I've got a time series which plots surface reflectance over time. Ideally surface reflectance is high in the winter and is low in the summer, and is fairly constant during both of those periods. I want to do several tests, but am not sure where to start:

  • First, for the time series, I want to test if there actually is a statistically significant high surface reflectance time period and a statistically significant low surface reflectance time period, without doing it visually/manually.
  • If there is this variation, I would like to break the data into sections of time based on high vs low surface reflectance. (ex Jan 20 - May 7 would be high surface reflectance, and May 16- Oct 12 would be low surface reflectance).

Are there any specific analytical techniques to do this relatively automatically?

Here is an example time series where visually it looks like there is strong variation:

time series example

  • $\begingroup$ You could use a paired t-test or a GLM to see whether the average low surface differs significantly from the average high surface reflectance. On a different note, your series (whether or not it is seen as one or two series), looks largely like white noise to me but you could apply a Ljung-Box test to find out for sure. $\endgroup$ – Digio Oct 24 '17 at 14:24
  • $\begingroup$ Hi, thank you for the help. The smaller variations on more of a daily/weekly scale are noise, are these the variations you are talking about that look like white noise? I mostly care about the larger trend of high reflectance time periods vs low reflectance time periods, do you think those look like white noise? $\endgroup$ – Ana Oct 25 '17 at 22:47
  • $\begingroup$ I do mean both, but it's hard to tell simply by looking at the series. Could you post pictures of the AC and PAC? $\endgroup$ – Digio Oct 26 '17 at 14:05

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.