两种反演地表净太阳辐射的参数化模式在北京地区的比较

Net surface solar radiation over Beijing area derived by two parameterization models:A comparison

地表净太阳辐射(NSSR)在地表辐射平衡、地-气能量交换、天气预报、气候变化和太阳能利用等研究方面具有重要的作用。利用卫星遥感技术能在广大的空间区域内快速地获取地表净太阳辐射的分布情况,特别是在无人观测的极地和海洋区域。本文基于CERES/SSF卫星数据,分别采用两种地表净太阳辐射(NSSR)的卫星反演参数化模式——Li模式与Masuda模式计算了北京市上甸子地区2005年1、4、7和10月的NSSR,并将反演结果与上甸子大气本底站的实测结果进行了对比分析,并对两种模式进行了修正以更好地反演北京地区的地表净太阳辐射。结果显示:Li模式和Masuda模式反演结果与实测结果间的平均绝对误差在晴空条件下分别为62.2和50.8 W/m2,在云天条件下分别为82.1和71.8 W/m2;Li模式相对于Masuda模式具有偏大的趋势,在晴天和云天条件下两者的平均绝对偏差分别为11.5和10.2 W/m2;对Li模式和Masuda模式进行线性拟合回归后能有效减小反演过程中的系统性偏差,修改后的Li模式和Masuda模式在所有天空状况下反演的月平均NSSR值与观测值间的平均绝对误差分别为-1.8和-3.4 W/m2,均方根误差分别为19.6和26.0 W/m2。

The Net Surface Solar Radiation （NSSR） plays a very important role in some research areas such as the radiation balance of earth＇s surface, the energy exchange of earth-atmosphere system, weather forecast, climate changes and solar energy utilization et al. Satellite Remote sensing technology can be used to quickly derive the distribution of NSSR over large areas, especially, over those areas where have no meteorological observation, for instance, polar and ocean areas. In this paper, Based on the CERES/SSF satellite data, two parameterized models named Li model and Masuda model are used to calculate the NSSR over Shangdianzi area in Beijing during January, April, July and October in 2005, and a series of analysis are performed to compare between the NSSR data derived from parameterized models and that measured at ground station. At the same time, two fitted models based on Li model and Masuda model are proposed in order to do better in deriving NSSR over Beijing area. The results show that the Mean Bias Error （MBE） of Li model and Masuda model are 62.2W/m2and50.8W/m2 under clear skies , while 82.1 W/m2 and 71.8 W/m2 under cloudy skies. Generally, the value of Li model is larger than that of Masuda model. The Mean Absolute Deviation （MAD） between the two results are 11.5 W/m2 and 10.2W/m2 under clear and cloudy skies, respectively. The systematic errors of the two models can be decreased evidently after the models are linear fitted. The MBE of monthly mean NSSR between observation and derived from models is 19.6 W/m2 for fitted Li model and 26.0 W/m2 for fitted Masuda model.