1:45 PM - 3:15 PM
[MGI26-P08] A sequential trajectory method with data assimilation
Keywords:Trajectory Analysis, Particle Filtering
Atmospheric transport pathways are generally estimated mainly by trajectory analysis, a method that predicts pathways using wind direction and wind speed (V-vector) or by following the movement of air masses tagged in a meteorological model. It is known that the accuracy of the former is greatly affected by the resolution of the objective meteorological data that serves as the background field and that the error increases exponentially with the forecast period. The latter method has no problem with the accuracy of the prediction of particle transport itself, but the computational cost is significant.
To solve this problem, this study applies the sequential data assimilation method to the trajectory analysis and improves the model by adding uncertainty to the trajectory itself. As a result, it is possible to add probability information to the trace line, which has been uniquely determined so far. The NIPR trajectory model (NITRAM, Tomikawa and Sato, 2005) is selected as the model to be improved here. The HYSPLIT model (Stein et al., 2021) and the FLEXible Particle Dispersion model (FLEXPART, Stohl et al., 2005, Pisso et al., 2019), which are representative trace analysis models, take dispersion into account and can also calculate frequency distributions. However, the emphasis is on solving the physical process with skewed turbulence parameterization. The improvement in this work applies a particle filter and adds uncertainty information to the trajectory in the time direction. It does not deal with changes in the material itself. We will introduce our prototype trajectory model with particle filtering in this presentation and discuss how many we need air parcels to calculate this model.
To solve this problem, this study applies the sequential data assimilation method to the trajectory analysis and improves the model by adding uncertainty to the trajectory itself. As a result, it is possible to add probability information to the trace line, which has been uniquely determined so far. The NIPR trajectory model (NITRAM, Tomikawa and Sato, 2005) is selected as the model to be improved here. The HYSPLIT model (Stein et al., 2021) and the FLEXible Particle Dispersion model (FLEXPART, Stohl et al., 2005, Pisso et al., 2019), which are representative trace analysis models, take dispersion into account and can also calculate frequency distributions. However, the emphasis is on solving the physical process with skewed turbulence parameterization. The improvement in this work applies a particle filter and adds uncertainty information to the trajectory in the time direction. It does not deal with changes in the material itself. We will introduce our prototype trajectory model with particle filtering in this presentation and discuss how many we need air parcels to calculate this model.