Control variables on pseudo-absence

[DelphineChabanne]

Hi,

I wonder if anyone in this community may be able to help with this. I used pseudo-absence which were defined as the cells with no observation of the target species and having the highest effort (systematic surveys while searching for the targeted species). The number of cells selected was equal to the number of presence cells for equal weight. When doing this way (which has been the most used for my target species (marine mammal)), I ended up with groups of pseudo-absence cells (instead of having them all over the study area). This is due to the study design (overlapping zigzag transects). If I were to use all available pseudo-absence cells (all with effort of search) or a random selection of those (with 10 repeats and regardless of the effort), the models (any algorithms) ended up with bad fitted and much lower than the threshold we would use to select the models for the ensemble modelling. When doing as initially mentioned (selection of the cells with highest effort), the prediction seemed biased toward the areas where the groups of pseudo-absence cells were selected from.

My question here is: if I use all pseudo-absence and I have information with effort and perhaps some probability of distance to presence cells for each of the pseudo-absence cells, how could I incorporate this information so the models could use it as a control variable?

Many thanks in advance for any tips you may have.

Delphine