Network structure in wet and dry seasons within each forest in relation to null models
For the dry forest, we found that most of the network metrics deviated from expected null distribution for the dry season, whereas there were no significant differences for the wet season during the El Niño year. However, some metrics showed deviation during the wet season of the normal year. During the dry season, the network was more compartmentalized (mean= 1.24; SD= 0.45; p-value= 0.00), less nested (weighted NODF) (mean= 15.29; SD= 2.95; p-value= 0.00), less connected (weighted connectance) (mean= 0.12; SD= 0.01; p-value= 0.05), and more modular than expected by the null models (mean= 0.50; SD= 0.03; p-value= 0.01). During the wet season of the normal year, the network was less nested (weighted NODF) (mean= 22.80; SD= 3.45; p-value= 0.01) and less connected (weighted connectance) (mean= 0.14; SD= 0.01; p-value= 0.00). Similar to the dry forest, we found that network metrics in the rainforest deviated from expected in only in the dry season. The network was less nested (weighted NODF) (mean= 12.73; SD= 3.03; p-value= 0.03) and more modular (mean= 0.37; SD= 0.04; p-value= 0.00) than expected by the null models. With regard to differences between seasons, we detected that the difference in robustness of bat species to plant extinctions was lower than expected by chance (p-value= 0.03) between seasons in the rainforest. In the dry forest, we found a higher niche overlap in the wet season of the normal year in comparison with the wet season of the El Niño year (p-value= 0.03) (Table 2).