AbstractMarine turtles are a group of imperiled marine megafauna particularly vulnerable to anthropogenic stressors. Most long‐term studies of marine turtles are based on nesting surveys which focus on numbers of eggs, hatchlings, and nesting females. However, we know less about long‐term abundance trends of immature and adult turtles in the marine environment. To address this data gap, we examined records from 35,000 underwater visual census (UVC) dives (1993–2019) and short‐term in‐water turtle survey data (2009–2014) at Cocos Island, Costa Rica. During UVCs, trained divemasters from UnderSea Hunter recorded observations of two species of marine turtles—green Chelonia mydas and hawksbill Eretmochelys imbricata. Our short‐term in‐water surveys revealed that most turtles at Cocos are greens, but both immature and mature greens occur at Cocos. We analyzed long‐term UVC data using a hierarchical modeling approach and we modeled a 26% decrease in the relative abundance of turtles observed on dives each year. Our model also revealed potential interactions between tiger sharks and turtles, finding that for each additional tiger shark present during a dive, the predicted relative abundance of turtles decreased by 43%. Lastly, our model suggested the influence of environmental variation on marine turtle relative abundance; a 1°C increase in sea surface temperature (SST) decreased the predicted relative abundance of turtles by 7%. Our results suggest that marine turtles are sensitive to long‐term environmental and oceanographic changes, and potentially avoid certain areas to reduce exposure to tiger sharks. Given our study area is already protected, there needs to be more focus on protecting adult turtles during their movements across the Eastern Tropical Pacific. Our work also highlights the importance of long‐term underwater surveys to monitor adult turtles.