Understanding the erosion process of sand will help improve dredging projects. Pickup functions are used to predict the amount of sand that can be eroded by a flow with certain parameters. The most common functions that are used have been validated to data in the low velocity regime (θ < 1). However, modern day dredging processes are in the high velocity erosion regime (θ > 1).An example of such a function is the widely used pickup function of Van Rijn (1984). It predicts the pickup of grains well at low flow velocities, but overestimates pickup at high flow velocities.Recently Bisschop (2018) has performed erosion experiments at high flow velocities. In 2015 Boskalis performed experiments in the same velocity regime during a preliminary study, however, a different setup was used. It was found that there is a large (factor 2) discrepancy between the two studies.The present study has performed experiments with a setup similar to that of Van Rijn (1984), however, the flow velocity was not limited to the low velocity regime. Erosion tests were done with bed shear stresses up to 60 Pa (θ<30). The setup in question uses a sediment lift to push a prepared sand bed up into the flow.The determination of the bed shear stress was questionable in preliminary study of Boskalis. Since it is a very important parameter in erosion predictions, it was a focus of the present study to determine the bed shear stress accurately. The bed shear stress was determined with three different methods which all agree within an acceptable range (up to 8%).The erosion velocity is determined by measuring the lift velocity, which is equal to the erosion velocity when the top of the sand bed is level with the floor of the setup. Whether the top of the sand bed is level or not has been determined with the use of a high speed camera.The produced data has been compared to existing data and pickup functions of other studies. From this comparison it is clear that no existing functions match the data of the present study.In the present study an analysis is done to find out why the new data does not match existing functions. Also, based on several existing functions, a new function is presented that has a good fit with the data of the present study.