Knowledge of the salinity of fluid inclusions, including the type of salts and their amounts, is of great importance for the interpretation of geological fluids and their role in rockforming processes like diagenesis, metamorphism, and hydrothermal processes. One of the most vital binary fluid systems for understanding geological processes is the NaCl-H2O system. The salinity of aqueous solutions in fluid inclusions is commonly determined with microthermometry. Moreover, the cryogenic Raman spectra of hydrates of chloride salts in fluid inclusions can be not only used to calculate salinities, but also used to distinguish different brine types. Theoretically, the salinity of one fluid inclusion can’t be calculated via the Raman spectrum in a single point because of the heterogeneity of the fluid inclusion at low temperature and the location of focus range of the laser beam. In order to have a better understanding of the characteristics of phase change of fluid inclusions at low temperature and reveal their Raman spectra responses to salinities, this paper studied the crystallization process and Raman spectroscopy of ice and hydrohalite in NaCl solutions with 5 different salinities. The result shows that in the process of freezing and heating, the crystallization of hydrohalite occurs from an interstitial, hypersaline liquid among the prior formed ice crystals. The forming of hydrohalite may depend on the concentration of brine. The relative intensities of p1 ［(3 402±1) cm-1］ and p2 ［(3 419±1) cm-1］, which are the two of four characteristic Raman peaks, are very stable, while the variation of the relative intensities of the other two peaks p3 ［(3 432±2) cm-1］ and p4 ［(3 535±4) cm-1］ increase with the salinity increasing because of different relative crystal orientations of hydrohalite compared to the polarized laser beam. Therefore, the characteristic ratios of Raman peaks at different points in the same sample are usually variable, and the traditional method of calculating salinities of the fluid inclusion using cryogenic Raman spectrum at one single point makes a big error. Based on large data statistics and analysis, the results indicate that the statistical average value of characteristic Raman peaks maybe reflects the salinity of the fluid inclusion better for the first time. With respect to the relative intensity and the width at half height of the characteristic Raman peaks, the total peaks’ area shows the best linear dependence on the salinity, which is the first choice for the calculation of the salinity of brine. The results of this study expounds the suitable experimental operation and data analysis methods of the salinity calculation of brine using cryogenic Raman spectroscopy, and illustrates the applied conditions of this method in fluid inclusions analysis. Although it’s a complex operation process to calculate salinity using cryogenic Raman spectra, this method is very important because of its anti-interference, large applied range and reliability.