A water district in Southern California supplies drinking water with a combination of local shallow well groundwater and imported raw and treated water. The well is located 120 ft away from a seasonal creek. Based on limited temperature data, the California Department of Health Services (DHS) determined that the groundwater was under direct influence from the creek. In 2000, DHS required the district to install a water filtration plant. Before expending upwards of $1 million on a new treatment plant, the district elected to set up a sampling and testing protocol to determine more conclusively whether or not groundwater was under direct influence of surface water, and if so, what treatment credit could be given to river bank filtration. This paper reports a natural filtration study conducted in 2003 and 2005 for the district. The study, supported on well and creek monitoring, was devised to establish how rapidly well water quality responds to surface water quality changes, and to quantify river bank effectiveness in removing turbidity, total and fecal coliform (TC and FC), Giardia and Cryptosporidium. Devising a meaningful well and creek monitoring program presented challenges because of the seasonal nature of creek flow and well production (i.e. the wells start production early in the rainy season before creek flow starts, and well production continues for a period after the creek stops flowing). The monitoring program included different sampling intervals, tests and procedures for the various flow conditions. The monitoring program captured several significant rain events that resulted in high creek flows and associated increased creek water turbidity. Tracer test data showed a 30-hour groundwater travel time from the creek to the production wells. Based on creek and well sample data, natural filtration provides removals as high as 3-log turbidity, 4.4-log TC, and 2-log FC. The highest creek Cryptosporidium concentration during the sampling program was 0.4 organisms/L and Giardia 0.03 organisms/L. Well samples showed none detected for both organisms. Data evaluated included turbidity, coliform bacteria, protozoa, temperature, pH, electro-conductivity, and particle counts. Based on the results, well water is under the influence of creek water; however, natural filtration provides removal of turbidity, bacteria and protozoa. Particle count data show that 5 to 15 micron particles, the size of pathogenic protozoa, although partially removed in the creek bank, still appear in product water. Includes tables, figures.