I am pleased to share with you that I successfully defended my PhD thesis at TU Delft on Monday, November 8 2021. It was an exciting experience and I enjoyed the day very much.
Summary
As I am writing this, parts of Central Europe are plagued by a series of intense rainfall events that, in less than two days, turn rivers into powerful streams, cause flooding, damage infrastructure and property, and harm people. The number of such extreme events, which are associated with high economic losses and casualties, has been increasing for decades. How is this happening? And, what is the relation between extreme convective precipitation events and increasing temperatures, such as which we are currently experiencing due to climate change? To tackle these questions, consider the following simplified version of a convective rain event. We imagine a column of air, a part of the atmosphere with a cloud inside of it. Near the surface, air streams into the column where it starts rising vertically. While gaining height, the air cools, until at a certain level the contained water vapor will condensate in the form of small cloud droplets. From this level, the cloud base, the air mass continues ascending while the amount of condensed water keeps increasing, so that the cloud droplets grow in size. Finally, when they grow sufficiently large, precipitation will set in, and, in the most extreme case, all the cloud water will reach the ground as rain. Following this conceptual model, one way to increase the amount of precipitation is to increase the moisture content of the air that enters the cloud.
In this case, the Clausius-Clapeyron (CC) relation plays an important role. It describes the water holding capacity of air with respect to its temperature. If relative humidity remains climatologically unchanged with global warming the amount of moisture contained in the atmosphere will increase at a rate of about 7% per degree warming. Indeed, it was found that, globally, extremes of daily precipitation amounts are bound by this rate of change. However, local precipitation extremes on hourly and subhourly time scales were found to exhibit higher rates than dictated by the Clausius-Clapeyron relation. This phenomenon is commonly called super CC-scaling. To explain this behavior, a common theory suggests that if the amount of moisture increases the amount of energy released through condensation in a convective updraft will intensify. This will lead to stronger updrafts and ultimately to a stronger moisture convergence into the cloud column.
Despite the fact that station observations of short-term precipitation extremes can display super CC-scaling, the exact process behind the phenomenon is still unknown. Moreover, it remains unclear how the characteristics of individual convective events, such as the size and precipitation intensity, will respond to a warmer and moister atmosphere. In this thesis we use rain radar data with a full coverage of the Netherlands and isolate individual convective events with a tracking algorithm. We then investigate the statistics of size and intensity of rain events at the time of their peak intensity. The results from this analysis show that the spatial extent and the intensity of extreme convective events are tightly coupled. Both characteristics jointly increase, meaning that stronger rain events are generally larger in size. Under warmer and moister conditions, the size of events rapidly increases at dew point temperatures above approximately 15 degree Celsius. Super CC-scaling is found when including all events in the analysis. However, the largest events are crucial to sustain super CC-scaling at higher temperatures. Enhanced precipitation rates appear over the whole area of the event.
To further study the dynamics of convective precipitation extremes in a controlled setting, we make use of a large eddy simulation (LES) model. A LES model can run at a high enough resolution to resolve processes which are necessary for the formation of small-scale thunderstorms. It is further possible to take time-dependent influences on the simulation domain into account, such as a large-scale wind profile and convergence. Here, we apply an idealized yet realistically forced case setup that is representative for heavy summertime precipitation in the Netherlands. We simulate the base case, as well as a number of experiments with changed temperature and moisture following the assumption of constant relative humidity. To study the characteristics of rain events we cluster continuous areas with precipitation to rain cells. Based on that, we can confirm the relationship between size and intensity of rain cells. In warmer and moister simulations, rain cells grow larger and become more intense over their whole area. The number of large events increases at the cost of smaller events. These results draw a consistent picture together with the previously mentioned observation-based study.
So far, we have considered precipitation events as rather isolated entities from each other. However, another factor that can influence the magnitude of rain events is the process of convective organization. For instance, compare an isolated convective cell to a mesoscale convective system that can be seen as a cluster of multiple thunderstorms. The latter is a more organized form of a convective precipitation event. In this case, the horizontal extent of the two types differs by around a factor of ten. Local dynamics induced by cold pools play a key role in the formation of more organized events. Cold pools originate from evaporation of rain drops in the atmosphere. This causes a cooling of air masses which intensifies the downdraft strength due to their relatively higher density. At the surface, the downdraft will spread out horizontally over a distance of tens of kilometers. This effect is tangible, for instance, when observing a nearby thunderstorm. The relatively dense moving air mass vertically dislocates the surrounding air and can trigger the formation of new convective clouds. The effect is even stronger when two or more cold pools collide. Also, cold pools transport and accumulate moisture into confined regions that are preferred locations of new convective events.
Besides the clear effect of cold pools on the formation and organization of convective rainfall events, little is is know about how these processes change in a warmer climate. Also, how cold pool dynamics respond under these conditions and how this relates to precipitation characteristics is an open question. To study this, we use a similar LES setup as in the above mentioned case. We also repeat the experiments with a warmer and moister setting. Similar to the previously mentioned method of rain cell clustering, we apply the same technique to low-level temperature and moisture fields to identify cold pools and relatively moist areas, so-called moist patches. We see that increased size and intensity of rain cells under warmer and moister conditions have a strong correlation with the size and spreading speed of cold pools. This, in turn, relates well to the size and amount of moisture contained in moist patches. These correlations suggest a feedback loop in which cold pools that originate from initially rather randomly distributed and weak rain events, cause a higher variability in the low-level moisture field, and trigger events that are more organized and have higher intensities. Thus, cold pool dynamics can enhance the local moisture availability and supply to newly forming rain events without additional large-scale convergence. From a separate group of experiments we deduct that enhanced large-scale convergence alone only affects the rain cell area but not the intensity. But, do these findings hold in a more realistic scenario of climatic warming? For instance, it is known that relative humidity will decrease with warming in many regions over land. Furthermore, it is expected that the atmosphere will not get warmer uniformly with height. Due to convection, the temperature profile will follow a change close to a moist adiabatic lapse rate, leading to a more stable atmosphere. Our experiment design applies these principles and separate simulation groups are created for both situations. We find that a reduced relative humidity with warming generally further amplifies precipitation events, cold pool dynamics and moist patches. This highlights relative humidity, besides precipitation intensity, as another factor that controls the evaporation rate of rain drops and ultimately the strength of cold pools. Despite of a dampening effect, more stable conditions with warming still lead to more intense and larger rainfall events that are associated with stronger cold pool dynamics and moist patches.
Finally, the results presented in this thesis show that extreme convective precipitation events exhibit a strong response to climatic warming and highlight the importance of small-scale dynamics in this context. This must be taken into account for better predictions of future extremes.