Managers making decisions may benefit from a well-informed understanding of a species' population size and trends. Given the cryptic nature and habitat characteristics of the western pond turtle (Emys marmorata), however, imperfect detection may be high and population estimates are frequently varied and unreliable. As a case study to investigate this issue, we used temperature dataloggers to examine turtle behavior at 2 long-term monitoring sites with different hydrological characteristics in Sequoia National Park, California, to determine if common stream-survey techniques are consistent with site-specific turtle behavior. Sycamore Creek is an intermittent stream that dries up every summer while the North Fork Kaweah River flows year-round. We found that while turtles spent most of the recorded time in the water (55% in Sycamore Creek and 82% in the North Fork Kaweah River), the timing of traditional surveys only coincided with the turtles' aquatic activity in the North Fork Kaweah River. At Sycamore Creek, turtles were most likely to be in the water at night. In contrast, failure to detect turtles in North Fork Kaweah River is likely owing to the larger size and complexity of the underwater habitat. In both streams, turtles were also more likely to be in the water in the weeks leading up to important changes in hydroperiods. Our findings illustrate the effects that differences in water permanence can have on turtle behavior within the same watershed and how phenotypic plasticity may then affect detection during surveys. Our study highlights the importance of tailoring survey practices to the site-specific behavioral traits of the target species.