Valley spacing is an important characteristic to describe the distance between adjacent valleys. It can reflect the structural feature and spatial distribution of valleys. In this study, we took the Heihe Zhengyixia region and Lanzhou Dalingxian region as examples. We used the ArcGIS to resample the digital surface models (DSM) at 0.12 m resolution which were obtained by unmanned aerial vehicle to different resolution DSM. The different resolution DSM data were further processed by Fourier transform in MATLAB software. We compared different spectrums to build the relationship between land surface spectral feature and valley spacing. Analytical results show that: (1) for the first-order valley region, the effective identification of valley spacing feature signal in the spectrum required a terrain resolution higher than 1/5 of valley spacing. The terrain resolution could directly affect the identification accuracy of valley spacing signal. While the space domain had less influence on the valley spacing signal identification. (2) for the multi-order valley region, the effective identification of valley spacing feature signal in the spectrum required a terrain resolution higher than 1/3 of valley spacing. Either the decrease in terrain resolution or the increase in the spatial domain could decrease the spacing signal of lower-order valleys and increase the spacing signal of higher-order valleys in the spectrum.