I'm calculating warming trends in specific places and periods of time, that are related to the time a migratory bird spends in a stopover. But using different methods I got different results and I'm trying to figure out which is the most accurate. I know it will depend on location, so my sites are Southeast Canada and the Arctic. How much warming is possible in 40 years for example in a period covering only one month in spring, like May? What would be a good value to compare my results? spring temperature warming trend?

There's a lot of fluctuation in temperatures, so when averaging them I get no changes, including in the Arctic, which cannot be true. I looked at trends of the maximum temperatures and they are increasing, also investigated the anomalies, but I would like to have an advice.

  • $\begingroup$ What? "so my sites are Southeast Canada and the Arctic. How much warming is possible in 40 years for example in one month in spring" are you asking for 40 years or a month in spring? $\endgroup$
    – Eevee
    Commented Feb 27, 2018 at 13:19
  • 2
    $\begingroup$ Perhaps warming of May average temperature in 40 years? $\endgroup$
    – Communisty
    Commented Feb 27, 2018 at 13:57
  • $\begingroup$ Thank you for your suggestion @Communisty, do you think this also works if I have only 15 days and I want to estimate a warming trend? $\endgroup$
    – MSS
    Commented Feb 27, 2018 at 15:54
  • $\begingroup$ If you pick only a specific location and/or small portion of a year the overall trend can be very different from the global average. $\endgroup$
    – Communisty
    Commented Feb 27, 2018 at 16:01
  • $\begingroup$ A meteorologist would most like be able to give you a better answer. For your month of interest you look at either the average of the daily maxima over time, or the average of each daily maximum minus the daily minimum over time. $\endgroup$
    – Fred
    Commented Feb 27, 2018 at 16:30

1 Answer 1


It is hard to help unless you clarify what methods have you tried. However, I would suggest to explore reanalysis weather models, as they are the most accurate source of estimated weather conditions worldwide in the past.

A recommended reanalysis is ERA-interim, starting in 1979 so it gives you 39 years, if you really need 40 you can use JRA-55 that starts on 1955.

The KNMI Climate Explorer is a great way to retrieve and plot data from those models.

If you get in there you can set it to plot monthly mean temperatures for a given location. Here is an example of somewhere in SW Canada: enter image description here

That will produce a plot like this: enter image description here

Were you can click the "raw data" to get the raw values, and that will produce an output like this:

enter image description here

Where each row is a year and each column is a month. So you can pick the column of the month of your interest and compute the trend from that.

Does it makes sense?

  • $\begingroup$ Yes, I'm sorry for my lack of detail, I was probably too cautious to don't be sent to the statistics forum. Yes I'm using NARR data (validated with some weather stations), I averaged the daily temperatures for each area according to each of the periods I have (which are not true months, but based on dates according to the bird migration) and from that I want to estimate a trend for each of them. $\endgroup$
    – MSS
    Commented Feb 27, 2018 at 18:38
  • $\begingroup$ @MSS That sounds as a reasonable methodology. So, what do you mean with "But using different methods I got different results". Which different methods? How different are the results? $\endgroup$ Commented Feb 27, 2018 at 18:47
  • $\begingroup$ I used the Sen's slope, the Mann-Kendal test (adjusted for autocorrelated data), Zhang method (from package ZYP in R) models gls (from package nlme, not able to remove the autocorrelation of the residuals), ARIMAX and with all of them I get different results. Only ARIMAX was able to remove the correlation in the residuals. Which method would you trust more? $\endgroup$
    – MSS
    Commented Feb 27, 2018 at 19:03
  • $\begingroup$ The problem with ARIMAX was that when I looked at the overall period trend (76 days) this was negative for the Arctic! (not much, but negative) $\endgroup$
    – MSS
    Commented Feb 27, 2018 at 19:09
  • $\begingroup$ I'm not familiar with those methods. But it would be informative if you add to your question a plot with the source data (mean temp of the period of the year of interest versus year). If there is no significant trend, maybe there IS no trend, and it is pointless to try to use a statistical method to find the significance of the trend. Keep in mind that NARR data have uncertainties, and any trend that lay within that uncertainty won't have any significance. $\endgroup$ Commented Feb 27, 2018 at 19:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.