Introduction

Transformation and fragmentation of natural landscapes, along with climate change, have among the most severe impacts on freshwater habitats (Maynou et al. 2017; Riad et al. 2020). Historic structure and quality of aquatic habitats are important for maintaining freshwater diversity (Nagy et al. 2019) and habitat loss and fragmentation lead to freshwater diversity decline and population loss (Cardoso et al. 2020; Lima et al. 2022). Habitat quality decreases when aquatic and marginal vegetation structure and water physiochemical properties are altered (Heino 2009; Hill et al. 2019). Landscape transformation also often leaves only narrow natural vegetation buffer areas around ponds, which interferes with habitat colonization dynamics of species that occupy these aquatic habitats (Harabiš 2017). These changes adversely affect both amphibiotic and aquatic insects (Thorp and Rogers 2014).

Odonata (suborders Anisoptera and Zygoptera) are amphibiotic insects, and aquatic habitat quality directly influences odonate occupancy (e.g., Kietzka et al. 2017; Briggs et al. 2019a), with breeding being limited to high quality freshwater habitats in most species. Yet, they cross the ecosystem boundary from aquatic to terrestrial as they mature (Knight et al. 2005). Odonate occupancy in terrestrial areas also increases during the dry season when surface water resources become scarce (Harabiš 2017; Harabis and Dolny 2011). Being highly mobile as adults, most odonate species occupy wide terrestrial areas surrounding freshwater habitats (Kalkman et al. 2007), with anisopterans overall having larger overall habitat ranges than zygopterans (Conrad et al. 1999).

Co-occurring species variably rely on secondary terrestrial areas that facilitate daily movements (i.e., frequent movement between water and the adjacent terrestrial matrix), and seasonal movements (i.e., movement between suitable habitats in a terrestrial matrix) (Raebel et al. 2012). Odonate movement in the longer-term is also determined by water permanency, with species that occupy unpredictable seasonal ponds and wetlands relying more on quality terrestrial spaces compared to species which occupy more permanent habitats such as lakes and rivers (Hof et al. 2006; Deacon et al. 2020).

An odonate’s ability to move among ponds and establish new populations is not exclusively dependent on their movement ability. It is also partly driven by their phylogenetic and behavioural limitations (McCauley 2006; Outomuro and Johansson 2019). Furthermore, successfully reaching previously unoccupied habitats does not necessarily translate into successful establishment, as the new habitat must also support the ecological and behavioral needs of odonates. For instance, patrolling odonates more regularly move between aquatic and terrestrial habitats as opposed to those species that spend most of their time perched close to the water’s edge (Bried and Ervin 2006). Patrolling odonate species with wide territories thus need larger areas to establish breeding populations. This means that functional connectivity among suitable habitats relies on combinations between flight performance, flight behaviour, competitor/mate density, and habitat preferences in odonates (Chin and Taylor 2009; French and McCauley 2019).

Given the great variation in habitat preference, mobility and behavioral traits among the two odonate suborders, as well as among the species within them, odonates show promise as indicators of both freshwater habitat and terrestrial matrix quality (Sahlén 2006; Nagy et al. 2019). Furthermore, it is recognized that the ‘pondscape’ (i.e., landscape configuration, spatial abundance of ponds, and connectivity among them) may be as important as hydrography, vegetation characteristics, or water chemistry in maintaining historic population levels and assemblage diversity (Kadoya et al. 2008; Hassall et al. 2016).

Here, we focus on odonates that occupy permanent and seasonal fishless ponds in the coastal sub-tropical region of KwaZulu-Natal, South Africa, where much of the landscape is transformed and fragmented by plantation forestry. Our overall aim was to determine the importance of the pondscape (i.e., the ponds relative to the terrestrial context) in driving odonate species richness, abundance, assemblage composition and functional diversity patterns in a fragmented landscape. Specifically, we (1) determined whether pond size/natural terrestrial patch size interaction affects overall odonate species richness, abundance, assemblage composition, and functional diversity relative to pond characteristics, and (2) determined whether anisopterans and zygopterans respond differently to landscape context. Although local pond characteristics are important for odonates (Briggs et al. 2019a), we hypothesized that local pond characteristics and size of natural terrestrial patches in which ponds are located share equal weight in determining odonate occupancy in the transformed landscape investigated here, with large ponds in large natural patches expected to have the highest number of species. Due to the difference in area requirements between the two suborders of Odonata (Conrad et al. 1999), we also hypothesized that anisopterans predominantly occupy ponds within large natural patches of terrestrial vegetation, while zygopterans preferentially occupy ponds within smaller natural patches.

Methods

Study area

The study was conducted in northeast KwaZulu-Natal, a summer rainfall area falling in the Maputaland-Pondoland-Albany Biodiversity Hotspot (Myers et al. 2000). Natural vegetation is dominated by Maputaland Coastal Belt and Maputaland Wooded Grassland vegetation types (Mucina et al. 2006). Sample ponds were selected on two neighbouring Eucalyptus spp. plantation estates (owned and managed by SiyaQhubeka Forestry) on the eastern portion of the study area, and in iSimangaliso Wetland Park, a World Heritage Site on the western portion of the study area (Fig. 1). Plantation estate ponds occur in networks of large conservation corridors of remnant natural patches among plantation compartments, with varying sizes and degrees of connectivity.

Fig. 1
figure 1

Spatial layout of sampled odonate ponds in northeast KwaZulu-Natal, South Africa. Pale yellow indicates natural terrestrial patches and bright yellow indicates Eucalyptus plantation forestry compartments. Blue indicates small ponds in small patches, green indicates small ponds in large patches, red indicates large ponds in small patches and gray indicates large ponds in large patches

Full size image

Ponds were fishless and used by large mammals roaming the study area, as the boundary between the plantation estates and the conservation area is unfenced. Ponds in the area are greatly influenced by El Niño Southern Oscillation events in the longer term, and their sizes and permanency are highly variable from one year to the next due to seasonal summer rainfall. These permanent and temporary ponds are obligate habitats for at least 29 Odonata species in the greater iSimangaliso Wetland Park region, and supplementary habitats for roughly 45 species, collectively supporting ~ 86% of all odonates in the region (Hart et al. 2014; Deacon et al. 2021).

Study design

We selected 27 sample ponds ranging in size and landscape context (Fig. 1). Ponds with a surface area smaller than 1999 m2 were classified as a ‘small pond’ while those with a surface area larger than 2000 m2 were classified as a ‘large pond’. As pond margins are highly dynamic in the study area, we considered pond size as the area covered by surface water at the time of sampling. The size of the natural patch of grassland and thicket vegetation in which each pond occurred was classified, and natural patch size was considered as the uninterrupted surface area covered by grassland and thicket vegetation, and excluding those areas covered by plantation trees or indigenous forest, where shading of pond habitats reduce odonate occupancy (Osborn and Samways 1996). Natural patches with a size smaller than 0.39 km2 were classified as a ‘small patch’, while natural patches with a size larger than 0.4 km2 were classified as a ‘large patch’.

Following this classification, eight ponds were classified as ‘small pond in small patch’, six ponds were classified as ‘small pond in large patch’, four ponds were classified as ‘large pond in small patch’, and nine ponds were classified as ‘large pond in large patch’ (Online Resource 1). The connectedness of each pond relative to surrounding terrestrial matrix was also determined. If a pond was totally surrounded by plantations, it was considered ‘disconnected’, but if a pond had shared grassland/thicket with a neighboring pond (i.e., ponds connected through a conservation corridor), it was considered ‘connected’. Spatial information was extracted using QGIS version 3.26 (Quantum GIS Development Team 2022), using a 10 m resolution Sentinel2 spatial dataset, captured on the 9th of February 2022 (Copernicus Sentinel2 Data 2022).

Field sampling for odonates was conducted from the 7th to the 27th of February 2022, during the peak unimodal flight season in the region. Each of the 27 sample ponds was visited in a random order on cloudless, windless days, between 09h30 and 16h00. As many odonate individuals, especially males, remain faithful to their territories (Jooste et al. 2020), each pond was sampled once to avoid double counts of individuals over a limited sampling period. At each sample pond, two observers walked along opposite pond margins for 45 min (regardless of pond size to equalize sampling effort), recording adult odonate species and abundance, considering both males and females. For small ponds where observers were closer than 30 m from one another, each record was vocally announced to ensure that the same individuals was not counted twice. If an individual could not be identified in the field, it was caught with an insect net, and identified ex situ.

Water pH, conductivity (µS/cm), temperature (ºC) and dissolved oxygen concentration (mg/L) were recorded at five randomly selected points along pond margins, using a Hanna HI98129 and a Hanna HI9142 portable water probe. A measure of turbidity was excluded from the study, as ponds in the areas are frequently used by roaming large mammals, and turbidity fluctuates dramatically over short periods. To estimate vegetation characteristics, five quadrats of 4 m2 were randomly selected along pond margins. Quadrats were equally aligned over aquatic and terrestrial space to represent fuzzy pond margins. Within quadrats, we recorded % grasses cover, % reeds cover, % herbaceous cover, and % bare ground. All obtained values relating to pond chemistry and vegetation characteristics were averaged for each study pond.

Data analysis

A sampling site-based species accumulation curve for observed odonate species richness was constructed using the ‘vegan’ package for R version 3.6 to evaluate sampling adequacy (Oksanen et al. 2020; R Core Team 2020). To supplement the species accumulation curve, we calculated the Chao2, Jackknife2 and ICE species estimator metrics using the ‘fossil’ package for R (Vavrek 2011). The species accumulation curve for observed odonate species richness reached a near asymptote and observed species richness neared the values obtained for the three calculated species estimator metrics (Online Resource 2). Consequently, we used raw species richness values as response variable in subsequent univariate analyses.

We calculated functional richness (i.e., the distribution of traits in the biological assemblage) and functional diversity (i.e., the distribution and abundance of species in trait space, simultaneously; represented by Rao’s Quadratic Entropy), using the ‘FD’ package for R (Laliberté and Legendre 2010). Functional richness and diversity were calculated for each study pond, and separately for the overall odonate assemblage, the anisopteran assemblage, and the zygopteran assemblage. In the case of Zygoptera, ten sample ponds were excluded, as no zygopterans were recorded from these ponds. In all cases, a Cailliez correction was applied to calculate each metric, and both functional richness and diversity were standardized to values ranging between 0 and 1. We considered ten traits from four broad categories: (1) habitat preference traits (relative number of biotopes occupied by adults, habitat preference (lentic vs. lentic/lotic), latitudinal range extent, elevation range extent, and Dragonfly Biotic Index score (DBI; Samways and Simaika 2016), (2) mobility traits (average body length, average fore wing length, and wing-to-body length ratio), (3) behavioural traits (categorized as ‘darter’, ‘flutterer’, ‘glider’, ‘hawker’, and ‘percher’), and (4) phenology traits (number of months active as adults (including overwintering). All odonate trait information (except DBI scores) were extracted from a previous continental-scale study (Deacon et al. 2020).

Eleven predictor variables were considered initially to test the relationships between environmental variables and species richness, abundance, functional richness, functional diversity and assemblage composition across all surveyed ponds. These variables were: water temperature, pH, conductivity, dissolved oxygen concentration, vegetation height, % grass cover, % reed cover, % herbaceous cover, % bare ground, pond category (large pond in large patch vs. small pond in large patch vs. large pond in small patch vs. small pond in small patch), and connectivity (binary; connected to nearby ponds vs. disconnected from nearby ponds).

Each response variable was tested for normality, and Mantel tests were performed to test all response variables for spatial autocorrelation. None of the investigated response variables were spatially autocorrelated, and no random spatial terms were considered in the subsequent modelling procedure. For the six normally distributed response variables (Odonata species richness, Anisoptera species richness, Anisoptera functional richness, Odonata functional diversity, Anisoptera functional diversity, and Zygoptera functional diversity), we used linear models (LM) to test the relationships between these variables and the environment. For the six non-normally response variables, we used generalized linear models (GLM) with Poisson distributions (in the case of Odonata species abundance, Anisoptera species abundance, and Zygoptera species richness and abundance), and Gamma distributions (in the case of Odonata functional richness and Zygoptera functional richness) to test for significant effects.

We performed model averaging to test the significant effects of predictor variables on each response variable (Bartoń 2020). Prior to model averaging, multi-collinearity (Variance Inflation Factor, VIF) was determined for all global models, using the ‘car’ package for R (Fox and Weisberg 2019). Only predictor variables with a VIF < 3 were considered during the final modelling procedure to reduce the statistical effects of strong correlations among predictor variables. Connectivity (strongly correlated with pond category; large ponds in large natural patches were mostly connected, while small ponds in small natural patches were mostly disconnected) and dissolved oxygen concentration (strongly correlated with water temperature) were the only variables to consistently have a VIF > 3 and was excluded from the model averaging procedure for all modelling scenarios. All global models built with the predictor variables with a VIF < 3 for each modelling scenario were tested for overdispersion, and candidate models were ranked from lowest to highest Akaike’s Information Criterion (AICc) and ΔAICc values, using the ‘MuMIn’ package for R (Bartoń 2020). Only the subsets of models where ΔAICc < 4 were considered for model averaging, and a predictor was considered as significant when the upper and lower 2.5% confidence intervals did not include 0, and if the relative variable importance under each modelling scenario was > 50% (i.e., the variable contributed significantly to explaining total variance under the modelling scenario; Bartoń 2020).

To determine odonate assemblage composition variation among study ponds, we performed a model-based analysis of multivariate abundance data in ‘mvabund’ for R (Wang et al. 2020). Only variables with a VIF < 3 (i.e., all environmental variables except pond connectivity and dissolved oxygen concentration) were considered for the multivariate component of the study. To supplement model-based analyses of odonate assemblage variation among pond categories, we used fourth corner analysis to analyze the direct relationships between pond category and the set of traits used throughout the study, implemented through ‘mvabund’. Fourth corner analysis was conducted for all odonates combined, supplying the modelling scenario with an abundance matrix, predictor variable matrix (containing only pond categories), and a trait matrix (containing all behavioural, phenological, habitat preference, and mobility traits). To account for species with low abundances, we applied a LASSO penalty to set association coefficients to zero when their statistical effects were small in the fourth corner model. To visually represent associations between traits and predictor variables, we constructed a fourth corner plot using the ‘lattice’ package for R (Sarkar 2008).

Results

Among 27 study ponds, we recorded ten lentic species and 21 lentic/lotic species across 20 Odonata genera (Online Resource 3). Twenty-five species were anisopterans, and six species were zygopterans. Palpopleura lucia was the most abundant species among anisopterans, with 328 individuals recorded, and Ceriagrion glabrum was the most abundant species among zygopterans, with 268 individuals recorded. Lestes plagiatus, Trithemis annulata, T. dorsalis, and Urothemis assignata had the lowest abundances overall, each with one individual recorded among all studied ponds.

Overall species richness increased with decreasing herbaceous cover along pond margins (Table 1). Overall abundance showed an increase with decreased water pH, water temperature, and herbaceous cover, while also showing an increase with increasing overall vegetation cover along pond margins. Pond size and patch size ranked among the top predictor variables, and odonate abundance was highest in large ponds in small patches compared to large ponds in large patches (z = 4.93, p < 0.001), small ponds in large patches (z = -9.46, p < 0.001), and small ponds in small patches (z =-8.91, p < 0.001; Fig. 2a). Large ponds in large patches had higher odonate abundance compared to small ponds in large patches (z = -8.67, p < 0.001) and small ponds in small patches (z = -6.18, p < 0.001). Small ponds in both small and large patches had similar odonate abundances. For odonate functional traits, highest functional richness was associated with increased water pH, while highest functional diversity was associated with decreased water pH levels (Table 2).

Table 1 Model averaging results showing model estimates, standard errors, 95% confidence intervals and relative model importance of predictor variables against odonate species richness and abundance in northeast KwaZulu-Natal, South Africa. Significant variables are indicated in cursive, and marginally significant variables are indicated in roman text
Full size table
Fig. 2
figure 2

Pairwise differences among pond categories relative to diversity metrics for odonates overall (a), anisopterans (b), and zygopterans (c-e) in northeast KwaZulu-Natal, South Africa. Dissimilar lower-case letters indicate statistically significant differences in grouping medians

Full size image
Table 2 Model averaging results showing model estimates, standard errors, 95% confidence intervals and relative model importance of predictor variables against odonate functional richness and functional diversity in northeast KwaZulu-Natal, South Africa. Significant variables are indicated in cursive, and marginally significant variables are indicated in roman text
Full size table

For Anisoptera, highest species richness was associated with decreasing herbaceous cover along pond margins (Table 1). Anisopteran abundance showed an increase with decreasing water pH, herbaceous cover and reeds cover. Anisopteran abundance was also highest in large ponds in small patches compared to large ponds in large patches (z = 2.76, p < 0.05), small ponds in large patches (z = -6.85, p < 0.001), and small ponds in small patches (z = -6.51, p < 0.001; Fig. 2b). Large ponds in large patches had higher anisopteran abundance compared to small ponds in large patches (z = -6.54, p < 0.001) and small ponds in small patches (z = -4.67, p < 0.001). Small ponds in small and large patches had equal anisopteran abundances. Although water pH, water temperature, and herbaceous cover along pond margins were important drivers of anisopteran functional richness, none of these predictor variables were significant (Table 2). No predictor variables were identified as important for anisopteran functional diversity.

For Zygoptera, pond category was the only variable selected in species richness models, although the effects were not significant. However, zygopteran abundance increased with decreasing water temperature, as well as with decreasing herbaceous and reeds cover along pond margins. Zygopteran abundance increased with an increase in overall vegetation cover (dominated by grasses) although decreased with increasing overall vegetation height. Zygopteran abundance was also highest in large ponds in small patches compared to large ponds in large patches (z = 2.65, p < 0.05) and small ponds in large patches (z = 5.12, p < 0.001; Fig. 2c). Large ponds in large patches had higher zygopteran abundance compared to small ponds in large patches (z = -4.72, p < 0.001). Small ponds in small patches had similar zygopteran abundance compared to large ponds in small and large patches. Zygopteran functional richness was higher in large ponds in small patches compared to small ponds in large patches (z = 2.54, p < 0.05; Fig. 2d). Small ponds in large patches had lower zygopteran functional richness compared to large ponds in large patches (z = 2.55, p < 0.05) and small ponds in small patches (z = -2.56, p < 0.05). Small ponds in small patches had equal zygopteran functional richness compared to large ponds in large and small patches. Although reed cover along pond margins was important for zygopteran functional richness, this relationship was not significant. Zygopteran functional diversity was lowest in small ponds in large patches compared to small ponds in small patches (z = 2.41, p < 0.05), and large ponds in both large (z = 2.36, p < 0.05) and small patches (z = 2.22, p < 0.05; Fig. 2e).

Out of the eleven pre-selected environmental predictors, changes in odonate assemblage composition across the 27 sample ponds was only driven by pond and patch size interaction (likelihood ratio test value (LRT) = 171.4, p < 0.01). Large ponds in large patches had an overall distinct odonate assemblage compared to small ponds in large patches (LRT = 67.3, p < 0.05) and small ponds in small patches (LRT = 81.4, p < 0.05), but shared assemblages with large ponds in small patches (Fig. 3). Large ponds in small patches shared assemblages with small ponds in large and small patches. The assemblages of small ponds in small patches could also not be distinguished from those of small ponds in large patches.

Fig. 3
figure 3

Constrained ordination results indicating odonate assemblage composition (dis)similarity among investigated pond categories in northeast KwaZulu-Natal, South Africa. Grouping circles represent model standard errors

Full size image

Fourth corner results indicated that pond category occupancy was significantly related to odonate traits (LRT = 74.78, p < 0.05), and was predominantly driven by mobility traits (Fig. 4). Odonate species with relatively short wings mostly occupied small and large ponds in small patches, while odonate species with long wings occupied small and large ponds in large patches. These trends were reversed for odonate body size, and species with large bodies predominantly occupied small and large ponds in small patches, while small-bodied odonates mostly occupied small and large ponds in large patches. Similarly, odonates with high wing-to-body length ratios (i.e., species with high relative dispersal ability) were associated with small and large ponds in small patches, while species with low wing-to-body length ratios (i.e., species with low relative dispersal ability) were associated with small and large ponds in large patches. Odonate traits related to behaviour, phenology, and habitat preference had overall weak relationships with pond category occupancy.

Fig. 4
figure 4

Fourth corner results indicating associations between pond category and traits of odonates in northeast KwaZulu-Natal, South Africa. Blue indicates negative associations, red indicates positive associations, and clear indicates no associations between traits and pond categories. Values indicated on the index bar are standardized coefficients of all trait-environment interactions obtained through GLM-Lasso modelling

Full size image

Discussion

The various pond categories supported comparable odonate species richness levels, though their abundance was greatly affected by landscape fragmentation, as well as by water chemistry and vegetation characteristics. These findings lend partial support to our first hypothesis that landscape configuration and pond characteristics share equal weight in determining odonate diversity patterns. Overall odonate functional richness and functional diversity were not driven by pond category, but rather by water pH gradients. Pond category was also an important driver of odonate assemblage variation, while also closely related to mobility traits of odonates. At the suborder level, although the drivers of anisopteran functional richness and diversity were unclear, for zygopterans they were driven by pond category and vegetation characteristics. These findings partially support our second hypothesis that anisopterans have high affinity to large terrestrial areas, while zygopterans have high affinity to small terrestrial areas. However, our results suggest that zygopterans (with overall lower movement ability) are more sensitive to the effects of landscape fragmentation, while anisopterans (with overall higher movement ability) are more resilient and select suitable habitats based on water chemistry and vegetation characteristics, as was found for other freshwater habitat types (Remsburg et al. 2008; Kietzka et al. 2017).

Relative importance of pond spatial characteristics

Various pond categories supported similar numbers of Odonata species. Yet, pond category was an important driver of individual abundance, with large ponds supporting highest overall odonate abundance, as well as for the two suborders individually. Small ponds are important steppingstone habitats for highly mobile lentic species (Hassall 2014; Maynou et al. 2017), though large ponds may support species often absent from small ponds, while also maintaining larger population sizes (Oertli et al. 2002). This is related to large ponds being more detectable to dispersing individuals, remaining in the landscape for longer in some cases, and having more niche space available, so reducing inter- and intraspecific competition for resources (Kadoya et al. 2004; Ruggiero et al. 2008).

Interestingly however, large ponds in small natural patches had overall higher odonate abundance compared to large ponds in large natural patches. While ponds in small natural patches had less terrestrial habitat available, small natural patches surrounded by tall plantation trees likely provide shelter against strong weather events such as wind and unpredictable downpours, while still receiving the benefits of sunning for most of the day (Harabiš et al. 2013; Timofeev 2016). Due to the windbreak caused by surrounding Eucalyptus spp. plantation trees, these small habitat islands may also better retain warmth and humidity, providing attractive microclimates to the odonates which occupy these habitats (du Toit et al. 2017). This means that the individuals that occupy ponds in small natural patches may be able to remain active around their territories for longer periods of the day.

On the other hand, temperatures in coastal KwaZulu-Natal may often exceed the optimal temperature for odonate activity and rainfall is unpredictable, and some species (e.g., Trithemis spp.) may take shelter during the wettest or warmest part of the day (Damm et al. 2009). Being closely surrounded by tall trees may enable shade tolerant species to avoid overheating or excessively wet conditions, but without the expense of leaving their territories entirely (May 1991; but see the relative importance of vegetation below).

Further emphasizing the overall importance of pond size relative to natural patch size in a fragmented landscape, results also indicated that each pond category supported dissimilar diversity in odonate assemblages. While a variety of ponds are important from an assemblage structuring perspective, overall odonate and anisopteran functional richness and diversity here were not significantly related to pond categories. The overall lentic odonate assemblages in the region are dominated by anisopterans (Samways and Simaika 2016; Deacon et al. 2020), with these findings suggesting that pond variety has a greater influence on the distribution of species as opposed to the distribution of traits across the landscape (Le Gall et al. 2018). However, in the case of zygopterans, functional richness and diversity was highest in large ponds, especially those in small natural grassland patches, suggesting that zygopteran assemblages rely more on large ponds rather than large remnant patches to remain in the landscape (Raebel et al. 2012).

Among the range of odonate traits investigated here, physical traits related to dispersal ability were most important relative to pond category occupancy. Long-winged species predominantly occupied ponds in large patches, while short-winged species largely occupied ponds in small patches. This seems to be related to the area requirements of species in that short-winged species are better adapted to move around in small areas, and consequently have smaller area requirements to persist in any given habitat. Conversely, long-winged species often traverse the landscape and require larger activity areas to persist (Wakeling and Ellington 1997; Rundle et al. 2007).

However, we found that large odonates and those with greater wing-to-body length ratios (as a proxy for greater overall dispersal ability) were more common in small natural patches compared to small species and those with smaller wing-to-body length ratios. This suggests that species with greater overall dispersal ability are better able to discover ponds in small remnant patches, and perhaps move around more freely across fragmented landscapes (Jenkins et al. 2007). This means species with low dispersal ability seldom reach small remnant patches, and are at higher risk when present (Vanschoenwinkel et al. 2013). It is possible that the presence of large-bodied zygopterans with relatively low dispersal ability (e.g., Lestes pallidus and L. plagiatus), and small-bodied anisopterans with relatively high dispersal ability (e.g., Diplacodes lefebvrii, Brachythemis leucosticta, Palpopleura portia, and P. lucia) in our dataset could have amplified these results.

Relative importance of vegetation characteristics and other environmental variables

Relatively higher species richness and abundance levels were associated with higher overall vegetation cover, but lower herbaceous and reeds cover along pond margins. Although vegetation along pond margins is important for odonates, tall and dense vegetation causes shading for long and continuous periods of the day (Remsburg et al. 2008). With some exceptions, most odonates in the region are sun-loving and avoid shaded areas (Samways and Simaika 2016). Ponds with intermediate-height grass cover along margins attract them, as these ponds provide many microhabitats where dragonflies and damselflies can perch, hold territories and hunt, while remaining relatively open and overall free of shading close to pond edges (Briggs et al. 2019a).

Higher abundance of both suborders, as well as high overall functional diversity, were both associated with relatively acidic ponds. With Eucalyptus spp. leaf litter leading to slightly more acidic soils and water (Soumare et al. 2015), and odonates being overall sensitive to changes in water pH (da Rocha et al. 2016; Jooste et al. 2020), these findings suggest that regionally common species are better able to occupy acidic ponds among Eucalyptus plantations. This is further emphasized by high overall functional diversity being associated with ponds with close-to-neutral pH.

High anisopteran abundance here was also associated with relatively cooler ponds. Although slightly warmer water shortens developmental time for most lentic insect immature stages (Suhling et al. 2015), water temperatures above optimal conditions significantly affect immature stage development and leads to lower oxygen levels and high mortality rates (Castillo-Pérez et al. 2022). Adult odonates are able to assess water conditions, and relatively sensitive species may avoid sub-optimally warm aquatic habitats (Kietzka et al. 2017).

Implications for conservation

From a terrestrial perspective, an effective strategy to mitigate landscape fragmentation is to instigate networks of conservation corridors of remnant natural vegetation among plantation compartments (Samways and Pryke 2016). While insects were not originally conservation targets in the area, they have benefitted greatly from conservation methods aimed at large mammals, birds, and vegetation types (Pryke and Samways 2012; Pryke et al. 2015). From an aquatic perspective, plantation forestry operations are usually performed around large freshwater habitats, meaning that large and permanent ponds are often left intact throughout conservation corridor networks. This is related to freshwater habitats in general maintaining hydrological processes, while also being important water sources for vertebrates that roam the area (Samways and Pryke 2016). Small and highly dynamic ponds, on the other hand, are not well protected due to their unpredictable formation and low detectability during the dry season. This means that small ponds are at risk, particularly through infilling (due to obscurity during dry seasons) or oversight to leaving sufficient buffer areas of natural vegetation around small ponds.

Our results emphasize previous findings that conservation of the pondscape is essential for maintaining regional odonate taxonomic and functional diversity patterns (e.g., Hill et al. 2018; Briggs et al. 2019b). Conserving whole pondscapes accounts for variation in local characteristics (e.g., physicochemical gradients among ponds, changes in vegetation characteristics among ponds), to which odonates (Janssen et al. 2018; Perron et al. 2021) and other aquatic insects (Oertli and Parris 2019) respond. However, pond conservation should be considered in the terrestrial context, with natural patch size and terrestrial matrix context also being important drivers of odonate occupancy. These are important considerations for management of landscape transformation, as amphibiotic insects such as odonates track suitable habitats across terrestrial space, especially in the case of wandering species and during pre-reproductive phases (Kadoya et al. 2008).

Both odonate suborders, and the species within each suborder, differ in their terrestrial area size requirements. Although mostly occupied by good dispersers, ponds in small remnant patches are important, as they contribute greatly to regional odonate diversity patterns. Despite ponds in small patches being somewhat isolated, they are also important habitats for those species seeking refuge against unfavorable conditions (Timofeev 2016). Ponds in small patches can be even more attractive to a range of odonate species where there is improved connectivity, and weaker dispersers have greater opportunity to move across the landscape. On the other hand, ponds in relatively large remnant patches may be subject to less artificial disturbance and are important habitats for odonate source populations (Oertli et al. 2002).

Pondscape variety is important, with pondscapes supporting a range of ponds in both small and large patches having high conservation value for regional odonate diversity (Briggs et al. 2019b). In fragmented landscapes, ponds across a dynamic range of sizes in varying sizes of remnant natural grasslands patches should be preserved to encourage full odonate assemblages. By ensuring that a range of pond/natural patch size combinations are represented in pondscape conservation at the landscape level, a range of environmental conditions and microclimates will also be preserved, so enriching regional odonate diversity despite the total landscape being transformed and/or fragmented by plantation forestry.

While our focus was on adult odonates, they are good sentinel organisms for conservation in a variety of landscape types, informing how co-occurring amphibiotic insect taxa are affected by large-scale transformation (Bried et al. 2007). We encourage future studies to incorporate a range of other amphibiotic insect taxa, and other regions where habitat fragmentation threatens biodiversity associated with freshwater habitats, to improve conservation efforts for maintaining regional aquatic and amphibiotic insect diversity patterns.