中文摘要：

传统计算大气总水汽量是利用改进的Langley方法，通过准确测定传感器的光谱响应函数结合辐射传输方程模拟大气中水汽透射比并反演总水汽量。而本文提出了一种根据天空辐射计940nm通道的太阳直接辐射资料反演晴空条件下大气柱总水汽量的算法，该算法直接根据天空辐射计的观测数据估算了描述大气中水汽透射比的参数（a和b），而不依赖于光谱响应函数的精确测量；反演得到的a和b值包含了观测站温度、气压和湿度垂直廓线的季节变化等信息，不受模式模拟误差的影响。利用2009年3—8月兰州大学半干旱气候与环境观测站（SACOL）天空辐射计资料，用该算法获得了观测时期内大气总水汽量，然后利用同期探空资料反演的水汽量验证天空辐射计反演和微波辐射仪观测的水汽量。结果表明，这两种方法得到的水汽总量都是可靠的。天空辐射计与微波辐射仪、CE318型太阳光度计的反演水汽量表现出较好的一致性，拟合斜率值分别为1．03和1．64，相关系数均〉0．95，相对误差在2．1％～11．3％范围内。该算法可广泛应用于东亚地区天空辐射计网（SKYNET）对总水汽量的反演。

英文摘要：

An inversion algorithm was presented, which can retrieve the atmospheric water vapor content from direct solar irradiance measurements of sky-radiometer at 940 nm under clear-sky conditions. For the traditional method, the atmospheric transmissivity and total columnar water vapor content are generally calculated by radia- tive transfer model with accurate spectral response function of sensor through a modified Langley plot method. The constants of a and b were calculated directly by using combination of this method and measurements from sky radiometer. The constants can describe water vapor transmissivity and not dependent on the accurate spectral response function. Meanwhile, the constants （a and b） include the seasonal variations of vertical profiles of air temperature, pressure, and moisture over observation site and may reduce the error of water vapor content re- trieved from the traditional method. This method is preliminarily applied to the sky-radiometer data collected at the Semi-Arid Climate and Environment Observatory of Lanzhou University （SACOL） from March to August 2009. And the water vapor content during this period was retrieved. Firstly, compared the results obtained from sky-radiometer and microwave radiometer with the simultaneously radiosonde, and the validations show that the results from different methods are both reliable. Secondly, the results derived from sky-radiometer agreed well with microwave radiometer and sun photometer. The fitting slope values are 1.03 and 1.64, respectively. While the correlation coefficients are all greater than 0.95 with the relative error ranging from 2.1% to 11.3 %. This method can be widely used to compute the water vapor content measured by sky-radiometer in SKYNET over East Asia.