MODELING OF LAND-SURFACE EVAPORATION BY 4 SCHEMES AND COMPARISON WITHFIFE OBSERVATIONS

We tested four land surface parameterization schemes against long-term (5 months) area-averaged observations over the 15 km x 15 km First In ternational Satellite Land Surface Climatology Project (ISLSCP) Field Experiment (FIFE) area. This approach proved to be very beneficial to understanding the performance and limitations of different land surfac e models. These four surface models, embodying different complexities of the evaporation/hydrology treatment, included the traditional simpl e bucket model, the simple water balance (SWB) model, the Oregon State University (OSU) model, and the simplified Simple Biosphere (SSiB) mo del, The bucket model overestimated the evaporation during wet periods , and this resulted in unrealistically large negative sensible heat fl uxes. The SWB model, despite its simple evaporation formulation, simul ated well the evaporation during wet periods, but it tended to underes timate the evaporation during dry periods. Overall, the OSU model ably simulated the observed seasonal and diurnal variation in evaporation, soil moisture, sensible heat flux, and surface skin temperature, The more complex SSiB model performed similarly to the OSU model. A range of sensitivity experiments showed that some complexity in the canopy r esistance scheme is important in reducing both the overestimation of e vaporation during wet periods and underestimation during dry periods. Properly parameterizing not only the effect of soil moisture stress bu t also other canopy resistance factors, such as the vapor pressure def icit stress, is critical for canopy resistance evaluation, An overly s imple canopy resistance that includes only soil moisture stress is una ble to simulate observed surface evaporation during dry periods. Given a modestly comprehensive time-dependent canopy resistance treatment, a rather simple surface model such as the OSU model can provide good a rea-averaged surface heat fluxes for mesoscale atmospheric models.