The water flow velocity over hillslope is an important hydraulic parameter. A series of comparative flume experiments were conducted to study the flow velocity with pulse boundary model method and dye tracer method, along both the frozen slopes and non-frozen slopes. The flume used in the experiments was 3.8m long, 0.2m wide, and 0.08m high, with soil taken from Koxkar Glacier basin in Akesu, Xinjiang Automous Region. The experimental soil was packed evenly into the flume to a bulk density of 1.0g/cm3 and a thickness of 5cm. The packed soil was saturated before frozen prior to each experiment. The experiments involved three flow rates (12, 24 and 48L/min) and three slope gradients (5°, 10° and 15°). The velocities measured by pulse boundary model method increased exponentially with distance from the solution injection location, before becoming steady. The distances for measuring flow velocity with pulse boundary model method on the frozen slope were calculated from the relationship of velocity and distance. The calculated results showed that the minimum distances for measured velocity with an error less than 5% were 1.7~2.7m and those for an error less than 10% were 1.4~2.1m from the injector under the given flow rates and slope gradients. The velocities, over frozen slopes, measured by the pulse boundary model method under the given experimental conditions ranged from 0.45m/s to 0.98m/s, increased with slope gradients and flow rates, which were 1.43 times of those over non-frozen slopes. The velocities over frozen slopes, as measured by the dye tracer method were 3%~20% higher than those measured by the pulse boundary model method under different slope gradients and flow rates. Those over non-frozen slopes measured by the dye tracer method were 6%~35% higher than those measured by the pulse boundary model method. The result is useful and important to understand the hydraulic dynamic process of frozen soil erosion.