【摘要】 选用1994-2016年四川省7个辐射站气象数据,在3个辐射区（川西高原Ⅰ区、川东盆地Ⅱ区和川西南山地Ⅲ区）中评价了 6种地表太阳总辐射（R_s）估算模型在3种天气类型（晴、多云、阴）下的适用性,并分析基于天气类型的组合模型在不同区域的模拟效果,以探寻最适宜全省不同区域的R_s估算方法。结果表明:（1）各经验模型在四川省整体表现良好（决定系数R²介于0.554～0.934,P<0.001）,Ⅰ区（甘孜和红原站）模拟效果最好的为日照时数模型A-P (平均绝对误差MAE为2.210±0.714MJ·m^-2·d^-1),Ⅱ区（成都、绵阳和泸州站）、Ⅲ区（峨眉山和攀枝花站）模拟效果最佳的均为混合模型Chen (Ⅱ区MAE为1.510±0.027MJ·m^-2·d^-1,Ⅲ区为1.930±0.006MJ·m^-2·d^-1);（2） 6个模型在四川省3种天气类型下的模拟效果呈晴天>多云>阴天的规律,日照时数模型（A-P和Ba模型）能更好地模拟晴天时的R_s,混合模型（Chen和Ab模型）则在多云和阴天时模拟效果更佳,Ⅰ区在晴天、多云、阴天3种天气下模拟效果最好的模型分别是A-P （整体评价指标 GPI 为 0.850）、Ab （1.294）、Ba （0.862）,Ⅱ区分别为 A-P （0.381）、Chen （1.358）、Chen（1.742）,Ⅲ区分别为 Chen （0.204）、Chen （0.857）、Chen （0.526）;（3）基于天气类型的组合模型（M_新）模拟各区R_s的效果均比未组合前各模型的效果好（3个区GPI分别为0.558、0.582、0.134）。因此,推荐使用基于天气类型的组合模型来估算四川省R_s。更多还原

【Abstract】 The global solar radiation（R_s） is of great significance in the fields of agricultural production system,hydrometeorological research and clean energy development.Due to the complicated topography and various climate in Sichuan province,the distribution of global solar radiation is uneven.In order to explore the most suitable R_s estimation models for different zones of Sichuan,Sichuan was divided into three solar radiation zones in this paper（zone Ⅰ:western Sichuan plateau,zone Ⅱ:eastern Sichuan basin,and zone Ⅲ:southwest Sichuan mountain）for research.Zone Ⅰ（Ganzi and Hongyuan stations） has low temperature,sufficient sunshine and strong solar radiation;zone Ⅱ（Chengdu,Mianyang and Luzhou stations） has high relative humidity,short sunshine duration and low radiation;zone Ⅲ（Emeishan and Panzhihua stations） is abundant in solar radiation resources,but the air temperature and humidity vary greatly within the zone.In three solar radiation zones,based on the meteorological data from 1994 to 2016,the applicability of six R_s estimation models were evaluated under three weather types（sunny,partially cloudy and cloudy）,and the simulation effect of combined models based on weather types in different zones were analyzed.The results showed that:（1） the empirical models performed well in Sichuan province（the coefficient of determination R² ranged from 0.554 to 0.934,P<0.001）.The most accurate simulation model in zone Ⅰ was sunshine-based model A-P with mean absolute error（MAE） 2.210±0.714 MJ·m^-2·d^-1.The hybrid model Chen was the best in zone Ⅱ(MAE was 1.510±0.027 MJ·m^-2·d^-1) and zone Ⅲ(MAE was1.510±0.027 MJ·m^-2·d^-1).（2） The statistical performance of the six models under three weather types in Sichuan showed sunny>partially cloudy>cloudy.The sunshine-based models（A-P and Ba models） could better simulate the R_s in sunny days,while the hybrid models（Chen and Ab models） had higher accuracy in simulating partially cloudy and cloudy days.The A-P（the global performance indicator GPI=0.850）,Ab（1.294） and Ba（0.862） models were the best models to simulate R_s of zone Ⅰ in sunny,partially cloudy and cloudy,respectively.The A-P（GPI=0.381）,Chen（1.358） and Chen（1.742） models were the best models to simulate R_s of zone Ⅱ under three weather types,respectively.The Chen model was the best model to simulate R_s of zone Ⅲ in sunny,partially cloudy and cloudy with GPI 0.204,0.857 and 0.526,respectively.（3） The R_s combined models(M_new) based on weather types had the best simulation accuracy in each zone（GPI were 0.558,0.582 and 0.134 in 3 zones,respectively）.Therefore,it is recommended to use the R_s combined models based on weather types to estimate R_s in Sichuan province.更多还原