Dependence of trajectory accuracy on the spatial and temporal densities of wind data

Abstract

An investigation has been carried out to evaluate the dependence of trajectory accuracy on the density of wind data. The present study has applied an observing system simulation technique to 3 meteorological episodes over eastern Asia. A mesoscale numerical model with a horizontal grid size of 20 km has been employed to generate the reference atmospheric fields. Particles are initially located at 3 isobaric levels (925, 850 and 700 hPa). The results indicate that the 3-dimensional trajectories are sensitive to both the temporal and spatial densities of wind data. The trajectories show relatively strong sensitivity to the temporal density change between 6- and 12-h intervals in general, and also to the spatial density change between 160- and 320-km intervals in one of the cases. The effectiveness of enhancing either the temporal or spatial resolution of wind data for the improvement of trajectory accuracy is found to vary with the meteorological situation. When the intervals of 12 h and 320 km are assumed to be the resolution of wind observations, enhancing the spatial resolution to 160-km interval improves the accuracy more than enhancing the temporal resolution to 6-h intervals in a case with an excursion of the Siberian High. However the opposite is true in the other 2 cases in which either a synoptic scale cyclone or a low pressure system with significant diabatic heating is important, although the difference in the accuracy improvements is small in 1 of the 2 cases. It is also found that the estimated accuracy of a trajectory model and its sensitivity to wind data density may vary significantly with the resolution of the reference wind data sets: the use of reference wind data with low resolution can result in a significant underestimation of transport error and may exaggerate the effectiveness of enhancing the temporal resolution for the improvement of trajectory accuracy. The accuracy of an isobaric model is generally comparable to that of a 3-dimensional model, when data sets with 12-h, 320-km resolution are used. Under a strongly baroclinic condition, however, the isobaric model can give a very large trajectory error.