Water flow velocity along an eroding rill over frozen and non-frozen slopes is of great importance for understanding rill erosion hydrodynamics, which is closely related to erosion of freeze thaw soils and sediment transport. A series of flume experiments were conducted with loess soil from Ansai, Shaanxi Province. Flow velocities were measured in a stainless steel flume which was 8 m long, 0.1m wide, and 0.12m high. The experiments involved five slope gradients (5°, 10°, 15° and 20°)，four flow rates (1L/min, 2L/min, 4L/min and 8L/min), with three replicates. Time intervals for water flow running from the top of the flume to the bottom end, covering a total length of 8m was measured to calculate the velocities under different slopes and different flow rates, for both frozen soil and non-frozen soil slopes. The peak velocities were also gained by the data of conductivity changed with time by the electrolyte tracer method during the experiment. The leading edge velocities over frozen soil were between 0.260m/s and 0.843m/s and that of non-frozen were between 0.175m/s and 0.552m/s. Leading edge velocities over frozen soil and non-frozen soil were both increased with the increase of slope gradients and flow rate, the increasing rate of leading edge velocity was decreased with the increase of slope gradient or flow rate under gentle slope（5°~10°, 10°~15°） and low flow rate（1L/min , 2L/min and 4L/min）. Over frozen and non-frozen soils, the velocity was increased faster on gentle slopes than that on steep slopes and under low flow rate than that under high flow rate. The power function fitted the leading edge velocity very well for all the experimental conditions over frozen and non-frozen soils. The slope and the flow rate had greater effect on velocity over frozen soil slopes than over non-frozen soil. The result demonstrated that leading edge velocities and peak velocities over frozen soil were faster than that of non-frozen soil under all slopes and flowrates, the ratio of flow velocities over frozen soil and non-frozen soil was 1.43 and 1.40. The ratio of peak velocities and leading edge velocities over frozen and non-frozen slopes were 0.61 and 0.63. The experiment showed that the method for flow velocity measurement was reasonable, and it can also be widely used. The results can provide a data set for flow dynamics research over frozen soil and non-frozen soil.