VLBI数据处理中的剩余钟行为与剩余大气效应建模

ON THE MODELING OF THE RESIDUAL EFFECTS OF THE CLOCK BEHAVIOR AND THE ATMOSPHERE EFFECTS IN THE ANALYSIS OF VLBI DATA

抽样选取了 2 7次上海天文台佘山站参与的天测与测地VLBI实验 ,分别进行了单次解算。通过分析解算参数随剩余钟行为和剩余大气效应分段拟合长度的变化 ,得到以下初步结论 :(1)选取不同分段拟合长度时 ,站坐标解算结果和时延残差加权均方根存在差异 ,最大分别至厘米级和数十皮秒 ,因而分段拟合长度不能随意选取。 (2 )分段拟合时段长度存在某一合理取值范围 ,它不宜过长 ,否则钟和大气的剩余效应短周期变化不能很好地模型化。为了保证待估参数解算时有足够的自由度 ,拟合时段不宜过短 ,否则将导致法方程近于或出现奇异 ,达不到较好控制噪声的效果。 (3)由于各次实验、同一次实验中的不同台站相应的钟和大气条件存在差异 ,有必要对每次实验以及每次实验中的各观测台站分别分析 ,寻找合适的分段拟合长度。这在实际操作中显然相当烦琐。 (4)一般而言 ,在剩余钟行为拟合长度缺省值 6 0min情况下 ,剩余大气效应分段拟合长度以介于 10min至 4 0min为宜 ;在剩余大气效应分段拟合长度缺省值 2 0min情况下 ,剩余钟行为分段拟合长度以介于 2 0min至 10

We checked the dependence of the estimation of parameters on the choice of piecewise interval in the continuous piecewise linear modeling of the residual clock and atmosphere effects by single analysis of 27 VLBI experiments involving Shanghai station (Seshan 25m). The following are tentatively shown: (1) Different choices of the piecewise interval lead to differences in the estimation of station coordinates and in the weighted root mean squares (wrms) of the delay residuals, which can be of the order of centimeters or dozens of picoseconds respectively. So the choice of piecewise interval should not be arbitrary. (2) The piecewise interval should not be too long, otherwise the shortterm variations in the residual clock and atmospheric effects can not be properly modeled. While in order to maintain enough degrees of freedom in parameter estimation, the interval can not be too short, otherwise the normal equation may become near or solely singular and the noises can not be constrained as well. Therefore the choice of the interval should be within some reasonable range. (3) Since the conditions of clock and atmosphere are different from experiment to experiment and from station to station, the reasonable range of the piecewise interval should be tested and chosen separately for each experiment as well as for each station by real data analysis. This is really arduous work in routine data analysis. (4) Generally speaking, with the default interval for clock as 60min, the reasonable range of piecewise interval for residual atmospheric effect modeling is between 10min to 40min, while with the default interval for atmosphere as 20min, that for residual clock behavior is between 20min to 100min.