To conserve or restore juvenile fish habitat, resource managers, restoration practitioners, and engineers need fine-scale information to understand what conditions they need to preserve or what specifications are needed to create new habitat. The objective of this study was to develop statistical models using a dataset of 18 coastal ponds in Southwest Florida, USA, to predict the abundance of juvenile common snook Centropomus undecimalis (hereafter snook) and presence of juvenile tarpon Megalops atlanticus, both of which are flagship species used for conservation and restoration of subtropical and tropical wetlands. Model predictors included water conditions and characteristics of vegetation, sediment, and geographic position. Modeling results indicated that juvenile snook used coastal ponds that were directly connected to tidal creeks. In contrast, juvenile tarpon were more likely to be present in coastal ponds that were separated from tidal creeks by dense mangrove forest (e.g., elevation > 0.48 m above Mean Low Water) characterized by highly organic sediment and low dissolved oxygen, to which tarpon are well adapted. Overlap between snook abundance and tarpon presence was greatest where connections between ponds and the nearest tidal creek were at a relatively low elevation (e.g., -0.12 m relative to Mean Low Water). Although these findings are most applicable to management and conservation of populations in areas along the Gulf coast of Florida (i.e., in areas of similar tidal range and coastal geomorphology), incorporating more detailed habitat surveys to better capture landscape context, geomorphology, and connectivity into fish sampling designs should be broadly applicable to estuarine fish ecology.