Hello,
I am currently developing a SDM for an invasive plant (Ulex europaeus) in Chile. I have 60 occurence points from the GBIF database and additional 40 occurence points placed by myself based on google earth satellite images. I developed four models: three models with 60 occurence points and different bias sampling methods (random background, target group sampling and presence thinning) and one model with all 100 occurence points (and random background). My idea is that the google earth analysis offsets the sample bias. To proove that I need to compare my models. Therefore, I use the continous boyce index.
The question I have now: Is it possible to compare the evaluation metrics of two models built with partly different occurence localities?
Interesting research design. I would add a model based on the 40 presence points from your imagery interpretation in the comparison as well.
It is indeed interesting. Agree with Hein, it would be interesting to also add the model on the 40 presences as a comparison.
Thank you for your contribution. However, I wont include a model with the 40 presences because that would not help to answer my research question if adding occurence points based on satellite images can offset the sample bias.
- Badges
- Science method
- Similar topics
- Biological Science
- Ecology