The drift distances and time spent in the drift by freshwater shrimps, Gammarus pulex, in a small stony stream, and their implications for the interpretation of downstream dispersal

1. The objectives were (i) to determine experimentally and to model the relationship between mean water velocity and both the mean distance travelled, and the mean time spent, in the drift by freshwater shrimps, Gammarus pulex; (ii) to develop a drift distance–water velocity model from the experimental study, and validate it with field data; (iii) to examine the relationship between drift rate, water velocity and benthic density with the latter expressed as a mean value for the whole stream and a mean value corrected for the distance travelled in the drift. 2. In field experiments at 10 water velocities (0.032–0.962 m s^?1), the significant relationship between the mean drift distance and mean water velocity was described both by a power function (power, 0.96) and a linear relationship. The mean drift time was fairly constant at 8.3 s (95% CL ± 0.4). A simple model estimated the drift distance and time spent in the drift by different percentages of the drifting invertebrates. This model predicted correctly the positive relationship between drift rate and water velocity for field data over a year. 3. The relationship between drift rate per hour and the independent variables, water velocity and benthic density, was well described by a multiple‐regression model. Adding temperature and date did not improve model fit. Variations in water velocity and benthic density explained 96% of the variation in nocturnal drift rate (65% to velocity, 31% to benthic density), but only 40% of the variation in diurnal drift rate (29% to velocity, 11% to benthic density). Correcting benthic density for the drift distances did not improve model fit. 4. The significance of this study is that it developed models to predict drift distances and time, values being similar to those obtained in another, larger stream. It also illustrated the importance of spatial scale in the interpretation of drift by showing that when drift distances were taken into account, the impact of drift on the population was higher (4–10% lost day^?1) than when drift distances were ignored (usually < 3% lost day^?1), especially at a local level.