Temperature Sensitivity of encoder

I have a weight/rate scale that measures pounds per hour, that has been giving me issues. I currently have it running with no material across it, but test weights hung on its loadcells to give me a constant rate(5500 lbs/hr).
The drive of the scale is run by a AB 1336 VFD...rather old, but so far has been trustworthy. The motor has an encoder installed on the aft end of the motor, an its signal is used for the rate on the scale head. (Schenk Accurate).
I have noticed a trend that in the cooler night hours (current lows about 27*F) that the scale will take on about +2,500 lbs/hr increase, and then on the warmer temp of the day (current highs about 60*F) that the scale goes back to reading the normal constant rate.
What could be driving this?
I thought that maybe the belting was expanding and constricting (31ft), but even if I adjust the belt jack screws very tight or loose, I cannot get that much deviation of the weight. The only other thing that would account for would be the pulses from the encoder.
Are standard hall effect quadrature encoders affected by temperatures. This is a desert here, so temperatures range from -20*F in wintertime to +110+ in summertime.

The encoder only gives you belt speed. How do you know that the problem is with the encoder, rather than with your strain gauge (or other device) that measures the actual weight?

The weigh scale is currently running empty, and I have set the VFD to a manual constant speed. In this case 38Hz. We also have a set of test weights for zero/tare of the scale, and when the scale is running 38Hz with the test weights on the loadcells (strain gauge), the scale should read a constant. It is currently showing about 5600 lbs/hr. As long as loadcells are not zeroed when damaged from a load-shock or some other incident, they tend to either work or not. There are 2 loadcells attached to a single roller on each side of the belting. I noticed when hanging one test weight that the scale jumped to 3000 lbs/hr, and when I hung the second, that it went to 6000 lbs/hr. While that is not a "great" test, it was sufficient to me to indicate that both are working. Even if one was slightly damaged, which I highly doubt due to the fact that the belting/scale is completely enclosed, the weight indicated would be fairly constant.

I attached a trend. The purple trend is PPH, and the aqua is the ambient temperature. The green line is the driven speed of the VFD.

Temperature based signal drift in shaft encoders was an issue in the past, and if your encoder is of the same general vintage as the 1336 (20+ years), you may have one that was not temperature compensated. Most new ones are now.

Might be time to think about upgrading your entire system. 1336s are no longer supported, no spare parts are available outside of Fleabay and other after-market sources where you are getting, at BEST, re-manufactured boards but at worst, someone else' junk. I keep getting clients telling me they don't want to spend the money to upgrade old VFDs, but when they DO fail, the downtime cost ends up far far exceeding whatever they thought they were saving. Then when I go in and replace the failed drive on a Saturday night, they pay me 3X my normal rate...

When I first took this job, many of the VFD's were either Woods drives, Powerflex 70 or 1336s. There were a handful of 700s. I slowly been migrating to 753s and 525s. I was seriously thinking of changng these 1336s a while back, and I have 753s on hand to do so, but I generally have little issue with them.
I talked with Schenck, and they think either a bad excitation voltage from the controller, or a possible moisture in the loadcell cable that shows up whether the moisture is frozen or not.
The controller gives current ft/sec rate bas d on the encoder, and may have operators log that through the night tme hours.