Large-scale wind farms are often located in the Gobi and desert regions. However, existing research for wind farms in Gobi-type terrain are still limited, and conventional fixed-point observational methods suffer from insufficient spatial representativeness. To address these gaps, this study employs a self-developed mobile platform, The Wind Runner, to conduct field campaigns at four representative large-scale Gobi wind farm. Doppler LiDAR in DBS mode is used to systematically characterize the winter-spring wind-field features. And the dominant wind direction, diurnal wind speed variation, wind twist angle and wind speed profile are analyzed. Additionally, a multi-angle, multi-point scanning strategy combining PPI and RHI modes is proposed and implemented to reconstruct the three-dimensional wake flow under the typical high wind speed and low turbulence conditions of the Gobi environment. Data are used to evaluate common analytical wake models. Results show that siting near mountain foothills or topographic corridors enhances wind speeds and directional stability. Pronounced diurnal variations peak from midday to early evening. Near-wake regions (x/D <= 2) exhibit significant velocity deficits and asymmetric bimodal structures, extending downstream up to similar to 8D. Gaussian-based wake models outperform conventional top-hat models in capturing wake expansion and velocity recovery, with the Bastankhah-Port & eacute;-Agel model showing best agreement. Wakes under Gobi conditions exhibit strong downstream persistence, potentially affecting neighboring farms up to similar to 1 km downstream. A minimum turbine spacing of 6D is recommended to mitigate wake losses and optimize energy yield.