The effects of aerosols on the climate system are a major source of uncertainty in past and future simulations of climate. The role played by carbonaceous aerosols is particularly uncertain. Black carbon, which absorbs shortwave radiation, exerts a positive radiative forcing of the climate system, warming the Earth. This warming is offset by co-emitted organic carbon, which scatters shortwave radiation back to space. However, a subset of organic carbon aerosols, called brown carbon, may be absorbing, further complicating the issue. Carbonaceous aerosols also influence cloud formation and properties, potentially contributing an additional cooling of Earth’s surface temperatures. This work aims at improving the simulation of carbonaceous aerosols in the UK Met Office’s HadGEM3-UKCA model with the GLOMAP-mode aerosol scheme. That model does not always compare well with regional multi-variable observations, often underestimating surface concentrations and aerosol optical depth. Although single scattering albedo, a measure of aerosol absorption, is well replicated in biomass burning regions, it is overestimated in regions where fossil- and bio-fuel emissions are dominant. To improve the comparison, we propose a range of changes to modelled carbonaceous aerosol emissions, refractive index and density. An analysis of the sensitivity of the model to these changesshows that emissions alone cannot be responsible for the discrepancies between observations and models. The analysis also shows that while the black carbon mass absorption coefficient needs to be high in fossil fuel combustion regions, black carbon absorption must be reduced in biomass burning areas to leave room for brown carbon absorption. A selected combination of changes leads to improvements in the simulation of carbonaceous aerosols in most regions compared with observations. Those improvements do not however lead to a significant change in the total aerosol effective radiative forcing, but reduces the contribution of aerosol-radiation interactions. Using an internally-mixed aerosol scheme means that changes in one aerosol species affects the simulation of other aerosol types. Therefore, aerosol model improvements require an integrated consideration of all aerosol species.

In this thesis, mountain wave breaking triggered by directional wind shear is investigated using numerical simulations of idealized and semi-idealized orographic flows. Idealized simulations are used to produce a regime diagram to diagnose conditions for wave breaking in Richardson number-dimensionless mountain height parameter space. It is found that, in the presence of directional shear, wave breaking can occur over lower mountains than in a constant-wind case. Furthermore, the extent of regions within the simulation domain where Clear-Air Turbulence (CAT) is expected increases with terrain elevation and background wind shear intensity. Analysis of the model output, supported by theoretical arguments, suggest the existence of a link between wave breaking and the relative orientations of the incoming wind vector and the horizontal velocity perturbation vector. This condition provides a possible diagnostic for CAT forecast in directional shear flows. The stability of the flow to wave breaking in the transition from hydrostatic to nonhydrostatic mountain waves is also investigated. Wave breaking seems to be inhibited by non-hydrostatic effects for weak wind shear, but enhanced for stronger wind shear. In the second part of the thesis, a turbulence encounter observed over the Rocky Mountains (in Colorado, USA) is studied. The role of directional shear in causing wave breaking is isolated from other possible wave breaking mechanisms through various sensitivity tests. The existence of an asymptotic wake, as predicted by Shutts for directional shear flows, is hypothesized to be responsible for a significant downwind transport of unstable air detected in cross-sections of the flow. Finally, critical levels induced by directional shear are studied by spectral analysis of the horizontal velocity wave perturbations. This is done for a fully idealized flow and for the more realistic flow corresponding to the investigated turbulence encounter. In these 2D power spectra, a rotation of the most energetic wave modes with the background wind and their selective absorption can be found. Such behaviour is consistent with the mechanism leading to wave breaking in directional shear flows.

Extreme weather and climate-related events can have devastating impacts on people's lives and livelihoods in developing countries, particularly in Africa. Understanding the influence of anthropogenic climate change on extremes is key when addressing the impacts of events now and in the future. Probabilistic event attribution aims to quantify the effect on individual events by analysing changes in probabilities of their occurrences. This thesis investigates this technique and its applications in an African context. It first assesses whether robust attribution results can be produced for events in West Africa. A case study of high precipitation in 2012 concludes that the probability of such an event was likely decreased due to anthropogenic climate change. The different climate model ensembles analysed produce complementary results, but the study highlights the importance of correctly modelling the anthropogenic climate change impact on sea surface temperatures if results are to be robust in regions such as this with strong teleconnections. The application of event attribution is then studied in two key contexts relevant to addressing the impacts of extreme weather events in Africa: national adaptation policy, focussing on addressing urban flooding in Senegal, and international loss and damage policy. In both cases there were suggestions for roles these scientific results could play, but there are barriers to their inclusion at present. In Senegal little climate information is currently used in decision-making, and loss and damage policy lacks clarity around what it will address and therefore how scientific information can support this. In both cases stakeholders demonstrated limited awareness of event attribution, highlighting the need to enhance understanding to encourage further dialogue about its relevance. To this end a participatory game focussing on event attribution was developed and is shown here to be a useful tool for stakeholders to consider how results could be relevant for their decisions.

The Antarctic Peninsula (AP) was the fastest warming region on earth during the 20th century. The eastern side of the Peninsula is typically 5-10K colder than at equivalent latitudes on the western side. Despite this cooler climate, a number of ice shelves have collapsed in the last two decades. In 1995 Larsen A collapsed, shortly followed by Larsen B in 2002. Larsen C Ice Shelf (LCIS), the largest remaining ice shelf, is now showing evidence of potential destabilisation, including melt ponding and rift acceleration. The `hydrofracture mechanism’ suggests that percolation of melt water into crevasses allows them to deepen and extend to the ice shelf base, which leads to destabilisation. Advection of warm, dry air onto the ice shelf from foehn winds is partly responsible for the melt water. Investigating foehn winds over LCIS was one aim of the Orographic Flow and Climate of the Antarctic Peninsula (OFCAP) project. The aim of this research is to investigate the spatial distribution and frequency of the foehn winds, and assess their impact on the LCIS. To investigate this, near-surface observational data at six locations is combined with archived regional model output at 5km horizontal-resolution from the Antarctic Mesoscale Prediction System (AMPS). A novel semi-automatic algorithm has been developed to detect foehn winds from near-surface observations. A relatively new algorithm has been adapted for use over the AP, to detect foehn conditions from the AMPS output. Foehn characteristics over the ice shelf have been identified to create a mini-climatology of the location and occurrence of foehn winds from 2009 to 2012. Foehn conditions have been observed as far south as ~68S for the first time. Foehn events are most frequent in spring, when over 50% of days can experience the warm, dry winds. The average length of foehn events is approximately 12 hours, but they can occur in quick succession to have a longer-lasting effect. Some of the spatial features within the foehn flow were investigated further using high-resolution (1.5km) Weather Research and Forecasting (WRF) model. Foehn jets, hydraulic jumps and localised foehn enhancement were simulated by WRF in four case studies. The presence of a statically-stable cold pool over the ice shelf appears to reduce the propagation of foehn air, and interrupt the near-surface signal. When foehn winds occur in late spring they prompt earlier melt onset, increase the number of melt days and lengthen the melt season. Foehn-induced surface melting has been observed over 130km from the mountains. This research has highlighted the potential destabilising effect of foehn winds, which may provide an insight into the stability of the LCIS.

Extensive research has been carried out to fully understand the mechanisms involved in the destruction of large amounts of ozone in the Arctic region. For this purpose, a novel light-weight, relatively cheap and low-powered ozone instrument has been developed for measurements in the lower stratosphere on long-duration balloon flights in the polar vortex, and is presented here. The instrument contains a solid-state semi-conducting sensor with a heated tungsten oxide surface. Adsorption of ozone onto the surface increases the resistance of the sensor, so that the value of the resistance gives a direct indication of how much ozone is present. The instrument has flown alongside ozone sondes to test its performance. The free flow configuration, with the sensor placed directly into the air flow caused by the ascent of the balloon, has shown that the instrument measures structures in the ozone profile better than the ozone sonde. Discrepancies between the instrument and sonde measurements are mainly related to changes in the diffusion near the sensor surface at lower pressures, as indicated by laboratory experiments. Smaller effects are related to changes in humidity and flow rate. A quantitative procedure, with a single collaboration fully explaining the sensor behaviour prior to a flight, is within reach. The behaviour of the long-duration balloon has been modelled, providing a test for the accuracy of a radiative transfer model and its sensitivity to several input variables. The radiative cooling rates in the lower stratosphere are a measure of the descent, which in turn determines the amount of ozone transported into the region with the highest destruction rates.

Remote sensing of the atmosphere using a Fourier transform infrared spectrometer with the sun as a source of radiation allows the measurement of vertical column amounts of trace species in the atmosphere. This technique has been applied during the winters of 1994/95 and 1995/96 from a mid-latitude site at Rutherford Appleton Laboratory (RAL), Chilton, Oxon and results from this research are discussed in this thesis. An overview of the theory of infrared spectroscopy is presented followed by a description of the spectrometer used to record the measurements. A 'forward model' that has been developed to analyse the measurements and the retrieval technique for reducing the measurements to vertical column amounts of HF, HNO<SUB>3</SUB>, HCI and C1ONO<SUB>2</SUB> is discussed. Variations in vertical column amounts of mid-latitude HF measured during the 1994/95 winter can be understood as a results of variations in tropopause height and movement of air from higher and lower latitudes over the measuring site. Columns of HC1 and C1ONO<SUB>2</SUB> indicate that chemically processed air with a depleted HC1 column and enhanced C1ONO<SUB>2</SUB> column were measured on the 23<SUP>rd</SUP> January 1995, associated with the polar vortex which meteorological analyses indicate was above RAL on this day. Similar measurements performed by a group from the National Physical Laboratory (NPL) from Aberdeen show depletions of both HC1 and C1ONO<SUB>2</SUB>, indicating that there is a strong latitudinal gradient in the C1ONO<SUB>2</SUB> column at the edge of the polar vortex. Comparisons with a three-dimensional off-line chemical transport model (CTM) show that dynamical variability in the mid-latitude stratosphere is well represented, but that chlorine partitioning and the effect of seasonal timescale diabatic descent at high latitudes are less accurately calculated.

Recent trends in global energy systems have seen a rapid uptake in renewable generation, however, few studies have investigated the impacts of inter-annual climate variability and climate change on power system operation. This thesis aims to explore these impacts for the GB power system. Multi-decadal re-analysis and climate model datasets are used to create demand and wind power time-series as inputs for a load duration curve based power system assessment. Using the MERRA reanalysis, it was found that all aspects of the GB power system are impacted by inter-annual climate variability, but the impacts are most pronounced for baseload generation. The impacts of climate variability are amplified by increasing onshore wind power capacity, and decreased by increasing offshore wind power capacity. The GB power system model is most sensitive to winter weather. A system with no installed wind power capacity is driven by inter-annual variability in temperatures. As the amount of installed wind power capacity is increased, the power system becomes increasingly sensitive to variability in winds. It was found that more than 10 years of climate data are required to adequately sample the impacts of inter-annual variability of climate on the power system. In the HiGEM 4XCO2 climate scenario, mean winter demand reduces (-6%) while mean summer demand increases (+5%) primarily due to warmer temperatures. These changes result in a reduction in the use of conventional generation (-30%) and peak load (-6%). Furthermore, suggesting that climate change may somewhat counteract the increases in inter-annual power system variability which would otherwise be associated with increasing installed wind power capacity.

Statistical models of atmospheric variability typically attempt to account for deterministic seasonal variations by constructing a long-term average for each day or month of the year. Year-to-year variability can then be treated as some form of stochastic process about this long-term average. In general, the stochastic processes are assumed to be statistically stationary (invariant under time translation). However, for a non-linear system such as the Earth’s atmosphere, multiple seasonal evolutions may be possible for the same external forcing. In the presence of such a multiplicity of solutions, the identification of a seasonal cycle with a long-term average may not be the optimal procedure. Previous research has suggested that multiple evolutions of the seasonal cycle of the Southern Hemisphere mid-latitude circulation may be possible. The central goal of this thesis is to build on this work and to present evidence for different seasonal evolutions of the Southern Hemisphere mid-latitude circulation. This evidence is initially presented by highlighting a low-frequency peak in an aspect of the Southern Hemisphere mid-latitude circulation that is viewed as a harmonic of the annual cycle (quasi-two year). Statistically stationary models of variability about a long-term average are argued to be unable to account for the presence of this harmonic. Following this, an alternative model of circulation variability is proposed that explicitly references various stages of the seasonal cycle in a deterministic manner. In particular, explicit reference is made to the downward shift and to the final breakdown of the stratospheric polar vortex. A re-interpretation of several previous results in the literature including Southern Annular Mode persistence timescales, Southern Hemisphere mid-latitude climate change and the semi-annual oscillation of the mid-latitude jet is subsequently presented using this alternative perspective.

Flooding from intense rainfall, resulting from convection, causes millions of pounds of damage each year. However, convection has limited predictability, often resulting in short lead times for warnings of such events. This research aims to determine the spatial scales of perturbation growth in convective-scale forecasts for different environmental conditions over the British Isles. The convective adjustment timescale identifies whether cases are in equilibrium with the large-scale forcing, so can be used to quantitatively classify convection into the regimes of quasi-equilibrium and non-equilibrium. A method is found to calculate the timescale based on criteria considering its variability and the environment in which it is calculated. The most appropriate method uses a Gaussian kernel to spatially smooth convective available potential energy and precipitation accumulation fields before the calculation. A model climatology is created over the summers of 2012-2014 (due to limited operational data from the United Kingdom Variable resolution, UKV, configuration of the Met Office United Model) to understand the characteristics of the regimes over the British Isles. In summer 85% of convection is in convective quasi-equilibrium, with more nonequilibrium events in the south and west of the British Isles. The UKV is perturbed with Gaussian buoyancy perturbations to create an ensemble. These perturbations represent unresolved processes within the boundary layer. The perturbation growth is examined across a spectrum of cases and it is shown that events at the non-equilibrium end of the spectrum have higher spatial predictability than those at the equilibrium end (O(1 km) vs. O(10 km)), implying more localised growth in nonequilibrium, than quasi-equilibrium, environments. This research has implications in the interpretation of forecasts for defining regions when issuing weather warnings associated with convection. The research also has implications for adaptive forecasting, in which high-resolution forecasts are used for nonequilibrium convection and large-member ensembles are used for events in convective quasi-equilibrium.

Recent studies have demonstrated that the representation of orography in models is highly uncertain. Motivated by the large spread in the climatological circulation and the circulation response to climate change seen among models, the primary aim of this work is to quantify the uncertainty introduced by their representation of orography. This is done through a number of experiments using different comprehensive atmospheric models across horizontal resolutions and timescales. First, it is shown that two of the models considered are unable to maintain an equivalent total (resolved plus parameterized) orographic drag across resolutions over the Northern Hemisphere (NH) mid-latitudes, leading to systematic biases at lower climate model resolutions. The suitability of substituting one drag parameterization scheme for another is also investigated. It emerges that there is a strong regional dependence of the model error on the drag parameterization scheme employed. High-resolution global and limited area models analysed over the Himalayan Plateau are used as a proxy for the truth. The non-robustness to resolution over this region is attributed to particular components of the orographic drag parameterization scheme and its formulation. It is shown that most of the reduction in short-range forecast error that occurs with increasing resolution is due to a reduction in the parameterized orographic drag, as opposed to the addition of resolved orographic drag. Finally, the impact of the uncertainty in the parameterized orographic drag scheme on the circulation and its response to climate change is investigated. The low-level parameterized orographic drag is found to be beneficial for the modelled stationary wave field over the NH and for the jet latitude in both hemispheres. Over the NH, the amplitude of the stationary wave response to climate change across the experiments is shown to scale with the magnitude of low-level parameterized orographic drag through its influence on the present-day climatological stationary wave amplitudes.