3 RESULTS
3.1 Model performance
Predicting suitable habitat for Francois’ langur with MaxEnt resulted in
high model performance. The AUC metric for the ten replicate runs was
0.973±0.006 (mean±SD), indicating a high level of predictive performance
in the current period. For future climate models of the Francois’ langur
distribution, the obtained AUC values in 2021-2040, 2041-2060,
2061-2080, 2081-2100 were 0.977, 0.977, 0.974, 0.975, respectively, all
of which were greater than the 0.9 AUC threshold, indicating good model
performance.
3.2 Environmental variable contributions
The Jackknife analysis showed the contribution of each variable and
their permutation importance in predicting the habitat of Francois’
langur. The current prediction was mainly affected by Dis_res (20.6%),
Bio6 (20.4%), Bio2 (15.6%), Bio19 (14.9%), Bio7 (7.4%) and Bio15
(5.1%) in decreasing percentage of their contribution towards the
prediction model. All other variables contributed less than 16% (Table
1). When considering the permutation importance, Bio19 (34%), Bio7
(15.3%), Dis_road (11.7%), Bio15 (10.7%), Bio2 (9.5%) showed the
highest percentage. The Jackknife test also revealed the importance of
the variables when used in isolation as well as when each variable was
excluded. The variables Bio6, Bio7, Bio2, Bio15, Bio19 registered the
highest gain when the variables were evaluated individually, whereas the
variables aspect and slop showed the negative gain (Figure 2).