mtss omission rate

[Rogério]

To discuss the omission error of a model, do I use the omission result from training or testing? In the ENMEVAL article, it talks about the test result (?) for MTP and 10%, but is it for maxSSS? What is the expected value of the omission error for maxSSS?

Example for this model in SDMtune package:

Maximum training sensitivity plus specificity 

Cloglog value:   0.574666206

Fractional predicted area: 0.15

Training omission rate : 0.44

Test omission rate: 0.05

Maximum test sensitivity plus specificity

Cloglog value:   0.59

Fractional predicted area: 0.14

Training omission rate : 0.45

Test omission rate: 0.05

Which result is correct to say the model omission error? Training or testing?

And to generate the binary map, I must use the test cloglog, correct?