Changes in survey implementation can bias traditional design-based indices of abundance, particularly for migratory species where local abundance can vary at short temporal scales. Spatiotemporal model-based estimators provide a flexible alternative that can better account for changes in the timing and location of surveys. Here we develop a geostatistical model, a generalized linear mixed effects model with Gaussian Markov random fields, that is sufficiently flexible to account for changes in survey design, as well as seasonal and interannual variability in abundance and spatial distribution. We then apply this model to a suite of migratory species—juvenile Pacific salmon—with survey data collected from southern British Columbia’s continental shelf over the previous 25 years. Juvenile Pacific salmon showed species-specific spatial distributions with considerable seasonal variability. While the location of species-specific hot spots varied among years, seasonal variation in distributions were greater than interannual variation. We found that a shift from standard transects and opportunistic sampling to a random stratified survey design resulted in lower estimates of abundance for four of five juvenile Pacific salmon species; however, these differences were not significant after accounting for other sampling attributes. Trends in abundance for several species shifted when models included survey and diel effects. Our approach provides a means of deriving indices of abundance for migratory species from surveys with variable effort in time and space. Ultimately such indices can be used to better understand the ecological mechanisms regulating productivity by linking distribution and abundance to environmental drivers.