We aimed to describe accurately the timing and site-specific recurrence pattern for surgical resected lung adenocarcinoma and develop genetic-pathological risk prediction models to guide individual postoperative surveillance strategies. We retrospectively analysed radiological, pathological and sequencing data concerning 9 common oncogenic driver mutations from 1531 patients with resected lung adenocarcinoma between 2008 and 2015. The first recurrence site and time-to-recurrence were recorded. Independent risk factors were identified by multivariable regression analysis and consequently incorporated into prediction models. With a median follow-up of 53.2 months, postoperative recurrences were noted in 483 (31.5%) patients. Bone and brain recurrence tended to occur early (median 11.7 and 17.0 months, respectively) while thorax recurrence occurred later (median 22.2 months), which was validated across different tumour stages. EGFR mutation was an independent predictor for brain and bone recurrence and KRAS mutation for early recurrence. Both internal and external validation of the nomograms for brain and bone recurrence prediction showed optimal discrimination (concordance index: internal, 0.75 and 0.81, respectively; external, 0.77 and 0.84, respectively) and calibration. Recurrence occurred relatively evenly during the follow-up period in low-risk groups but mainly occurred within 2 years in high-risk groups. Unique biological differences exist among lung adenocarcinoma leading to distinct patterns of recurrence. These user-friendly genetic-pathological nomograms may help physicians to better stratify patients and make individual postoperative follow-up plans.