Observation and simulations of water vapor turbulence profiles in convective boundary layers over the ARM Southern Great Plains and Tropical Western Pacific sites

Poster PDF

Description

This study explores water vapor turbulence in convective boundary layers (CBL) using a combined strategy of observations and fine-scale modeling. We found some striking differences between observations of mid-latitude, tropical monsoon and tropical non-monsoon cases. Further, with the help of LES simulations we found and developed understanding for discrepancies between temporal and spatial statistics, which can help to better interpret observational data. Water vapor turbulent profiles were computed from the TWP Raman lidar located at Darwin, Australia. We identified 62 cases, comprising 13 monsoon and 49 non-monsoon events, over the 5-year data record. The seasonal variations of the vertical profiles of integral scale, variance, structure function coefficient, and skewness have been analyzed and compared with similar observations from the Raman lidar at the Southern Great Plains (SGP) site. The median of the water vapor variance profiles is larger in monsoon events than in the non-monsoon, and neither agrees well with median variance profile from the SGP. The coefficient of the structure function profiles, which is related to molecular dissipation, in monsoon events is also found to be larger than during the non-monsoon. A common assumption in observational strategies is that the temporal statistics (average, variance, etc.) converge with the spatial statistics for sufficiently large domain and sufficiently long time series, as long as the flow is spatially homogeneous and in quasi-steady state. This assumption was tested in the large-eddy simulation (LES) model. For the quasi-steady-state period of the CBL we observe a discrepancy between the humidity distributions based on spatial versus temporal statistics, especially in the higher-order statistics. Most notably, the sub-gaussian kurtosis at the boundary-layer top that was reported from observations (McNicholas and Turner, 2014) can be reproduced in the temporal LES statistics, but not in the spatial statistics. We introduced two passive scalars into the simulation to characterize the bottom-up thermals and the top-down entrainment in the CBL. We show that the difference between spatial and temporal statistics can be fully explained by high kurtosis values (around 5 or larger) in the spatial distribution of the top-down scalar. We attribute this to relatively rare but long-lasting entrainment events (dry tongues, Couvreux et al., 2004), which cause a bias in the temporal statistics.