Effects on the suitability maps of orchid SDMs
Ecological suitability maps of target species generated by SDMs are commonly an invaluable reference for biodiversity conservation planning. The effects of human influences and physiological differences on the accuracy of orchid SDMs are one of the objects of our exploration. However, more vital we wonder about their effects on orchid suitability maps. In our study, this variation was evident (see Appendix S2.2 for the detailed predicted suitability area by different modeling strategies).
At first, considering the orchids lifeforms classification, we established the spatial comparison between the overall layer (modeled by all-data) and the total layer (generated by superposition of the epiphytic, mycoheterotrophic, and terrestrial orchids suitability map), which showed that there was an enormous variance between them. In the GLM models, the changed area was 39499 km2 (6.59% of the total study area). In the MaxEnt models, the changed area was 28018 km2 (4.68%). In the RF models, the changed area was 19974 km2 (3.33%). Moreover, it was usually shown that the overall modeling predicted suitability area was more than the total classification modeling results. To explore these discrepancy areas, we plotted the changed area map. Geopolitically, we counted the changed area for all counties separately in our study area. On this basis, we did the double-ranking to find the critical regions in the change. The results showed that the eastern, western, and southeastern regions of the study area were the vital changed regions, especially Zayü County, Longyang District, and Tengchong City, which always was at the forefront of considering the sequential and area weights no matter the modeling strategies (see Fig. 3).
Secondly, in exploring the effect of human interference on suitability maps, we obtained 12 groups of comparison variables based on different models with varying data sets (see Appendix S2.2). The results showed that models containing the HI factor predicted fewer suitability areas than without this factor (more than 66%), with only four groups showing the opposite, namely, G-m, M-e, R-e, and R-m. To further explore the effect of the HI factor on different lifeforms orchids, we calculated the changed area in each county for all models separately and did the same double-ranking exercise (see Appendix S3.2 for details) and plotted Fig. 4. This figure indicated that when the HI factor was included in the model variables, the change in the predicted area caused by it had the greatest impact on terrestrial orchids, followed by mycoheterotrophic orchids and epiphytic orchids. Region analysis displayed that Zayü, Tengchong, and Yangyuan counties were the most changed in all comparisons (see Appendix S2.3, we have mapped the differences of all model strategies).