A practical, inexpensive computer model for estimating the level of blood carboxyhemoglobin (percent COHb) as a function of time for measured carbon monoxide concentrations (ppm CO) was developed from published studies on the assimilation of CO into the blood of human subjects. The model was designed to consider realistically the dynamic characteristics of urban CO concentrations measured continuously at air monitoring stations, and it was applied to a year's CO data measured at the San Jose, CA. air monitoring station (8760 hourly values). Accounting to the model, the National Ambient Air Quality Standards (NAAQS) for CO sometimes were violated in San Jose without exceeding 2% COHb, as well as the converse: 2% COHb was exceeded without violating the standards. The model's estimated COHb levels also provided an advance warning of impending violations of the 8-hr CO NAAQS, and analysis of the model's response to CO 'spikes' suggests that averaging periods as short as 10 or 15 minutes are necessary to serve completely the dynamic characteristics of ambient CO monitoring data. These findings suggest that the margin of safety included in the current CO NAAQS would not be the same if the actual time variation of measured CO concentrations is taken into account.