Sensitivity of mesoscale simulations of land–sea breeze to boundary layer turbulence parameterization

The sensitivity of mesoscale simulations of land and sea breeze circulation on the south east coast in the Chennai region of India to boundary layer turbulence parameterizations is studied using the community based PSU/NCAR mesoscale model MM5. High-resolution simulations are carried out with nested domains. Four widely used boundary layer turbulence parameterization schemes are selected for the study. Two of these schemes (Blackadar (BK) and medium range forecast (MRF)) are simple first-order non-local schemes and the other two (Mellor–Yamada (MY) Eta planetary boundary layer (PBL) and Gayno–Seaman (GS)) are more complex 1.5 order local schemes that include a prognostic equation for turbulence kinetic energy. The other physics used in the model are identical. The micro-meteorological tower and flux observations, GP sonde and radiosonde observations from the study region are used to compare the simulated mean variables. In spite of differences in complexity, all the schemes show similar capability in simulating the boundary layer temperature, humidity and winds. The land–sea breeze circulation and internal boundary layer formation, which are special phenomena at the coastal site, could be simulated by all the schemes. The BK, MRF schemes produced boundary layers that are more mixed than those produced with the MY and GS schemes. While all the schemes underestimated surface sensible heat fluxes during stable night conditions they are reasonably simulated during daytime by MRF and BK schemes. Among the tested schemes, the BK scheme has reasonably produced the PBL height, humidity, wind fields and proves suitable for operational forecasting. The results suggest that there is little improvement in overall accuracy of predictions with more complex turbulence parameterizations.