Understanding the distribution and abundance of organisms can be exceedingly difficult for pelagic fish species that live in estuarine environments. This is particularly so for fish that cannot be readily marked and released or otherwise tracked, such as the diminutive delta smelt, Hypomesus transpacificus, endemic to the San Francisco Estuary. The environmental factors that influence their distribution operate at multiple scales, from daily tidal cycles and local perceptual fields to seasonal and annual changes in dominant environmental gradients spanning the entire San Francisco Estuary. To quantify scale-specific patterns and factors shaping the spatiotemporal abundance dynamics of adult delta smelt, we fit a suite of models to an extensive, spatially resolved, catch survey time series from 13 annual cohorts. The best model included cohort-specific abundance indicators and daily mortality rates, a regional spatial adjustment, and haul-specific environmental conditions. The regional adjustment identified several density hot spots that were persistent across cohorts. While this model did include local environmental conditions, the gain in explained variation was relatively slight compared to that explained by the regional adjustment. Total abundance estimates were derived by multiplying habitat volume by catch density (design-based) and modeled density (model-based), with both showing severe declines in the population over the time period studied. The design-based approaches had lower uncertainty but potentially higher bias. We discuss the implications of our results for advancing the science and improving management of delta smelt, and future data collection needs.