ESR prepress abstract   -  DOI: https://doi.org/10.3354/esr01386

ABSTRACT: Development of habitat occupancy models for protected species with low detection rates can be difficult; however a possible solution is to use proxy species. In the San Francisco Bay-Delta estuary, detection of the endemic and endangered Hypomesus transpacificus is extremely rare despite extensive survey efforts. We applied a tree-based machine learning algorithm to evaluate habitat characteristics associated with detection of H. transpacificus using paired data from a pelagic fish survey and a lower-trophic monitoring study. Preliminary analysis using only H. transpacificus data produced a model with limited predictive ability; models combining H. transpacificus with a closely related species, Hypomesus nipponensis, or a broader set of potential surrogate pelagic fishes were better supported. The catch per unit volume (CPUV) of the small pelagic fish community (SPFC) was the strongest explanatory variable for both H. transpacificus and combined H. transpacificus - H. nipponensis detections, indicating shared habitat use among multiple species in the estuary. Salinity and zooplankton prey density had the highest relative influence on SPFC CPUV. These results indicate that H. nipponensis and the SPFC are utilizing similar habitats as H. transpacificus and may therefore serve as appropriate proxies for H. transpacificus in habitat modeling or other monitoring and analysis. This finding also suggests the potential that H. transpacificus may compete with multiple species for limited food resources within small patches of high-quality habitat; the consequences of such dynamics on species recovery is identified as an important topic for future research.