Item

Abstract

How will marine low-level cloudiness change in a warming climate? To answer this ques-

tion a better process understanding of low-level cloudiness is needed. This dissertation

uses a multitude of observations and large-eddy simulations to explore how meso-scale

patterns of shallow convection relate to this challenging question. This study focuses

on the downwind trades and its meso-scale patterns that only recently raised interest

based on the work of Stevens et al. (2020) who supplemented the traditional classes of

meso-scale patterns of the upstream trades. These new classes are named based on their

visual impression Sugar, Gravel, Flowers and Fish. Here they are further investigated in

terms of their climatic relevance, physical characteristics, atmospheric environment and

emergence.

The core of these investigations consists of deep neural networks that have been

trained to identify these patterns in satellite images. At the same time, the deep neural

networks proved to be a valuable tool to derive a common perception of subjectively

defined classes that do not have a ground truth. Although the crowd-sourced labels were

therefore very noisy, the neural networks ranked among the highest in inter-annotator

agreements.

The classification of the neural network reveals that the patterns are common to the

trades beyond the winter season in the western North Atlantic and can represent more

than 40 % of the observed variability depending on season and region. This variability

expresses itself not only in changes of the visual appearance but also physically in the

cloud cover. A linear relationship between the cloud cover and the cloud radiative effect

makes the processes leading to the patterns relevant for the climate.

The underlying physical processes of each meso-scale pattern are related to the air-

mass origin with an influence of diurnal variations that are potentially modulating the

large-scale factors. One large-scale factor that is most distinct among the patterns is wind

speed. Other factors are only related to a particular pattern but can be a necessity for the

pattern to form. Fish for example is associated with anomalously strong convergence.

Sugar favors warmer surface temperatures. Both the forcing of Fish and Sugar are related

to air-masses intruding from outside the trades, leaving Gravel and Flowers be the only

native trade-wind patterns.

Large-eddy simulations reveal that they are in general capable of replicating the

observed variability in meso-scale cloud patterns. However, they are unable to match

the observed vertical distribution of cloudiness in both their absolute values and their

variability in particular for Flowers and Fish. Nevertheless, the distribution of moisture

and the presence of meso-scale circulations indicates that the responsible processes for

the formation of the different patterns are captured and the simulations are a valuable

tool to complement the observations to gain a better process understanding.

Based on the relationships between large-scale forcing and mesoscale patterns found

in this dissertation, conditions preferred by patterns with higher cloud amount and

more negative cloud radiative effect are expected to occur less frequently in a warming climate.