Iverse Modelling with a Lagrangian Particle Disperion Model: Application to Point Releases Over Limited Time Intervals

Conclusions

A new method for inverse modelling of atmospheric trace species emissions, using a backward-running LPDM, has been introduced. It has been successfully applied to the first ETEX tracer release, and could be used to reconstruct the timing or the spatial location of the release. However, regularisation methods and weighting of observations need to be further developed, including nonlinear regularisation terms, in order to improve the reconstruction of non-smooth shapes and to eliminate the detoriating influence of observations that are either wrong or cannot be explained well by the dispersion model. Introducing different weights for the observations is desirable also because of the correlation among the measurements. However, possibilites to estimate covariance matrices of both measurements and their errors are limited as compared to 4DVAR in numerical weather prediction, where a large climatological data set is available.

For point releases, other strategies for finding the minimum of the cost function (e.g., searching all possible grids as a source while using the inversion to determine the optimum temporal evolution for this grid) should be tried.

It is planned to extend the method to cases with deposition or decay. Indeed, the scaling presented here should be valid in this more general case, too, it appears only to be more difficult to provide the proof.

Potential applications include nuclear accidents and nuclear bomb testing (verification of the Comprehensive Test Ban Treaty³). Furthermore, it may be applied to any source determination problem (point or area source) where nonlinearities are not too important.

For updates of this work, please visit http://boku.ac.at/imp/envmet/invmod.html.