This paper presents analysis and spatial channel modeling based on a series of 930 MHz measurements conducted along the “Zhengzhou-Xi’an” high speed rail (HSR) of China. Raw data are collected in five cuttings in rural and suburban environments subject to the consideration of the geometry of cuttings, including crown width and bottom width. A comparison among different distributions of the small-scale-fading is made by using a root mean squared errors (RMSE) goodness-of-fit test, which shows that the Ricean distribution offers the best fit. Moreover, it is found that the rich reflection and scattering components in the cutting scenario reduce the effect of the dominant component of the received signal, which results in a Rayleigh distribution, validating a good fit in this line-of-sight (LOS) scenario. Due to the rich multipath components and the directional transmitting antennas in HSR, the Ricean K-factor is modeled as a piecewise function of distance associated with the median K-factor and a standard deviation. Our analysis shows that a “wide” cutting scenario, such as with shallower slopes and wide crown and bottom widths, can reduce the impact of channel fading.