Deep-sea sponge grounds provide structurally complex habitat for fish and invertebrates and enhance local biodiversity. They are also vulnerable to bottom-contact fisheries and prime candidates for Vulnerable Marine Ecosystem designation and related conservation action. This study uses species distribution modeling, based on presence and absence observations of Geodia spp. and sponge grounds derived from research trawl catches, as well as spatially continuous data on the physical and biological ocean environment derived from satellite data and oceanographic models, to model the distribution of Geodia sponges and sponge grounds in the Northwest Atlantic. Most models produce excellent fits with validation data although fits are reduced when models are extrapolated to new areas, especially when oceanographic regimes differ between areas. Depth and minimum bottom salinity were important predictors in most models, and a Geodia spp. minimum bottom salinity tolerance threshold in the 34.3-34.8 psu range was hypothesized on the basis of model structure. The models indicated two currently unsampled regions within the study area, the deeper parts of Baffin Bay and the Newfoundland and Labrador slopes, where future sponge grounds are most likely to be found.