Résumé
A multistate capture-recapture model was developed to estimate movements of brown trout (Salmo trutta) between a main stem and its headwater tributary and their survival and recapture probabilities in each stream. As all individuals entering or leaving the tributary were captured by trapping, the studied ecological system was fully controlled. The performance of multistate models combining two sources of data (trapping and electrofishing) available for 6 years was first evaluated. Realistic estimates were obtained to infer the average spawning behaviour of trout: (i) 58% returned to their original site after spawning, (ii) 9% returned to their natal site for reproduction, (iii) 55% of the ascending individuals performed natal homing. Because less informative systems are pervading, we eventually assessed the sensitivity of multistate models to the level of trapping data integration. A lack of such data led to an underestimation of movement probabilities, and we found that this effect could be compensated by electrofishing samplings.