Convection in the Earth's mantle is a complicated phenomenon that causes various tectonic activities and affects mantle evolution on geologic time scales (billions of years). It is a subject as yet not fully understood. The early success of the high Rayleigh number constant viscosity theory was later tempered by the absence of plate motion when the viscosity is more realistically strongly temperature dependent. A similar problem arises if the equally strong pressure dependence of viscosity is considered, since the classical isothermal core convection theory would then imply a strongly variable mantle viscosity, which is inconsistent with results from postglacial rebound studies. We consider a mathematical model for Rayleigh-Bénard convection in a basally heated layer of a fluid whose viscosity depends strongly on both temperature and pressure, defined in an Arrhenius form. The model is solved numerically for extremely large viscosity variations across a unit aspect ratio cell, and steady solutions are obtained. To improve the efficiency of numerical computation, we introduce a modified viscosity law with a low temperature cut-off. We demonstrate that this simplification results in markedly improved numerical convergence without compromising accuracy. Continued numerical experiments suggest that narrow cells are preferred at extreme viscosity contrasts. We are then able to determine the asymptotic structure of the solution, and it agrees well with the numerical results. Beneath a stagnant lid, there is a vigorous convection in the upper part of the cell, and a more sluggish, higher viscosity flow in the lower part of the cell. We then offer some comments on the meaning and interpretation of these results for planetary mantle convection.

The Olavi Mountain Land is a 10 000 km2 mineral province located at the eastern extremity of the exposed Northern Platform of the Damara Pan African orogenic belt. The Olavi Mountain Land is tbe most important mineral province on the Northern Platform. Exploitation of tbe Cu-Pb-Zn-V province has been on-going since the possession of the territory by the German colonial authority in 1890. Production has been mostly from four mines which in order of importance are Tsumeb, Kombat, Berg Aukas and Abeoab. A second mineral province on the Northern Platform located in the west is centred on Sesfontein where as yet only insignificant mineralization has been noted. Besides these localities, the Northern Platform is conspicuously devoid of notable mineralization. The aim of this thesis has been to document tbe Berg Aukas deposit, an important end-member type of mineralization in the Otavi Mountain Land. The basic premise bas been to show tbat the derivation and localization of the mineralization is a consequence of two broad controls which can be simply summarised as features of the basement and of the carbonate sequences. The geodynamic evolution of the Damara Belt commenced with intra-continental rifting approximately 900 Ma ago. Rift grabens trending north-east were filled by the Nosib Group which comprises mostly clastic lithologies but also some volcanics. The earliest and largest rift is referred to as the Northern Rift. Separation of the Congo, Kalahari, and proto-South American cratons resulted in rifting and rapid downwarping so that an encroaching sea and an Olavi Group carbonate shelf developed along the northern margin of the Northern Rift. Significantly, the carbonates only covered the Northern Rift in the area of the Otavi Mountain Land where a basinal dome, referred to as the Grootfontein Basement High, marked the basin edge. In the west, the carbonates covered the less important Sesfonfein Rift, and it is only in these two areas where Nosib sequences underlie the carbonate platform. Carbonate sedimentation was interrupted by a major period of crustal readjustment and the deposition of an extensive mixtite throughout the geosynclinal Swakop Trough and Northern Platform. This is referred to as the Chuos Formation and subdivides the Olavi Group into a lower Abenab and an upper Tsumeb Subgroup. Reversal of spreading led to plate collision and subduction of tbe Kalahari craton beneath the Congo craton. It was accompanied by orogenesis which resulted in F1 folding of the Northern Platform into a series of north-easterly trending intermontane basins into which a molasse sequence known as the Mulden Group was unconformably deposited. Following this major north-south deformation mild east-west compression initiated F2 folding and the formation of doubly plunging synclines. The Berg Aukas Syncline represents a primary depositional basin which was subsequently folded. The original basin was formed by late Nosib rifting wben spreading caused the Swakop geosynclinal Trough to form. Carbonates of the basal Berg Aukas Formation were deposited in a lagoonal setting typified by reef and fore-reef facies witb peri-platform conditions. Rapid subsidence caused these sediments to be overlain by deep water carbonates of the Gauss Formation. Two styles of mineralization known as the Tsumeb-type and Berg Aukas-type are stratigraphically, isotopically, and mineralogically distinct. The Tsumeb-type is a cupriferous variety of discordant bodies confined to the upper sequences beneath the Mulden unconformity. The Berg Aukas-type is a Zn-Pb variety confined to tbe basal unconformity. The Berg Aukas deposit comprises three ore bodies known as the Northern Ore Horizon, the Central Ore Body, and the Hanging Wall Ore Body. Sphalerite and galena constitute the bypogene ore. Willemite, smithsonite, cerussite, and descloizite are important supergene ores. A review of genetic models concludes that a magmatic origin initially proposed for tbe Tsumeb deposit is entirely rejected and a basin dewatering model in line with Mississippi Valley-type deposits is proposed. The syntectonic nature of mineralization at Berg Aukas and elsewhere in the Otavi Mountain Land indicates that orogenesis encouraged dewatering and leaching of metals from a broad mineralizing front along the margin of the Swakop Trough. These were transported by acidic saline brines which migrated along the clastic aquifers and structural conduits provided by the Northern Rift. Fluid inclusion studies indicate that the hydrothermal fluids at Berg Aukas were very saline (23% TDS) and were transported at temperatures ranging between 92° to 210°C. Hydrothermal fluids which mineralized Berg AukaS-type deposits migrated along the basal unconformity towards the basement high and were responsible for hydrothermally altering the basement granites and gabbros and the Nosib clastic rocks. Tsumeb-type deposits resulted by migration of fluids through the carbonate pile and along north-easterly trending basement geofractures. As a consequence of variation in transport, the Berg Aukas-type and Tsumeb-type fluids leached different sources and therefore derived mineralogically and isotopically seperable characteristics. The localization of the Berg Aukas ores was controlled by the carbonate stratigraphy and structure. Hydrothermal karsting and ore deposition took place on the contact between Massive Grey and Light Grey Dolostones which represents a permeability contrast. The movement of the hydrothermal fluids was controlled by north-south trending vertical fractures caused by F2 folding which resulted in a peric1inal structure. Hydrothermal karsting was accompanied by ca1citic, dolomitic and silicic alteration. The heated acidic fluids initiated solution collapse and a variety of breccia types. Supergene processes resulted in oxidation and upgrading of the ore. Vanadium derived indirectly from gabbros in the basement complex were transported as calcium metavanadate complexes and deposited on contact with the oxidizing base metal sulphides.

Stratiform and stratabound ore deposits commonly show a direct relationship with rifts. This association is studied by developing a geodynamic model of mantle processes and crustal responses. The geodynamics of the earth can be modelled by the process of mantle advection, which involves the episodic generation and segregation of low density mantle diapirs and their rise and subsequent interaction with the crust. The theory of mantle advection explains the genetic association between rifting, magmatism, basin development and subsequent orogeny and metamorphism. Global evolution has passed through a number of major stages of non-uniformitarian development in which each cycle was characterized by fairly uniform behaviour terminated by intense geodynamic upheaval. The relationship between geological evolution and mantle advection is examined by reviewing the major characteristics of each of the cycles, which correspond to the Archean, Early Proterozoic, Mid Proterozoic, Late Proterozoic-Palaeo2oic, and Mesozoic - Cainozoic eras. Although mentle advection has controlled crustal processes throughout time, the decrease in the thermal energy of the earth has caused >the major evolutionary changes in response to thickening and a greater rigidity of the sialic crust. Rifts are penetrative taphrogenic faults in the earths crust which act as major conduits for the transfer of magmas, from the mantle and lower crustal levels, to the upper crust and the surface. Rifts are also permeable zones for the migration of metalliferous brines, generated by magmatic differentiation. These metalliferous brines would either be exhaled at surface to form stratiform volcanogenic and volcanosedimentary ore deposits , or would interact with preferential host horizons to form stratabound ore deposits . The associat ion between rifting and stratiform and stratabound ore deposits is illustrated by examining :he tectonic setting, and st ratigraphic relationships of typical ore deposit types .

The Coast Mountains in British Columbia contain over 10,000 km2 of glacial ice. While these glaciers have lost significant mass since the Little Ice Age (LIA; around 300 years before present), the melting rate has significantly increased over the past decade, likely due to the effects of climate change. The purpose of this study was to develop an approach to quantifying the isostatic response to LIA glacier change and investigate how it can further our understanding of the Earth’s rheology through GIA modelling. The Coast Mountains in southwestern British Columbia were chosen due to their significant ice mass loss since the LIA, their location in a tectonically active region, which includes a volcanic arc, and the presence of information of vertical land motion. The GIA models in this study use a wide range of Earth rheological parameters that are then constrained through comparison to observations of vertical land motion in the region. The study used available Global Navigation Satellite System (GNSS) vertical velocity data as the observable from seven GNSS sites in southwestern BC, using a combination of Western Canada Deformation Array (WCDA) and British Columbia Active Control System (BCACS) GNSS stations. Raw data were analyzed using the GIPSY 6.4 software following the Precise Point Positioning processing strategy. Two ice load histories were developed based on gridded estimates of present-day ice thicknesses in the region in order to simulate the change in the surface loading as the glacial ice mass fluctuates over time. Ice Load A used a simple uniform thickness change profile over 3 time-steps based on extrapolated modern melt rates. Ice Load B is more complex and utilized a published profile of glacier change through time basing the magnitude of volume changes on the volume-area scaling relationship with a range of coefficient values. This allowed for a range of ice change magnitudes to be tested. The Earth models used were spherically symmetric Preliminary Reference Earth Models (PREM). Their viscosity structure is based on VM5a for the transition zone and lower mantle, but with variable lithospheric thickness and asthenospheric viscosity. The goodness of fit for the modeled velocities were compared to the observed velocities using a normalized RMS (NRMS) statistic. Ice Load A models had a best fitting lithospheric thickness of 50 km and an asthenospheric viscosity of 2×1019 Pa s. For all variations of Ice Load B, the best fitting model parameters had lithospheric thicknesses ranging from 45 km to 55 km and asthenospheric viscosities between 6×1018 Pa s and 3×1019 Pa s. Corrected GNSS vertical velocity observations were tested to check the effects of interseismic vertical signal and assumed residual GIA from the Cordilleran Ice Sheet. However, the corrections did not improve the NRMS fit. Overall, the asthenospheric viscosity results from this study overlap with all the ranges found in the previous studies while lithospheric thicknesses agree with some past studies. The results of this study generally align with previous work and the current understanding of the Coast Mountains region and can inform a future round of sea-level projections for the region as ice mass loss continues in the Coast Mountains. This study serves to further refine constraints on Earth rheology and can be used to guide future work on GIA in the region. / Graduate

The research presented in this thesis seeks to address meaningful geodynamic problems related to the viscosity structure of the Earth’s interior. Isostatic adjustment is a process which is dependent upon the mechanical properties of the lithosphere and mantle. By performing computational simulations of the isostatic response for various surface-loading scenarios and numerous viscosity structures, insight can be gained into the mechanical structure of the Earth and geodynamic processes related to that structure. The modelled isostatic signal for a given set of Earth model parameters can be compared to real-world observational data in order to identify valid Earth model configurations. In Chapter 2, the “Transition Zone Water Filter” theory is tested by modelling the geophysical effects of a low-viscosity melt-rich layer atop the 410 km discontinuity. The thickness and viscosity of this layer, and the surrounding mantle, is constrained using observations of relative sea level and the geodetic J ̇_2 parameter, as well as multiple ice-loading scenarios by which the isostatic adjustment process is driven. The relative sea level data, being most sensitive to the upper mantle and the theorized melt-rich layer it contained, constrain layer properties more effectively than the J ̇_2 observation, which is strongly dependent on the lower mantle. Constraints on the viscosity of the melt-rich layer vary according to thickness, with thicker layers requiring stiffer viscosities to satisfy observations. For instance, a 20 km thick layer would require a viscosity of 10^17 Pas or greater, but any of the considered viscosities could be possible for a 1 km thick layer. Similarly, a broad range of upper mantle viscosities are possible, but they must be balanced by variations in the lower mantle. However, J ̇_2 results show a strong preference for a high-viscosity lower mantle (≥10^22 Pas). For every evaluated Earth model parameter, there is evidence of ice-model sensitivity in the inversion results. Although the results of this study demonstrate that observables related to glacial isostatic adjustment can provide constraints on the properties of this theorized melt-rich layer, the confounding effect of parameter trade-off prevents a more definitive test of this model of mantle geodynamics. The purpose of the study presented in Chapter 3 is to analyze the nature of solid-Earth deformation beneath the Lower Mississippi River, most crucially in the Mississippi Delta region where subsidence is an ongoing and costly problem. The study uses the displacement of the long profile of the Lower Mississippi River over the last 80 kyr to constrain isostatic deformation and determine constraints on the mechanical structure of both the mantle and lithosphere. Deformation recorded in the northern portion of the long profile is dominated by the effect of glacial isostatic adjustment, whereas the southern portion is governed by sediment isostatic adjustment. However, the southern portion is also potentially affected by past fault displacement, and to account for this the observational data are corrected using two distinct faulting scenarios. Displacement of the long profile is modelled using either an entirely elastic lithosphere or a lithosphere with internal viscoelastic structure, the latter of which is derived from two end-member geothermal profiles. Between the elastic and viscous lithosphere models, the viscous models are better able to replicate the observational data for each faulting scenario – both of which prefer a viscous lithosphere corresponding to the warmer geotherm. The chosen faulting scenario exerts no control over the optimal mantle model configuration, however the optimal mantle for the viscous lithosphere models is much stiffer than was determined for their elastic counterparts, reflecting significant parameter trade-off between mantle and lithosphere mechanical structure. These study results demonstrate the utility of the long profile displacement data set for constraining Earth viscosity structure, as well as the importance of considering more-complex models of lithosphere mechanical structure when addressing surface-loading problems similar to those encountered in the Mississippi Delta region.

No description available.

Les processus de déformation contrôlant l'amincissement de la lithosphère continentale sont encoremal contraints. Nos connaissances sont principalement basées sur la modélisation thermomécaniqued'extension à l'échelle de la lithosphere—utilisant des lois rhéologiques derivées expérimentalement,l'imagerie géophysique et l'analyse de xénolithes provenant de rift continentaux actifs à ce jour, tels quele Rift Est-Africain. L'originalité de ce travail reside dans l'étude des deux plus grands massifs depéridotites sous-continentales ayant enregistrées des conditions primaires du facies à diamant: lesmassifs de Beni Bousera au nord du Maroc et de Ronda au sud de l'Espagne, respectivement. Lesstructures et la zonation petrologique et métamorphique —impliquant une évolution polybarique etpolythermique— préservéees dans ces massifs offrent une opportunité unique pour étudier l'évolutionthermo-mécanique du manteau sous-continental dans un contexte extensif.Dans ce travail, nous avons étudié les mécanismes de déformation des péridotites et despyroxénites afin de contraindre les modes d'exhumation du manteau lithosphérique sous-continental,depuis des conditions du facies des lherzolites à grenat, jusqu'au facies à spinelle et enfin à plagioclase.Nous avons combiné la cartographie des faciès tectono-métamorphiques et des structures ductiles dedéformation, l'analyse des microstructures, la mesure d'orientations préférentielles de réseau (OPR), etla géothermobarométrie conventionelle couplée à la modélisation thermodynamique (PerpleX) afin decontraindre les conditions de pression et température de la déformation. Nous avons montré quel'exhumation précoce du facies à grenat au facies à spinelle était accomodée par une faille transtensiveaffectant le manteau lithosphérique. Dans ce contexte, la zonation tectono-métamorphique et legradient thermique important (ca. 100ºC/km) préservés à Beni Bousera résultent de la juxtapositionmécanique de domaines lithosphériques initialement équilibrés à différentes pressions et températures,fossilisée à une profondeur de ca. 60 km durant l'Oligocène supérieur (ca. 25 Ma). L'exhumation finaledu facies de lherzolite à spinelle au facies à plagioclase et l'emplacement final dans la croûte, mieuxenregistrés dans Ronda, se sont produits par inversion et plissement de la section lithosphériquefortement amincie dans un contexte arrière-arc, probablement lors du retrait vers le sud de lalithosphère subduite et la collision de l'arc avec les paléo-marges maghrébines au Miocène inférieur(21-23 Ma). / The mantle deformation processes that control the thinning and break-up of continentallithosphere remain poorly understood. Our knowledge is restricted to either lithospheric scalethermo-mechanical models —that use experimentally derived flow laws—, geophysicalimaging and/or rare xenoliths from active continental rifts, such as the East African Rift System.The originality of this work relies on the study of the two largest outcrops of diamond faciessubcontinental lithospheric mantle in the world: the Beni Bousera and Ronda peridotite massifsin N Morocco and S Spain, respectively. The structures and petrologic and metamorphic zoningpreserved in these massifs —implying a polybaric and polythermal evolution— provide aunique opportunity to investigate the thermo-mechanical evolution of thick subcontinentallithospheric mantle in extensional settings.In this thesis we studied the deformation mechanisms in both peridotites andpyroxenites to constrain the modes of exhumation of subcontinental lithospheric mantle fromgarnet-, to spinel-, and finally, to plagioclase lherzolite facies conditions. We combined fieldmapping of tectono-metamorphic domains and structural mapping of ductile structures,microstructural analysis, crystal preferred orientations (CPO) measurements and conventionalthermobarometric calculations and thermodynamic modeling (Perple_X) to unravel the pressureand temperature conditions of deformation. We showed that exhumation from garnet- to spinellherzolite facies conditions was accommodated by fast shearing —in thermal disequilibrium—along a lithospheric scale transtensional shear zone. In this context, the petrological zoning andthe large temperature gradient (ca. 100ºC/km) preserved in the Beni Bousera massif representthe mechanical juxtaposition of progressively deeper and hotter lithospheric levels at depths ofca. 60 km in the latest Oligocene (ca. 25 Ma). Final exhumation from spinel- to plagioclasefacies lherzolite and emplacement into the crust is best recorded in the Ronda massif where itoccurred by inversion and lithospheric scale folding of the highly attenuated continentallithosphere in a back-arc region, probably in relation with southward slab rollback andsubsequent collision with the palaeo-Maghrebien passive margin in the early Miocene (21-23Ma).

Caractériser les réseaux de fractures dans les réservoirs de subsurface est un challenge majeur. En effet, les fractures contrôlant en grande partie la perméabilité dans les réservoirs, il est nécessaire aujourd’hui d’en prédire l’organisation. Or, la prédiction des fractures en subsurface est difficilement contraignable du fait (1) de données limitées en termes de résolution (sismique), ou de continuité spatiale (imagerie de puits, carottes), et (2) d’un manque de compréhension des facteurs contrôlant la fracturation. Les analogues de terrain sont une bonne alternative aux données de subsurface, permettant d’accéder à la complexité du réseau de fractures, et aux hétérogénéités des carbonates en 3D. Pour chaque affleurement sélectionné, une étude pluridisciplinaire est réalisée, combinant géologie structurale, sédimentologie, diagenèse et pétrophysique. Ainsi, les conditions, le « timing » et les facteurs contrôlant la fracturation peuvent être précisés à travers l’histoire de l’encaissant carbonaté. La fracturation diffuse dans les carbonates est régie par (1) le faciès de dépôt qui contrôle les hétérogénéités mécaniques et la susceptibilité de la roche à la diagenèse, (2) la diagenèse que contrôle l’acquisition précoce du caractère cassant et l’inhibition de la fracturation tardive, et (3) l’enfouissement, responsable des variations de contraintes verticales et de l’apparition des fractures et stylolites. / Characterizing fracture networks in Naturally Fractured Reservoirs (NFR) is a major challenge for hydrocarbon exploration and production. Because fractures control most of the permeability through the reservoir, there is a necessity to predict their organization. However, at present-day, fracture prediction in the subsurface is poorly constrained because: (1) of limited data from seismic, well imaging and cores, and (2) of lack of understanding of controls on fracture occurrence. Outcrop analogues are good alternative to subsurface data for a full 3D fracture characterization. It enables to access the spatial complexity of fracture patterns, and to the 3D heterogeneities of carbonates. In each selected outcrop, a multidisciplinary approach has been undertaken, There, the conditions, timing and controlling factors on fracturing can be precised through the geodynamic history of the host carbonate. Diffuse fracturing in carbonates is governed by (1) the depositional facies which controls mechanical heterogeneities and rock diagenetic susceptibility, (2) the diagenesis which controls the early rock embrittlement and the late fracture inhibition, and (3) the burial, responsible for vertical stress variation and fracture–stylolite occurrence.

La tectonique actuelle et récente (~40 Ma) de la région Egée-Anatolie en fait un laboratoire naturel privilégié pour l'étude de la déformation continentale. L'analyse de la structure détaillée et de la déformation du manteau de cette zone de collision continentale immature constitue une étape indispensable pour comprendre les relations entre la cinématique de surface et la dynamique du manteau. La structure du manteau supérieur a été étudiée à l'échelle de la région en réalisant un modèle 3-D de la vitesse des ondes S par tomographie télésismique en ondes de surface. Les résolutions latérale (~100-200 km) et verticale (~50 km) des images obtenues, grâce à l'utilisation d'un réseau sismologique large bande composé de ~150 stations (permanentes et temporaires (expérience SIMBAAD), apporte de nouvelles contraintes sur la structure du manteau de la Grèce continentale à l'Anatolie centrale. Cette tomographie révèle notamment des corrélations verticales claires entre les géométries des déchirures affectant le slab Hellénique et celles des zones de cisaillement trans-tensives qui accommodent le mouvement rapide du bloc Egéen vers le SW. Le slab Chypriote est clairement imagé dans le modèle 3-D comme plissé et déchiré en plusieurs segments sous l'Anatolie. L'analyse de l'anisotropie azimutale par méthode de réseau sur les ondes de Rayleigh a permis de proposer l'existence de deux couches anisotropes dans le manteau supérieur. L'étude de variations latérales de l'anisotropie sous la région suggère que la déformation des slabs et la cinématique de surface sont contrôlées par des flux mantelliques toroïdaux de différentes échelles. / The last ~40 M.y. tectonics of the Aegean-Anatolia region has shaped a natural laboratory ideal to study the continental deformation processes. The detailed analysis of the mantle structure and deformation of this non-mature continental collision zone constitutes an essential step to investigate the contribution of mantle dynamics to surface kinematics. The upper mantle structure is investigated over the entire region through a new 3-D S-wave velocity model from surface-wave tomography. The model reveals remarkable vertical correlations between geometries of the Hellenic slab tears and geometries of shear zones which accommodate the rapid SW movement of the Aegean bloc. The Cyprus slab is clearly identified in the 3-D model as fold and torn in tree segments beneath Anatolia. The observed azimuthal anisotropy from Rayleigh-wave array analysis suggests the existence of two anisotropic layers in the upper mantle. The lateral variations of anisotropy beneath the region are interpreted as the indication of toroidal mantle flows at different scales governing the slabs deformation and the surface kinematics.

The Icelandic mantle plume has had a profound influence on the development of the North Atlantic region over its 64 Myr existence. Long-wavelength free-air gravity anomalies and full waveform tomographic studies suggest that the planform of the plume is highly irregular, with up to five fingers of hot asthenosphere radiating away from Iceland beneath the lithospheric plates. Two of these fingers extend beneath the British Isles and southern Scandinavia, where departures from crustal isostatic equilibrium and anomalous uplift have been identified. In this study, the spatial extent of present-day dynamic support associated with the Icelandic plume is investigated using receiver function analysis. Teleseismic events recorded at nine temporary and 59 permanent broadband, three-component seismometer stations are used to calculate 3864 P-to-S crustal receiver functions. The amplitude and arrival time of particular converted phases are assessed, and H-k stacking is applied to estimate bulk crustal properties. Sub-selections of receiver functions are jointly inverted with Rayleigh wave dispersion data to obtain crustal VS profiles at each station. Both inverse- and guided forward modelling techniques are employed, as well as a Bayesian, trans-dimensional algorithm. Moho depths thus obtained are combined with seismic wide-angle and deep reflection data to produce a comprehensive crustal thickness map of northwestern Europe. Moho depth is found to decrease from southeast (37 km) to northwest (26 km) in the British Isles and from northeast (46 km) to southwest (29 km) in Scandinavia, and does not positively correlate with surface elevation. Using an empirical relationship, crustal shear wave velocity profiles are converted to density profiles. Isostatic balances are then used to estimate residual topography at each station, taking into account these novel constraints on crustal density. Areas of significant residual topography are found in the northwestern British Isles (1400 m), southwestern Scandinavia (464 m) and Denmark (620 m), with convective support from the Icelandic plume as its most likely source. Finally, the irregular planform of the Icelandic plume is proposed to be a manifestation of radial viscous fingering due to a Saffman-Taylor instability. This fluid dynamical phenomenon occurs when less viscous fluid is injected into a layer of more viscous fluid. By comparing the thermal and convective characteristics of the plume with experimental and theoretical results, it is shown that viscous fingering could well explain the present-day distribution of plume material.