Search

Extratropical cyclones produce extreme surface wind speeds and heavy rainfall which can individually and jointly influence impacts and potentially produce large aggregate impacts. Within this study, we assess the UKCP 12-member ensemble of local convection-permitting 2.2 km climate projections. We quantify the likelihood of cyclones producing large footprints of both extreme winds and rainfall over the UK in a control (1981-2000) and future (2061-2080, RCP8.5) climate simulation. Following this, we characterise the convective and frontal drivers of wet and windy conditions within cyclones, and identify the characteristics of cyclones, their tracks and interactions with the jet stream that contribute to the occurrence of large, combined footprints in the control and future simulations. The future simulations project an increased probability of extratropical cyclones producing extremely wet and windy conditions in the same storm, as well as an increase in the land area covered by such conditions. In both the control and future simulations, combined wet and windy extremes largely occur close to cold and warm fronts, likely due to the warm conveyor belt which produces heavy rainfall (due its ascent over the frontal boundaries) and high winds (when occurring within a region of tight pressure gradients). Cyclone composites reveal that the largest changes in joint extremes are closely located within the sector of cyclones where we expect to see the warm conveyor belt, suggesting their change arises partly through the response of this shared driver rather than being a simple consequence of increased rainfall due to thermodynamics. In further analysis, we identify favourable conditions and cyclone characteristics that lead to cyclones producing large rainfall and wind footprints over the UK.

Short-duration (1 to 3 hour) rainfall extremes can cause serious damage to infrastructure and ecosystems and can result in loss of life through rapidly developing (flash) flooding. Short-duration rainfall extremes are intensifying with warming at a rate consistent with atmospheric moisture increase (~7%/K) that also drives intensification of longer-duration extremes (1day+). Evidence from some regions indicates stronger increases to short-duration extreme rainfall intensities related to convective cloud feedbacks but their relevance to climate change is uncertain. This intensification has likely increased the incidence of flash flooding at local scales, particularly in urban areas, and this can further compound with an increased storm spatial footprint to significantly increase total event rainfall. These findings call for urgent climate-change adaptation measures to manage increasing flood risks, including rethinking the way climate change is incorporated into flood estimation guidance.

High resolution rainfall projections are useful for planning for climate change [1] but are expensive to produce using physical simulations. We make novel use of a state-of-the-art generative machine learning (ML) method, diffusion models [2], to more cheaply generate high resolution (8.8km) daily mean rainfall samples over England and Wales conditioned on low resolution (60km) climate model variables. The downscaling model is trained on output from the Met Office UK convection-permitting model (CPM) [3]. We then apply it to predict high-resolution rainfall based on either coarsened CPM output or output from the Met Office HadGEM3 general circulation model (GCM). The downscaling model is stochastic and able to produce samples of high-resolution rainfall that have realistic spatial structure, which previous methods struggle to achieve. It is also easy to train and should better estimate the probability of extreme events compared to previous generative ML approaches.The downscaling model samples match well the rainfall distribution of CPM simulation output. We use as our conditioning variables We obtained further improvements by also including high-resolution, location-specific parameters that are learnt during the ML training phase. We will discuss the challenges of applying the model trained on coarsened CPM variables to GCM variables and present results about the method’s ability to reproduce the spatial and temporal behaviour of rainfall and extreme events that are better represented in the CPM than the GCM due to the CPM’s ability to model atmospheric convection.References[1] Kendon, E. J. et al. (2021). Update to the UKCP Local (2.2km) projections. Science report, Met Office Hadley Centre, Exeter, UK. [Online]. Available: https://www.metoffice.gov.uk/pub/data/weather/uk/ukcp18/science-reports/ukcp18_local_update_report_2021.pdf[2] Song, Y. et al. (2021). Score-Based Generative Modeling through Stochastic Differential Equations. ICLR.[3] Met Office Hadley Centre. (2019). UKCP18 Local Projections at 2.2km Resolution for 1980-2080, Centre for Environmental Data Analysis. [Online]. Available: https://catalogue.ceda.ac.uk/uuid/d5822183143c4011a2bb304ee7c0baf7

How lightning and hail will change in the future with global warming is highly uncertain, mainly due to the small space and time scales of these phenomena, which requires high-resolution climate simulations for their accurate representation. Recently km-scale climate simulations have become available that are ‘convection-permitting’ and give the possibility of studying how convective storms may change. We use a pan-European convection-permitting (2.2km grid spacing) model to analyse future changes in deep moist convection, lightning, and hail across Europe under a high emissions-scenario. This includes a physically based lightning prediction scheme, that uses graupel and ice-fluxes that are explicitly represented within the model. It also provides convective-scale outputs, not previously available from conventional climate models, exploited here to study future changes using an ingredients-based approach inherited from severe storm forecasting. For hail, we derive a new severe hail potential metric that represents the likelihood of hail >2cm at the surface, accounting for hail growth and subsequent melting. Overall, we find an increasing frequency of deep moist convection across Europe mainly due to more instability with warming, but partially limited by greater convective inhibition. However, the increase in storms does not directly translate to an increase in lightning, mainly due to much higher melting level heights resulting in less cloud ice and hence less electrification. Weaker circulation and a northward shift of favourable weather regimes in the future simulation also affects the results, which vary regionally. While lightning increases in future over the mountains (due to enhanced Alpine Pumping) and in the north, lightning decreases over much of central Europe and over the sea. The potential for severe hail generally decreases in future, although storms locally produce higher amounts of graupel in the south. We also identify a new, warmer type of thunderstorms in the future, which may have important implications for changes to the frequency of the largest hailstones.

Convection permitting climate models (CPMs) are nowadays increasingly used in climate change assessment. These models have shown to have vastly improved convective rainfall statistics compared to parameterized regional climate models (RCMs). Here, we analyse hourly rainfall extremes within the framework of scaling, investigating the dependencies on temperature, dew point temperature as measure of absolute humidity, and dew point depression as a measure of relative humidity. We compared 7 RCM simulations and 5 CPM simulations to observations from The Netherlands (a moderate moist climate) and Southern France (a warm and dryer climate). Although present-day scaling is no simple predictor of climate change, reproducing observed dependencies on the various temperature and humidity measures provides evidence that these models are trustworthy in a climate change setting. We find that RCMs display a large spread in outcomes, in particular in their dependency on relative humidity and usually strongly biased towards too strong suppression of extremes in low relative humidity conditions. CPMs have (unsurprisingly) much better overall rainfall statistics, show much less inter-model spread, and temperature and humidity dependencies more consistent with the observations. Yet, most CPMs have a climatology biased towards too low relative humidity, affecting also the rain statistics, and underestimate the frequency of rain, in particular for conditions with high relative humidity. Our results suggests that, while CPMs are clearly better in convective rain processes, improvement are needed in weakly surface forced convection as well as the overall climatology/water balance of the models.

The future change in dry and humid heatwaves is assessed in 10 year pan-African convective scale (4.5km) and parameterised convection (25km) climate model simulations. Compared to reanalysis, the convective scale simulation is better able to represent humid heatwaves than the parameterised simulation. Model performance for dry heatwaves is much more similar. Both model configurations simulate large increases in the intensity, duration and frequency of heatwaves by 2100 under RCP8.5. Present day conditions that occur on 3 to 6 heatwave days per year will be normal by 2100, occurring on 150-180 days per year. The future change in dry heatwaves is similar in both climate model configurations, whereas the future change in humid heatwaves is 56% higher in intensity and 20% higher in frequency in the convective scale model. Dry heatwaves are associated with low rainfall, reduced cloud, increased surface shortwave heating and increased sensible heat flux. In contrast, humid heatwaves are predominately controlled by increased humidity, which is associated with increased rainfall, cloud, longwave heating and evaporation, with dry bulb temperature gaining more significance in the most humid regions. Approximately one third (32%) of present day humid heatwaves commence on wet days, suggesting the potential for compound flood-humid heat climate extremes. Moist processes are known to be better represented in convective scale models. Climate models with parameterised convection, such as those in CMIP, may underestimate the future change in humid heatwaves, which heightens the need for mitigation and adaptation strategies and indicates there may be less time available to implement them to avoid future catastrophic heat stress conditions than previously thought.

Thunderstorms are the most common source of hazardous weather at local scales, but are poorly represented in conventional climate models, resulting in high uncertainty in future changes. How these changes manifest in terms of lightning is even more uncertain, with previous studies giving conflicting projections. Here, for the first time, we use a km-scale model that explicitly resolves convection across Europe; suggesting more convection by 2100 under RCP8.5. Our ice-based lightning scheme indicates a highly altered lightning climatology-the consequence of general increases in instability, partly limited by convective inhibition, along with huge increases in melting level height and less cloud ice. A northward shift in favourable weather regimes increases lightning frequency at higher latitudes, and favours more thunderstorms over the Alps, but lightning decreases over lower terrain elsewhere and over the sea. Our results suggest the need to re-evaluate lightning risk to wildfires, properties, and human life across Europe.

The need for improved understanding of how a warming climate may change precipitation variability and extremes has focused model developers' attention on the inability of convection parameterizations to represent the observed range of deep convective processes. Under particular scrutiny are the consequences of poorly simulated sub-daily, grid-point precipitation variability on rainfall distributions at longer (e.g., daily, seasonal, decadal) timescales and larger spatial scales. Lack of knowledge or understanding of the spatial and temporal variability in rainfall, in observations and models, hampers model development and can undermine our confidence in projections. A major challenge in advancing our understanding is a lack of comprehensive diagnostics and metrics for analysing the characteristics of both observed and modelled preciptitation across time and space scales. The ASoP diagnostic package (Analysing Scales of Precipitation; Klingaman et al. 2017; Martin et al., 2017) has been developed and applied to various model and observation datasets over the past few years. ASoP can be applied to data ranging from the gridscale and time-step to regional and sub-monthly averages, and measures the spectrum of precipitation intensity, temporal variability as a function of intensity, and spatial and temporal coherence. When applied to time-step, gridscale tropical precipitation from a range of models, the diagnostics reveal that, far from the "dreary" persistent light rainfall implied by daily mean data, most models produce a broad range of time step intensities that span 1-100 mm/day. Averaging precipitation to a common spatial (km) or temporal (3h) resolution substantially reduces variability among models, demonstrating that averaging hides a wealth of information about intrinsic model behaviour. ASoP analysis of tropical rainfall variability in MetUM simulations at a range of horizontal resolutions shows that the behaviour of the deep convection parametrization in this model on the native grid and time step is largely independent of the grid-box size and time step length over which it operates. There is also little difference in the rainfall variability on larger/longer spatial/temporal scales. Tropical convection in the model on the native grid/time step is spatially and temporally intermittent, producing very large rainfall amounts interspersed with grid boxes/time steps of little or no rain. Spatial and temporal averaging smoothes out this intermittency such that, on the km scale, for oceanic regions, the spectra of 3-hourly and daily mean rainfall in the MetUM agree fairly well with those from satellite-derived rainfall estimates, while at 10-day timescales the averages are overestimated, indicating a lack of intra-seasonal variability. Over tropical land the results are more varied, but the model often underestimates the daily mean rainfall (partly as a result of a poor diurnal cycle) but still lacks variability on intra-seasonal timescales. ASoP diagnostics have also been applied to European rainfall (Berthou et al., 2018) and in high-resolution rainfall projections for the United Kingdom (Kendon et al., 2020). Such work is shedding light on how uncertainties in modelling small-/short-scale processes relate to uncertainty in climate change projections of rainfall distribution and variability, with a view to reducing such uncertainty through improved modelling of small-/short-scale processes.

At the UK Met Office we have recently completed climate change simulations at convection-permitting resolution (2.2km grid scale) across a pan-European domain, which are feeding into the CORDEX-FPS and European Climate Prediction System (EUCP) projects. At such high resolution, the model gives a much better representation of convection and is able to capture hourly precipitation characteristics, including extremes, as well as better representing the influence of mountains, coastlines and cities. In this talk, I will present results from these new convection-permitting climate simulations, looking at future changes in high impact events, including hourly precipitation extremes and severe winds. I will also discuss remaining outstanding issues, such as the deficiencies in land-surface-atmosphere coupling, and work underway to try and address these.

This study compares the representation of extreme precipitation events between a 2.2km resolution Convection Permitting Model (CPM) and a 12km Regional Climate Model (RCM) from the UK Met Office over a pan-European domain from 1999-2008. The tracking algorithm DYMECS is applied to the data (which has been regridded to a common 12km grid), to compare evolution in size and intensity, and the spatial and temporal occurrence of extreme events between the two models. Extreme events within a northern European case area are categorised in terms of maximum intensity and maximum size reached within their lifetime. The CPM and RCM data show disagreement in the spatial distribution, movement direction and seasonal occurrence of maximum intensity storms. On the other hand, the two models show good agreement in both the spatial and seasonal distribution of the largest storms in terms of spatial area. A comparison with the COSMO REA6 reanalysis model with continuously nudged observations, suggests that the CPM performs better in capturing the spatial and seasonal distributions of the maximum intensity storms. The study concludes that the CPM seems to improve the representation of extreme events in terms of their spatiotemporal characteristics, with more realistic intensities.