This study explores the relationship between impact severity and resulting pulmonary contusion (PC) for four impact conditions using a rat model of the injury. The force-deflection response from a Finite Element (FE) model of the lung was simultaneously matched to experimental data from distinct impacts via a genetic algorithm optimization. Sprague-Dawley rats underwent right-side thoracotomy prior to impact. Insults were applied directly to the lung via an instrumented piston. Five cohorts were tested: a sham group and four groups experiencing lung insults of varying degrees of severity. The values for impact velocity (V) and penetration depth (D) of the cohorts were Group 1, (V = 6.0 m · s(-1), D = 5.0 mm), Group 2, (V = 1.5 m · s(-1),D = 5.0 mm), Group 3, (V = 6 m · s(-1), D = 2.0 mm), and Group 4, (V = 1.5 m · s(-1), D = 2.0 mm). CT scans were acquired at 24 h, 48 h, and 1 week post-insult. Contusion volume was determined through segmentation. FE-based injury metrics for PC were determined at 24 h and 1 week post-impact, based on the observed volume of contusion and first principal strain. At 24 h post-impact, the volume of high radiopacity lung (HRL) was greatest for the severe impact group (mean HRL = 9.21 ± 4.89) and was significantly greater than all other cohorts but Group 3. The concurrent optimization matched simulated and observed impact energy within one standard deviation for Group 1 (energy = 3.88 ± 0.883 mJ, observed vs. 4.47 mJ, simulated) and Group 2 (energy = 1.46 ± 0.403 mJ, observed vs. 1.50 mJ, simulated) impacts. Statistically significant relationships between HRL and impact energy are presented. The FEA-based injury metrics at 24 h post-contusion are ε(max) · ε(max) exceeding 94.5 s(-1), ε (max) exceeding 0.284 and ε(max) exceeding 470 s(-1). Thresholds for injury to the lung still present at 1 week post-impact were also determined. They are ε(max) · ε(max) exceeding 149 s(-1), ε (max) exceeding 0.343 and ε(max) exceeding 573 s(-1). A mesh sensitivity study found that thresholds based on strain rate were more sensitive to changes to mesh density than the threshold based on strain only.