中文摘要：

受到大气折射的影响，天文观测上通常回避仰角15°以下的目标的观测，但作为大气折射的完整理论研究，低仰角下的大气折射仍然是值得分析探究的．特别是对某些工程应用方面，低仰角的目标有时必须要观测．提出了一套新的利用较差方法测定大气折射的思路．利用一台较大视场的望远镜从天顶开始，在不同高度上对星空作一系列观测，计算不同天顶距处大气折射函数的各阶导数，最后经数值积分可给出大气折射实测值．该方法不依赖于严格的地方参数和复杂精密的观测仪器，并且观测原理相对简单．2007年底，利用一台简易的大视场望远镜在兴隆观测站进行了试验观测，根据较差方法实测得到真天顶距44．8°至87．5°的大气折射值，初步证明了大气折射较差测量方法的可行性．受到观测条件的限制，本次实测结果精度有限，偶然误差最大约为6″，并且存在一定的系统差．在天顶距84°时，与普尔科沃大气折射表的差值约为15″．如何消除因积分模型误差引入的累积误差是今后需要解决的关键问题．

英文摘要：

Because of the effect of the objects with altitude lower than 15 degree. atmospheric refraction, it is seldom to observe But for the complete theoretical research of the atmospheric refraction, it is necessary to observe such objects, even with altitude close to 0 degree. Especially in some special situation, some objects with low altitude need to be observed. A new differential method is presented to measure the atmospheric refraction. Starting with the zenith, a series of star fields at different altitude are observed by a tele- scope with larger field of view, the derivatives of various orders are calculated at different zenith distances, and finally the observational atmospheric refraction can be obtained by numerical integration. The method does not depend on strict local parameters and precise instruments, while the observational principle is relatively simple. An experimental obser- vation was carried out with a simple telescope with large field of view at Xinglong station of NAOC at the end of 2007. The atmospheric refraction from the zenith distance of 44.8 degree to 87.5 degree is gotten by the differential method, which proves the feasibility of the method tentatively. Limited by the observation condition, the precision of the result is not high. The maximal accidental error is about 6 arc second, and there is also some systematic error. The difference between the observation atmospheric refraction and Pulkovo table is about 15 arc second at the zenith distance of 84 degree. How to eliminate the accumulated errors introduced by the errors of the integration model is the key problem that needs solving in the future.