Abstract:Accurate simulation of grain protein accumulation is of considerable importance for the effective regulation of high-quality wheat production on drylands in hilly loess regions. Using field experimental data of Anjiagou Village, Fengxiang Town, Anding District, and Dingxi City from 2016 to 2017 and meteorological data of Anding District and Dingxi City from 1971 to 2017, a dryland wheat grain protein content model was established using the Agricultural Production Systems Simulator (APSIM) and tested using correlation analysis. Quantitative analysis was conducted on the effect of different tillage methods and sowing dates on wheat grain protein content. Four tillage methods were used:conventional tillage (T), conventional tillage with straw cover (TS), no tillage (NT), and no tillage with straw cover (NTS). Further, three sowing dates were set:early sowing date (ESW), normal sowing date (NSW), and late sowing date (LSW). The findings revealed that under the tested tillage methods and sowing dates, the root mean square errors between the simulated and measured yield and grain protein content were 66.4-121.9 kg·hm ^-2and 0.2%-1.1%, respectively, and the normalized root mean square errors were 1.23%-9.66% and 1.31%-9.94%, respectively. These results indicate a satisfactory precision. Sowing date had a significant effect on the wheat grain protein content for drylands. The highest grain protein content was found with NSW, but the content decreased significantly with LSW. The relationship between wheat yield and grain protein content for the four tillage methods showed a quadratic curve in an open downward direction. As the grain protein content increased, the yield first increased and then decreased. TS and NTS (straw cover) were more conducive to the increase in the wheat grain protein content than T and NT (no straw cover).