Virus removal from groundwater by soil passage often appears to be much higher during the first few metres due to the presence of more favorable sites for attachment than thereafter. A model is presented which interprets virus removal as a function of collision efficiencies alpha(beta) and alpha(lambda), inactivation rate coefficient mu(t) and rate parameter gamma. Initial high removal is determined by ab, which decreases exponentially at a rate g to a constant base removal rate that is determined by alpha(lambda) and mu(t). A hypothetical worst case was simulated to calculate the travel distance and time required for 9 log10 protection against virus contamination of groundwater wells in anoxic sandy aquifers. Unfavorable conditions for attachment were assumed. Virus was constantly leaking from a sewage pipe lying at the groundwater table. Mixing reduced virus concentration by 3.1 to 4.0 log10. For an additional 5.0 to 5.9 l log10 protection against virus contamination by attachment and inactivation, residence times of about three to seven times longer than the current guideline of 60 days are needed, depending on abstraction rates, aquifer thickness and grain size of the sand.