Investigating Shallow Convection with Observation and Modeling

Abstract

Shallow convection is essential to the climate system. It determines the vertical structure of convective boundary layer and provides a good starting point for deep convection by facilitating the vertical transport of heat, moisture and momentum. The stratocumulus and cumulus clouds from it are ubiquitous and strongly modulate Earth's radiation budget. However, due to our incomplete understanding of shallow convection, its representation in climate models contributes the largest uncertainty to climate projections. This dissertation investigates shallow convection from the perspectives of observation and modeling. Observation-wise, we study the aircraft measurements from the RACORO and ACE-ENA campaigns in a transformed Paluch diagram. Previous studies have interpreted samples in Paluch diagrams in terms of two‐point mixing and buoyancy sorting. The two‐point mixing pattern is frequently seen in RACORO observations and can be unambiguously distinguished from buoyancy sorting. The latter is not found in RACORO observations and is also rare in a large‐eddy simulation of a RACORO case. As opposed to the often assumed gradual dilution of updrafts, around half of the instances of two‐point mixing show no clear evidence of updraft dilution. We also find substantial spread in the properties of the parcels that have adjusted to their level of neutral buoyancy and that a LES with horizontally homogeneous surface underestimates this spread. The inclusion of surface heterogeneity in the LES improves the agreement of its horizontal moisture variations with the aircraft measurements. We also demonstrate another mechanism to generate the two-point mixing pattern through drizzling as a cautionary tale of the interpretation of the Paluch diagrams. Lastly, we validate two ground-based remote sensing techniques with the RACORO aircraft measurements. Modeling-wise, we introduce a new testing framework for entrainment and detrainment parameterizations in cumulus schemes using large-eddy simulations (LES). To demonstrate the framework in shallow cumulus, we select individual clouds from LES of the Barbados Oceanographic and Meteorological EXperiment (BOMEX), where a cloud is defined as a spatially and temporally connected cluster of cloudy grid points. We then add in small horizontally uniform perturbations on temperature and humidity at the birth point of each cloud to generate a set of perturbed runs. The clouds from the perturbed runs are identified with two approaches. The conventional approach defines cloud based on fixed thresholds on vertical velocity and liquid water content. This definition artificially detrains clouds falling short of the vertical velocity or liquid water content thresholds and consequently influences the average cloud properties. We also introduce an alternative definition that a grid point with the vertical velocity and total water mixing ratio above certain percentiles is considered as part of a cloud. The differences between the original clouds and the perturbed clouds in temperature, humidity, buoyancy ($B$), cloud fraction, vertical velocity ($w$), average distance to the cloud edge, critical mixing ratio ($\chi$) and bulk entrainment ($\epsilon$) and detrainment rates are analyzed and contextualized. We evaluate important entrainment and detrainment parameterization schemes such as $\epsilon \sim B/w^2$, $\epsilon \sim 1/w$ and $\epsilon \sim \chi^2$ by comparing their predictions of the anomaly profiles of the entrainment and detrainment rates against the diagnosed profiles. Under the conventional cloud definition, no entrainment parameterizations are able to correctly predict both cases. Under the percentile cloud definition, the $1/w$ scheme is able to account for both cases of the perturbation. Two detrainment parameterizations based on the critical mixing ratio are also found effective in both cases.