In assessing lumbar spine injuries of civil aircraft occupants under vertical impact, existing anthropomorphic test devices cannot accurately capture the injury mechanisms. Digital human body models, such as THUMS, are primarily developed for automotive crashes, and their default seated posture differs from that of real civil aircraft occupants. Although the THUMS model can be adjusted to better represent civil aircraft occupants’ seating posture, the mechanical response of the lumbar spine after adjustment has not been sufficiently validated. This limits the model’s reliable use in assessing lumbar spine injuries of civil aircraft occupants. To address this limitation, the THUMS lumbar spine model was modified to reflect aircraft seating posture and physiological lumbar lordosis, and the stiffness of key lumbar ligaments was systematically optimized. Following optimization, the model’s mechanical response curves under flexion, anterior shear, tension, and compression all fell within the post-mortem human subject (PMHS) response corridors, with deviations from the mean PMHS curves significantly reduced. Under the typical 14 g vertical impact specified by FAR 25.562, the optimized lumbar spine exhibited lower peak flexion moments at all lumbar levels, while the axial compressive load and predicted injury risk became more concentrated in the upper lumbar spine. These findings indicate that directly using the original THUMS model may underestimate injury risk in the upper lumbar region while overestimating it in the lower lumbar segments. Posture adjustment, lumbar curvature correction, and ligament stiffness optimization enhance model biofidelity, providing a more reliable basis for the assessment of occupant lumbar injuries.