Understanding and modeling the release of manure constituents is important for estimating the risk of microbial contamination of water sources that are used for irrigation, recreation, aquaculture, and other human activities.  The objective of this work was to compare contents of E. coli, enterococci, total coliforms, fecal coliforms, and total organic carbon in manure to initial concentrations of these components in runoff and infiltration water during rainfall simulation using partitioning boxes.  The boxes were designed to have manure applied on a mesh-covered frame (70cm x 70cm), and both runoff and infiltration were collected from troughs above and below the mesh, respectively.  Initial concentrations in runoff and infiltration were on average 0.7 and 1.1 log units less for manure from grain-fed and grass-fed cattle, respectively, than microbial concentrations in the manure liquid phase. The synchronous concentrations of microorganisms in runoff and infiltration did not differ significantly.  Regressions of 'time - log concentration' for all microorganisms did not differ significantly in most cases within each trial.  The microbial release kinetics followed the piece-wise, log-linear shape, beginning with a precipitous, approximately 2 unit, log-linear drop in concentration of released constituents during the first four to eight minutes of rainfall simulation, followed by a much slower log-linear release.  First-phase release rates of E. coli, fecal coliforms, and total coliforms were significantly lower in manure from grass-fed cattle than from grain-fed cattle.  The nutrient release kinetics displayed exponential release for both manure types.  This study shows that manure release kinetic models currently oversimplify the release description and can provide erroneous release estimates when extrapolating their parameters.  The results will be used for improvement of microbial fate and transport models that are critical for risk assessment of microbial contamination in the environment.