function to count extreme events #556
Conversation
| #for following events (t loop) or longer events (w loop) | ||
| daily_data[{dim: slice(t, t+w)}] = ( | ||
| daily_data[{dim: slice(t, t+w)}].where(~new_event) | ||
| ) |
There was a problem hiding this comment.
Just to confirm that I understood the method... this means that, for example, if the threshold is 100 mm over 3 days and I reach that accumulation on days [1st, 2nd, 3rd] (which would be detected as an extreme), and assuming the event continues, then the next extreme should be detected on [4th, 5th, 6th] instead of [2nd, 3rd, 4th]?
If above is correct, how is [2nd, 3rd, 4th] classified — which also has 100 mm in 3 days but isn't detected because days 2 and 3 are masked? Would this be considered a 'false extreme alert' and therefore discarded?
There was a problem hiding this comment.
That's correct, [2nd, 3rd, 4th] is not an extreme event because [1st, 2nd, 3rd]. Similarly, if all the rain fell in 1st, then [1st, 2nd, 3rd] is not an event, but [2nd, 3rd, 4th] could be. That's why I had to use loops rather than rolling or something alike.
I tried to think of a vectorial solution but couldn't. Probably because there is a "memory" aspect of the analysis: once a day is part of an extreme event, it can't be part of another one.