An experimental study was undertaken to explore the evolution of flow structures and their characteristics within a randomly packed porous bed with particular attention to evaluating turbulent scalar dispersion. A low aspect ratio bed of 4.67 (bed width to spherical solid phase particle diameter) with fluid phase refractive index matched to that of the solid phase was used in order to obtain time resolved two component particle image velocimetry data. Results are based on detailed velocity vector maps obtained at selected pores near the bed center. Pore, or large scale, regions that are associated with the mean flow were identified based on Reynolds decomposed velocity fields, while smaller scale structures embedded within pore scale regions were identified and quantified by combining large eddy scale decomposition and swirling strength analysis. The velocity maps collected in distinctive pore geometries showed presence of three types of flow regions that display very different mean flow conditions, described as regions with tortuous channel like flow, high fluid momentum jet like regions, and low fluid momentum recirculating regions. The major portion of pore space is categorized as tortuous channel flow. Time series of instantaneous velocity field maps were used to identify mean and turbulent flow structures based on their spatial scales in the different regions. Even though regions exhibit varied Eulerian statistics, they show very similar eddy characteristics such as spinning rate and number density. The integral scale eddy structures show nearly a linear rate of increase in their rotation rate with increasing pore Reynolds number, indicating a linear decrease in their time scales. The convective velocities of these eddies are shown to reach an asymptotic limit at high pore Reynolds numbers, unique for each flow region. Detailed Eulerian statistics for the identified flow regions are presented and are used to predict mechanical dispersion through the use of estimated Lagrangian statistics. Contributions from each of the flow regions are presented and the recirculating regions are shown to contribute most to the overall longitudinal dispersion, whereas the tortuous channel regions contribute most to the transverse dispersion. The overall dispersion estimates agree well with global data in the limit of high Schmitt number.