An algorithm for converting the Beckman dissolved-oxygen probe variables with other data from a WHOI/Brown CTD/O2 system to oxygen concentration is presented. Improvements over earlier oxygen algorithms are inclusion of an oxygen current bias and time-lag correction using a nonlinear, least-squares calibration to the titrated oxygen bottle samples. The calibration technique uses a quasi-Newtonian minimization scheme available in scientific subroutine libraries such as the “IMSL” library. Oxygen probe parameters are found to be stable over several (as many as 25) stations and give typical rms errors in the deep water of 0.005 ml l−1 which is roughly the expected error for the bottle samples. Oxygen current biases can be equivalent to concentrations as large as 0.8 ml l−1. Typical oxygen lag correction values, associated with diffusion times through the probe membrane are in the range of 4–10 seconds. Examples from the North Pacific where the shallow oxygen minimum severely tests the algorithm and ... Abstract An algorithm for converting the Beckman dissolved-oxygen probe variables with other data from a WHOI/Brown CTD/O2 system to oxygen concentration is presented. Improvements over earlier oxygen algorithms are inclusion of an oxygen current bias and time-lag correction using a nonlinear, least-squares calibration to the titrated oxygen bottle samples. The calibration technique uses a quasi-Newtonian minimization scheme available in scientific subroutine libraries such as the “IMSL” library. Oxygen probe parameters are found to be stable over several (as many as 25) stations and give typical rms errors in the deep water of 0.005 ml l−1 which is roughly the expected error for the bottle samples. Oxygen current biases can be equivalent to concentrations as large as 0.8 ml l−1. Typical oxygen lag correction values, associated with diffusion times through the probe membrane are in the range of 4–10 seconds. Examples from the North Pacific where the shallow oxygen minimum severely tests the algorithm and ...