Abstract

A major challenge in groundwater and petroleum modeling is how to represent the diagenesis effect "sensu lato" and its associated features (e.g., karst cavities, porosity and permeability characteristics). Both geometrical descriptions and genesis are necessary for modeling diagenetic features, because they allow the restitution of phenomenon in subsurface settings where they cannot be fully observed.

This study focuses on flank-margin caves developed on carbonate islands (as in the case of the Bahamian islands). Over the last twenty years numerous measurements and observations have been carried out on Quaternary carbonate islands. These, together with reasonable hypotheses of genesis (fresh-water lens dissolution), make island karst an appropriate subject for developing concepts and modeling methods which can be then applied to other diagenetic features, such as reflux dolomitization and hydrothermal fluid effects.

The speleogenesis of flank-margin caves was modeled following deterministic and stochastic methods to achieve a description of dissolution genesis. A fresh-water lens surface was built based on the topography of Long Island, Bahamas, and was then used to simulate virtual development of flank-margin caves at several formation phases, including uncertainty coefficients applied to both fresh-water-lens location and stability time by using 3D probability cubes. The cave simulations were compared with over sixty Bahamian cave surveys, in order to substantiate our modeling processes and results.

Finally, the model was exposed to a cave collapse situation. The likelihood of preservation and the extent of the area affected were measured to provide an evaluation of the spatial distribution of associated petrophysical characteristics (e.g., porosity and permeability).

The results reveal close resemblance and coherence between modeled and field data (including collapse features). Therefore the optimum method for restituting diagenetic features is by stochastic modeling actively driven by process concepts and most up-to-date methods.