The aim of this thesis is to identify recent inter-annual patterns of variability and long- term changes in the North Atlantic storm-track using ERA-40 reanalysis data. These results are also compared with variability in simulations by the newly developed Met Office Hadley Centre global coupled-model HadGEMl. The storm-track and storm activity is represented by both a variance and a feature- tracking analysis. The former involves (2-6 day) bandpass filtering the wind fields u',v' to calculate Transient Eddy Kinetic Energy. Feature-tracking involves the identification of positive vorticity (ε) extrema from a background field. These methods are found to be complementary. The analysis indicates that the leading inter-annual North Atlantic storm-track pattern (PI) over the ERA-40 study period involves meridional shifts of the storm-track exit region. This pattern is prominent in both the upper- and lower-troposphere and is found to predominantly be associated with shifts in the typical paths taken by medium and strong storms. PI appears to be unrelated to any key mean-flow teleconnection patterns such as the North Atlantic Oscillation (NAO). PI is strongly associated with local mean- flow changes to geopotential and temperature that support more intense storm activity, which indicates that this pattern is self-maintaining. Two less important patterns are identified in the TEKE teleconnection analysis which represent storm-track intensity in the west (P2) and east (P3) mid-latitude North Atlantic. There is clear evidence of a northward shift in medium and strong storm paths at lower-levels in ERA-40, whilst in the upper-troposphere the strongest storms have not shifted northwards, whilst medium storms have. The HadGEMI storm-track was found to have large decadal variability comparing two forty year periods, which in turn suggests that the ERA-40 results are perhaps just a 'snapshot' of North Atlantic storm-track behaviour. The inclusion of historic greenhouse gas forcings does not influence storm-track variability greatly in comparison with the internal variability of HadGEMl.

The predictability of individual convective storms over the mid-latitude ocean is investigated by quantifying the divergence of pairs of perturbed forecasts in a cloud-resolving model. The Met Office non-hydrostatic Unified Model is used in an idealised configuration to simulate moist convection initiating under homogeneous destabilisation. All convection is represented explicitly since a convective parameterisation scheme is not used. The growth of potential temperature perturbations at a single height is quantified as a function of time and spatial scale. The perturbations are found to grow in two distinct stages. Firstly, changes in the regime diagnosed by the boundary layer parameterisation scheme lead to rapid but limited perturbation growth before growth by convective instability becomes dominant. Both error growth mechanisms are found to contribute independently to the total error growth in the forecast. The range of predictability in this perfect model framework is quantified for different spatial scales and initial condition error. The upper limit (provided by O.OO2K perturbations) is shown to be around 200 minutes at scales of 10km. Initial condition perturbations of similar magnitude to those of typical analysis errors (i.e. of order I K) were found to saturate almost immediately at all scales. The short time taken for the forecasts to become uncorrelated in all cases indicates that individual showers will always be unpredictable beyond approximately four hours. The asymmetry in the evolution of initially equal and opposite perturbations highlights the nonlinear nature of the growth, which could prove problematic for convective scale data assimilation and the design of 'optimal' perturbations for convective scale ensemble forecasting.

The North Atlantic storm track is a key component of the climate system and dominates the weather and climate of north western Europe. Its unique characteristics are thought to be controlled by a combination of surface features including orography, land-sea contrast and sea surface temperature anomalies, although a consistent understanding of the individual and collective roles of these forcings and their feedbacks remains elusive. Such an understanding is vital for confident predictions of climate change in Europe. This thesis presents a comprehensive approach using a state-of-the-art atmospheric general circulation model, HadAM3, to perform a series of simulations with both idealised and semi-realistic boundary conditions. This hierarchy enables the behaviour of the more complex simulations to be interpreted in terms of simpler configurations. In this manner, the effects of the idealised northern hemisphere continents, South America, the Rocky mountains and the sea surface temperature anomalies associated with the Gulf Stream and the North Atlantic Drift are investigated both in isolation and in various combinations. Each of the features is found to be capable of influencing the North Atlantic storm track. In particular, the deflection of the flow caused by the Rocky mountains is important in generating the southwest -northeast tilt of the track. The storm track is suppressed over continental land masses and over the oceans it is highly sensitive to changes in the midlatitude SST gradients on the poleward side of the subtropical jet, with the Gulf Stream enhancing and the North Atlantic Drift suppressing storm activity. Physical mechanisms for these signals are discussed. Finally, the atmospheric response to the sea surface temperature patterns of the North East Atlantic ocean are discussed in the context of the ocean's thermohaline circulation, both in terms of the impacts on Europe and the potential feedbacks on the ocean.

Extratropical cyclones are important to the day to day weather of the midlatitudes. They provide essential rainfall for human activities such as agriculture, but can also cause large amounts of damage by their strong winds and heavy precipitation. It is therefore very important to predict these cyclones as accurately and as far in advance as possible. A new storm tracking approach to forecast verification is developed and implemented to obtain detailed information about the prediction and predictability of storms. The approach is applied to observing system forecast experiments constructed using the European Centre for Medium Range Weather Forecasts (ECMWF) 40-year reanalysis system (ERA40). It is also used to analyse the ECMWF and National Centers for Environmental Prediction (NCEP) Ensemble Prediction Systems (EPS). An Internet Service ha" been developed to allow users to run a storm tracking program from a web browser with remote datasets and distributed computing. The service has been used in this thesis to help with the large amounts of data processing involved. Results show that the intensity and propagatiun speed of storms are more difficult to predict than the direction the storms take. There is a small bias for forecast cyclones to propagate too slowly. The importance of a high vertical resolution of observations is highlighted. This is particularly true for northern hemisphere winter storms, which have much larger growth rates. It is shown that the ECMWF EPS has a slightly higher level of performance than the NCEP EPS in terms of the prediction of storms. The results also illustrate a number of benefits an ensemble forecast can offer over a single deterministic forecast. It is suggested that in the future operational forecast centres should consider using a storm tracking verification approach in addition to the standard approaches.

In an effort to improve upon rainfall forecasts produced by simple storm advection methods (nowcasts) and to broach the gap between them and the forecasts of complex Numerical Weather Prediction (NWP) models, in terms of the spatial detail and length of lead-time each provides, the research presented explores the possibility of combining elements of each into a physically-based algorithm for rainfall forecasting. It is an algorithm that uses as its foundation the rainfall prediction model of Mark French and Witold Krajewski, developed in 1994. Their model was designed to take advantage of the high resolution rainfall observations and tracking abilities provided by weather radar and to achieve a rainfall forecast by augmenting extrapolation techniques with a representation of storm dynamics in the form of "rising parcel" theory. The new algorithm/model retains those features but incorporates NWP data to assist with forecasting, using it as a means to enable an informed choice of algorithm pathways and, more specifically, to identify the ingredients of precipitation, namely ascending air of high moisture content. A case study application of the new rainfall forecasting model to storms in Northern England shows its performance, at a lead-time of one hour, compares favourably with respect to extrapolation and persistence techniques and also NWP forecasts, and that it is able to provide more assured forecasts than persistence and nowcasts at longer lead-times. The robustness of the model is tested and confirmed by way of another case study, this time using Mediterranean storms and with predictions made in the context of urban hydrology. The case studies help to identify aspects of the model that need improvement, with representation of orographic forcing being a key one. Both the model's encouraging performance and its pinpointed weaknesses provide impetus for further research in the area of integrated mesoscale-hydrometeorological modelling for flood forecasting.