There were only about 100 orca observations over the past few decades, but I could resist taking them through Ben's tutorials and seeing what the output looks like. Attached a screenshot here.
Some notes...
- Only the boosted regression tree gave a reasonable model. Others were quite bad. MaxEnt had an accuracy of 0.08. (!)
- I only tuned using the true skill statistic (tss_max) because I suspected accuracy would skew predictions away from presence because of so few observations. Plus it sped things up a bit.
- I'm still working through some errors, but the basic mapping worked... I wouldn't put too much into these maps given the low number of sightings, but I bet it's picking up on some realistic features, such as avoiding shallow waters and hanging out along the continental slope.
There were only about 100 orca observations over the past few decades, but I could resist taking them through Ben's tutorials and seeing what the output looks like. Attached a screenshot here.
Some notes...