ToonVanDaele / trias-test

test repository ofr trias
0 stars 0 forks source link

Best covariate for occupancy to compensate research effort bias #12

Open damianooldoni opened 4 years ago

damianooldoni commented 4 years ago

In formulas for GAM on occupancy ((lines 118-119 in 5c_method_GAM_short.R)) I see you use s(ncobs) instead of s(cobs). I have still doubts about this. Although cobs is often similar to ncobs (e.g. in 70's- 80's), it does not always hold true: e.g. birds, the last years number of observations are very often higher than number of cells.

Moreover, in case of occupancy we have to correct a spatial research effort, i.e. number of square kilometers. Using number of observations instead of number of cells we risk to introduce ia bias instead of removing it, as we can have a class where native species are massively observed but in a very limited part of Belgium, in other words:. high cobs and low ncobs.

What do you think?

damianooldoni commented 4 years ago

I implemented idea I explained above in pipeline: See https://trias-project.github.io/indicators/occurrence_indicators_modelling.html, section 3.3.2.