Microorganism removal efficiencies in deep bed filters vary with time and depth in the filter bed as the filter collects particles. Improved knowledge of such dynamics is relevant for the design, operation and microbial risk assessment of filtration processes for drinking water treatment. Here we report on a high-resolution spatio-temporal characterization of virus and bacteria removal in a pilot-scale dual-media filter, operated in contact-filtration mode. Microorganisms investigated were bacteriophage Salmonella typhimurium 28B (plaque assay, n=154)), fRNA phage MS2 (plaque assay/RT-qPCR, n=87) and E. coli (Colilert-18, n=73). Microscopic and macroscopic filtration models were used to investigate and characterize the removal dynamics. Results show that ripening/breakthrough fronts for turbidity, viruses and E. coli migrated in a wave-like manner across the depth of the filter. Virus removal improved continuously throughout the filter cycle and viruses broke through almost simultaneously with turbidity. Ripening for E. coli took longer than ripening for turbidity, but the bacteria broke through before turbidity breakthrough. Instantaneous log-removal peaked at 3.2, 3.0 and 4.5 for 28B, MS2 and E. coli, respectively. However, true average log-removal during the period of stable effluent turbidity was significantly lower at 2.5, 2.3 and 3.6, respectively. Peak observed filter coefficients λ were higher than predicted by ideal filtration theory. This study demonstrates the importance of carefully designed sampling regimes when characterizing microorganism removal efficiencies of deep bed filters.