Abstract:

The performance of APSIM model in simulating soil moisture dynamics and plant growth and production for lucerne (Medicago sativa L.) crop was assessed against the crop observations from three sites in semi-arid Ningxia region of Western China. The model parameters for characterizing plant growth and development for local cultivar, Guyuan lucerne (Medicago sativa L. ‘Guyuan’), were specified or estimated using try-and-error methods based on the detailed observations at one site (Guyuan). Then the model performance in simulating crop phenological development and dry matter production was compared with the observations at another two sites (Tongxin and Yinchuan) to validate the model parameterization. The simulations used local meteorological data and soil profile description at each sites. The results showed that modeled plant phenological development (number of days from sowing to emergence, floral initiation and flowering) matched excellently with the observed at both sites (R²=0.99，D=0.99，NRMSE=2.4% at Tongxin and R²=1.00, D=0.99, NRMSE=7.8% at Yinchuan, respectively); modeled plant dry matter production agreed well with the observed (R²=0.82, D=0.99, NRMRE=7.1% at Tongxin and R²=0.98, D=0.99, NRMSE=9.3% at Yinchuan, respectively), though a slight overestimation (7.1% on average); and modeled soil moisture also agreed reasonably to the observed (R²=0.82, D=0.85, NRMSE=10%), also with a overestimation (10.1% on average). Slightly overestimation of plant production is most likely related with the overestimation of soil moisture, which suggests more detailed measurements being needed for describing soil profiles and improving model predictions. This model parameterization and validation provides the confidence for using the model to optimize crop management, and assess climate change impacts in the semiarid area of Ningxia.

Key words: APSIM-lucerne model, Guyuan lucerne, Model parameterization and validation, Adaptability, Semiarid area of Ningxia

摘要：

基于宁夏3个试验点的数据和同期气象资料，对APSIM模型在宁夏半干旱地区的适应性进行研究。根据试验条件建立相应模型，通过“试错法”对模型相关参数进行校准，实现参数的本地化；并运用多个检验统计指标验证紫花苜蓿（Medicago sativa）土壤含水量及其分别在同心和银川试验点的生育期和干草产量。结果表明：各生育期模拟值和实测值显著相关，决定系数R²分别为0.99和1.00，D值均为0.99，NRMSE分别为2.4%和7.8%；干草产量表现出良好的相关性和一致性，决定系数R²分别为0.82和0.98，D值均为0.99，NRMSE分别为7.1%和9.3%。模拟值比实测值平均高估了7.1%；土壤含水量整体表现出较好的相关性和一致性，决定系数R²为0.82，D为0.85，NRMSE为10%，模拟值比实测值平均高估了10.1%。产量的高估很有可能与水分的高估相关，需要更详细的测量田间试验数据，改善模型土壤参数描述，提高模型预测准确度。校准后的APSIM模型在宁夏半干旱地区具有较好的适应性，可以用于指导该区半干旱区苜蓿生产及栽培措施优化管理。

关键词: APSIM苜蓿模型, 固原苜蓿, 参数本地化, 适应性, 宁夏半干旱区