TMondain
commented
2 years ago I really like this idea as creating the score after the models is relatively quickly done. We should chat about this.
Maybe a measure Vs predicted comparison
We could have something fairly simple like if the mean decide score of the newly visited locations was greater than the 75th percentile of decide scores for that species then rerun the model? Might have to fiddle with what percentile to use but could work... I'm also sure there's a better way of doing it though.
I'm sure there's lots to think about, but one thing is that the score in the app is across the whole taxa but then we'd be looking at a species by species basis. Could it be that the models for some species never get run if we go down this route? I guess we could force it to run once every so often regardless of new observations.
This is a random Friday thought that I'd thought I'd put here before I forget it.
How about we write a script that can be run per species with the species' existing SDM and any new data that has been recorded since the SDM was produced. This script then assesses how much difference the new data would make the SDM, if it assesses the new records as significant contributions then it triggers something to rerun the SDM for that species. This would mean that we're running SDMs as and when they are needed, rather than running all of them as set intervals. Essential prioritising models given limited computational capacity.
I don't really know how this script would work. Maybe a measure Vs predicted comparison. Eg. Lots of records of butterflies in location X, what does the SDM for location X? If they weren't expected mayb the model needs to be rerun.
It could be a nice milestone to notify recorders, "your records have provided significant new information that we need to rerun our models, well done!". This is obvs a nice to have but