Hidden Markov models reveal ontogenetic plasticity in green and loggerhead sea turtles

Conservation of long-lived marine species depends on understanding how individuals use habitats as they grow. Yet management often relies on size- or place-based stage classifications that overlook behavioral plasticity. We apply a behavior-first, data-driven approach to 47 years of capture-recapture for green (Chelonia mydas) and loggerhead (Caretta caretta) turtles in southeast Florida. Hidden Markov Models using standard straight carapace length and seasonal photoperiod from nearly 20,000 capture events estimated the probability that each capture belonged to one of three latent behavioral states: Nomadic, Resident, or Transient. We interpreted states using associations with body size, seasonality, and recapture frequency. Nomads were generally smaller and rarely recaptured, consistent with exploratory movement; Residents showed high site fidelity and frequent recapture; Transients were larger and appeared seasonally, consistent with migratory use. In green turtles, states showed relatively distinct size distributions but still overlapped, challenging rigid size-based life-stage assignments. Loggerheads displayed greater behavioral plasticity, with state membership weakly predicted by size. Across species, recaptured turtles were associated with the Resident state, supporting ecological validity. These results challenge the assumption that body size alone defines life stage or site use and show how probabilistic state models reveal hidden behavioral structure and availability in developmental habitat use.