A predictive, physically based model for pumping water from a well using air injection (air-lift pumping) was developed for the range of flow rates that we explored in a series of laboratory experiments. The goal was to determine the air flow rate required to pump a specific flow rate of water in a given well, designed for in-well air stripping of volatile organic compounds from an aquifer. The model was validated against original laboratory data as well as data from the literature. A laboratory air-lift system was constructed that consisted of a 70-foot-long (21-m-long) pipe, 5.5 inches (14 cm) inside diameter, in which an air line of 1.3 inches (3.3 cm) outside diameter was placed with its bottom at different elevations above the base of the long pipe. Experiments were conducted for different levels of submergence, with water-pumping rates ranging from 5 to 70 gallons/min (0.32--4.4 L/s), and air flow ranging from 7 to 38 standard cubic feet/min (0.2--1.1 m3 STP/min). The theoretical approach adopted in the model was based on an analysis of the system as a one-dimensional two-phase flow problem. The expression for the pressure gradient includes inertial energy terms, friction, and gas expansion versus elevation. Data analysis revealed that application of the usual drift-flux model to estimate the air void fraction is not adequate for the observed flow patterns: either slug or churn flow. We propose a modified drift-flux model that accurately predicts air-lift pumping requirements for a range of conditions representative of in-well air-stripping operations. ¿ American Geophysical Union 1996