tobac
tobac copied to clipboard
tobac-project
different values for n_min_threshold dependent on thresholds
After some discussions, @freemansw1 and I came to the conclusion that would be useful and rather easy to allow different minimum areas for different thresholds in the multithreshold feature detection. More specifically, that would mean to make n_min_threshold a dict/list that has to have the same number of elements as the provided threshold values (see suggestion @freemansw1 below).
The idea behind it is that more extreme threshold values often are inherently associated with a smaller areas (e.g. the convective core/ the region with highest rain rates/ the region with coldest cloud temperatures of a cloud system).
I think these are both excellent ideas. For 1, we could make n_min_threshold in feature_detection_multithreshold_timestep optionally a dict or list, with each threshold value assigned to a different number of minimum points. I think this would be a relatively easy change, although we would want to make sure to do some input validation to ensure that we have a n_min_threshold for each threshold value.
For 2, this should be a relatively easy change to add and would be of value especially for those with large datasets. I can start working on a PR to add that once we eventually get #57 (and #60 ) merged.
Great, that sounds good! Agree about your point on input validation. We could allow either a single value in which case n_min_threshold will the same for all threshold values or a list/dict in which case it has to match the number of given thresholds. I can have a look at this.