The growing amount and complexity of particle data in various fields of study (e.g. molecular dynamics) poses an increasing demand of high-quality and high-performance visualizations of such data. Ray tracing is known for the high image quality it can produce and its well-suitedness to global illumination effects, e.g. shadows or ambient occlusion. Effects like these can serve as helpful cues for understanding three-dimensional structures within the data. As an optimization for the rendering of particle data, Wald et al. [WKJ+15] suggest the Particle-k-d (P-k-d) tree, a ray tracing acceleration structure for particles which does not require any memory overhead. Ibrahim et al. [IRR+21] propose a novel, probabilistic method to accelerate particle rendering by culling particles which are likely to be occluded. The goal of this thesis is to examine the viability of probabilistic occlusion culling in the context of a P-k-d-based ray tracer. I contribute by adapting the method proposed by Ibrahim et al. for the use within a ray tracing application and implementing the adapted method as an extension of a P-k-d ray tracer by Gralka et al. [GWG+20]. I go on to evaluate my approach with respect to visual quality and performance. My findings show that probabilistic occlusion culling can provide a performance improvement during super sampling. However, the magnitude of the improvement and the effect on visual quality are greatly dependent on the size of and features within the data set among other factors.