Residual and cumulative frequency graphs for determining the optimal number of fitted exponentials. (Left) In this example, untreated all zones data are shown (squares) fitted to an increasing number of exponentials (dashed lines). (Right) For the corresponding fit, we show the residuals (fitted values minus data values). A larger amplitude on the y axis means a poorer fit, and it can be seen that the amplitudes decrease as the number of fitted exponentials increases. However, a clear deviation from the data is seen (arrows on left graphs) and the systematic noise (arrows on right graphs) remains until three exponentials are used. In the latter case, the maximum noise after fitting is <0.5%.