Evaluating effects of changing sampling protocol for a long-term ichthyoplankton monitoring program

Abundance indices play a crucial role in monitoring and assessing fish population dynamics. Fishery-independent surveys are commonly favored for deriving abundance indices because they follow standardized or randomized designs, ensuring spatiotemporal consistency in representative and unbiased sampling. However, modifications to the survey protocol may be necessary to accommodate changes in survey goals and logistic difficulty. When the survey undergoes changes, calibration is often needed to remove variability that is unrelated to changes in abundance. We evaluated a long-term monitoring program, the Long River Survey (LRS) in the Hudson River Estuary (HRE), to illustrate the process of calibrating survey data to account for the effects of changing sampling protocol. The LRS provided valuable ichthyoplankton data from 1974 to 2017, but inconsistencies in sampling timing, location, and gears resulted in challenges in interpreting and comparing the fish abundance data in the HRE. Generalized Additive Models were developed for five species at various life stages, aiming to mitigate the impact of sampling protocol changes. Model validation results suggest the consistent performance of the developed models with varying lengths of time series. This study indicates that changes in the sampling protocol can introduce biases in the estimates of abundance indices and that the model-based estimates can improve the reliability and accuracy of the survey abundance indices. The model-estimated sampling effects for each species and life stage provide critical information and valuable insights for designing future sampling protocols.