Global bee decline and its impact on pollination in agricultural and natural ecosystems have attracted public attention. However, more data are needed to show their generality and intensity in different ecosystems, especially in the tropics. For centuries, the tropical dry forest (TDF) of Costa Rica underwent intense deforestation, but in the last four decades, a large part of this forest has entered a recovery process. Using data of orchid bees generated by Janzen et al. (1982) in TDF in the Santa Rosa sector of the Guanacaste Conservation Area (ACG), we posed a general question: What changes have occurred in diversity, composition, and seasonality of the euglossine bee community in the TDF during the last 40 years and how are these changes related to the current recovery of this forest. We sampled euglossine bees during 2018-2019 using methods similar to those applied previously. To characterize the response of euglossine bees to habitat loss, we extended the sampling to pastures adjacent to the protected area. With the loss (n=4) and gain (n=3) of bee species, we did not find significant changes in overall species richness between now and 40 years ago. However, the composition of the euglossine community in the protected area in 1977 was more similar to that found in current pastures than to the current community in forests, where the presence of forest-dependent species has been favored. It is possible that TDF regeneration in Santa Rosa has led to changes in the composition of the community of these bees.

We carried out this study in the Santa Rosa sector of the Guanacaste Conservation Area and adjacent areas in the northwest part of the province of Guanacaste, Costa Rica (10° 45’ -11° 00’N, 85° 30’ - 85° 45’W). We selected three sites located at 300 m a.s.l. on the Santa Rosa Plateau sampled by Janzen et al. (1982), which were forest fragments surrounded by pastures in 1977-1979 and that today are embedded in a continuous forest matrix. To determine the effect of forest cover on the euglossine bee community and the species better adapted to altered habitats, three sampling sites outside the protected area were included in this study. Sampling was performed during the years 2018-2019 (four sampling periods per year, see below). We sampled during the same months as in the previous study (Janzen et al., 1982): in the middle of the dry season (March) and beginning, middle, and end of the rainy season (June, August, and December, respectively). ). We maintained the same sampling effort used by Janzen et al. (1982), one day at each sampling site (three in forests and three in pastures). We sampled from 0700 h to 1100 h using the same five chemical attractants: cineol, eugenol, methyl cinnamate (solid dissolved in 95% ethanol), benzyl acetate, and methyl salicylate. Unlike the previous study, we identified most species in the field with a 40x, 25 mm hand lens and then released them. To avoid recounting the same individual, we marked bees on the wing. More details are in the manuscript.