As automated vehicles evolve, seating designs must accommodate a wider range of postures, particularly for non-driving-related activities such as relaxing and sleeping. This study aims to model human back shapes in seated and reclined positions to improve ergonomic seat designs. Human back contour data were collected from 36 participants using a custom measurement device in two setups: a 25° backrest angle and a seat pan angle of 15°, simulating a driving posture, and a 50° backrest angle with the same seat pan angle, representing a reclined posture. Statistical Shape Models (SSMs) were developed to analyze the variability of back contours. The 25° setup exhibited a flatter spinal curve and higher compactness, capturing 79.7 % of the variance with the first principal component (PC1), compared to 74.6 % in the 50° setup. The combined setup balanced these differences, providing a comprehensive model for diverse postures. Overall, PC1, PC2, and PC3 together captured more than 96 % of total contour variance, indicating that variations in back height, neck bending, and lumbar prominence constitute the dominant sources of geometric diversity. These findings offer actionable dimensions for designing ergonomic backrests that support diverse users and postures. Future research should investigate whether implementing these guidelines enhances comfort and should include more diverse populations and a broader range of postures.