Does anyone have any thoughts on the use of these methods and when both might be used to explore trends in the data.

I think most people who sample monthly, quarterly or biannually would jump straight into using the Seasonal Mann-Kendall test but I think both tests can provide useful information.

For example, I used the Mann-Kendall test to evaluate trends in monthly groundwater levels. This revealed information about the behavior of trends that would not be apparent using the seasonal kendall test. Because each month is evaluated separately you can assess and compare changes across each season. In my case this revealed summer groundwater levels are declining faster than winter. I also observed a far greater number of trend detection over summer months.

The pattern in trend detection and rates of decline are almost certainly linked to increased irrigation over time during summer periods. However, evaluating monthly trends separately reduces the power of the tests, this is where the collective assessment of the seasons is useful for testing the statistical significance of the trend. Therefore I think much can be added by performing both tests where applicable.

I'd appreciate your thoughts?

  • $\begingroup$ Shouldn't the Seasonal Kendall test be applied only when trends occur in the same direction in all seasons, since its purpose is the detection of an overall trend across seasons? According to vsp.pnnl.gov/help/vsample/Design_Trend_Seasonal_Kendall.htm, the Mann Kendall test "should be used when seasonality is not expected to be present or when trends occur in different directions (up or down) in different seasons." In the latter case, one clearly should not be performing both tests! $\endgroup$ – Isabella Ghement Apr 11 '19 at 2:21
  • $\begingroup$ Yes that is what I understand too. The SKT is only applicable if trends are in the same direction. But even if they are in the same direction, applying the Mann-Kendall test to each month can still reveal temporal processes that you wouldn't otherwise detect using the seasonal kendall test. If you apply the Mann-kendall test separately to each season, your data set will be reduced which may affect whether you detect a statistically significant trend or not, but where results are significant, this may help understand what's driving these changes. Individually, each trend maybe down but not ... $\endgroup$ – Simon Apr 11 '19 at 4:18
  • $\begingroup$ ... significant, however collectively (using the seasonal kendall test) they maybe highly significant. I'm sure someone will enlighten us. Thanks for posting your comment $\endgroup$ – Simon Apr 11 '19 at 4:18
  • $\begingroup$ "I'd appreciate your thoughts" is too vague and broad as a question, and your post seems mostly to consist of you saying what you think, which appears to be more of staking out a position than you asking a specific question. This isn't a discussion-board. $\endgroup$ – Glen_b -Reinstate Monica Apr 11 '19 at 12:59