中文摘要：

海底地震仪（OBS）广角地震探测技术已广泛应用于研究边缘海和大洋深部构造,与多道地震相比,其优势主要在于OBS通常使用4个分量,可以记录转换剪切波（S波）;另外,在作业时还可以同时记录到天然地震.但对于不同科学目标（如洋陆过渡带或拆离断层）和不同工区（2D剖面或3D工区）,其野外数据采集、室内数据处理、模型正反演及其误差分析都有差异且较为复杂,规范这个过程十分重要.本文根据多年的实际经验,结合2010年西南印度洋中脊开展的OBS地震试验的某测线数据采集、处理和模型正反演过程,详细介绍了OBS地震调查的海上作业（仪器投放,炸测和仪器回收）、数据处理（炮点和OBS位置校正,钟漂校正和滤波技术）、建模反演（RayInvr正演模型和阻尼最小二乘法反演）和模型分析评价（误差统计,模型分辨率和不确定性测试）的方法和流程,并总结了提高OBS回收率、数据处理、震相识别、2D与3D工区模型参数化、确定模型分辨率及其不确定性的实用技巧.

英文摘要：

Ocean bottom seismometer（OBS）has been widely used in wide-angle seismic explorations on structures of marginal sea and deep ocean,because of the four components（three geophones and one hydrophone）,which can record turned Shear wave that carrys more information from velocity anomaly,and as well as earthquake and micro-earthquake can be recorded at the same time.However,due to the different scientific aims,like focus on continental ocean transition or detachment fault,and different experiment data set,like single2 Dprofile or 3Ddata set area,the operation of data collection,processing,forward/inversion modeling and error analysis is complicated,it is very important to write a paper to explain the standard flow on the practice.With previous work experiences,and in particular the 2010 Southwest Indian Ridge OBS seismic experiment data,this paper introduced the normal process on OBS data collection（the details on OBS deploying,airgun shooting and OBS recovering）,processing（shots and OBSs relocation,time drift correction and filter methods,et al.）,modeling（forward modeling method by using RayInvr）,inversion（damped least square method）and model assessment（statistical misfit, resolution and uncertainty of model parameters）.Finally,we summarized 1）the right way to use acoustic release is important to get a better OBS recover ratio,2）a serial of steps of processing are needed in order to get clear signals,3）double check by using the forward model is needed for the seismic phases identification,4）forward model is helped,even for 3D projects,direct and indirect model assessments are need to get a more reliable velocity model.