramatic improvement in the Kostroma simulations by using the actual field capacity of 28 cm of plant-available water in the top 1 m of soil rather than the standard 15 cm used in the GCM version of the bucket model. Not only are the winter soil moisture values accurately simulated, but the summer drying is accurately modeled in wet and dry summers. In fact, this improved simulation of soil moisture for Kostroma, produced by changing only one parameter in the bucket model, the field capacity, appears to be superior to the