riparias/gbif-alert

the forest for the trees

Closed this issue · 2 comments

timadriaens commented

Hi, in the ever ongoing thought process of trying to improve the early alert tool, I still think that for species with a lot of observations (so species that are not too rare), the alerts are kinda ineffective. An example is Ludwigia grandiflora for which you can easily get several hundreds of records in one month from various databases.

Thinking about a number of potential options to make the alerts more targeted:

  • Solution 1: Move from "observation alert" to "place-based alert" by grouping records based on a spatial unit to avoid repetition of the species being observed by several persons in the same location. Of course, this should partly be solved when people can issue alerts for a specific spatial delimiter. To go place-based or not could actually be a user option when tailoring the alerts since some users will probably want everything at least for some species.
  • Solution 2: increase the number of areas that can be used for setting alerts (going to smaller spatial scales)
  • Solution 3: allow users to make a polygon themselves
  • Solution 4: include a "time delimiter" in the alerts, for instance you want weekly alert emails but you only want in that alert email the observations of that last week, or even last day (on other words, focus more on recent reports)

I'll keep thinking...

niconoe commented

Thanks @timadriaens, that might deserve a discussion to understand the details better and plan appropriately!

I have the feeling that most of it could be solved with a good "custom polygon management" system, which is in place but currently being polished and improved.

We also have to see how to do this best given the timing constraints.

@damianooldoni, do you want to set a chat soon? Or you have another preference?

damianooldoni commented

Solution 3 has been deployed.

We are already speaking about solution 4 in #227. The user has to choose how old the occurrences can be, e.g. max 1 year, or 1 month. This allows us to mobilize more and more datasets in the future without having people disturbed by old occurrences they are not actually interested. Default value could be 1 year. See #227 for more info.

So, I think that 3 + 4 will solve this issue. And as 4 is already discussed in an issue, this one can be closed.