A parameterization that describes in-cloud oxidation of S(IV) by hydrogen peroxide and ozone, has been developed for use in large scale models. This parameterization, which is based on the reaction rate equations and basic cloud characteristics, is an explicit function of the concentration of ambient chemical species and some gross cloud parameters. Comparisons of the parameterization scheme with a well-established three-dimensional cloud chemistry model, and also with the cloud chemistry module of a regional model have been used to formulate and test this parameterization scheme. Results show that the parameterization agrees with the 3-D chemistry model very well and that the parameterization holds considerable potential for application in large-scale models. / Preliminary application in a regional climate model confirms that the parameterization is able to improve the agreement of the mass budget and spectrum distribution of sulphate aerosol with observations.

Bistatic Doppler radar networks have become in the last five years a viable and inexpensive alternative to multiple-Doppler networks. Operational experience with a bistatic network at McGill University showed many cases in which data quality seemed heavily affected. Study of those situations suggested sidelobe contamination from the transmitter antenna pattern to be the principal cause. / To confirm these findings a sidelobe simulation model (SISI model) was constructed. Comparison between simulations and actual data showed a good reproduction of the observed effect. It is also shown that this effect may have damaging consequences in Doppler fields in both convective and stratiform precipitation events. An index of contamination that can be obtained either with the SISI model or directly using the reflectivity bistatic data is introduced to detect areas of low quality data. Recommendations for the effective use of bistatic data are presented. / These findings are taken into account when the optimization of the layout of a bistatic network is analyzed. Sidelobe contamination was found to be a serious problem irrespective of the receiver's location. More than one passive receiver increases the extent of the dual Doppler area but unfortunately does not significantly reduce the problem of sidelobe contamination within a predetermined area. A rule-of-thumb for the deployment of a bistatic network is presented. Some suggestions for improvements of the network are given.

The purpose of this research is to further our understanding of the hydrostatic adjustment process due to diabatic heating. This is achieved by the use of an atmospheric model based on Euler's non-hydrostatic, fully elastic set of equations resolved by the semi-Lagrangian, semi-implicit marching scheme. Such a model thus allows a tri-dimensional, numerical approach of a problem which could, until now, only be addressed within a simplified context (one-dimensional and analytical; bi-dimensional and numerical). / The hydrostatic adjustment is the mechanism by which the atmosphere tends to reach a new hydrostatic balance when this one has been upset. This can be the case through a diabatic forcing, such as the sudden latent heat release during the formation of a thunderstorm cell. This is actually the kind of forcing we chose for this study. The problem of the hydrostatic adjustment is tackled through two different contexts: an idealized and a real case. Both are simulated by three different model's versions: fully compressible (PE), quasi-hydrostatic (QH) and quasi-anelastic (QA). These versions differ from each other only by the formulation of the diabatic heating terms. / The idealized case makes use of a prescribed heating source. A high resolution, in time and space, allows to show (PE version) how the mass, pressure and energy redistribution process is accomplished by the elastic, Lamb and gravity disturbances. The QH version, which consists in a spreading of the heating over the whole column above the source, inhibits vertically propagating elastic perturbations. The QA version, which considers the omission of the diabatic term in the pressure prognostic equation, shows that only the gravity mode is maintained. / The real case simulates a highly convective summer situation. The choice of a long time step brings out only the gravity mode; the hydrostatic adjustment by the fast modes (elastic and Lamb) is taking place quasi instantaneously. Thus this large-scale case doesn't exhibit significant differences between the PE, QH and QA versions.

The evolution of a long-lived mesovortex and its transformation into a tropical storm are studied by a three dimensional mesoscale model. The major aspects of the work are: (1) to demonstrate the mesoscale predictability of a long-lived mesoscale convective system (MCS) from a mid-level vortex over land to a tropical storm over ocean, (2) to understand the processes involved in the transformation of a mid-level continental vortex into a tropical storm, and (3) to perform sensitivity experiments to assess the impact of model physics on an idealized simulation. / The major conclusions are: The 90-h real-data simulation shows that the mesoscale model reproduces very well much of the meso-beta-structures and the evolution of the long-lived MCS. These include the development and dissipation of the continental mesovortex, the initiation of a new MCS both in time and in space, the genesis of a surface mesolow over the warm Gulf Stream water, the track and the deepening of the surface cyclone into a "tropical storm", the maintenance of a mid-level mesovortex system, and the propagation of a large-scale cold front with respect to the surface cyclone. / The simulation also shows that the mid-level mesovortex provides persistent convergence at its southern periphery for the continued convective development, whereas the convectively enhanced low-level flow increases surface energy fluxes over the warm water causing further conditional instability. Such feedback processes lead to the rapid deepening of the "tropical storm". / An idealized simulation was performed to eliminate the possible influence of the cold frontal system on the genesis of the tropical storm. The initial conditions resemble the basic structure of the subtropical high in the real-data case, but without the embedded frontal system. The simulation reproduces almost all of the essential features in the real-data simulation. In particular, the eye-like warm core structure of the tropical storm is well simulated. By decomposing the vertical relative vorticity into the curvature vorticity and the shear vorticity, it is shown that the amplification of a low-level vortex after 36 h arises mainly from the increase of curvature vorticity. The quasi-Lagrangian theta budget calculation shows that the descending motion in the center of the surface cyclone contributes to the formation of the warm core at 800 hPa. / The results of sensitivity experiments demonstrate the impact of model physics on the idealized simulation. Using the Betts-Miller scheme, we showed that the model fails to simulate the multiple-episodes of convective activity which are present in observational studies of tropical cyclogenesis and in the simulation using the Kain-Fritsch scheme. We further found that the surface latent heat fluxes represent a dominant factor in the production of CAPE to maintain the persistent convection. Finally, we discussed the cumulative cooling effect by long-wave radiation on the destabilization of the environment of the storm.

This thesis explores the relationships between multifractal measures, multiplicative cascades and correlations. A review of fractal geometry, multifractal formalism and multiplicative cascades is offered. The importance of the Legendre transformation in multifractal formalism is highlighted, especially for multifractal spectrums which are not convex or twice differentiable. By reconsidering the scaling assumption $N sb{p}( alpha$) d$ alpha$ = $ sigma( alpha) rho sp{-f( alpha)} d alpha$, we show that subsets of a measure may offer a different multifractal spectrum. A gradation of self-similarity and scaling vis-a-vis canonicity is offered. Localized and generalized correlations are introduced, and two-point correlations are revisited for multiplicative cascades. A three-point correlation function is presented and discussed. The presence of an integral scale is shown to produce a more involved correlation scaling behaviour.

Entrainment, extreme inhomogeneous mixing, in the presence of wind shear, and their effect on cloud droplet spectra are investigated. A dynamical model in conjunction with a microphysical model designed to predict evolution of cloud droplet spectra, is employed to perform a two-dimensional simulation of a small nonprecipitating cumulus cloud in the presence of wind shear. / Results show that vortex circulations and penetrative downdrafts are responsible for entrainment of clear air into the cloud structure. Entrainment and mixing are more severe on the downshear side of the cloud leading to a more fragmented structure and often to total dissipation of cloudy air rather than partial dilution as is the case on the upshear side. Mixing followed by uplifting leads to fresh activation of cloud droplets and results in multimodal spectra. In areas where mixing has occurred, the spectra exhibit smaller average radius and larger standard deviation.

Data collected from zenith pointing radar is used to study the range dependence of some meteorological radar errors associated with different sampling methods between radar and gauges in stratiform precipitation. Errors due to the vertical variation of reflectivity such as those related with the bright band or with snow growth cause a much larger bias in radar estimates than those due to beam filling or gradients. The maximum useful range varies with the bright band height and the elevation angle program used, a CAPPI giving superior results especially for snow. The sudden changes in bright band height over short distances and the large scatter of its thickness limit the accuracy of current corrections for the vertical variation of reflectivity based on scanning radar data. The possibility of using a zenith pointing radar to obtain this correction is discussed.

A thorough literature review of cloud algorithm validation strategies is presented, with particular emphasis on the problems of detecting cirrus clouds. To further advance our cirrus detection capabilities, a new cloud detection technique is proposed for the Multi-angle Imaging Spectro-Radiometer (MISR), which is scheduled to be on the first platform of the Earth Observing System. Radiative Transfer simulations have been used to develop a Band-Differenced Angular Signature technique. This new technique takes the difference between two solar spectral reflectances as a function of view angle. The resulting angular signature is used to discriminate between high- and low-level clouds and surface reflectance anomalies. This technique, coupled with a Predetermined Clear Sky Threshold developed in this study for MISR, can detect cirrus clouds of visible optical thickness $>$0.5 without any a priori knowledge of atmospheric conditions. Other techniques that can be used with MISR (i.e. stereo) are also discussed.

Hurricane Hazel struck North America on 15-16 October 1954, leaving a pattern of heavy rainfall and flooding in its wake. A complete analysis of the synoptic-scale conditions associated with the transformation of Hazel from its tropical phase into an extratropical cyclone was first undertaken to discern the dynamic and thermodynamic elements crucial to the intensification of this storm. / An analogue search was then conducted for Hazel using linear correlations of anomaly sea level pressure and 1000-500 hPa thickness. Three cases were found in 1985, 1995 and 1999. A comparison of these analogues to Hazel yielded the conclusion that Hazel is a unique event in recent meteorological history, as none of the analogues produces the extreme precipitation values in Hazel. The lack of significant hurricane circulations in all of the analogues is the important difference, as Hazel provides important moisture and latent heating that are absent in the analogues. / Finally a mesoscale modeling study was carried out to test the sensitivity of Hazel to improved surface vortex structure and increased horizontal resolution. Specification of the vortex led to a dramatic improvement in the simulation results, as precipitation and track closely mimicked the observed values. Enhancing the horizontal resolution to 12 km did not improve upon the 36 km specified vortex simulation. The movement of the storm slowed considerably as the development of an upper tropospheric cutoff circulation was diminished in the 12 km run. The parameterizations governing the interaction between the diabatic outflow from Hazel and dynamics of the midlatitude trough are poorly modeled in this situation, and lead to the severe time lag in the path of Hazel.

Ensemble of seasonal forecasts done with four atmospheric models over 26 winters and springs (1969-1994) are used to investigate the Northern Hemispheric model responses to SST anomalies and to compare them with observations. The response, for winter (DJF) and spring (MAM), is obtained as the regression of the 500hPa geopotential heights against the main modes of sea surface temperature (SST) variability over two different regions of the world oceans. The regression against the main mode of the equatorial Pacific Ocean SST, takes the expected form of the Pacific-North-America (PNA) pattern both in models and observations. The response to the North Atlantic SST is an AO/NAO-like signal for the observations and also, to varying degrees, for the models. The similarity of the regression to the third mode of equatorial Pacific SST and to the first mode of North Atlantic SST during the spring season is discussed, along with the nonlinear features of the ENSO influence.