The study area was located in the Lousã mountains (40°3'N, -8°15'W, central Portugal; Fig. 1a), covering a surface of 170 km² with an elevation ranging between 100 and 1,205 m a.s.l. The climate in the area is classified as Mediterranean. However, the geographic location (i.e., close to the Atlantic Ocean) and topography (i.e., steep hills and valleys) in the Lousã mountains lead to higher precipitation levels than those that are typical for classic Mediterranean regions (Mora and Vieira 2020). The annual mean temperature is 16 °C, the annual mean minimum and maximum temperatures are 10 °C and 23 °C, respectively, and the annual precipitation varies between 650 and 1000 mm (climatic data from 2000 to 2022, from the nearest meteorological station (IPMA 2025). In recent decades, this area has undergone remarkable changes due to the natural regrowth of forests on abandoned agricultural lands, the establishment of monoculture plantations, forest fires, and the spread of non-native Acacia species, including A. dealbata and A. melanoxylon. A. dealbata has bipinnate compound leaves, and contrasts morphologically with A. melanoxylon, which possesses thick phyllodes (i.e., widened and flattened petiole with the appearance of a leaf blade) (Bentham 1875; Murphy and Maslin 2023). These Acacia species were originally planted along roadsides for ornamental reasons and erosion prevention but have expanded beyond their original locations. In most of the Lousã mountains area, A. dealbata is the dominant species, but A. melanoxylon is also common. Importantly, A. dealbata is much more widespread, possibly due to its invasion history and higher propagule pressure, as it was more widely used (Montesinos et al. 2016; da Silva et al. 2019; Garcia et al. 2023).

Figure 1.

A. Location of the study area within the Mediterranean Basin and Portugal; B. Sample area, showing the grid with 25 regular square segments and dashed lines showing the contour lines (i.e., connecting points of equal elevation) in 100 m intervals, and C. Exact location of the 25 sampling points (red dots) within the 25 km² grid.

In spring 2023, a 25 km² grid was overlaid on the study area and further divided into 25 individual square segments, each measuring 1 km² (Fig. 1b, c). The grid covered an invasion gradient of A. dealbata, and A. melanoxylon to a lower extent, across a diverse range of vegetation types that included: conifer stands with monoculture plantations of maritime pine (Pinus pinaster Aiton; native), European black pine (Pinus nigra J.F.Arnold; alien, non-invasive), Scots pine (Pinus sylvestris L.; alien, non-invasive), Lawson cypress (Chamaecyparis lawsoniana (A.Murray bis) Parl.; alien, non-invasive) and Douglas fir (Pseudotsuga menziesii (Mirb.) Franco; alien, non-invasive); planted birch stands (Betula pubescens Ehrh.; alien, non-invasive), broadleaf natural forests of Orocantabrian oak (Quercus orocantabrica Rivas Mart., Penas, T.E.Díaz & Llamas; native), chestnut (Castanea sativa Mill.; alien, non-invasive) and strawberry tree (Arbutus unedo L.; native); and Mediterranean shrublands with evergreen shrubs such as bramble broom (Genista tridentata L.; native), bristle-leaved heath (Erica umbellata L.; native) and gorse (Ulex minor Roth; native).

A total of 25 sampling points were established within the grid, with one point allocated to each 1 km² square; in each of these, data on vegetation structure, litter and soil quality, and springtail functional structure were gathered. Sampling points were selected as near as possible to the center of each sampling square, with small adjustments due to accessibility (Fig. 1c).

The coordinates, information on topographic variables (elevation, slope, and aspect; LNEG 2025) and Acacia invasion status (invaded or non-invaded) of each sampling point were recorded (Table 1). Of the 25 sampling points, nine were invaded (with a cover of Acacia, including A. dealbata and/or A. melanoxylon, of at least 5%; Table 1) and the remaining 16 showed no signs of invasion.

At each sampling point, three quadrats of varying dimensions were used to sample vascular plants, arranged in a nested structure from smallest to largest. Specifically, a 2 × 2 m quadrat was used for herbs, a 3 × 3 m quadrat for shrubs, and a 10 × 10 m quadrat for trees (also used to assess Acacia invasion status). The total number of quadrats of each size was 25, with the smaller quadrats nested within the larger ones. The cover of each species was visually estimated in 5% intervals, and when lower than that it was assessed in shorter intervals. We followed the plant nomenclature established by POWO (2025). The cover of each vegetation layer (trees, shrubs and herbs) was measured at each sampling point as the sum of the cover of all plant species within each growth form. These covers can exceed 100% due to overlapping species within the same growth form. Furthermore, we calculated the overall plant species richness at each sampling point by adding up all plant species observed. To quantify the extent of Acacia invasion, we calculated a continuous Acacia invasion gradient as the proportion of Acacia relative to total tree abundance by dividing the combined cover of Acacia species (A. dealbata + A. melanoxylon) by the total tree cover at each sampling point. Due to the remarkably lower proportion of A. melanoxylon compared to A. dealbata, and its limited distribution to specific sampling points (n = 3, Table 1), we combined the two Acacia species for our analyses, thereby ensuring sufficient statistical robustness. This calculated proportion provided insight into Acacia’s dominance relative to other tree species, helping to assess its invasion status within the fragmented forest landscape. Furthermore, by examining this proportion, we gained a better understanding of how varying levels of Acacia invasion might influence the understory (i.e., herb and shrub cover).

Since the litter C/N ratio is a reliable predictor of litter turnover (Delgado-Baquerizo et al. 2015) and soil organic carbon (SOC) gives information about the ability of soils to store carbon (Lal 2004), they were used as litter and soil quality indicators, respectively. Two quadrats of 25 × 25 cm were used at each sampling point to collect litter down to the mineral soil (i.e., the entire decomposing forest floor, excluding the underlying mineral soil; Gillespie et al. 2021), totaling 50 samples. This collection method focused on the litter layer specifically, which is composed of decomposing plant material, rather than the soil organic matter that is integrated with the mineral soil. The two 25 × 25 cm quadrats per sampling plot were established at the exact coordinates (latitude and longitude) listed in Table 1. Litter samples were oven-dried at 65 °C and ground to powder (~1 mm) for homogenization. Total carbon and nitrogen contents (%) were determined using a CN 802 Carbon Nitrogen Elemental Analyzer (VELP Scientifica, Usmate Velate, Italy) at the Centre for Functional Ecology, University of Coimbra (Portugal).

After removing the litter from each 25 × 25 cm quadrat, one soil sample was collected with a PVC core (Ø 5 cm x 5 cm depth), totaling again 50 samples. Soil was oven-dried at 105 °C and subsequently sieved through a 2 mm mesh. Soil organic carbon content (%) was measured following digestion with HCl using the same analyzer as above.

At each of the same sampling points, two samples were collected using PVC cores (Ø 5 cm x 5 cm depth), including both soil and litter, to sample springtail communities, totaling 50 samples. Samples were placed in plastic bags and kept cool (~4 °C) until extraction (i.e., within 24 hours). Springtails were extracted in a Macfadyen high-gradient extractor during 4 days (Sousa et al. 2004). Afterwards, specimens were preserved in 96% ethanol in plastic vials and counted with a binocular stereo microscope. Each springtail specimen was classified into one of the following functional types: epigeic, hemiedaphic, or euedaphic. For this classification, a range of eco-morphological traits reflecting springtail adaptations to the soil profile was used: i) the presence or absence of ocelli, ii) antennae length, iii) furca characteristics (presence/absence and length), iv) the presence of scales or thick hairs along the body, and v) body pigmentation (Parisi et al. 2005; Vandewalle et al. 2010; Martins da Silva et al. 2016, 2023; Reis et al. 2016). To specifically classify each specimen into a functional type, each morphological trait was assigned a score from 0 to 4 (see Suppl. material 1: table S2 for detailed information). After assigning scores to each trait, a composite life-form trait was calculated by summing the partial scores of each functional type, and then this was divided by the maximum possible score of 20 so that each life-form trait has a continuous value varying between 0 and 1. This normalization allowed each individual to have a single life-form trait score ranging from 0 to 1 (Martins da Silva et al. 2016, 2023), with higher scores indicating a greater adaptation of springtails to the soil profile (i.e., from 0 to 0.3 for epigeic springtails; from 0.4 to 0.6 for hemiedaphic springtails; and from 0.7 to 1 for euedaphic springtails; Suppl. material 1: table S2). The number of specimens of each functional type found in each sample was divided by the surface of the core (19.625 cm²), and the abundances of epigeic, hemiedaphic, and euedaphic springtails were then expressed as individuals per square meter of soil.

We conducted two preliminary analyses to assess any possible impact of spatial patterns and the physical environment on Acacia invasion within the fragmented forest landscape. First, we applied a Moran’s I test using the “ape” package (Paradis and Schliep 2019) to evaluate spatial autocorrelation in the proportion of Acacia (A. dealbata + A. melanoxylon) relative to total tree abundance across sampling points. Second, we used Spearman correlation tests to examine the specific relationships between the proportion of Acacia relative to total tree abundance with latitude, longitude, elevation, slope, and aspect. After detecting significant spatial autocorrelation in the proportion of Acacia cover (observed Moran’s I statistic: 0.064; expected Moran’s statistic: -0.042; p-value < 0.001), and significant correlations of latitude and elevation with the proportion of Acacia cover (Suppl. material 1: table S3), we tested our four hypotheses (Fig. 2) as detailed below. H₁, H₂, and H₃ considered the entire landscape, which included both invaded and non-invaded areas, while H₄ focused specifically on the sampling points invaded by Acacia.

Figure 2.

Conceptual flow chart showing the hypotheses of the study. Expected effects of Acacia invasion: (H₁) reduce understory cover (herb and shrub layers) and plant species richness; (H₂) alter litter and soil characteristics, with a lower litter C/N ratio and increased soil organic carbon (SOC); (H₃) shift in springtail functional structure, affecting the abundances of epigeic, hemiedaphic, and euedaphic groups; and (H₄) cascading effects, where Acacia-invaded areas become more homogeneous and dominated by nitrogen-fixing trees, leading to changes in litter and soil quality that influence springtail resource-acquisition strategies.

To test H₁, we ran independent generalized linear models (GLMs) with a negative binomial family, using herb cover, shrub cover, and plant species richness as response variables. The proportion of Acacia (A. dealbata + A. melanoxylon) cover was the predictor variable, and latitude and elevation were included as covariates. Additionally, using the covers of each plant species, we compared plant community assemblages between Acacia-invaded and non-invaded sampling points by conducting non-metric multidimensional scaling (NMDS) based on the Bray-Curtis similarity index, and we subsequently evaluated the dissimilarities observed with a similarity percentage analysis (SIMPER).

To test H₂, we conducted independent GLMs with a negative binomial family for the litter C/N ratio and SOC as response variables, and the proportion of Acacia (A. dealbata + A. melanoxylon) cover was the predictor variable, with latitude and elevation included as covariates.

To test H₃, we fitted a single GLM with a Gaussian family to evaluate how springtail abundances responded to changes in the proportion of Acacia cover (A. dealbata + A. melanoxylon) across functional groups (epigeic, hemiedaphic and euedaphic). Springtail abundances were transformed to log(x+1) to meet normality criteria, and we used a zero-inflated model to account for the high number of zero values in the abundance data. Functional group was included as a categorical grouping factor, and the interacton term between functional group and the proportion of Acacia cover was used to test whether the effect of Acacia varied among groups. Latitude and elevation were also included as covariates.

To test H₄, we performed two analyses. To evaluate how changes in vegetation structure affect litter and soil characteristics, we used independent GLMs with a negative binomial family, with litter C/N ratio and SOC as response variables. Each predictor variable (i.e., herb cover, shrub cover and plant species richness) was used independently in separate models. Moreover, to evaluate how changes in litter and soil characteristics affect the functional structure of springtail communities, we performed independent GLMs with a Gaussian family. In these models, the abundances of epigeic, hemiedaphic, and euedaphic springtails were the response variables, while the litter C/N ratio and SOC were used as independent predictor variables in separate models. As in H₃, we log-transformed springtail abundances.

All generalized linear models (GLMs) were conducted using the “glmmTMB” package (Brooks et al. 2017). We assessed the significance of latitude and elevation as covariates across all models for H₁, H₂, and H₃ by performing step-by-step model selections (Crawley 2012). Specifically, we started with a full model that included the proportion of Acacia cover, latitude, and elevation as predictor variables, and we then progressively simplified the model by removing non-significant variables (reduced model). Model comparisons were based on Akaike Information Criterion (AIC), selecting the model with the lowest AIC, and using ANOVA likelihood-ratio tests to ensure that the selected models significantly improved the fit (p-value < 0.05). Graphical assessments were then performed to confirm the normality and homogeneity of the residuals for each final model. All analyses were conducted using R (R 3.6.2, R Core Team 2024) except NMDS and SIMPER, which were run with Primer v6 (Clarke and Warwick 2001).

As the proportion of Acacia (A. dealbata + A. melanoxylon) cover increased, herb cover decreased significantly (Estimate = -0.037, z = -3.837, p-value < 0.001; Fig. 3a). Additionally, latitude (Estimate = 124.760, z = 2.450, p-value = 0.014; Suppl. material 1: fig. S2a) and elevation (Estimate = 0.006, z = 2.337, p-value = 0.019; Suppl. material 1: fig. S2b) were found to positively affect herb cover. While the full model showed marginally more significance than the model with only the proportion of Acacia cover, it also had the lowest AIC among all models tested (Suppl. material 1: table S4). In contrast, the proportion of Acacia cover did not significantly affect shrub cover (Estimate = -0.016, z = -1.033; p-value = 0.301) in either the full model or the reduced models (Fig. 3b; Suppl. material 1: table S4).

Figure 3.

Scatterplots showing the relationships between the proportion of Acacia (A. dealbata + A. melanoxylon) relative to total tree abundance with vegetation structure (H₁): A. Herb cover; B. Shrub cover, and C. Plant species richness. The regression lines of best fit are derived from each GLM analysis. The solid line shows the model prediction, and the grey shaded region represents the 95% confidence interval for those predictions. P-values highlighted in bold show a significant effect of proportion of Acacia cover (p-value < 0.05).

When assessing plant species richness, we found that the proportion of Acacia (A. dealbata + A. melanoxylon) cover had a significant negative effect, with higher proportion of Acacia associated with lower plant richness (Estimate = -0.007, z = -2.012, p-value = 0.044; Fig. 3c). The model using only the proportion of Acacia cover as a predictor variable was identified as the best fit, showing the lowest AIC (129.7) among all models tested (Suppl. material: table S4).

A. dealbata and A. melanoxylon were found growing together at two sampling points, while A. dealbata was recorded alone at six points and A. melanoxylon at only one (Table 1). The lowest proportion of Acacia (A. dealbata + A. melanoxylon) cover was 18%, and A. melanoxylon was found in lower proportions than A. dealbata across the whole fragmented forest landscape (Table 1). Moreover, we found considerable differences in plant species assemblages between invaded and non-invaded sampling points (Suppl. material 1: fig. S1). The SIMPER analysis (Suppl. material 1: table S5) showed an average dissimilarity of 90.41% between these two groups. The species contributing most to this dissimilarity was A. dealbata, accounting for 18.33%. Following this, P. pinaster contributed with 12.56%, and several other species contributed with more than 5%, including P. nigra, the fern Pteridium aquilinum (L.) Kuhn, and the shrubs U. minor and G. tridentata. Detailed information on all surveyed vascular plants is shown in Suppl. material 1: table S1.

As the proportion of Acacia (A. dealbata + A. melanoxylon) cover increased, the litter C/N ratio decreased significantly (Estimate = -0.003, z = -2.214, p-value = 0.027; Fig. 4a). In contrast, latitude (Estimate = 22.941, z = 2.671, p-value = 0.008; Suppl. material 1: fig. S3a) and elevation (Estimate = 0.001, z = 3.079, p-value = 0.002; Suppl. material 1: fig. S3b) were found to positively influence the litter C/N ratio. Model comparisons showed that the full model provided a significantly better fit than the reduced models (Suppl. material 1: table S4). Contrarily, as the proportion of Acacia cover increased, soil organic carbon (SOC) increased significantly (Estimate = 0.003, z = 2.020, p-value = 0.043; Fig. 4b). Although the full model with latitude and elevation showed the lowest AIC, it did not improve the fit, as these two covariates were not significant (Suppl. material 1: table S4).

Figure 4.

Scatterplots showing the relationships between the proportion of Acacia (A. dealbata + A. melanoxylon) relative to total tree abundance with litter and soil quality (H₂): A. Litter C/N ratio, and B. SOC. The regression lines of best fit are derived from each GLM analysis. The solid line shows the model prediction, and the grey shaded region represents the 95% confidence interval for those predictions. P-values highlighted in bold show a significant effect of proportion of Acacia cover (p-value < 0.05).

The proportion of Acacia (A. dealbata + A. melanoxylon) cover did not significantly influence the overall springtail abundance (Estimate = 0.002, z = 0-62, p-value = 0.533; Fig. 5), and the interaction between functional group and proportion of Acacia cover was also non-significant, with any group showing responses to Acacia invasion. However, independently of Acacia presence, the model revealed significant differences among functional groups, with both hemiedaphic (Estimate = 0.84, z = 5.95, p < 0.001) and euedaphic springtails (Estimate = 0.79, z = 5.61, p < 0.001) exhibiting greater abundances than epigeic springtails. Model comparisons showed that the model including elevation as a covariate had the lowest AIC, but it did not significantly provide a better fit than the most reduced model (Suppl. material 1: table S4).

Figure 5.

Scatterplot showing the relationships between the proportion of Acacia (A. dealbata + A. melanoxylon) relative to total tree abundance across springtail functional groups (H₃): epigeic (dotted line), hemiedaphic (solid line), and euedaphic (dashed line). The regression lines of best fit are derived from the GLM analysis. The dotted, solid and dashed lines show the model predictions, and the grey shaded region represents the 95% confidence interval for those predictions. There were not any significant effects of proportion of Acacia cover on any springtail functional group.

Shrub cover significantly increased the litter C/N ratio and SOC (Table 2a), while herb cover and plant species richness had no significant effect on either litter C/N ratio or SOC (Table 2a). Moreover, when we evaluated the functional structure of springtail communities in relation to litter and soil quality in Acacia-invaded sampling points, we observed that for epigeic springtails, neither the litter C/N ratio nor SOC significantly influenced their abundance (Table 2b). However, hemiedaphic springtails showed a significant negative response to increasing litter C/N ratio, but not to SOC levels (Table 2b). Contrarily, euedaphic springtails responded positively to higher SOC levels, although the litter C/N ratio did not significantly affect their abundance (Table 2b).

The occurrence of A. dealbata and A. melanoxylon at the landscape scale was associated with a significant reduction in herb cover and plant species richness. These results align with da Silva et al. (2019), who found that even moderate Acacia invasion can alter understory vegetation structure in the same study area. Additionally, the impact of Acacia presence on herb cover varied with topographic factors such as elevation and latitude, likely due to their influence on light availability for A. dealbata (Lorenzo et al. 2010b) and other species with phyllode-like leaves like A. melanoxylon (Unwin et al. 2006). These spatial patterns highlight the interaction between invasion-driven biological changes and broader physical landscape characteristics (Ohlemüller et al. 2006; Heringer et al. 2020). Nevertheless, replication within specific vegetation types was limited (Table 1), preventing us from including this category as a factor in our analyses. Therefore, future studies should ensure adequate replication across vegetation types at the landscape level to detect any specific invasion effects.

We expected a negative effect of Acacia proportion (A. dealbata + A. melanoxylon) relative to total tree abundance on shrub cover in the fragmented forest landscape. Although we observed a generally negative relationship between Acacia invasion and shrub cover, this was not statistically significant. This may be because several sampling points lay in transitional zones where Acacia trees and native shrubs coexisted (Suppl. material 1: table S1). In such mixed zones, shrub cover values were more variable, which diluted the contrast with fully Acacia-invaded or fully shrub-dominated sampling points and reduced statistical power to detect a clear effect. Importantly, U. minor and G. tridentata were identified as key contributors to the differences between invaded and non-invaded areas. Although their contributions to overall dissimilarity were relatively low (< 7%), these species are important for understanding the observed patterns. Both are nitrogen-fixing and were less abundant in invaded areas, suggesting that they may occupy similar niches such that Acacia pre-empts niche space that would otherwise be available (Vieites-Blanco and González-Prieto 2020). However, it might also be the case that these native shrubs are inhibiting Acacia invasion, rather than being outcompeted, and this highlights the difficulty in the determination of causality in such interactions (Cassini 2020). The coexistence of multiple nitrogen-fixing species in Acacia-invaded areas may be constrained, as increased Acacia dominance boosts nitrification rates, soil nitrogen availability, and canopy closure, thereby reducing microclimatic variability (i.e., a pattern observed in A. dealbata; Lorenzo et al. 2017). Interestingly, higher shrub cover in invaded areas was associated with improved litter and soil quality, as indicated by higher litter C/N ratios and increased SOC (Table 2). This is in line with a previous study showing that natural shrublands have been shown to act as barriers to Acacia invasion in the northern Iberian Peninsula (Rodríguez et al. 2017), and our results suggest that the density and identity of the shrub layer are key factors determining shrub vulnerability to Acacia’s competitive dominance.

The Acacia invasion had contrasting effects on litter and soil characteristics in the fragmented forest landscape. The reduction in litter C/N ratio and increase in SOC were likely driven by the chemical composition of Acacia litter because, as nitrogen-fixing species, Acacia can significantly affect soil properties (Liao et al. 2008; Lee et al. 2017), and persistent litter inputs may drive long-term alterations in soil conditions (Vivanco and Austin 2008; Souza-Alonso et al. 2015). Early in decomposition, Acacia litter contributes with labile organic carbon, while recalcitrant compounds resist rapid breakdown (Souza-Alonso et al. 2017, 2024). Species such as A. dealbata (Madureira and Ferreira 2022) and A. melanoxylon (González et al. 1995; Marchante et al. 2019; Pereira et al. 2021; Pereira and Ferreira 2022) produce litter with high lignin content, slowing down organic matter decomposition. However, these changes do not immediately translate into shifts in nutrient cycling or soil carbon pools, as organic matter turnover depends on thresholds of litter input and accumulation (Prescott and Zukswert 2016; Souza-Alonso et al. 2024). Furthermore, topographical gradients within the fragmented forest landscape played a significant role not only in shaping herb cover but also in influencing litter C/N ratio, likely through parallel mechanisms that modulate the niche requirements of the invader (Warren et al. 2011). Additionally, differences in the age and aboveground carbon stocks of Acacia stands may have further influenced the magnitude and timing of these effects (Matos et al. 2023). It is important to note, however, that A. dealbata litter generally does not accumulate as much as A. melanoxylon litter due to its more fragmented structure. Acacia dealbata has compound leaves, while A. melanoxylon produces sclerified phyllodes that decompose more slowly (Renner et al. 2021). Thus, their litter may influence C/N ratios and SOC accumulation in different ways. Distinguishing between these species in future studies at a landscape level could provide further insights into the mechanisms driving changes in litter and soil quality.

Although Acacia invasion caused notable shifts in vegetation structure and litter and soil characteristics, its direct effects on springtail functional groups across the fragmented landscape were not significant. This broader pattern at the landscape level, however, contrasted with the more pronounced cascading effects observed in invaded sampling areas, where springtail functional structure was notably influenced by changes in litter and soil characteristics. Hemiedaphic springtails, in particular, showed a significant negative correlation with the litter C/N ratio. Their reliance on the formation of thick organic layers makes them sensitive to changes in abiotic conditions, and as intermediate dwellers in the soil-litter interface, they are highly dependent on litter quality, which shapes their microhabitats and provides essential food resources (Marian et al. 2018; Xie et al. 2024). A lower litter C/N ratio indicates higher proportional nitrogen content and greater litter palatability, making it richer in terms of nutrient availability (Gerlach et al. 2014; Nascimento et al. 2019). Therefore, the observed negative association between hemiedaphic springtail abundance and higher litter C/N ratios suggests that these springtails are favored by improved litter quality associated with lower C/N ratios. Meanwhile, euedaphic springtails, which inhabit deeper soil layers and are relatively independent of food or substrate quality (Krab et al. 2010), showed a positive response to increased SOC levels in the invaded sampling points. Predominantly parthenogenetic (i.e., reproducing without fertilization by a male), they are well-adapted to microhabitats with stable abiotic conditions like deeper soils, allowing for rapid population growth, particularly following disturbances (Lindberg and Bengtsson 2005; Pollet et al. 2022). Moreover, the accumulation of SOC in these deeper layers was likely driven by slower decomposition rates and reduced microbial activity, conditions often associated with Acacia invasion (Lazzaro et al. 2014). Consequently, the interplay between their reproductive strategy and the altered soil environment likely underpinned their higher abundance in invaded areas with higher SOC levels.

Unlike hemiedaphic and euedaphic life forms, which are closely linked to variations in litter and soil conditions and were found in much higher numbers throughout the landscape, surface-dwelling epigeic springtails seemed to exhibit a reduced dependence on these factors. This reduced sensitivity may stem from their greater mobility, which allows them to exploit a wider range of microhabitats and makes them less constrained by microhabitat soil and litter characteristics (Martins da Silva et al. 2016). This result contrasts with the findings of Long et al. (2023), who showed that epigeic springtails were the only springtail functional group significantly affected by bamboo invasion. This discrepancy may be due to differences in the specific alterations originated by Acacia versus bamboo invasion, likely reflecting that the impacts of invasive alien plant species on belowground decomposers are context- and species-dependent (Luan et al. 2021).

It is important to note that we explored potential cascading effects of Acacia invasion on springtails through correlation analyses framed within a causal structure defined a priori. However, we acknowledge the limitations of this approach: while correlations are informative, they do not establish causality, and other processes not captured in our study, such as shifts in microbial activity or complex interactions within the soil food web (Potapov et al 2016; McCary and Wise 2019), may also influence springtail functional structure. These factors should be considered in future studies aiming to disentangle the full set of mechanisms underlying belowground responses to Acacia invasion.

Some methodological constraints may also partly explain the weak direct effects of Acacia invasion on springtail communities observed here. Although we sampled the upper 0–5 cm of soil (where springtails are typically concentrated) with a standardized 5 × 5 cm core, deeper layers were not assessed and might harbor additional species or invasion-driven shifts. Moreover, our functional trait-based approach captures the ecological consequences of community shifts without requiring the specialized expertise and resources needed for full species-level identification (Vandewalle et al. 2010; Moretti et al. 2017), but it cannot reveal taxonomic turnover in detail. Future work should consider deeper soil sampling and, where feasible, combine trait-based and taxonomic approaches to provide a more complete picture of invasion impacts on springtail diversity and function.

Our results showed that even low levels of Acacia invasion influenced vegetation structure and litter and soil quality within a fragmented forest landscape. These changes underline the profound and interconnected impacts of invasive species on both aboveground and belowground ecological processes, highlighting the cascading effects of Acacia invasions on ecosystem functioning.

Management efforts typically focus on the invasive species or the invaded community, but integrating the broader landscape context can improve intervention strategies (Vilà and Ibáñez 2011). While A. dealbata currently causes stronger impacts due to its wider distribution, prioritizing areas with smaller infestations of either A. dealbata or A. melanoxylon is more effective for preventing further spread and impacts. Early intervention increases the likelihood of successful control, reduces costs, prevents negative effects and contributes to greater ecosystem stability. Regardless of the specific measures applied, follow-up management is essential, as both species have persistent, long-lived seed banks and can resprout after disturbances such as cutting or fire. Additionally, restoring native habitats is recommended. This is crucial not only to prevent the establishment of Acacia species in areas where they are absent, but also to detect them at early invasion stages and control them promptly, minimizing the risk of more severe ecological consequences (IPBES 2024).

Our results underscore the need for proactive measures to manage the Acacia invasion in the Mediterranean Basin, specifically in the case of A. melanoxylon, which remains more localized (Souza-Alonso et al. 2017), but also of A. dealbata, and other Acacia species in areas where they start to invade. In particular, our studied fragmented forests in Central Portugal, where A. dealbata is already widespread but also found in more localized areas, and A. melanoxylon is still confined to smaller patches, present an opportunity for early intervention to prevent further spread and impacts of both species.

The authors have declared that no competing interests exist.

No ethical statement was reported.

No use of AI was reported.

This work was funded by the project “F4F – Forest for Future – Pilot project for the constitution of a regional network for the valorisation of the forest sector in the Central Region. PP6 – MyFORESt (Ref. CENTRO-08-5864-FSE-000031)”, co-financed by the Regional Operational Programme Centro 2020, Portugal 2020, and the European Union, through the European Social Fund (ESF). R. Juan-Ovejero holds a postdoctoral contract (Ref. ED481B-2022-006) funded by the Consellería de Cultura, Educación, Formación Profesional e Universidades from the Xunta de Galicia. J. Alves was funded by FCT – Fundação para a Ciência e a Tecnologia, I.P. under Scientific Employment Stimulus Individual support (2022.05848.CEEIND; DOI: https://doi.org/10.54499/2022.05848.CEECIND/CP1714/CT0008). This study was partially funded by the FCT through the strategic projects of the Centre for Functional Ecology – Science for People and the Planet (CFE) (UIDB/04004/2025), and Associate Laboratory TERRA (LA/P/0092/2020; DOI: https://doi.org/10.54499/LA/P/0092/2020).

Conceptualization: RJO; PMS. Formal analysis: RJO; PMS. Data curation: RJO; FR; FG; FC; MCD. Visualization: RJO; EM; PMS. Funding acquisition: JA; JPS; HF. Investigation: RJO; FR; FG; FC; MCD, AAS. Methodology: AAS; JPS, JA. Supervision: JA. Project Administration: JA. Writing – original draft: RJO. Writing – review and editing: EM; PMS; FR; FG; MCD; FC; HF; JA.

Raquel Juan-Ovejero https://orcid.org/0000-0002-1686-4167

Filipa Reis https://orcid.org/0000-0001-6297-8862

Pedro Martins da Silva https://orcid.org/0000-0002-5930-264X

Elizabete Marchante https://orcid.org/0000-0003-1303-7489

Fernanda Garcia https://orcid.org/0000-0001-8470-0693

Maria Celeste Dias https://orcid.org/0000-0002-3083-6218

Filipe Covelo https://orcid.org/0000-0003-0432-2843

António Alves da Silva https://orcid.org/0000-0001-9544-3936

Helena Freitas https://orcid.org/0000-0002-1907-9615

José Paulo Sousa https://orcid.org/0000-0001-8045-4296

Joana Alves https://orcid.org/0000-0003-2858-7803

All of the data that support the findings of this study are available in the main text within tables and figures and in the Supplementary material.