Clouds play an important role in the weather prediction and global climate changes. Cloud liquid water path (LWP) is one of the important cloud microphysical parameters. LWP can be extracted from the millimeter wave radar. To investigate the dependence of the LWP retrieval on retrieval methods and cloud particle spectrum assumptions and improve the LWP retrieval,based on the cloud particle spectrum parameters from the aircraft detection data over China,the retrieval experiments are carried out by using the ground-based millimeter wave radar data in Shouxian from AMF-ARM Mobile Facility in 2008. Based on different cloud particle spectrum parameter assumptions,a physical iteration method and a statistical regression method are respectively used to retrieve the stratiform cloud LWP. And the results are evaluated with the products from the collocated microwave radiometer products. It is shown that although the results are strongly dependent on the spectrum assumption of the cloud particles, the physical iteration method less depends on the cloud spectrum assumption than the statistical regression method. Overall,the retrieval results from the physical iteration method agree better with the radiometer based LWP than those from the statistical regression method. Particularly,the spectrum of the clouds from the aircraft based measurements over China favor the LWP retrievals. The further analysis indicates that the ignorance of the large size particles could lead to the overestimation of the LWP retrievals,which increases with the increase of the radar reflectivity. Finally,a spectrum parameter selection scheme based on different cloud characteristics is proposed. The results show that the new scheme can significantly improve the agreement of the LWP retrievals with the microwave radiometer results.