A model representing the trends and variations in the Mauna Loa data is presented. The model uses an autoregressive process for low-frequency variations, which enables the residuals to pass two tests for whiteness. The analysis of the noise residuals provides a measure of uncertainity for estimates of nonfossil fuel emission of CO2 to the atmosphere. Finally, an adaptive Kalman filter is used to estimate measurement errors contained in the Mauna Loa data. The results obtained agree with previous estimates of the precision and accuracy of the measuring instruments. ¿American Geophysical Union 1987