Abundance, occurrence and time series: long-term monitoring of social insects in a tropical rainforest

The magnitude of worldwide insect decline is hotly debated, with multiple examples of stable or increasing insect populations. In addition, time series data for tropical insects are scarce, notably in rainforests where insect di-versity is poorly known but reaches a peak. Despite social insects (ants, termites, bees and allies) being key organisms in these habitats, long-term monitoring data for these groups are crucially lacking. For many of these insects, the difficulty of locating nests in rainforests could be one reason. In this context, species occurrence in samples is often used as a surrogate for abundance to evaluate species distribution in space/time, but the loss of information is difficult to assess. In a tropical rainforest in Panama, we employed various sampling methods to examine the time series of seven insect assemblages with differing degrees of sociality: termite workers and soldiers, termite alates, bess beetles, litter ant workers, army ant alates, orchid bees, and nocturnal sweat bees. We used five community variables and six models related to occurrence and abundance, to test for significant trends in assemblages over a 13-year period (2009-2021). While assemblages of bess beetles increased, those of termite workers and soldiers, army ant alates, and orchid bees remained relatively stable. Termite alate, litter ant worker, and nocturnal bee assemblages showed signs of decline, demonstrating the need for monitoring distinct assemblages. Significant trends in generalized additive mixed models (GAMM) were observed in three out of five assemblages that could be tested. Our study indicates that trends in assemblages may be more informatively reported with abundance than with occurrence. We recommend (1) monitoring multiple insect assemblages as ecological indicators responsible for diverse ecosystem services; and (2) reporting species richness, changes in faunal composition, occurrence, and, when possible, using time-explicit analyses (such as GAMM models) for evaluating population trends over time.