Biomass samples from marine scientific surveys are commonly used to investigate spatial and temporal variations in stock abundances. Biomass records are often characterized by a high proportion of zeros on the one hand, and occasional large catches on the other. These features induce a modeling challenge when trying to understand the state of populations and their ecological associations with one another and with habitat. We develop a hierarchical Bayesian model to represent the spatial structure of biomass and analyze the spatial distribution and habitat associations of three species of macro-invertebrates sampled in the southern Gulf of St. Lawrence (Canada). A zero-inflated distribution based on a compound Poisson with Gamma marks is used for the observation layer, and a linear model with spatial correlated errors accounts for the role of habitat variables (temperature, depth and sediment type) in the process layer. Maps of quantities of interest (e.g. probability of presence, quantity of biomass) are produced, taking into account the uncertainty of the estimated parameters and observation errors. This hierarchical Bayesian modeling approach provides a useful tool for spatial management of human activities that may affect living resources that may affect living resources, such as marine protected areas. Crown Copyright (C) 2013 Published by Elsevier B.V. All rights reserved.