Analyse des donnees gravimetriques en forage d'un gisement de sulfures massifs volcanogenes dans un contexte geologique complexe

A forward modeling and an inversion code have been developed to study the use of the borehole gravity method for exploration of volcanogenic massif sulphides (VMS) deposits in the Abitibi region of Quebec. Two problems are associated with the gravity method: acquiring data can be a long and costly method in the context where there is a limited quantity of boreholes and the separation of the response caused by the immediate or local geology and the response of deeper and farther formations called the regional. The principal objective of this master's project is to analyse those two major problems by modelling and inverting synthetic data. The specific objectives are the optimisation of the data acquisition settings and the regional-residual anomaly separation.;The forward modeling method is based on the prismatic method described by Li and Chouteau (1998). A stochastic approach developed by Shamsipour et al. (2010) is chosen for the inversion and was adapted for borehole data. A density model of a typical VMS ore deposit was designed based on a number of well-known mines in the region of Rouyn-Noranda, Val-d'Or and Matagami.;The data acquisition settings include the number of boreholes, their location and data collection sampling in the boreholes. Since the borehole gravity method is a costly geophysics method, it is best to know well the influence of the data acquisition settings to be able to optimise them. A minimum of three boreholes within appropriate distance from the target is required to locate any structure. When four boreholes situated at the detectability range of the deposit are used, it can be located with precision. In the scenario where the borehole gravity method is used to calculate the excess mass of a deposit and to define its structure, at least four boreholes should be used with one intercepting the deposit and fixed densities or gradient constraints must be applied. A 10m sampling interval is recommended. If the position of the deposit is known, a good compromise is to use a larger interval far from the deposit and use a 10m interval when the borehole is closer to the deposit.;The regional-residual anomaly separation is a very important aspect of the data interpretation. Up to now no technic is effective in performing an optimal separation. Three different methods are used in this project; these methods are the vertical gradient, a non-linear filter and a wavelet filter. Once the data has been treated, the inverted density model is compared to the initial model. Though the different methods do not calculate the same residual, the results are fairly similar. All the methods can position the deposit well, but the shape differs from the initial model. Also, the excess mass calculated are similar to each other, but they are a bit underestimated compared to the real excess mass. Non-linear and wavelet filtering were proven to be the best methods to calculate results closest to the actual model. The vertical gradient grossly underestimates the density contrasts when no borehole intercepts the deposit. Furthermore, when the method is applied to evaluate the deposit, the shape is not recovered and the excess mass is underestimated even when constraints are used.;Real borehole gravity data was acquired over and in the vicinity of the Virginia Gold's Coulon deposit (Quebec, Canada). A model was built using borehole electromagnetic data and geological data. This model is used to compare the inverted results. The three regional-residual anomaly separation methods were applied to the Bouguer anomaly of the Coulon data as well as the graphical method. Like for the synthetic data the residual calculated differed from method to method but the results resembled one another. The shapes of the structure calculated by the graphical, non-linear filtering and wavelet filtering methods were essentially the same. It was also observed that the excess masses calculated by graphical method and non-linear method were similar. Finally, the excess masses calculated by vertical gradient and wavelet filter were a bit lower.