Search

Offline atmospheric transport models are normally driven with meteorological analyses. However, subsequent analysis fields are dynamically not consistent with each other, because they are produced in independent data assimilation cycles that lack strong dynamical constraints between each other. In this paper, it is shown that when these data are used with Lagrangian transport models, spurious mixing results from the dynamic inconsistencies. As a consequence, quantities such as potential vorticity or specific humidity that tend to be conserved along trajectories are found to be significantly less well conserved when analysis data are used than when forecast data are used for the trajectory calculations. This leads, for instance, to enhanced stratosphere--troposphere exchange. It is also shown that the dispersion of initially neighboring particles occurs more rapidly with the analysis than with the forecast data. It is therefore concluded that small-scale tracer structures develop too quickly in Lagrangian models, due to the inconsistencies between the driving wind fields.