Problem Statement. The lockdown against coronavirus 2019 pandemic has led to a decrease in physical activity (PA), but any study has analyzed PA’ evolution during this transition following the social-ecological theoretical framework of human development. It suggests that PA engagement depends on the interaction of individual, interpersonal and environmental determinants. This model is essential to better understand PA engagement and to prevent future crises. Purpose. This study investigated the evolution of adolescents’ PA with their protective and inhibitory barriers according to ecological model of PA engagement. Methods. A longitudinal follow-up of 808 French adolescents between before and during lockdown were questioned from PA and socio-demographic data. Wilcoxson signed rank test and Boruta forest algorithm were used to quantify the evolution of each PA intensity and to identify protective and inhibitory socio-ecological’s determinants on PA behaviour. Results. Our results showed a significant decrease in the overall PA (-231 Mets.min.week-1, p<.05). More specifically, low PA decreased significantly (-480 Mets.min.week-1, p<.05), while moderate and vigorous PA has remained significantly stable during the lockdown (+0 Mets.min.week-1, p<.05). The three determinants of the socio- ecological framework are included in the model of evolution of each PA’ intensity. Individual (PA initial profile, 58.8 median importance score) and environmental (garden: 10.42; housing: 7.93 median importance score) determinants are the two most important. Interpersonal determinants have not a significantly importance to explain PA’ evolution. Conclusions. This study highlighted the complex and non-linear process of PA engagement at a critical period where the determinants are being disrupted. Socio-ecological tools are identified to rethink prevention strategies adapted to future crisis periods. Future interventions to promote PA should not target one determinant of the system but should consider the interactions of all socio-ecological variables.