The Applicability of Mechanism Phenology Models to Simulating Apple Flowering Date in Shaanxi Province

China's apple growing area and production rank first in the world, and thus apple is one of important economical crops in China. Meteorological disaster occurring in apple critical phenology stage is one of the main disasters impacting yield and quality, especially in the dominant planting provinces such as Shaanxi. Accurate forecasting on flowering date in Shaanxi can provide scientific support for taking applicable defensive management and improving the ability to resist meteorological disasters, and therefore benefit to apple yield and quality. Taking apple flowering stage as an example, the applicability of 4 typical phenology models is evaluated, including Sequential Model (SM), Parallel Model (PM), Deepening Rest Model (DRM), and Thermal Time Model (TTM). There are 4 apple planting divisions in Shaanxi Province. In each division, there are two phenology observation sites. Internal validation and cross validation (Leave One Out Cross Validation) of 4 models are done using sites with longer observations, while shorter record sites are used to evaluate the effect of model extrapolation application. In 4 divisions, 4 sites performing internal validation and cross validation are Xunyi, Luochuan, Liquan and Baishui, respectively, while four sites to conduct extrapolation application are Changwu, Baota, Fengxiang and Tongchuan, respectively. Model performance is assessed according to the root mean square error (RMSE) of modelled flowering date. Internal validation results show that optimal models are different in different sites and generally TTM and SM give similar accuracy (3.30 d). Cross validation also verifies and there is no particularly prominent model. The average RMSE for all four models is 4.52 d. TTM is then extrapolatively applied to other sites with two methods (extrapolation based on values in a single site, and extrapolation based on average values of 4 sites). The accuracy of both methods is higher than that of similar studies abroad (10.0 d), while the accuracy of extrapolation based on values in a single site (5.90 d) is higher than that of extrapolation based on average values of 4 sites (7.21 d). Considering the complexity and simulation accuracy, TTM is recommended to be used to simulate the flowering period in each apple planting division in Shaanxi Province.