A research version of the Canadian regional finite-element (RFE) model is used to evaluate the capability of the operational model in reproducing the meso-$ beta$-scale structure and evolution of three different types of midlatitude squall-line systems, and to advance our understanding on the development of these features. / In this thesis, we use the well-documented 10-11 June 1985 squall line as a test bed to examine the appropriate incorporation of various physical representations and their coupling with RFE's model components. It is demonstrated through a series of sensitivity studies that the operational prediction of squall lines can be improved if more realistic model physics, reasonable initial conditions, and high resolution are used. It is shown that subgrid-scale moist convection and grid-scale moist physics must be adequately treated in order to reproduce the internal structures of the squall line. / Then, the improved version of the RFE model is used to study the role of gravity waves in the development of a prefrontal squall line associated with the 14 July 1987 Montreal flood. It is found that the gravity waves and convection propagate in a "phase-locked" manner and that the wave-CISK mechanism accounts for the maintenance and intensification of the system. It is also found that frontogenetical processes and release of conditional symmetric instability are responsible for the development of a trailing stratiform rainband associated with the July 1987 Montreal flood. Numerous sensitivity experiments are conducted, and they show that the meso-$ beta$-scale structures and the wave-convection system are very sensitive to the interaction of the parameterized convection with grid-scale physical processes. / In the last part of the thesis, the along-line variability 26-27 June 1985 squall line during PRE-STORM is examined. It is found that the three-dimensional structures of the squall's circulations are determined by both a large-scale trough and convectively generated disturbances. In particular, it is shown that rear inflows in the stratiform region tend to be more intense to the south of the mesolow and neat the base of the large-scale trough.

The dynamics of low-frequency fluctuations (periods between 10 days and a season) is investigated. While those fluctuations are known to be forced, at least in part, by the underlying surface, the emphasis in this thesis is placed on processes taking place within the atmosphere. / The thermal interaction between the high-frequency (periods 2 to 10 days) and the low-frequency flow is first investigated. The temporal and spatial relationship between the heat flux convergence by the synoptic-scale eddies and the low-frequency temperature field is identified. It is shown that the low-frequency temperature fluctuations are negatively correlated with the heat flux convergence by the synoptic-scale eddies, implying the damping effect of high frequency eddy heat flux on the slowly varying temperature field. The damping effect is not homogeneous in space, however, and different temperature patterns have different damping rates. The low-frequency temperature patterns over the North American and Siberian inland areas, where the strongest low-frequency temperature variances are observed, are associated with weak damping from the high-frequency eddies. A stronger dissipation is found for the low-frequency temperature variations in the storm track regions. / We then examine the El Nino-related interannual variations of the transient eddy activity and the associated multiscale interactions. A dataset from the U.S. National Meteorological Center (NMC) of 24 years is used. The purpose is to determine when the seasonal mean flow is less dependent on processes internal to the atmosphere and therefore more dependent on external forcing and thus more predictable. The results show that during El Nino winters the low-frequency eddy activity is reduced over the North Pacific and the high-frequency baroclinic waves are shifted south-eastward of their normal position in the Pacific. Over the North Pacific less kinetic energy is supplied to the low-frequency eddies both from the large-scale seasonal mean flow and from the synoptic-scale eddies. Thus the atmosphere in that region is more "stable" during El Nino winters, and its state depends more on the external forcings. / Finally, the atmospheric predictability is studied explicitly. A large number of numerical experiments are performed to determine whether the forecast skill is dependent on the weather regimes. The predictions are made with a T21 three-level quasi-geostrophic model. The relationship between the forecast behaviour and the "interannual" variation of the Pacific/North American (PNA) anomaly is investigated. Comparison of the error growth for the forecasts made during the positive and negative PNA phases indicates that little differences of error growth can be realized before about a week. After that period the forecast error grows faster during the negative PNA phases. The forecast skill for the medium- and long-range predictions over the North Pacific, the North American and the North Atlantic regions is higher during the positive PNA phase than that during the negative PNA phase. A global signal of this relationship is also observed. The physical mechanism for the difference of error growth is discussed.

Two classes of warm-season meteorological events in Southern Quebec have been studied. (1) thunderstorms and (2) mesocyclones detected by radar. / Characterization of the synoptic-scale flow accompanying warm-season mesoscale vortices and isolation of the moisture flow are documented. A comparison of both 500-hPa height and dynamic tropopause potential temperature composite fields between each set is completed. Precipitable water content and mixing ratio composites of the two classes are diagnosed. / The storm sample associated with mesocyclones show an organized trough-ridge couplet 3 to 4 days before the onset time. Strong precursors of coherent dynamic tropopause forcing seen several days in advance for the mesoscyclones are not as apparent in the thunderstorm sample. During several days prior to the events, the Atlantic Ocean, the Gulf of Mexico and the Great Lakes form a wider and more active area feeding moisture into the mesocyclonic systems than into the thunderstorms.

The Northern Aerosol Regional Climate Model (NARCM) is being developed in order to better understand how aerosols affect the Canadian climate. NARCM uses microphysical and chemical parameterizations that enable it to predict the evolution of the aerosol spectrum using the concentration and composition in 12 size-segregated bins as prognostic variables. In this study, two aerosol species are considered as prognostic variables within NARCM: sulphate and sea-salt. The sulphur species concentrations and aerosol distributions simulated by NARCM are evaluated against clear-sky in-situ aircraft measurement taken off the coast of Nova Scotia during the North Atlantic Regional Experiment (HARE). NARCM demonstrates skill at predicting the column burdens of the sulphur species concentrations at the NARE site throughout the NARE period. Simulations of the average column burdens of sulphur dioxide and sulphate showed differences of 57% and 28% respectively from the observed values. The accumulation mode in the aerosol distributions simulated by NARCM has a tendency to be smaller than the measured mode. There is evidence that the absence of other aerosol species may account for the too small sizes of the simulated aerosol when sulphate volumes are small.

In this thesis, methods based on the wavelet transform are used to extend spectral methods for studying turbulent mixing. These methods are then applied to momentum and heat flux for fast response measurements taken above and within a deciduous forest at Camp Borden, Ontario. / Multiscale distributions of flux event intensities are derived, presented and interpreted. Results show intensification of mixing efficiency near the canopy top, dominated by turbulent structures near a dominant scale, as reported in previous studies. There is no indication of a distinct, clearly separable population of intense 'coherent structures', as is often assumed, but rather an overall increase in intermittency near a particular scale. / The multiscale flux event distributions are simplified to provide component cospectra for down-gradient and counter-gradient fluxes. Dimensional arguments are used to explain observed scaling, and differences between upward and downward cospectra of momentum above and within the forest are used to understand the influence of different terms in the Reynolds stress budget. / A multiscale parameterization of the Reynolds stress budget is developed using the wavelet results for momentum flux. The physical meaning of the scale-dependent parameters is examined. Empirical values derived from the Camp Borden observations show considerable consistency. Changes in parameter values within the canopy are consistent with the effect of neglected Reynolds stress budget terms which are known to become significant within the forest.

The Empirical Orthogonal Function (EOF) analysis technique has proven to be one of the most powerful methods to analyze data in meteorology and many other fields. However, this method is statistical only and has no physical basis. Brunet (1994) has introduced Held's (1985) concept of conservation of wave activity and orthogonal functions into the EOF analysis and called it the "Empirical Normal Mode" (ENM) analysis technique. This new method uses both statistical concepts from the classical EOF analysis method and a dynamical constraint from the generalized Eliassen-Palm theorem to ensure that the functions that we obtained are orthogonal to each other and are the solutions of linearized dynamical equations. / In this thesis, we use the ENM analysis to analyze data from both a (2D) shallow water model integration and from 3-D atmospheric observations, with an emphasis on stratospheric sudden warming events. / For the shallow water model case, the results of the ENM analysis are evaluated by testing against the theoretical (numerical) normal mode solutions provided by Longuet-Higgins (1968). It is shown that the ENM analysis can recover the spatial structures and the frequencies of the normal modes with a great degree of accuracy if the temporal record is sufficiently long. The average errors in the periods for 2000 and 100 day time series are found to be 1% and 4.6%, respectively. From the eigenvalues (percentage of the total variance) and sharp frequency peaks associated with normal modes, the ENM analysis shows that the model generates only a few modes with monochromatic frequencies. The method can be used to test a new or modified shallow water model integration or to study other Hough modes generated by different kinds of forcings. / Having shown the value of the ENM technique in a barotropic context, we advance further by performing an ENM analysis on an 11 year atmospheric data set. In this study, we focus on stratospheric warming events. The winter (DJF) data set is partitioned into warming and non-warming periods in order to characterize the flow differences between the regimes. The stratospheric quasi-potential vorticity or wave activity structure in the warming period is found to be much stronger, as expected, than in the non-warming periods. The ENM analysis clearly shows the tropospheric difference between the two periods, e.g., a higher wave activity in the main tropospheric structure as well as in the tropospheric polar regions in the warming periods. The analysis also reveals that there is a higher level of stratospheric wave activity during the warming periods in the second normal mode of zonal wave number 1 but the tropospheric structures of the quasi-potential vorticity are the same as during non-warming periods. This suggests that there is/are (a) mechanism(s) associated with the stratospheric warming other than the upward wave propagation. All the common features of the stratospheric warming event are captured by the first two normal modes of zonal wave numbers 1 and 2, such as wave-mean flow interaction leading to the deceleration of the zonal mean wind, the polar vortex being displaced by the northward movement of the Aleutian High, as well as wave amplitude enhancement/reduction during the growing/decaying stages.

Traditionally, the diffusional growth of a cloud droplet population is calculated using values of the environmental conditions that represent averages over large volumes, the so called macroscopic conditions (Srivastava 1989). However, it is apparent that the growth rate of an individual droplet is a function of the temperature and the vapor pressure in its immediate environment. These quantities vary from droplet to droplet and with time in a turbulent medium such as a cumulus cloud. In most theoretical and numerical studies of clouds, the hypothesis is made that these variations are unimportant when calculating the growth of an ensemble of droplets. The objective of this work is to determine the validity of this hypothesis. In order to do so we use a 3D turbulence model coupled with a cloud droplet growth model which solves for the trajectories and growth of several tens of thousands of individual droplets as a function of their local conditions (microscopic approach). / A series of experiments with various initial size distributions were conducted using no turbulent flow conditions or one of three turbulent flows with increasing eddy dissipation rate. The results show that in the absence of any turbulent flow or sedimentation of droplets, the non-uniform distribution of cloud droplets in space results in significant variance of the distribution of the supersaturation perturbation over all droplets (DSP) and the distribution of the degree of growth (DDG), defined as the Lagrangian integral of the supersaturation perturbation along each droplet's trajectory. The variance of the DDG is directly responsible for the broadening of the microscopic size distribution relative to the macroscopic size distribution. However, in the presence of turbulence and sedimentation of droplets, the variance of the DSP is significantly reduced. Furthermore, the average, over all droplets, of the decorrelation time of the supersaturation perturbation decreases as a function of increasing level of turbulence. Consequently, the variance of the DDG is significantly reduced compared to the no turbulence and no sedimentation experiments and furthermore, it decreases as a function of increasing level of turbulence. / We have found that for the typical levels of turbulence found in adiabatic cloud cores, the spatial distribution of the larger cloud droplets can significantly deviate from a Poisson distribution. The increasing preferential concentration as a function of increasing level of turbulence does contribute to an increase in the DSP as a function of increasing level of turbulence. However, the DDG decreases as a function of increasing level of turbulence. / These results are at odds with those in the idealized studies of Pinsky et al. (1996) and Shaw et al. (1998). These authors specified and maintained very significant preferential concentration artificially rather than obtaining it by solving explicitly the trajectories of the droplets. / Comparison of our results with the observations of Brenguier and Chaumat (096) made in adiabatic cloud cores lead us to the conclusion that the microscopic approach, even under the most favorable condition of no turbulence, produces too little broadening to explain the observations.

A four dimensional variational (4D-Var) formulation is developed for the assimilation of radar reflectivity, Doppler velocities, and near-surface refractivity index data into a non-hydrostatic fully compressible limited-area atmospheric model coupled with a simplified warm microphysics scheme. The cloud-model is used as a weak constraint so the model error is explicit in the 4D-Var formulation. The ultimate goal is to provide initial conditions to a high-resolution numerical weather prediction model. The environmental flow around storms is modelled by a linear wind in a moving frame using Doppler velocity measurements over a given assimilation window. A three-stage procedure is established to solve the assimilation problem. The background-, observationand model-error statistics are adaptively estimated by comparison with a posteriori residuals and they converge after only a few minimizations. During the adaptive procedure, a smoothing constraint is applied to the analysis variables. The smoothing constraint diminishes towards zero for the last minimization while still leading to a smooth analysis. / Experiments with synthetic data from model outputs at 1 km horizontal resolution show that the method is able to retrieve unobserved variables. An assimilation period of 10 minutes is shown to be optimal for the analysis of clouds. All the other variables of the model are rather insensitive to the assimilation period. Here the model time step has been varied from 1 to 5 minutes. Most of the a posteriori residual distributions have a high kurtosis while the velocity and the near-surface refractivity index residual distributions are nearly Gaussian. The data assimilation for the case of a shallow hailstorm suggests that the model used for the assimilation is able to forecast the system for 30 minutes within the estimated observational errors. Application of the method to the initialization of the MC2 model leads to a better forecast of a convective system over 40 minutes than the nowcasting technique based on Lagrangian persistence.

The purpose of this thesis is to evaluate the downscaling ability of one-way nesting regional climate models (RCM). To do this, a rigorous and well-defined experiment for assessing the reliability of the one-way nesting approach is developed. This experiment, baptised the Big-Brother Experiment (BBE), is used for addressing some important one-way nesting issues. / The first part of this work is dedicated to the development of a scale decomposition tool employed for the BBE. This tool involves a new spectral analysing technique suitable for two-dimensional fields on limited-area domains, and is based on the discrete cosine transform (DCT). It is used for degrading the spatial resolution of the lateral boundary conditions (LBC) used to drive the Canadian RCM (CRCM), for extracting mesoscale features from the atmospheric fields, and for regional validation, and producing power spectra. / The second part of the thesis describes the BBE framework and its first results. The BBE consists in first establishing a reference virtual-reality climate from an RCM simulation using a large and high-resolution domain. This simulation is called the "Big Brother". This big-brother simulation is then degraded toward the resolution of today's global objective analyses (OA) and/or global climate models (GCM) by removing the short scales. The resulting fields are then used as nesting data to drive an RCM (called the "Little Brother") which is integrated at the same high-resolution as the Big Brother, but over a sub-area of the big-brother domain. The climate statistics of the Little Brother are then compared with those of the big-brother simulation over the little-brother domain. Differences between the two climates can thus be unambiguously attributed to errors associated with the dynamical downscaling technique, and not to model errors nor to observation limitations. The results for a February simulation shows that the Canadian RCM, using a factor of 6 between the model and the LBC spatial resolution, and an update interval of 3 hours, is capable to pass the BBE test; thus showing the reliability of the one-way nesting approach. / In the third and last part of the thesis, the BBE is used to investigate the sensitivity of an RCM to the spatial resolution and temporal update frequency of the LBC. It is shown that spatial resolution jumps of 12, and an update frequency between twice a day (every 12 hours) and four times a day (every 6 hours), are the limits for which a 45-km RCM yields acceptable results.

Despite being ubiquitous in the Mackenzie River Basin (MRB) of Canada, the role of snow in its energy and water budgets are still open to much speculation. This thesis presents a multi-scale analysis of the contribution of blowing snow to the hydrometeorology of the MRB. A climatology of adverse wintertime weather events is first presented and demonstrates that blowing snow events are rare within the forested sections of the MRB but become more frequent in the northern parts of the basin covered by Arctic tundra. It is these areas which experience the largest impacts of blowing snow transport and sublimation due to large-scale processes. To further assess the mesoscale and microscale effects of blowing snow to the northern regions of the MRB, the development of a bulk blowing snow model is then described. The single- and double-moment versions of the PIEKTUK blowing snow model are shown to produce equivalent results as a previous spectral version of the numerical model while operating about 100 times faster. The application of the double-moment PIEKTUK model (PIEKTUK-D) to a Canadian Arctic tundra site near the northern tip of the MRB reveals that blowing snow sublimation depletes ≈3 mm snow water equivalent (swe) from the snowpack over a period of 210 days during the winter of 1996/1997 at Trail Valley Creek, Northwest Territories. Various assumptions on the state of the background thermodynamic profiles and their evolution during blowing snow, however, can yield significantly higher (>300%) rates of sublimation over the same period. PIEKTUK-D is then coupled to the Mesoscale Compressible Community (MC2) model for an interactive simulation of a ground blizzard at Trail Valley Creek. This coupled mesoscale simulation reveals that moistening and cooling of near-surface air associated with blowing snow sublimation is observed but mitigated in part by advective and entrainment processes. Combined, blowing snow sublimation and mass divergence are then shown to rem