中文摘要：

以一个江淮梅雨锋低涡的48h模拟结果为基态计算了奇异矢量，并利用奇异矢量构造了假反扰动。分析表明，假反扰动能在一定程度上描述初始误差，可以借助其进行初始误差分析；在假反扰动中，发展缓慢的奇异矢量分量是主体，描述了绝大部分模拟误差且增长迅速的奇异矢量所占比例却很小；分析假反扰动的具体形势发现850hPa上初始误差主要是低涡西南方向的低压槽偏弱，而在500hPa，渤海湾低压槽向西南方向的伸展不够，低压槽东边的高压脊应进一步向北伸展，这些初始误差都分布在关键系统的周围，并且具有中尺度的特征。另外，假反扰动中不同物理量场改进预报质量的贡献各不相同， 各种物理量场间存在着使其空间分布趋于一致的动力相关关系。

英文摘要：

The singular vectors （SVs） is calculated with one Meiyu front mesoscale low simulation trajectory as base state, then the initial error is discussed with the pseudo-inverse perturbation： one kind of special perturbation formed with SVs. The pseudo-inverse perturbation is similar to initial error to certain extent; hence the initial error analysis could be implemented using pseudo-inverse perturbation. The slow-growing SVs dominate pseudo-inverse perturbation, while the quick-growing SVs, which are responsible for most of simulation error, are relatively weaker. The pseudo-inverse perturbation indicates some initial error： on 850 hPa the pressure trough to the west of the low is not deep enough, while on 500 hPa the pressure trough on Bohai sea should have extended to the south-west further and the pressure ridge to the east of pressure trough on Bohai sea should have reached to the north further. All these error scatter around the key system and have the mesoscale feature. The efficiency of different physical fields in pseudo-inverse perturbation of improving forecast are different, and one kind of dynamic correlation exists among different physical fields, which makes the spatial distribution of pseudo-inverse perturbations tend to be con sistent.