Fig. 7. (a) Temperature variation of sensor CP-4, showing the day-night temperature cycle (about 15 cycles per day), for the interval 1988.8.26 to 1988.8.28. The line segments connect successive l0 minute averages. (b) Clock phase based on integrating the synthesized clock rate inferred from the combination temperature data, for the same interval shown in panel (a). The wave at about 15 cycles per day shows the magnitude of the variation in phase due to the day-night temperature cycle. This figure is similar to Fig. 6a, with the line segments giving the difference between the temperature-inferred clock phase and the phase obtained from linear interpolation between successive clock measurements. The crosses at zero phase indicate the times of actual clock measurements.