Little is known about the successional dynamics of insects in the highly threatened tropical dry forest (TDF) ecosystem. For the first time, we studied the response of carabid beetles to vegetal succession and seasonality in this ecosystem in Colombia. Carabid beetles were collected from three TDF habitat types in two regions in Colombia: initial successional state (pasture), early succession, and intermediate succession (forest). The surveys were performed monthly for 13 months in one of the regions (Armero) and during two months, one in the dry and one in the wet season, in the other region (Cambao). A set of environmental variables were recorded per month at each site. Twenty-four carabid beetle species were collected during the study. Calosoma alternans and Megacephala affinis were the most abundant species, while most species were of low abundance. Forest and pasture beetle assemblages were distinct, while the early succession assemblage overlapped with these assemblages. Canopy cover, litter depth, and soil and air temperatures were important in structuring the assemblages. Even though seasonality did not affect the carabid beetle assemblage, individual species responded positively to the wet season. It is shown that early successional areas in TDF could potentially act as habitat corridors for species to recolonize forest areas, since these successional areas host a number of species that inhabit forests and pastures. Climatic variation, like the El Niño episode during this study, appears to affect the carabid beetle assemblage negatively, exasperating concerns of this already threatened tropical ecosystem.

Climatic variation, ENSO, environmental variables, natural recovery, seasonality

Tropical dry forest (TDF) is considered the most threatened tropical ecosystem in South America and Africa (Janzen 1988; Miles et al. 2006) and is simultaneously one of the lesser-known ecologically (Sánchez-Azofeifa et al. 2005). Recent studies have shown that TDF has high levels of beta diversity and endemism, which could be lost if inappropriate conservation strategies are applied (Banda-R et al. 2016). In the Americas, Colombia hosts extensive dry forest areas (Miles et al. 2006). However, with high anthropogenic pressure, these forests are historically characterized by high levels of fragmentation (Etter et al. 2008), which in turn results in a reduction in habitat connectivity (Balzotti et al. 2020), ecosystem functions and services (Balvanera et al. 2011). Today, only 8% of the original TDF in Colombia remains (Pizano et al. 2016). An example is TDF in the Valley of the Magdalena River in Colombia, which has experienced intensive agriculture resulting in a heterogeneous landscape (Fernández-Méndez et al. 2014), with most forest patches reduced to less than 30 ha in size (Pizano et al. 2016). This landscape mosaic is characterized by areas at various stages of succession, similar to other countries (Quesada et al. 2009), which generate potential sources for spontaneous succession (Prach 2003). Spontaneous succession is considered a cheap and rapid recovery tool (Prach and Pyšek 2001), which in the case of TDF could improve connectivity, at least for poorly dispersing insects species (Aparício et al. 2018). As such, the application of conservation strategies and ecological restoration is of paramount importance (Vargas and Ramírez 2014), especially since TDF succession is slow compared to wet tropical forest (Murphy and Lugo 1986), because water is a limiting factor when it comes to recovery (Fajardo et al. 2013).

Knowledge on succession, defined as species turnover with time (Walker and del Moral 2003), is necessary to understand changes experienced by communities due to anthropogenic disturbances as well as the activities necessary for their recovery (Prach 2003; Prach and Walker 2011). Since insects are a major component of terrestrial ecosystems (Samways 1994; Schowalter 2006; Scudder 2009), knowledge on how they respond to succession is paramount. Yet, tropical insects do not show consistent patterns during succession and appear to be highly dynamic; their response to succession depends on the community variable evaluated (species richness, diversity or abundance), the region, and type of disturbance (e.g., Hilt et al. 2006; Neves et al. 2010a; Hernández et al. 2014; Nyafwono et al. 2014). For TDF insects, successional pathways are also difficult to predict, since changes are related to seasonal variability (Neves et al. 2010a). Seasonal fluctuations in insects in TDF are well document (e.g., Noguera-Martínez et al. 2007; Pérez and Zaragoza 2016; Corona-López et al. 2017; Noguera et al. 2018), however, peaks in species richness are not clear when successional stages are included (e.g., Neves et al. 2010b). Additionally, taxa occupying different forest strata respond differently to succession; e.g., dung beetles and hypogaeic ant richness change with succession (Neves et al. 2010a; Marques et al. 2017), while arboreal and epigaeic ant do not (Neves et al. 2010b; Marques et al. 2017). The mechanisms of these differences are not completely understood, but may be related to specific resource use and abiotic requirements (Neves et al. 2014).

Carabidae is a large coleopteran family (ca. 34,000 species) (Bousquet 2010), extensively distributed and with high abundance (Larochelle and Larivière 2003), making them a prevalent model organism, especially in the temperate region (Koivula 2011). The well-documented information about its taxonomy and biology, and its response to environmental change helped in its wide used as bioindicators (Koivula 2011; Kotze et al. 2011). However, in the Neotropics, little is known about this group (Lucky et al. 2002; Rainio and Niemelä 2003; Martínez 2005; Maveety et al. 2011). One considerable hurdle is taxonomy with few identification keys and a great number of undescribed species (Lucky et al. 2002; Maveety et al. 2011; Erwin et al. 2015). Knowledge on carabid beetles in TDF is scarce, with most studies dealing with assemblage characterization (Arenas et al. 2013; Uribe and Vallejo 2013; Arenas and Ulloa-Chacón 2016). Baseline information on how carabid beetle assemblages respond to succession, and their seasonal dynamics in TDF, is urgently needed both from an ecological and conservation perspective. The present study is the first to investigate these issues in this highly threatened ecosystem.

The overall aims of our study were to investigate carabid beetle assemblage changes during early succession in TDF in Colombia, and their response to environmental variables along this successional process. Furthermore, given the strong seasonality experienced in this ecosystem, and that the data were collected during an El Niño event (Varotsos et al. 2016; Whitfield et al. 2019), we evaluate the response of this group to wet and dry periods. The El Niño/Southern Oscillation (ENSO) is a periodic climatic event that affects inter-annual rainfall regimes. In Colombia, it consists of a dry episode with a precipitation deficit and raising air temperatures (Poveda et al. 2000). It can produce severe droughts as has happened in Colombian TDF during the study period (Montealegre 2014).

We hypothesize that the carabid beetle assemblage in early successional TDF is speciose with high abundance compared to forest (e.g., Magura et al. 2015; Barber et al. 2017), due to the arrival of open-habitat and habitat generalist species (similar to temperate ground beetles) (e.g., Nagy et al. 2016). If the majority of carabid species of the TDF matrix are polyphagous predators (Lövei and Sunderland 1996) and habitat generalists (Rainio and Niemelä 2003) (as in other ecosystems), they will benefit from exploiting resources in the initial stages of TDF recovery, where resources are heterogeneous (Lebrija-Trejos et al. 2009). We also expect carabid beetles to respond to architectural attributes of the habitat, like vegetation cover and leaf litter (e.g., Molnár et al. 2001), since cover and litter influence environmental conditions (Facelli and Pickett 1991; Lebrija-Trejos et al. 2011), and are considered important in structuring carabid assemblages (Koivula et al. 1999; Antvogel and Bonn 2001). However, we expect that soil humidity will be a major factor that affects ground beetles in this ecosystem, because i) it has been showed to influence carabid assemblages (Niemelä et al. 1992; Kaizuka and Iwasa 2015; Fidan and Sirin 2016), ii) water is a structuring and limiting factor of dry forest (Maass and Burgos 2011), and iii) dry habitat carabid beetles synchronize their life cycle to optimal soil humidity conditions (Paarmann 1979; Paarmann et al. 1986). Other insects in TDF have shown to be influenced by soil/litter humidity, which are linked to precipitation (García et al. 2001; Rangel-Acosta and Martínez-Hernández 2017).

This study was performed during an El Niño event (2015/16), which means that the TDF ecosystem experienced harsh conditions, reflected by a considerable decrease in rainfall and an increase in air and soil temperatures (Montealegre 2014). We showed that the carabid beetle catches in both Armero and Cambao did not reach species saturation. Rarefaction/extrapolation curves approached an asymptote in pasture only, and pasture also showed a homogenous assemblage structure, distinct from forest. Early succession assemblage structure was highly heterogeneous, encompassing both forest and pasture, sharing 28% of species with both habitat types. Canopy cover, litter depth and soil and air temperatures were influential variables in structuring the carabid assemblage. Surprisingly, neither soil humidity nor seasonality affected assemblage structure significantly. Finally, C. alternans dominated pasture in both Armero and Cambao, while species of the genus Megacephala (M. affinis and M. cribrata) were more abundant in the early succession habitat type. None of the other species was collected abundantly enough to be analyzed individually.

The epigaeic carabid assemblage in tropical dry forest was species poor but with high dominance, like in other tropical carabid communities (Paarmann et al. 2002; Vieira et al. 2008; Rosero 2010). Only 12% of the species collected had more than 10 individuals, while 46% were singletons or doubletons. This low abundance was reflected in the rarefaction/extrapolation curves, which did not reach an asymptote, suggesting that epigaeic carabid beetle diversity in the TDF is certainly higher than presented here. Although rarity may be common in tropical ecosystems, the pattern we observed could appear as a consequence of an inappropriate sampling method and/or intensity (Magurran and Henderson 2011). Vennila and Rajagopal (1999) recommended more than 35 pitfall traps per site for quantitative studies in tropical agroforests, and Boetzl et al. (2018) showed that the use of guidance barriers could improve the efficiency of the catch. Furthermore, Liu et al. (2007) indicated that pitfall traps do not permit a complete inventory and recommended using light traps as a compliment. Many tropical carabid species live in the canopy (Erwin 1979), and will not be collected using pitfall trapping (see Kotze et al. 2011; Boetzl et al. 2018). To collect TDF carabid beetles more efficiently, we recommend increasing the number of traps and the length of the survey (one complete year of continuous trapping at minimum), use guidance barriers if possible, and include other methods to sample arboreal species. Additionally, high soil and air temperatures during the dry season (see Suppl. material 3: Table S3) result in the rapid evaporation of pitfall trap collecting fluid (Ariza 2016; pers. obs.), and we recommend using deeper pitfall traps with more collecting fluid. Apart from the method used here, the El Niño event, which coincided with our sampling, could be an important component to the depauperate carabid community in TDF. During a previous El Niño event (1996) in the Amazonian rainforest, carabid beetle richness decreased drastically compared with other periods (Lucky et al. 2002). In Mexican TDF, a decrease in Cantharidae beetle richness was also observed during the 1997/8 El Niño event (Pérez and Zaragoza 2016). Finally, the historic use of forest fragments and agricultural practices in pastures may explain the poor carabid community in this landscape. Harvey et al. (2008) indicated that beside severe fragmentation experienced in the TDF, contamination by agro-chemicals and illegal logging could be additional drivers of biodiversity loss. The degree to which these aspects affect carabid beetles in the region remains to be investigated.

The carabid beetle assemblage in the early succession habitat overlapped with assemblages in pasture and forest habitat types, a pattern not observed for dung beetles (Neves et al. 2010a) or arboreal ants (Neves et al. 2010b) in Brazilian TDF, where communities were more distinct between early succession and forest habitat types. Reasons for this discrepancy may be related to habitat complexity and the biology of these groups (Neves et al. 2010a, b). Dung beetles depend on ephemeral resources produced by larger animals that may be more sensitive to changes in the landscape (Hanski 1991), while ground-nesting ants in early successional TDF forage in trees, but do not do so in late successional stages (Neves et al. 2010b). For carabid beetles, early successional TDF sites may provide a mixed environment with elements from both pasture and forest, thus providing a heterogeneity of resources (Lebrija-Trejos et al. 2009) to be exploited by a subset of carabid beetle species. The lack of knowledge of the natural history of tropical species prevents us from identifying the habitat preferences of species found in the TDF early succession habitat type. Despite this, the little knowledge that do exist is reflected in the heterogeneous resource used by the carabid species collected from forest and early successional sites in Armero; Apenes prasinus Ball & Shpeley Clivina sp., and Galerita sp. are related to leaf-litter (Erwin 1991; Larochelle and Larivière 2003; Martínez 2005), while species shared between early succession and pasture (Athrostictus paganus (Dejean), M. affinis and Tetragonoderus sp.) prefer open areas and/or pasture (Larochelle and Larivière 2003; Vitolo 2004; Shpeley et al. 2017). Calosoma alternans (in Armero and Cambao) and M. affinis (Cambao) were present in all habitat types, suggesting that these species are habitat generalists. Indeed, C. alternans can be found in a wide variety of habitats and ecosystems (Gidaspow 1963), but also seems to prefer pastures (Bruschi 2010) as confirmed by our results. Nevertheless, it is difficult to attribute a particular habitat preference to species occurring in the early succession habitat type, partly because ground beetles readily disperse at the local level, even to suboptimal habitat (e.g., Niemelä and Halme 1992; Boetzl et al. 2016; Schneider et al. 2016; Knapp et al. 2019). Resources are heterogeneous and patchy, both at the fine and coarse scale (Wiens 1976; Pickett and Rogers 1997), thus carabid species of different preferences could occasionally occur in early successional stages, even if such habitat is suboptimal to them.

Soil and air temperatures were the only microclimatic variables that influenced the structuring of the carabid assemblage in our study. Carabid beetles, similar to other insects, are poikilothermic and sensitive to temperature (Beck 1983; Neven 2000; Bowler and Terblanche 2008), mainly during egg and larvae stages (Lövei and Sunderland 1996; Potter et al. 2009). Surprisingly, soil humidity did not affect the beetle assemblage, even though moisture is considered important in these dry forests (Balvanera and Aguirre 2006; Espinosa et al. 2011). During 2015/6, one of the strongest El Niño episodes occurred in Colombia (UNGRD 2016), producing a severe rainfall deficit that affected soil humidity negatively. This was reflected in our measurement of soil humidity, where differences between habitat and season were minimal. It is expected that in normal years, seasonality will result in more contrasting differences in soil humidity between open and forest habitats (Ceccon et al. 2006; Zhang et al. 2010). This could explain why variation in soil humidity did not affect carabid beetles significantly. Canopy cover and litter depth significantly influenced carabid assemblage structure in our study (see also Koivula et al. 1999; Antvogel and Bonn 2001; Taboada et al. 2008; Yu et al. 2008; Ogai and Kenta 2016). These structural variables can reflect and influence microclimatic conditions (Gardner 1991; Sanderson et al. 1995). In TDF, structural and environmental variables are related to succession but also depend on season, thus strong environmental gradients are not observed during the dry season (Lebrija-Trejos et al. 2009). For instance, TDF forests consists of deciduous trees that shed their leaves during the dry season, creating an open canopy (Murphy and Lugo 1986; Holbrook et al. 2009) thus minimizing differences between habitat types in terms of climatic gradients. On top of that, due to an extremely dry period because of the El Niño phenomenon (even during the wet season), microclimatic conditions (apart from soil and air temperature, see above) likely varied little between habitat types. Alternatively, conserving water is a challenger for small organisms in these harsh environments (Chown and Klok 2003), and as such, litter depth become an important environmental variable to these beetles in providing shelter (Koivula et al. 1999; Magura et al. 2005). Hopp et al. (2010) found that litter quantity was a better predictor of beetle assemblage recovery than soil humidity. Litter improve the environment in the soil (Facelli and Pickett 1991; Magura et al. 2004), offer habitat structure for organisms (Magura et al. 2000; Kalinkat et al. 2013) and supplies prey for carnivorous species (Guillemain et al. 1997).

Seasonality did not significantly influence the carabid beetle assemblage, even though numerous studies have shown seasonality to be important in dry forest beetles (e.g., Novais et al. 2016; Pérez and Zaragoza 2016; Rangel-Acosta and Martínez-Hernández 2017; Noguera et al. 2018). However, there are exceptions; for example dung beetle richness differences between wet and dry seasons in the Caatinga forest in Brazil were small (Medina and Lopes 2014). A reason for the lack of a seasonality signal in our data may, again, be due to a deficit in precipitation during the 2015/6 El Niño southern oscillation. Many carabid species diapause as an adaptation to harsh environmental conditions (Lövei and Sunderland 1996). Some observational studies have suggested that moisture could be a trigger for diapause development (Tauber et al. 1998; Hodek 2003). Either because of low humidity or environmental signals generated by the El Niño phenomenon, carabids species could display a lengthened diapause (during a drought event), thus resulting in the absence of seasonal peaks in their numbers (see Hanski 1987; Matsuo 2006). At the species level, C. alternans and the “rest of species” group did respond to the wet season. Some Calosoma species are associated with open habitats and dry soils (Larochelle and Larivière 2003), so a slight improvement of environmental conditions in the wet season during El Niño could be sufficient for this species to express seasonal peaks (e.g., Jacobs et al. 2011).

Our study showed that early successional areas in TDF have a prominent role in the conservation of carabid beetles since it can act as a temporal habitat for a number of species that occur in forest and pasture. The loss of connectivity between dry forest patches limits the dispersal of species (Kindlmann and Burel 2008; Balzotti et al. 2020). Early successional stages could act as habitat corridors for carabid beetles, including some stenotopic species (e.g., Noordijk et al. 2008, 2009, 2011; Eggers et al. 2010), and promote the recolonization of forest patches. Land-use intensification homogenizes carabid assemblages (Meng et al. 2012), which is also the case for pasture in our study. In Armero and Cambao, pasture was dominated by C. alternans, which seems well-adapted to dry soils. We showed the importance of restoration to the recovery of this ecosystem. We demonstrated that climatic variation, like the El Niño episode, impacts the abundances and species richness of TDF carabid beetles markedly, necessitating the call for long-term studies to evaluate recovery in this landscape.

We dedicate this paper to the late Terry Erwin, whose devotion to the study of life on earth and beetles in particular, inspired a generation of scientists, young and old, in the pursuit of understanding the creatures with whom we share this planet. Students of Biology and Agricultural Engineer from the Universidad del Tolima helped with sorting the samples. This research was partially supported by funding from the Universidad del Tolima. We want to thank workers of the Universidad del Tolima North Regional University Center, and from La Batalla farm at Cambao, Dr. Nelson Canal, and Pedro Galeano for helping in one way or another in carrying out the field collections and laboratory activities. We thank the reviewer and subject editor of ZooKeys for their constructive comments that have greatly improved this manuscript. The authors declare that there were no conflicts of interest in conducting the research.