Hello everyone, I‘m modeling distribution maps for about 1000 bird species to make the distribution pattern for conservation planning. But it is difficult for me to select a suitable threshold in MaxEnt software. There are 9 thresholds calculated by MaxEnt software for users to select.
My question is: which one is more better for my research?
1) Balance training omission, predicted area and threshold value;
2) Maximum test sensitivity plus specificity;
3) Equal test sensitivity and specificity;
4) Equate entropy of thresholded and original distributions;
5) Average value of all 9 threshold;
Many thanks for any suggestion.
Hi Ruocheng,
Since your research is about conservation, I suggest you to use number (2): Maximum test sensitivity plus specificity. This threshold maximises the cases where the model erroneously assigns unsuitable habitat (true negative) and misses suitable habitat (false positive). This approach is very common when using Maxent and other SDMs (see papers attached).
Good luck.
Manuel
Article Environmental suitability of a highly fragmented and heterog...
Article Mammal defaunation as surrogate of trophic cascades in hotspot
Article Predicting species distribution at range margins: Testing th...
Article Predicting seasonal habitat suitability for the critically e...
I agree with the previous answers. But all your models (for every species) will have their one specific value for the threshold (Maximum test sensitivity plus specificity). You just have to extract that information. I am assuming you are using R.
I think the reference provided by Yoan Fourcade is incorrect. I believe the paper by C. Liu that compares multiple threshold methods (N=12) is:
Liu, C., P. M. Berry, T. P. Dawson, and R. G. Pearson. 2005. Selecting thresholds of occurrence in the prediction of species distributions. Ecography 28:385–393. Article Selecting Thresholds of Occurrence in the Prediction of Spec...
- Badges
- Science topic
- Similar topics
- Filing