We analyzed monthly web-based depression surveys from 84 high school students in the Tokyo Teen Cohort before and during the COVID-19 pandemic using energy landscape analysis. Adolescents showed a reduced population-level tendency toward depression during the pandemic. Stratification analyses identified low-stable and high-unstable groups with distinct transition dynamics, and longitudinal MRI suggested the brain development contributions to depression susceptibility, which may enable early identification of at-risk individuals during societal disruptions.