Isoscapes of remnant and restored Hawaiian montane forests reveal differences in biological nitrogen fixation and carbon inputs

Introduction

Forests provide essential ecosystem services including carbon (C) storage and nutrient cycling. However, human-induced disturbances, deforestation, land conversion, and invasive species have led to devastating losses in biodiversity and potentially irreversible alterations to biogeochemical processes (Cramer et al., 2004; Asner et al., 2008; Handa et al., 2014; Veldkamp et al., 2020). While active forest restoration has been touted as a tool to spur secondary succession and regain desirable ecosystem states, it remains unclear if this practice will lead to the recuperation of the biogeochemical properties provided by primary forest reference ecosystems (Aerts & Honnay, 2011; Sullivan et al., 2019; Yelenik et al., 2021). As land management strategies focused on restoration are developed to reconcile losses of ecosystem services, there is a need to understand the spatial extent and long-term impacts of these restoration strategies on restoring biogeochemical properties and C and nutrient cycling.

Nitrogen (N) is a key component of biogeochemical cycles, and N is often one of the most limiting nutrients for plant growth and photosynthesis—especially in early successional or secondary tropical forests (Vitousek & Farrington, 1997). Furthermore, the conversion of tropical forests to pasture and other agricultural crops can lead to significant changes in soil properties such rates of N cycling and N availability, as well as C sequestration and storage (Veldkamp et al., 2020). Trees capable of atmospheric nitrogen (N₂) fixation via symbiotic interactions with root nodule inhabiting bacteria (ex. Rhizobia spp.) can be integral to restoring forests by catalyzing ecological succession, reversing the effects of destructive land-use practices on soil N availability, and aiding global efforts to mitigate climate change through increasing C sequestration (Chazdon, 2003; Batterman et al., 2013; Chazdon et al., 2016; Levy-Varon et al., 2019; but see, Kou-Giesbrecht & Menge, 2019). Specifically, as the density of N₂-fixing trees, such as leguminous species in the genus Acacia (Resh, Binkley & Parrotta, 2002) increases, N in the soil and neighboring plants can also increase (Sitters, Edwards & Olde Venterink, 2013). Therefore, biological nitrogen fixation (BNF) has the capacity to positively influence the performance and N-budgets of both N₂-fixing and non-N₂-fixing plants.

In the tropics, N₂-fixing trees are often used as ecosystem engineers for the restoration of degraded landscapes (Fisher, 1995; Scowcroft, Haraguchi & Hue, 2004). However, the outcomes of these efforts are mixed, with some restored areas stalled in apparent alternative stable states (Yelenik, 2017), while others progress toward ecosystem targets and restoration goals (Fisher, 1995; Rhoades, Eckert & Coleman, 1998; Koutika et al., 2021). These differing effects of N₂-fixing trees may be owed to heterogeneous N inputs across landscapes (Dixon et al., 2010; Sullivan et al., 2014), as well as ecosystem-specific biotic and abiotic factors (Pearson & Vitousek, 2002; Staddon, 2004; Wynn & Bird, 2007; Dixon et al., 2010; Barron, Purves & Hedin, 2011; Sitters, Edwards & Olde Venterink, 2013). For example, active restoration that relies on establishing forests of N₂-fixing trees may have unintended consequences, such as promoting weedy or invasive species (Funk & Vitousek, 2007), thereby altering successional trajectories toward non-target states (Stinca et al., 2015). A better understanding of the spatiotemporal effects of N₂-fixing trees on ecosystems is needed, especially in the context of forest restoration.

Stable isotopes are time-integrating markers that can provide insight into biogeochemical processes and shifts in ecosystem services occurring across multiple landscapes and spatiotemporal scales (Cheesman & Cernusak, 2016). Spatially explicit, geo-referenced isotope landscapes (termed ‘isoscapes’) allow for the measured isotope values of individual replicates (i.e., within or among species, or sample types) to be interpolated to landscape scales (Bowen, 2010). While only a few studies have used isoscapes in plant ecology (Hellmann et al., 2011; Rascher et al., 2012), their application may provide new and important insights into plant-microbe and plant-soil feedbacks, such as BNF by N₂-fixing trees, which can vary at relatively smaller spatial scales.

BNF results in N stable isotope values (δ¹⁵N) that are more similar to the atmosphere (~0‰) relative to N assimilated from soil N pools (Craine et al., 2015). Soil δ¹⁵N values represent an integration of N inputs and outputs in an ecosystem, and are influenced by processes that lead to isotope fractionation (i.e., nitrification, denitrification, ammonia volatilization), atmospheric deposition, leaching, as well as the pedogenic and environmental factors that shape these processes (Natelhoffer & Fry, 1988; Austin & Vitousek, 1998; Martinelli et al., 1999; Burnett et al., 2022). Variability in soil δ¹⁵N can complicate the interpretation of leaf δ¹⁵N values, especially for non-N₂-fixing plants (Robinson, 2001). However, the low δ¹⁵N values associated with BNF are distinct enough to be traced through soils and vegetation (Hellmann et al., 2011). For instance, δ¹⁵N isoscapes revealed the contribution of an invasive, non-native N₂-fixing shrub (Acacia longifolia) to the N pools of surrounding plants and the corresponding modification of nutrient cycling and community function due to this invasive species (Rascher et al., 2012).

Along with standard comparisons of C to N concentrations, leaf and soil C stable isotope values (δ¹³C) can provide additional evidence on the rates of nutrient cycling, the relative contribution of plant functional types (i.e., C₃ vs. C₄ photosynthesis) to soil C pools, and the integrated plant water use efficiency (WUE_i) of plants in an ecosystem (Cernusak et al., 2013; Driscoll et al., 2020). Combined, δ¹³C and δ¹⁵N values sampled across a landscape can provide spatially explicit information on plant performance and ecosystem processes, which can be compared among ecological community members or between communities (Hellmann et al., 2016). While it is clear that isoscapes provide new perspectives on the spatial relationships of biochemical processes and biological interactions fundamental to ecology (Cheesman & Cernusak, 2016; McCue et al., 2020), they have yet to be widely applied to understand plant-soil feedbacks in the context of restoration.

Native Hawaiian montane mesic and wet forests have significantly declined due to over a century of deforestation and extensive human- and livestock-mediated disturbance (Pau, Gillespie & Price, 2009; McDaniel et al., 2011; Yelenik, 2017). These forests are dominated by two foundational endemic tree species—N₂-fixing Acacia koa (koa, Fabaceae) and the non-N₂-fixing Metrosideros polymorpha (‘ōhi‘a lehua, Myrtaceae)—and provide critical ecosystem services, including habitat for many endangered bird species in Hawai‘i (Paxton et al., 2018). To restore montane mesic forest habitats on Hawai‘i Island, over 390,000 nursery-grown A. koa trees (i.e., outplants) have been planted in the Hakalau Forest National Wildlife Refuge since 1987 (hereafter, Hakalau; Fig. 1). This effort started in the greenhouse, where individual trees (one per pot) were grown in 15-cm-long cone-tainers, reaching a height of 30-cm in <6 months. In the plantation, a bulldozer with a miniblade scoured the soil to remove grass competitors and make plots (1 × 1 m) every 3 m, with rows every 4 m a part. Individual greenhouse-grown koa were planted in holes created by an auger power planter (the same size as the root plug) with fertilizer added to the auger hole (see Jeffrey & Horiuchi, 2003).

Acacia koa outplants were introduced to 5,000 acres of fenced, abandoned, previously grazed pastureland in close proximity to remaining forested habitats and endangered bird populations (Jeffrey & Horiuchi, 2003; McDaniel & Ostertag, 2010). Forest patches are critical in supporting biodiversity (Wintle et al., 2019), and patchiness and/or fragmentation can lead to dramatic differences in forest climate that support rates of litter decomposition and nutrient cycling (Crockatt & Bebber, 2015). Therefore, the active outplanting of A. koa in discrete corridors near more intact forest patches was leveraged to facilitate the passive restoration and expansion of koa and other native plant species recruitments, with the goal of ultimately turning isolated patches into intact, contiguous forest habitat (Scowcroft & Yeh, 2013). However, after 30 years the understories of A. koa plantations remain grass-dominated, native woody plant recruitment is low, and endangered birds are sparse (Yelenik, 2017; Paxton et al., 2018). The causes of this stalled forest recovery are uncertain, but they likely include the influence of A. koa BNF on soil chemistry, nutrient concentrations, and nutrient cycling (Scowcroft, Haraguchi & Hue, 2004). As a N₂-fixing tree, A. koa produces abundant low C:N leaf litter (high N), which can lead to greater supply of N to the soil surface under A. koa relative to M. polymorpha, which has high litter C:N (Scowcroft, 1997; Yelenik, Rehm & D’Antonio, 2022). These differences in litter N can increase rates of decomposition and ultimately affect the return of nutrients to soil N pools (Baker, Scowcroft & Ewel, 2009; Zhou et al., 2018). Differences in canopy composition and associated litter decomposition can have important implications for forest nutrient cycling and the composition of forest understories. In the plantations, the ability for A. koa to increase soil nutrient, coupled with the high densities of this canopy tree in monoculture corridors, may be favoring the growth of exotic nitrophilous grasses from the historic pastures and inhibiting native seedling establishment (Yelenik, 2017).

To examine the interactions between BNF, soil and leaf N content as well as C inputs at the ecological community scale, we applied a spatially explicit sampling approach within a mixed M. polymorpha/A. koa native montane forest (hereafter, the “remnant forest”) and an A. koa plantation (hereafter, “koa plantation”). The remnant forest is largely dominated by native understory plants, whereas the koa plantation understory is dominated by C₄ exotic grasses interspersed with a C₃ non-native shrub (Rubus argutus). While the land-use history and present-day plant communities of these two forest types differ, they occupy the same climate and parent soil types (http://rainfall.geography.hawaii.edu, https://gis.ctahr.hawaii.edu/SoilAtlas). We used δ¹⁵N and δ¹³C values of soil and foliar samples to construct isoscapes and examine the degree of BNF between the two forest types and test the influence of BNF by A. koa on δ¹⁵N values in soils and neighboring plants (Rubus spp.). We hypothesized that the higher density and more even distribution of A. koa in the koa plantation compared to the remnant forest will result in lower δ¹⁵N values in soil and plants from the koa plantations with greater and spatially more homogeneous BNF inputs than the remnant forests. In addition, we expected higher foliar δ¹³C values in the koa plantation as a result of greater water demand in the young, dense A. koa forests driving higher WUE_i (Kagawa et al., 2009). Introduced C₄ grasses are abundant in the understory of the koa plantation but are uncommon in the remnant forest where there are no known native C₄ plants (Yelenik, 2017). Therefore, we also predicted that soil δ¹³C would be higher in the koa plantation relative to the remnant forest, indicating greater contributions of C₄ grasses to soil C pools. The results of this study will clarify the effects of tropical N₂-fixing monoculture plantations on the distribution of N and soil nutrient cycling and how these may differ from the target ecosystem for restoration.

Materials and Methods

Results

Soil and foliar isotope values

δ¹⁵N values differed according to sample type (p < 0.001) and forest (p < 0.001), with overall lower δ¹⁵N values in the koa plantation (Table S2). δ¹⁵N values of soils were not different among the forests (p = 0.085). Acacia koa and Rubus spp. δ¹⁵N values were both significantly closer to zero (less ¹⁵N-enriched by ~1‰) in the koa plantation compared to remnant forest samples (post-hoc: p ≤ 0.029) (Fig. 2A). As expected, δ¹⁵N values of the N₂-fixing A. koa were significantly lower compared to Rubus spp. in both the koa plantation (post-hoc: p = 0.020) and remnant forest (post-hoc: p = 0.029). Overall, there was a greater range and standard deviation in δ¹⁵N values in the remnant forest A. koa (range: 2.19‰, SD: 1.03‰) and Rubus spp. (range: 3.54‰, SD: 0.89‰) compared to plantation A. koa (range: 0.85‰, SD: 0.65‰) and Rubus spp. (and 2.55‰, SD: 0.75‰). This larger range and variance in plant δ¹⁵N values is consistent with heterogeneous N sources in the remnant forest, which is being expressed in the leaves of canopy N₂-fixing trees and understory plants. Soil δ¹⁵N values were ~3‰ higher than those in foliar samples, but similar among the koa plantation and the remnant forest (post-hoc: p = 0.085) (Fig. 2A).

δ¹³C values were different among sample types (p < 0.001) and between forests (p < 0.001), with significantly lower δ¹³C values in all sample types from the remnant forest relative to those from the koa plantation (Fig. 2B; Table S2). However, when compared within the same forest, the δ¹³C values of A. koa and Rubus spp. were not different from each other in the remnant forest (post-hoc: p = 0.057) or the koa plantation (post-hoc: p = 0.281).

Isoscapes and model variograms

Variograms provided insights into the spatial correlation within isoscapes and the semivariance-by-distance relationship. δ¹⁵N and δ¹³C isoscape variograms for both soil and leaves showed equivalent ranges in spatial autocorrelation in both forests (~3–5 m) (Table S3). These results indicate similar relationships between semivariance-by-distance decay (i.e., less autocorrelation with increasing distance) in the koa plantation and the remnant forest at the scale of our experimental sampling. However, these estimates should be interpreted with caution since (1) the minimum sampling distance between samples for soils was 4 m and (2) in several modeled variograms the increased nugget values coupled with low sill values may indicate greater influence of random variation. Nevertheless, we found significantly different distributions of δ¹⁵N and δ¹³C isotope values between forest types for both soil and foliar isoscapes (p < 0.001) (Figs. 3–5; Table S1).

In both soil and foliar δ¹⁵N isoscapes, we observed lower δ¹⁵N_predicted values in the koa plantation (p < 0.001) (Table S1), with mean δ¹⁵N values for all interpolated data points lower by ~0.5‰ (soil isoscape) and 1.2‰ (foliar isoscape) in the plantation relative to the remnant forest (Figs. 3A–3C). Both forests showed hot spots of low δ¹⁵N values (<4‰ soil Figs. 4A and 4B, <2‰ foliar Figs. 5A and 5B) that tended to be in areas proximate to A. koa or sampling points representing A. koa leaves (and to a lesser extent Rubus spp.) (Figs. 4 and 5). However, low δ¹⁵N areas were most pronounced in the koa plantation, where A. koa density was greater than the remnant forest (18 KP vs. 10 RK mature trees), particularly through the middle of the koa plantation where A. koa trees are numerous and planted in parallel rows. Similarly, in the δ¹⁵N soil-isoscape, the koa plantation had more uniform low δ¹⁵N areas within this central A. koa corridor (Fig. 4A). The influence of A. koa on soils in the remnant forest was more variable, with a localized region of low δ¹⁵N values in the southern portion of the forest plot (Fig. 4B).

δ¹³C isoscapes of soil samples showed lower δ¹³C values throughout the remnant forest relative to the koa plantation (Figs. 4C and 4D), with the distribution of δ¹³C_predicted values for soil samples (Fig. 3B) being lower in the remnant forest (p < 0.001) (Table S1). These differences represent ~1.5‰ lower mean δ¹³C values for interpolated values in the remnant forest compared to the plantation. Hotspots of low δ¹³C values in both δ¹³C soil isoscapes corresponded to areas within the A. koa corridor in the plantation and areas where juvenile A. koa were located (Figs. 4C, 4D and S1).

Discussion

Assessments of biogeochemical landscapes can provide insight into key ecosystem properties, plant-soil feedback, and plant-plant interactions. As the degree of BNF increases, there is a well-documented decrease in foliar δ¹⁵N in N₂-fixing plant species (Craine et al., 2015). Differences in BNF between ecological communities often have corresponding effects on other ecosystem properties such as soil and neighboring non-N₂-fixing plant δ¹⁵N values and nutrient status, as well as rates of photosynthesis and nutrient cycling. In addition to δ¹⁵N analyses, δ¹³C values coupled with information on C and N concentrations can aid in inferences about each of these ecosystem-level processes occurring at different spatial scales (Moyer-Henry et al., 2006; Rascher et al., 2012; Hoogmoed et al., 2014; Craine et al., 2015). The use of large “super-plots” in discrete habitats, which are sampled at small spatial scales and/or grids to perform spatial interpolation, is common in isoscape experimental designs (Rascher et al., 2012; Hellmann et al., 2016). Earlier studies sampled A. koa across elevation and rainfall gradients on Hawai‘i Island, mapping δ¹⁵N and δ¹³C values to determine environmental impacts on plant performance and suitability of sites for forest restoration (Lawson & Pike, 2017). We find that spatially-explicit sampling at smaller spatial scales is equally useful in identifying difference is plant-soil-microbe interactions and the influence of BNF. Here, using δ¹⁵N isoscapes to determine differences in BNF between two forest patches that differ in their land use histories and management, we found strong evidence for greater, and more homogeneous contribution of newly-fixed N in the soil and plants of the afforested koa plantation compared to the remnant native forest. While these forests plots represent but a subset of the plantation and remnant forest habitats, each plot covered a significant area (700 m²) and was specifically chosen because the vegetation was representative of each unique forest type, while the abiotic conditions were similar (climate, slope, aspect, soil parent material). Therefore, the results of our study should be considered representative of forest community-level dynamics. However, increasing plot-level replication in future studies may allow for greater biogeochemical heterogeneity to be quantified within and among forest types (plantations and remnant forests) and the influence of environmental factors (such as elevation, rainfall, volcanic organic gases) on BNF and forest restoration.

In the koa plantation, δ¹⁵N soil and foliar isoscapes revealed a more evenly distributed signature of low δ¹⁵N values in the plantation relative to the remnant forest, indicative of a greater contribution of BNF in the koa plantation that could be related to the demography and density of A. koa (Figs. 4 and 5). The finding that both A. koa and understory Rubus spp. also had lower δ¹⁵N values in the koa plantation relative to the remnant forest (Fig. 2) further suggests greater BNF by A. koa and a greater contribution of newly-fixed N to neighboring plants. While soil δ¹⁵N values did not statistically differ significantly between forest types, the δ¹⁵N soil isoscape showed overall lower δ¹⁵N values in the koa plantation compared to the remnant forest (Fig. 4). The marginal differences in soil δ¹⁵N between the two forest types may be due in part to differences in water availability between forest types (Austin & Vitousek, 1998), but further efforts are needed to determine if the soil water status differs among these forests (Bothwell et al., 2014). In both forests, however, the high soil δ¹⁵N values indicate significant losses of N relative to the size of the nitrogen pool, possibly from rapid N turnover and fractionation or leaching (Natelhoffer & Fry, 1988; Austin & Vitousek, 1998; Burnett et al., 2022). Nevertheless, our suite of biogeochemical metrics (δ¹⁵N, δ¹³C, N and C concentrations) provide new insights on plant-soil-water relations, nutrient and C cycling, and offer clues as to why the plantations continue to foster a non-native grass dominated state even after three decades post reforestation (Yelenik, 2017; Yelenik et al., 2021).

In addition to higher density of A. koa in the plantation, the demographics of A. koa outplants and saplings in the plantation (<30 y old) compared to the remnant forest may be influencing C and N cycling in this system. Nodule biomass and N₂-fixation rates are life-stage dependent, with younger stands (6 y) exhibiting an order of magnitude higher nodule biomass and N₂-fixation rates than older stands (20 y) (Pearson & Vitousek, 2001). The growth rate of A. koa across the Hawaiian Islands is variable, with dbh-based growth rates of 10–15 mm/y in sunlit areas where the crown is exposed and 6–7 mm/y estimated across a range of sites in the Hawai‘i Department of Forestry and Wildlife long-term forest plots (Baker, Scowcroft & Ewel, 2009). In remnant forests of Hakalau, growth rates of 4 mm/y for A. koa have been observed (Hart, 2010), but no dbh-age estimates exist for the koa plantation. If we assume A. koa in the sunlit plantation have growth rates similar to other sunlit areas across Hawai‘i (~12 mm/y) and A. koa in the remnant forest growth rates are 4 mm/y, we estimate mean (±SE) ages of trees from the plantation to be significantly younger (p = 0.049, Table S1) than the remnant forest (30 ± 2 KP and 55 ± 12 years RK) (Fig. S3), with two trees from the remnant forest being >110 years old. These estimates agree with the known age of the planting of the plantation (~1990) and support the hypothesis that forest demographics may be important to affecting the degree of N₂-fixation in A. koa and its contribution to plants and soils of Hakalau.

In our study system, the high densities A. koa in the plantation, along with immature A. koa recruits in thickets adjacent to planted trees, may be contributing to the higher rates of N₂-fixation we detected based on lower δ¹⁵N values (Figs. 2A, 4 and 5). The high density of A. koa has the potential to contribute more leaf litter of lower C:N to the forest floor relative to the mixed-canopy remnant forest. The greater number of multi-stemmed A. koa in the plantation (80% in KP vs. 20% RK)—which may be due to the spacing of outplants and the lack of competition in the plantation—also emphasizes the differing growth patterns in the outplanted A. koa relative to naturally recruited trees in remnant forests. These conditions may shape litterfall to the forest floor, which can be dynamic in both abundance and nutrient concentrations and reflective of shifts in plant community composition (Lanuza et al., 2018). Taken together, we suggest the greater densities, faster growth rates, and younger demographics of A. koa in the koa plantation provide a context for low C:N litterfall (and greater N contributions) in the early stages of reforestation to lead to persistent changes in nutrient cycling.

Despite a strong signal of significantly higher BNF in the plantation (Figs. 2A, 4A and 5A), we found no statistically significant differences in foliar or soil N concentrations (Figs. 2C and 2D). While this result is puzzling, one likely explanation is that the plantation may have relatively more well drained soils, and plantation A. koa may have higher water demands compared to remnant forests (Meinzer, Fownes & Harrington, 1996; Brauman, Freyberg & Daily, 2015). We note A. koa foliar δ¹⁵N values reported here are high relative to other studies of A. koa (Burnett et al., 2022; Lawson & Pike, 2017); however, these values are in-line with N₂-fixing plants from dry-forests and grasslands (Heaton, 1987), suggesting water limitations may be influencing A. koa δ¹⁵N in these forests. Even though we did not measure water holding capacity of each soil type, the significantly higher foliar δ¹³C values in the plantation—coupled with the lack of differences in soil and foliar N despite evidence of higher BNF—is indicative of water being an important factor limiting A. koa productivity and affecting soil N properties (Ares & Fownes, 1999; Burnett et al., 2022).

Assuming water is more limited in the plantation, higher δ¹³C foliar values among A. koa and Rubus sp. individuals in the plantation may indicate higher rates of photosynthesis (reduced ¹³CO₂ discrimination) and/or higher WUE_i relative to the remnant forest (Farquhar, Ehleringer & Hubick, 1989; Cernusak et al., 2013). Considering the similarity in soil and foliar N in the two forests, we expect the differences in δ¹³C values are more likely to relate to WUE_i and not increased rates of photosynthesis. In support of this, (Ares & Fownes, 1999) found A. koa foliar δ¹³C values increased across a gradient of decreasing rainfall, and WUE_i increased in greenhouse grown seedlings experiencing drought-stress. Furthermore, it has been documented that remnant native Hawaiian forests conserve more water (Kagawa et al., 2009) and contribute more to aquifer recharge compared to plantation forests (Brauman, Freyberg & Daily, 2015). More stable water availability and less draw down of water resources in the remnant forest is consistent with patterns of greater stomatal conductance, higher ci/ca, and lower WUE_i (Farquhar, Ehleringer & Hubick, 1989). Therefore, lower foliar δ¹³C values in A. koa and Rubus sp. in the remnant forest may be an effect of less water demand in native/remnant forests, better soil water holding capacity, or both (Fig. 2). This finding is important, as the greater abundance of C₄ grasses in the understory of the plantation (Yelenik, 2017) may relate to the greater competitive ability of these grasses for water resources compared to native understory plants.

Photosynthetic fractionation is less in C₄ than C₃ plants, resulting in ¹³C-enrichment and higher leaf δ¹³C values ranging from −10‰ to −15‰ compared to C₃ plants (−21‰ to −30‰) (Farquhar, Ehleringer & Hubick, 1989; Ehleringer, Buchmann & Flanagan, 2000). In the koa plantation soils, a clear pattern emerges of higher soil δ¹³C values, consistent with a greater contribution of C₄-derived C to soils (Staddon, 2004). This finding is also supported by the soil δ¹³C isoscape (Fig. 4), where we observed increasing soil δ¹³C values in areas where grasses were dominant. N availability also influences rates of decomposition and has a direct influence on soil C pools (Averill & Waring, 2018); therefore, greater BNF in the plantation could accelerate soil C turnover, which may result in soils enriched in ¹³C (Choi et al., 2005). Accordingly, future studies should examine isotope values in understory grasses and how low C:N leaf litter, such as A. koa, is assimilated into soil and understory plant biomass, especially in the nitrophilous kikuyu grass (Cenchrus clandestinus). Follow up studies are on-going to determine rates of soil nutrient cycling in Hakalau, especially in order to disentangle how soil C, N, and leaf litter C:N influence soil δ¹³C values.

Conclusion

Recuperating biogeochemical properties in secondary tropical forests is vital to supporting plant growth and post-disturbance recovery (Sullivan et al., 2019). In some cases N₂-fixing plants may therefore be useful in facilitating nutrient and plant community recovery (Chaer et al., 2011). However, our data suggest that monoculture plantations of a native N₂-fixing tree can lead to an increase in BNF and a more homogeneous distribution of fixed N in plants and soils that is unlike the distribution of N in a primary forest—the target habitat for restoration (Figs. 4 and 5). If this “plantation-effect” promotes undesirable species such as the non-native grasses in our plantation site, then creating a homogeneous area of high BNF may hinder restoration goals. Therefore, restoration efforts should consider how the abundance and density of N₂-fixing trees may influence plant-plant and plant-soil interactions and the potential for restoration to produce ecosystem states that differ from reference ecosystem counterparts. Considering the wide application of N₂-fixing trees in tropical forest restoration, both in Hawai‘i and abroad, there remains a need to better understand how BNF by canopy trees varies within-and-among forest areas, across environmental conditions and habitat types, and its role in affecting nutrient cycling and forest communities. Addressing this uncertainty will support effective restoration strategic planning (i.e., outplanting density, multiple species planting) and management goals (i.e., habitat restoration, seedling and avian recruitment). Based on our findings in Hakalau, we suggest future restoration efforts might include a greater diversity of N₂-fixing and non-N₂-fixing canopy tree species to generate greater variation in understory and soil conditions more typical of remnant or intact forests.

Supplemental Information