Environmental monitoring is required for permitting, consenting, and responsibly developing the marine hydrokinetic (MHK) industry. One method of detecting cumulative effects of MHK devices on fishes is long-term observation of fish abundance at a tidal energy site using discrete hydroacoustic surveys. However, fish abundance in such dynamic environments is highly variable, and survey timing can greatly affect observed trends. We present an approach using survey timing to limit the effects of natural environmental variation on observed long-term trends in fish abundance. We examined three months of a long-term, continuous hydroacoustic dataset from a tidal energy site in Cobscook Bay, Maine, USA. Temporal analyses using Fourier and wavelet transforms indicated non-stationary, cyclic components in fish abundance related to tidal, diel, and semi-lunar cycles. Changes in these patterns over time were likely related to changes in the local fish assemblage. Four discrete sampling schemes were then simulated by sub-sampling the continuous time series: 1- and 24- hour surveys, first randomly spaced over time and then timed to hold tidal, diel, and lunar stages constant. The best discrete sampling design consisted of 24-hour surveys that held lunar stage constant. This design is likely to be effective at similar tidal sites worldwide.