The dispersion model Pello is used, amongst other applications, for estimating and tracking dispersion of radioactive nuclides and gases. Pello is a stochastic particle model, where the source is represented by emission of model particles which are then transported by the wind field and diluted by turbulence. Today Pello is accessible to the Swedish radiation emergency preparedness system via an implementation in ARGOS through the Match framework at SMHI. To visualise the result from Pello, the particles are traditionally counted grid-boxwise, box-counting, and thus the concentration field is estimated and then visualised on a map. Visualisation is one of several purposes of post-processing the result from the dispersion model, and box-counting is one post-processing method amongst many. In this report we have investigated whether kernel density functions (KDE) may serve as a good alternative method for post-processing of particle model dispersion results. All post-processing methods aims at redistributing the mass of each particle to its neighbourhood. The advantage with KDEs is that this redistribution can be done more delicately in order to reduce both noise as well as over smoothing in the model results. In this report we present a number of different alternative algorithms to compute KDEs. Two methods "Integrated turbulence" and "Partition varying bandwidth" are then singled out for benchmarking against box-counting. The test case is a model run with Pello of the Fukushima Daiichi accident. The post-processed results, deposition fields and air concentration fields, are then compared both visually and statistically (mean square error). Given a fixed number of particles, the KDE-methods generate results that are better (less noise, less over smoothing) than box-counting. We also show that, given a certain quality threshold, the KDE methods may reduce the relative number of particles that need to be simulated: KDE methods can yield equivalent results as box-counting, but with fewer particles. Our results indicate that the number of particles can be reduced by at least one order of magnitude. A reduction in the number of released particles will in turn reduce the time it takes to run the model.