Our bear article (Italy) and one of our elephant papers (Tanzania) contain a method to establish such radius that may apply to your case as well. Hope it helps.

There are many different strategies to select pseudo-absences (you can find many reviews on this topic). In my opinion, instead of arbitrary selecting a buffer area (or radius of circles), I would suggest to randomly select pseudo-absences from your study area (but if you have precise estimation of species movement ability, then you can use this radius). That being said, I strongly suggest to choose a proper extent of study area.

Hein Van Gils thank you for the help!

Could you possibly share a pdf version of the papers? I can't access them online...

Francesco Garzon - Is your study species closely associated with the littoral zone (inshore fishes)? Beaches (turtles)? Offshore / nearshore transitions (some sharks)? Or shoreline associated terrestrial species?

Not knowing what you are studying, makes it a little hard to suggest solutions. Assuming you are discussing a marine species, and you mention that most occurrences for telemetry data are near the shore, you could simply buffer a terrestrial layer a reasonable distance that encompasses most of your presence points.

If you are studying a terrestrial species, I would likely keep all land areas, unless you know for sure that your species never / rarely occurs greater than X metres from the high tide line.

Ian