The San Pablo Song Sparrow (Melospiza melodia samuelis) is one of three morphologically distinct Song Sparrow subspecies in tidal marshes of the San Francisco Bay estuary. These subspecies are rare, because as the human population has grown, diking and development have resulted in loss of 79% of the historic tidal marshes. Hundreds of projects have been proposed in the past decade to restore tidal marshes and benefit endemic populations. To evaluate the value of these restoration projects for Song Sparrows, we developed a population viability analysis (PVA) model to examine persistence of samuelis subspecies in relation to parcel size, connectivity, and catastrophe in San Pablo Bay. A total of 101 wetland parcels were identified from coverages of modern and historic tidal marshes. Parcels were grouped into eight fragments in the historical landscape and 10 in the present landscape. Fragments were defined as a group of parcels separated by >1 km, a distance that precluded regular interchange. Simulations indicated that the historic (circa 1850) samuelis population was three times larger than the modern population. However, only very high levels (>70% mortality) of catastrophe would threaten their persistence. Persistence of populations was sensitive to parcel size at a carrying capacity of <10 pairs, but connectivity of parcels was found to have little importance because habitats were dominated by a few large parcels. Our analysis indicates little risk of extinction of the samuelis subspecies with the current extent of tidal marshes, but the vulnerability of the small-est parcels suggests that restoration should create larger continuous tracts. Thus, PVA models may be useful tools for balancing the costs and benefits of restoring habitats for threatened tidal-marsh populations in wetland restoration planning.