Plastic film mulching amplifies yield resilience and stabilizes water use efficiency of winter wheat across wet and dry years on the semi‐arid Loess Plateau

BACKGROUND Mulching is usually evaluated on multiyear average yield, which hides how its benefit varies with rainfall. A three‐year split‐plot experiment at Tianshui, Gansu, on the semi‐arid Loess Plateau crossed four fertilization levels (no fertilizer, CK; sheep manure, M; mineral, NPK; manure plus NPK, MNPK) with three mulching types (conventional tillage, CT; plastic film with soil cover, FM; film without soil cover, FMS) in three replicates, across two wet years (2022/2023, 2023/2024) and one dry year (2024/2025). RESULTS Across the 12 combinations, grain yield fell from 6826 kg ha −1 in wet years ( n = 72) to 3376 kg ha −1 in the dry year ( n = 36), −50.5%. Year type explained 80.3% of yield variance, fertilization 7.2%, and mulching 5.5% (each P < 0.001). Averaged across fertilization levels, the FM and FMS advantage over CT rose from 16.2% and 9.7% in wet years to 38.3% and 37.6% in the dry year, a 2.4‐ to 3.9‐fold amplification; MNPK × CT declined most steeply (−59.1%), a high‐fertility trap. Mulching flattened the precipitation‐based water use efficiency (WUEp)–precipitation slope from 0.059 (CT) to 0.042 kg ha −1 mm −2 (FMS); year × mulching was significant for WUEp ( P = 0.042) but not for yield ( P = 0.376). CONCLUSION Amplification arose from a widened soil‐water advantage of mulching over CT (3.7–9.7% in wet years, 11.1–15.8% in the dry year) and a steeper soil‐water–yield response under water limitation. MNPK × FM gave the highest three‐year mean yield (6656 kg ha −1 ) at a coefficient of variation of 29.2%, making mulching a precondition for safe fertilizer intensification.