Low confidence heart rate at the end of the night

[Ella36]

Hi, 

I am a researcher analysing Fitbit Charge 5 heart rate data during sleep. I have noticed a very consistent increase in heart rate during the end of the sleep period across hundreds of nights of data, which coincides with low confidence values (0, 1, or 2). Does anyone have an explanation for why this might happen? Or any information about how heart rate data is generated?

For reference, here is an example of my own testing data (note that the bottom red line shows the onset of sleep and the top red line, the offset):

Thank you!!

[Gondwana]

Did you use the Web API to obtain these data? (You can't have used the SDK if you've got a Charge device.)