This study discusses an improved technique for the assimilation of carbon monoxide retrievals from the Measurements of Pollution in the Troposphere (MOPITT) instrument in a chemistry-transport model using a suboptimal Kalman filter. An online bias estimator algorithm is employed to identify systematic biases in the model and account for them during the assimilation. Results suggest a large decline (both locally and globally) in the observation minus forecast diagnostics and provide insights about possible model deficiencies by enabling explicit examination of model biases.