Urban heat island–induced circulation and convection in three dimensions are investigated theoretically and numerically in the context of the response of a stably stratified uniform flow to specified low-level heating that represents an urban heat island. In a linear, theoretical part of the investigation, an analytic solution for the perturbation vertical velocity in a three-dimensional, time-dependent, hydrostatic, nonrotating, inviscid, Boussinesq airflow system is obtained. The solution reveals a typical internal gravity wave field, including low-level upward motion downwind of the heating center. Precipitation enhancement observed downwind of urban areas may be partly due to this downwind upward motion. The comparison of two- and three-dimensional flow fields indicates that the dispersion of gravity wave energy into an additional dimension results in a faster approach to a quasi-steady state and a weaker quasi-steady flow well above the concentrated heating region in three dimensions. In a nonlinear, numerical modeling part of the investigation, extensive dry and moist simulations using a nonhydrostatic, compressible model with advanced physical parameterizations [Advanced Regional Prediction System (ARPS)] are performed. While the maximum perturbation vertical velocity in the linear internal gravity wave field exists in the downwind region close to the heating center, the maximum updraft in three-dimensional dry simulations propagates downwind and then becomes quasi stationary. In three-dimensional moist simulations, it is demonstrated that the downwind upward motion induced by an urban heat island can initiate moist convection and result in downwind precipitation. The cloud induced by the downwind upward motion grows rapidly to become deep convective clouds. Heavy rainfalls are localized in a region not far from the heating center by a convective precipitating system that is nearly stationary. The differences in results between two and three dimensions are explained by the presence of (moist) convergence in an additional dimension. The numerical simulation results indicate that the intensity and horizontal structure of the urban heat island affect those of circulation and convection and hence the distribution of surface precipitation.

The impacts of urban aerosols on clouds and precipitation are investigated using a spectral (bin) microphysics cloud model. For this purpose, extensive numerical experiments with various aerosol concentrations are performed under different environmental moisture conditions. To take into account the urban heat island and urban air pollution, it is considered that there is low-level heating in the urban area and that the aerosol concentration in the urban area is higher than that in the surrounding rural area. Simulation results show that a low-level updraft induced by the urban heat island leads to the formation of a low-level cloud and then a deep convective cloud downwind of the urban area. The onset of precipitation produced by the low-level cloud is delayed at higher aerosol concentrations. This is because when the aerosol concentration is high, a narrow drop size distribution results in a suppressed collision–coalescence process and hence in late raindrop formation. However, after the deep convective cloud develops, a higher aerosol concentration generally leads to the development of a stronger convective cloud. This is mainly due to increased release of latent heat resulting from the enhanced condensation process with increasing aerosol concentration. The low collision efficiency of smaller cloud drops and the resulting stronger updraft at higher aerosol concentrations result in higher liquid water content at higher levels, leading to the enhanced riming process to produce large ice particles. The melting of a larger amount of hail leads to precipitation enhancement downwind of the urban area with increasing urban aerosol concentration in all moisture environments considered.

The sensitivity of a cumulus parameterization scheme (CPS) to a representation of precipitation production is examined. To do this, the parameter that determines the fraction of cloud condensate converted to precipitation in the simplified Arakawa–Schubert (SAS) convection scheme is modified following the results from a cloud-resolving simulation. While the original conversion parameter is assumed to be constant, the revised parameter includes a temperature dependency above the freezing level, which leads to less production of frozen precipitating condensate with height. The revised CPS has been evaluated for a heavy rainfall event over Korea as well as medium-range forecasts using the Global/Regional Integrated Model system (GRIMs). The inefficient conversion of cloud condensate to convective precipitation at colder temperatures generally leads to a decrease in precipitation, especially in the category of heavy rainfall. The resultant increase of detrained moisture induces moistening and cooling at the top of clouds. A statistical evaluation of the medium-range forecasts with the revised precipitation conversion parameter shows an overall improvement of the forecast skill in precipitation and large-scale fields, indicating importance of more realistic representation of microphysical processes in CPSs.