In response to marine habitats destruction and living population depletion, marine spatial planning (MSP) proposes to regulate uses of the marine environment. Practically, MSP seeks an ocean zoning to meet both ecological and socio-economic objectives eventually aiming at a sustainable development. In particular, an identified conservation answer to mitigate marine biodiversity erosion consists in the development of a comprehensive network of marine protected areas (MPAs) accompanied with global benefits.To avoid ad-hoc and opaque conservation choices, systematic reserve selection procedure quickly became a worldwide research and operational stake. Mathematically, it can be understood as a resource allocation optimisation problem and modeled as a binary programming problem thanks to a minimum set formulation. Conservation science research field extensively studied the deterministic reserve site selection problem. Yet, data uncertainty can lead to deprecated reserve solutions and potential irreversible damages towards marine ecosystems, and/or useless constraints on marine stakeholders. Although a probabilistic approach was successfully proposed based on presence/absence data, this hypothesis is too restrictive and we would like to account for uncertainty when abundance data is available. Therefore, we propose a risk-averse approach to incorporate the parametric uncertainty expressed through a chance constraint and the associated risk-level parameter. We propose to approach this chance constraint framework using a sampling approximation. Indeed, we benefit from geostatistics theory and especially conditional simulations to generate a user-defined number of scenarios with an occurring probability. Such model gives a simple and efficient procedure to incorporate parametric uncertainty in the reserve site selection problem with probabilistic abundance data. A sensitivity analysis towards the risk-level parameter provides great insights with respect to the robustness of the reserve solution. It gives a simple way to properly account and represent parametric uncertainty in the reserve site selection problem. Numerical application is illustrated on the real case of Fernando de Noronha Brazilian archipelago in Tropical Atlantic.