No correlation between temperature and phase residuals, cold or warm weather. Strong (sometimes direct, sometimes inverse) correlation between temperature and pseudorange residuals during the winter months; shows some variation during summer months, but nowhere near as strong as wintertime.

Hypothesis: When there is a large 'clock jump' between days at NRC2, below-zero temperatures somehow affected the equipment. The results of equipment sensitivity to changes in temperature resulted in a time-varying, non-zero behavior of the pseudorange residuals, which introduced enough 'noise' into the daily solutions to introduce a significant clock bias at midnight between days.

With respect to phase residuals, Elosegui found no correlation between residuals and time, and that holds true in the few phase examples I've examined so far. However, in the case of NRC1/2 that is not true for pseudorange residuals, which seem to show a heavily time-varying behavior in the winter. No trend/correlation seems to exist for between residuals and the elevation and azimuth angles.

The charts below attempt to establish a correlation between the clock jumps at day boundaries and residual differences (residual patterns and/or differences between daily average residuals).
Plot descriptions:

• Daily residuals - for each day, there are actually two plots to represent the residuals:
  1. 2-D plot of pseudorange residuals, versus time
  2. pseudocolor plot of gridded average residuals (over the course of the entire day)
• Differences - these plots are created by taking the difference between the gridded pseudorange residuals on consecutive days. For example, a plot of 'May05 minus May06' would be the gridded residuals from May 6th subtracted from the gridded residuals from May 5th. These plots loosely establish a direct correlation of large daily differences in residuals to a large 'clock jump'. The clock bias between days is given in the title of each plot.

Winter - NRC2

Winter - ONSA

I took a brief look at Onsala's pseudorange residuals behavior, since the January to June plot of pseudorange residual sigmas was so steady. As expected, there is no time-varying behavior to the residuals for a random wintertime week at ONSA. ONSA is in Sweden and therefore should be pretty cold, but the temperatures may be moderated by a coastal setting, or a significant equipment difference.

Summer - NRC2