Toward a parameterization of mesoscale fluxes and moist convection inducedby landscape heterogeneity

A growing body of evidence from modeling and observations indicates that me soscale circulations generated by land-surface wetness heterogeneities resu lt in substantial vertical fluxes of sensible heat and moisture. These flux es are believed to have a strong impact on large-scale mean atmospheric var iables such as temperature, humidity, cloudiness, and precipitation. Curren tly, however, this type of mesoscale convective process is not considered i n general circulation models (GCMs). In this study, we develop a parameteri zation for these landscape-forced fluxes, similar to what might eventually be implemented into a GCM. In addition, we investigate the relationship bet ween the parameterized mesoscale flux and the convective condensation assoc iated with these circulations as a first step toward directly including clo uds and precipitation forced by surface heterogeneity effects as one compon ent of a comprehensive GCM convective scheme. To generate the data necessar y for this development, we perform a number of simulations with a state-of- the-art mesoscale model to determine the sensitivity of the fluxes and cond ensation to a variety of background atmospheric conditions and land-surface wetness distributions. We use similarity theory to determine the dependenc e of the mesoscale sensible heat and moisture fluxes on the parameters rele vant to the problem, and we create parameterized vertical flux profiles by fitting with Chebyshev polynomials. The parameterized fluxes are tested aga inst an independent, three-dimensional (3-D) simulation of mesoscale develo pment over a heterogeneous landscape, and general good agreement is found. We propose an empirical form for domain-averaged condensation on the basis of a Linear relationship with parameterized mesoscale moisture flux and als o demonstrate a reasonable agreement with the results from the 3-D simulati on. The methodology of this study, i.e., the use of a numerical model in th e preliminary stages of parameterization development, is advantageous for s ituations where the necessary observational data set is nonexistent. The us e of model simulations to fully explore the parameter space of this type of problem should then lead to observational campaigns that focus on only tho se key processes and variables which are relevant for the further refinemen t of a given parameterization.