Spatio-temporal characteristics of precipitation in very high-resolution climate models

Emma Dybro Thomassen

Research output: Book/ReportPh.D. thesis

78 Downloads (Pure)

Abstract

Precipitation events with a small spatial and temporal scale can cause flooding of cities with huge socio-economic consequences. Cities are especially vulnerable to this type of event, as response time is quick due to impervious surfaces and limited capacity in the sewer system. The Intergovernmental Panel on Climate Change (IPCC) projects an increased risk of heavy precipitation and urban flooding in northern Europe. Climate adaptation of cities relies on estimates of changes in precipitation patterns in the future to avoid increasingly frequent and severe floodings. Climate models are today the primary tool for assessing future precipitation pattern changes. High-resolution climate models, so-called convection-permitting models (CPMs), have been available for around a decade, benefitting from the development in computational power. The kilometre scale resolution of CPMs is expected to provide a better representation of processes important for heavy precipitation events and hence better predictions of changes in precipitation patterns in the future climate.

This PhD project aims to analyse the representation of extreme precipitation events in CPMs to understand how precipitation is resolved and assess such model simulations' applicability in an urban drainage context.

CPM simulations were together with coarser resolution model simulations and reanalysis datasets compared against available observational datasets, such as rain gauges, weather radar and gridded observational datasets. Well-known metrics to the model or urban hydrology community, such as wet days, IDFcurves, extreme event thresholds, moments and spatial correlation, were applied to analyse the added benefit of high-resolution CPM simulations. Metrics to describe characteristics of extreme precipitation were divided into two subgroups; within-event and inter-event metrics. Within-event metrics concern the event development over the lifetime of an event and require post-processing using a tracking algorithm to get the necessary data. Inter-event metrics concern spatio-temporal characteristics of extreme events analysed on time series or between grid points of interest.

A tracking algorithm was applied to a CPM simulation and a coarser resolution Regional Climate Model (RCM) simulation to analyse within-event metrics. A new method was developed in this PhD project, which simplifies area-intensity diagrams for studying the event-evolution of extreme precipitation events across events with many different durations and across models. The simplified event-evolution diagrams showed that the representation of extreme precipitation fundamentally differed between the CPM and RCM analysed in this study. Differences in movement patterns, geographical location, maximum intensity and event sizes were shown.

Metrics to analyse inter-event characteristics of extreme precipitation events showed an improved representation of these in CPM simulation compared to coarser resolution RCM simulations. The differences between models were especially clear for hourly precipitation extremes due to the coarse resolution RCMs which do not allow convection to be modelled explicitly. Despite the improved representation of precipitation events in the CPM simulations, intensities for hourly precipitation and spatial extent of extreme events do still not match extreme events in observational datasets. Spatio-temporal characteristics
analysed and compared between a dense network of rain gauges and biasadjusted weather radar data showed a large consistency between the two observational products. This indicates that bias-adjusted weather radar data can represent intensities measured at rain gauges and that a dense network of rain gauges can represent the same spatial extent as seen in the radar data.

Analysing sub-hourly extremes showed a less promising performance of CPM simulations when benchmarked against observations. The results showed a difference in conclusion between hourly and sub-hourly performance, emphasising the importance of including analysis of sub-hourly precipitation extremes when assessing climate model performance. Due to limitations in observational datasets and the availability of CPM simulations with a sub-hourly output resolution, there are very few studies within this field. This PhD-project has contributed substantially to this field.

The PhD project presented a unique 35-month long sub-kilometre (750m) CPM simulation covering five cloudburst seasons in Denmark. The study showed that with the current model setup, there is a limited added benefit of moving towards sub-kilometre resolution. This benefit does not seem to weigh off the extra computational effort. The study, however, provides a good basis for discussing model improvement concerning the representation of precipitation extremes.

This PhD project has shown that the applied metrics are useful for assessing climate model simulations focusing on extreme precipitation of scales important for urban drainage. The project has shown that both sub-hourly and possible sub-kilometre scales must be considered to understand rainfall representation and improve extreme rainfall representation in climate models.
Original languageEnglish
Place of PublicationKgs. Lyngby
PublisherTechnical University of Denmark
Number of pages146
Publication statusPublished - 2023

Fingerprint

Dive into the research topics of 'Spatio-temporal characteristics of precipitation in very high-resolution climate models'. Together they form a unique fingerprint.

Cite this