The fertile island effect is stronger for larger plants irrespective of ecosystem productivity

Abstract

Productivity and function of arid and semiarid ecosystems (drylands) are optimised when critical resources such as water, organic matter and nutrients are assembled into discrete patches known as fertile patches or fertile islands. The strength of these islands is greater when the focal island plants are larger, but it is unclear whether this size effect holds true within similar-sized island species or across gradients in productivity. We examined the strength of the fertile island effect beneath three shrub species of similar structure but varying size across a gradient in productivity from ranges (low productivity), to footslopes (intermediate) and plains (high productivity). We predicted that fertility effects would be greater for larger plants and under lower productivity. There was no evidence of a productivity effect, but the fertility island effect increased with shrub size, irrespective of species. Furthermore, the fertile island effect was stronger for plant-related attributes (plants and litter) than functions associated with either soil fertility or soil stability. Our results reinforce the notion that plant size is an important determinant of the extent to which perennial plants enhance their understorey environment, irrespective of productivity or species. This has important implications for restoration programmes involving shrub plantings where the aim is to enhance the functionality of degraded dryland systems.

Introduction

An innate feature of dryland ecosystems globally is the partitioning of the landscape into two discrete patch types. One patch type (run-on zone, fertile patch and fertile island) accumulates resources such as water, nutrients, seeds and organic matter, while the other (runoff zone and interspace) supplies resources to these accumulation zones. Fertile islands are a form of geomorphic patterning that is driven by the movement of water and wind (Aguiar and Sala, 1999). Water-driven morphologies tend to be linear shaped, such as in Brousse tigrée woodlands in Africa (Thiery et al., 1995) and banded mulga groves in Australia (Tongway and Ludwig, 1994). Where resources arrive from multiple directions, generally through the action of wind, fertile islands tend to assume a spotted, isodiametric shape and are often associated with long-lived woody plants (Schlesinger et al., 1990). Fertile patches or islands typically form around large individual perennial plants or patches of plants, creating geomorphic hotspots of enhanced productivity, diversity and function (Eldridge et al., 2024). Perennial woody plants are foundation species in drylands and play pivotal roles in the functioning of ecosystems (Maestre et al., 2016; Rey et al., 2018). Woody plants act as critical habitat for biota, provide important drought reserve for livestock and native herbivores, protect the soil against wind and water erosion and contribute to the creation of patchiness in the landscape, which can lead to the formation of fertile islands (Garner and Steinberger, 1989; Lozano-Parra et al., 2018).

We have a relatively good understanding of islands created by woody plants and how they influence biotic and abiotic processes (Schlesinger et al., 1990; Ward et al., 2018; Ding and Eldridge, 2024), but studies have often conflated differences in species and their morphologies with different geomorphic positions, which typically represent different productivity levels. These two factors would be expected to influence the extent of resource capture and retention by plants and therefore island strength. Larger plants have been shown to produce stronger island effects (Eldridge et al., 2011, 2024; Ward et al., 2018), but it is unclear whether this holds true within a particular species, among a group of closely related species, or across differences in ecosystem productivity. For example, global analyses have demonstrated increases in soil fertility and surface stability with increasing shrub size across a wide range of sizes from 0.2 m (Nassauvia glomerulosa) to more than 13 m (Colutea buhsei) in height (Eldridge et al., 2024). Differences in function could be driven by height and species-level morphological effects, and their interaction, but it is unknown whether size effects still hold within similar-sized island species. Despite the many studies of woody plants and their relationships with fertile islands (Eldridge et al., 2024), we know of no studies that have specifically examined how the fertile effect might vary in relation to productivity. Under conditions of low productivity, we might expect a stronger relative fertile island effect simply due to the larger difference between conditions beneath the focal island plants and those in the corresponding non-island interspaces. Conversely, under high levels of productivity, with deeper more fertile soils and greater plant cover in the interspaces, we might expect a lower fertile island effect for the opposite reason. Sites with intermediate levels of productivity would then be expected to exhibit levels of the fertile island effect somewhere between sites in high and low productivity. An examination of how productivity might influence the extent to which plants augment their understorey and below-ground effects is important because it can shed light on the effectiveness of restoration programmes that attempt to reinstate perennial nurse plants as drivers of ecosystem structure and function in areas of different productivity.

We used three con-specific, similar-sized woody shrubs to test the relative effects of species size (volume), and ecosystem productivity (using NDVI as a proxy of productivity), on the relative ability of these plants to function as fertile islands. The three species (Maireana pyramidata, Maireana astrotricha and Maireana aphylla) are all members of the family Chenopodiaceae, a genus that is widely distributed in drylands (Kadereit et al., 2005). By selecting shrubs from the same genus, we are able to remove potential effects associated with a comparison of multiple genera on the ability of plants to supplement surface soils. Our productivity gradient was driven by differences in landform, soils and runoff–run-on relationships. Previous studies have shown that these shrubs have enhanced hydrological function (Dunkerley, 2002), trap sediment beneath their canopies (Emmerson and Facelli, 1996) and act as important nurse plants for a range of understorey protégé species (Weedon and Facelli, 2008). The link between size and the ability to intercept wind, and therefore geomorphically mobilised sediments, is well established (Field et al., 2012).

Here we examine two predictions of the fertile island effect using an arid shrubland as our model system. First, we predicted that larger shrubs would support a stronger fertile island effect, irrespective of species, and second, that the relative fertile island effect would be more pronounced in areas of low productivity (ranges) than high productivity (plains). Notwithstanding the level of productivity, the fertile island effect would be expected to increase with increasing plant size, largely because shrubs facilitate understorey protégé species and provide habitat for plant resident arthropods that exude detritus and organic material in the area at the base of the shrub. Our study contributes ecological understanding of how islands are formed beneath shrubs and how this translates into differences in soil function, soil stability and understorey plant conditions. It also has practical implications for the restoration of degraded drylands where plant species and size may be a consideration.

Methods

The study area

The research was conducted at Fowlers Gap Arid Zone Research Station in northwest New South Wales, Australia (−31.30^o, 141.71^o), an area classified as arid (Aridity Index = 0.21). The mean annual precipitation is approximately 242 mm, with most rainfall occurring in the summer (BOM, 2023). The region’s average temperatures range from 11.5 °C in winter to 27.3 °C in summer. The variability in weather patterns is influenced by the El Niño-Southern Oscillation, significantly affecting rainfall distribution, particularly in arid areas of eastern Australia (Nicholls, 1991).

The productivity gradient was represented by differences in geomorphology from ranges and hills, through footslopes to plains (Figure 1a). The ranges are characterised by steep slopes (> 20%) with shallow sandy loam soils. The plains occur in the lower elevation areas (<1% slope) that are subject to periodic flooding and characterised by relatively level areas with a mixture of clay soils and red duplex clay loams (Eldridge, 1988). Plains are intermittently flooded and receive runoff water from the surrounding hills and ranges. They are characterised by deep, clay-rich soils. Footslopes represent positions of intermediate productivity, with slopes from 5 to 10% and the soils characterised by stony-covered gibber landscapes of variable depth (Table S1). These three landscape positions represent a gradient in net primary productivity (NPP; linear model: F _2,147 = 241.0, P < 0.0001; Figure 1b) with a strong increase from ranges (low) to plains (high), though differences in NPP between low and intermediate productivity were not significantly different. The lack of a significant difference between low and intermediate productivity is likely due to the diffuse boundary between the two landforms (Eldridge, 1988). The selected species were chosen because they are widespread and representative of this region (Cunningham et al., 1992). M. pyramidata is found across the productivity gradient, Maireana astrotricha in areas of low and intermediate productivity and M. aphylla confined to areas of high productivity. These species have relatively similar shapes and sizes, and all have been described as being hemispherical perennial plants to about 1.5 m high (https://plantnet.rbgsyd.nsw.gov.au.). The three species vary in their morphology and leaf traits. For example, M. pyramidata has prominent dense fleshly leaves and can grow into a large sprawling shrub up to 1.5 m tall. Though closely related, Maireana astrotricha is a lower-growing shrub with dense and smaller leaves, while M. aphylla is largely leafless, with a twig-like appearance (Cunningham et al., 1992). All three shrubs are relatively long-lived, moderately palatable to livestock and often promote circular-shaped mounds beneath their canopies as a result of resource capture (Eldridge et al., 1990; Emmerson and Facelli, 1996). Our study site is grazed by unrestricted populations of kangaroos (Macropus spp.) and goats (Capra hircus) across the entire productivity gradient.

Figure 1. (a) Map of sampled sites within Fowlers Gap Research Station. Larger circles with red outline indicate field samples. Smaller circles show Normalized Difference Vegetation Index (NDVI) sampling points. Annual integral of NDVI values for 2022 was obtained from MOD13Q1 product in Google Earth Engine platform. Basemap was created by combining a Digital Earth Australia (DEA) Surface Reflectance (Sentinel-2) true colour composite image of 16/05/2022 and a hillshade derived from Geoscience Australia’s 5-metre Digital Elevation Model (DEM) of Australia. (b) Bar chart of mean NDVI for three landforms that are surrogates of productivity. These are ranges (low productivity – orange), footslopes (intermediate productivity – yellow) and plains (high productivity – green). Error bars depict the standard deviation from the mean; different letters show significant difference in mean NDVI (α < 0.01).

Site data collection

We selected a total of six sites, two each in low (ranges), intermediate (footslopes) and high (plains) productivity areas, separated by distances of at least 3 km (Figure 1a). At each site, we selected 10 individual plants of M. pyramidata (which occurred at all six sites) and 10 individuals of a second species, either Maireana astrotricha (in low and intermediate productivity) or M. aphylla (high productivity only). Plants were selected to encompass a range of sizes (height and diameter) and only aboveground biomass was measured. It was not possible to have all species present at all sites.

We placed a 30-cm-diameter circular quadrat beneath each shrub between the midline and the canopy edge within which we recorded the morphological and biological features of the soil surface according to the method of Tongway (1995). This was paired with a location in the interspace about 50 cm from the canopy edge of each plant. Within each quadrat, we assessed the following: 1) Surface roughness or microtopography is important because rougher surfaces have a greater ability to retain abiotic and biotic resources and to act as microsites for seeds. Surface roughness was assessed on a four-point scale from 1 (average roughness <3 mm) to 5 (deep depressions or cracks). 2) Crust resistance provides a measure of mechanical strength and therefore the likelihood that the crust will break up. It was assessed on a four-point scale from 1 (uncrusted or loose surface) to 5 (very hard and brittle). 3) Soil stability assesses how well the soil surface resists erosion and was assessed on a five-point scale of 1 (very unstable in water, as assessed using the Emerson slake test – Emerson, 1967) to 5 (extremely stable in water). 4) The absence of erosion was assessed as 1 (>50% evidence of erosion) to 5 (<10% erosion; see Tongway, 1995). The plant-related attributes were visually estimated within the same quadrat. Litter cover and depth, and plant foliage and basal cover were estimated from within the circular quadrat.

After measuring surface condition, we collected a soil sample from the surface 5–10 cm from beneath each of the 10 plants of both shrub species at each of six sites and a sample from the paired interspace, resulting in a total of 240 soil samples (two shrub species by two positions by 10 samples by six sites). Shrub volume was used as our measure of size and was calculated using the formula for an ellipsoid (volume = 4/3 π abc) where a, b and c are the semi-principal axes of the ellipsoid, i.e. a = half of the longest diameter through the centre, b is half of the diameter perpendicular to the longest diameter and c is half of the height.

Laboratory analyses

Soil samples were analysed for pH on a 1:5 soil water extract using the smartCHEM – Lab multi-parameter analyser (TPS Pty Ltd., Brendale Australia; Rayment and Lyons, 2011). Soil labile carbon determination followed the methods of Weil et al. (2002) and was measured using spectroscopy at 550 nm wavelength (UV mini-1240 spectrophotometer, Shimadzu, Japan). Nitrates and ammonium soil extraction followed methods by Keeney and Nelson (1982). Available phosphorus (Olsen method) was calculated following Rayment and Lyons (2011), best suited for alkaline soils (pH ranging from 7 to 10). XRF analysis (50 kV Vanta M-series, Olympus, USA) was used to measure the content of trace minerals Ca, Fe and K.

Statistical analyses

We compared the effects of three shrub species (M. pyramidata, M. and M. aphylla) on 16 soil physical, chemical and functional attributes with values in the adjacent unvegetated interspaces. The input data represented the relative difference between an attribute beneath the canopy and the value in the interspace for each shrub-interspace pair, i.e. (shrub − interspace) / (shrub + interspace). A positive (or negative) value of this standardised value, referred to as the relative interaction intensity (RII, sensu Armas et al., 2004), indicates a greater (or lesser) value of that attribute, respectively, beneath the shrubs. RII values range from −1 to 1, with positive values indicating greater levels of a given attribute beneath the island and vice versa. We refer to this RII as the relative fertile island effect. Evidence of the fertile island effect (either positive or negative) is based on whether the 95% confidence intervals (95% CIs), calculated using ‘Rmisc’ package in R (R Core Team, 2018), cross the zero line.

The RII values for the 16 attributes were assembled into three broad functional proxies related to soil fertility (based on soil NH₄, NO₃, Ca, Fe, labile C, pH, P and K, hereafter ‘fertility’), 2) ecosystem stability (based on the absence of surface erosion, surface resistance, roughness and stability, hereafter ‘integrity’) and 3) plant attributes, based on plant basal and foliage cover, and litter cover and depth (hereafter ‘plant’).

We examined how the fertile island effect differed in relation to productivity (using the three categorical variables of ranges, footslopes and plains), shrub species and shrub volume using a Bayesian hierarchical linear mixed model. Our RII values were modelled with a Gaussian (normal) distribution, with all individual ecosystem attributes (n = 16) estimated simultaneously in a single model. The standardised response variable (RII) was modelled hierarchically as a function of landform (ranges, slopes and plains), shrub species (M. pyramidata, M. astrotricha and M. aphylla), shrub volume (continuous data) and functional category (fertility, integrity and plant). The model fitted individual ecosystem functional attributes as groups (random intercepts) with varying slopes associated with landform and shrub species. We also included interactions between the three functional categories, and landform and shrub species to account for potential differences in the effects of each covariate within each ecosystem function category. Site was included as a random intercept to account for potential nonindependence from the same plot.

We specified weakly informative normally distributed priors for the intercept and all regression coefficients (mean = 0 and scale = 2.5). Default priors were used for sigma (exponential, rate = 1) and variance–covariance matrix of the varying intercepts and slope parameters (shape and scale of 1). We used posterior simulations of model parameters with the No-U-Turn Hamiltonian Monte Carlo sampler (Stan, Carpenter et al., 2017). Posterior distributions were estimated from four chains, each with 1000 iterations, after discarding the preliminary 1000 iterations. Visual diagnostics (autocorrelation, trace plots and posterior predictive checks) and effective sample size examination (min. 1000) and r^ values (<1.01) were used to assess model convergence. Models were fitted using the package ‘rstanarm’ (Goodrich et al., 2020) within R (R Core Team, 2018). A hierarchical model provides several benefits over simple averaging of standardised indicators or multiple separate models (McElreath, 2020): (i) simultaneous modelling of multiple attributes improves precision and estimates of uncertainty for each ecosystem function category; (ii) nonindependence of multiple attributes within sites is explicitly accounted for; (iii) enables simultaneous estimation of overall fertile island effect for each ecosystem functional category and the individual soil attributes within these (Eldridge et al., 2024).

Results

We found strong evidence of an increase in the strength of the fertile island effect with increasing plant size (H₁) but no effect of productivity (H₂). There were strong fertile island effects for all species in all productivity levels, though the sign of the effects varied among attributes (Figure 2). Seventy percent of a possible 96 attributes by productivity-level effects were significant, and of these, 54% (52) were significantly positive and 16% (14) significantly negative (Figure 2). Trends across levels of productivity were generally similar, i.e. either consistently significantly positive (e.g. surface roughness, foliage cover, basal plant cover and soil carbon) or consistently nonsignificant (e.g. soil resistance and soil stability). Only two attributes (e.g. NH₄ and cover of erosion) showed inconsistent effects across productivity levels. The fertile island effect for the plant-related function (estimate = 0.82, P < 0.001) was considerably greater than functions related to surface integrity or soil fertility, and this was irrespective of productivity or shrub species (Figure 3).

Figure 2. Mean (± 95% CI) relative fertile island effect for the 16 attributes in relation to productivity and shrub species. Attributes whose confidence intervals do not cross the x = 0 line are significant. Blue and red colours represent positive and negative fertile island effects, respectively. Grey colours represent attributes where the fertile island effect was nonsignificant.

Figure 3. Mean (± 95% CI) fertile island effect for the three productivity positions averaged across all species (overall) and effects for individual species.

Overall, we found that larger shrubs had stronger fertile island effects overall (estimate = 0.36, P = 0.001; Table S2), though this was largely due to shrub size effects under moderate productivity (footslopes), as increasing plant size was associated with declines in the effects under high (plains, model estimate = −0.26) and low (model estimate = −0.22) productivity compared with intermediate productivity (footslopes, Figure 4). Plant basal cover (F _1,118 = 5.90, P = 0.017), soil nitrate (F _1,94 = 12.68, P = 0.0006), soil ammonium (F _1,94 = 17.94, P = 0.0016) and soil labile carbon (F _1,114 = 5.17, P = 0.025) increased significantly with increasing shrub volume, but there were no significant effects for the other 12 attributes (Figure 5). Increases in the volume of M. pyramidata were associated with significant increases in the overall fertile island effect (model estimate = 0.10, PD = 98.1%, Figure 6, Table S1).

Figure 4. The fertile island effect (RII) for the three shrub species in relation to shrub volume (m³).

Figure 5. The relative fertile island effect (RII) for 16 attributes in relation to shrub volume (m³).

Figure 6. The fertile island effect (RII) for shrub volume (m³) across different levels of productivity.

Discussion

We sought to test two predictions related to the fertile island effect in an arid shrubland, examining the effects of three conspecific shrubs growing across a gradient in productivity driven by changes in geomorphology. Our first hypothesis was upheld, with clear increases in the strength of the fertile island effect with increasing shrub size, though this effect varied weakly among functions and depending on productivity (Figure 5). Contrary to prediction, we found that the strength of the island effect was consistent across the productivity gradient, irrespective of shrub species or specific function (Figure 3). Overall, our results reaffirm the view that larger plants have a stronger enrichment effect on their understorey biotic and abiotic environment, irrespective of productivity, and even within a narrow range of plant sizes. This knowledge advances our understanding of resource distribution in drylands and provides important information for practitioners to focus on the conservation and restoration of larger plants when targeting the restoration of degraded shrublands.

Shrubs promote fertile islands irrespective of productivity

Overall, the fertile island effect was apparent in 70% of the shrub–attribute combinations, irrespective of productivity, species identity or shrub size (Figure 2). However, island effects were stronger for fertility, principally due to increasing ammonium, nitrate and phosphorus enrichment (Figure 5). Nitrate enrichment has been demonstrated in fertile islands (e.g. Iwaoka et al., 2018) due to both increased nitrogen mineralisation resulting from microbial decomposition in plant mounds (Li et al., 2021), ammonium-oxidising bacteria beneath the plants (Trivedi et al., 2019) and/or the presence of nitrate-reducing bacteria such as Stenotrophomonas (Palleroni, 2015; Sly et al., 2015) in the interspaces. Furthermore, bacterial communities have been shown to differ in their nutritional strategies between interspaces (oligotrophs) and islands (copiotrophs). For example, Actinomycetota and Bacteroidota tend to occur beneath plant canopies, whereas Acidobacteriota (oligotroph) tend to occur in the bare interspaces (Xiang et al., 2018; Doniger et al., 2020). The process of resource capture and retention is critically important in drylands where resources are scarce and have a patchy distribution. Our results highlight the contribution of shrubs to the maintenance of stability and resilience of dryland ecosystems, and this contribution will likely increase as climates become hotter, drier and more variable.

Greater function associated with larger plants

Plant volume, our proxy of size, was the attribute most strongly associated with resource supplementation, with larger plants associated with more functional islands, but only for soil fertility, and mostly in areas of moderate productivity (footslopes). Larger plants have a stronger effect on resource enrichment (Garner and Steinberger, 1989; de Soyza et al., 1996; Ward et al., 2018; Ochoa-Ochoa‐Hueso et al., 2018; Ding and Eldridge, 2024) due to more litter shedding, the production of larger pools of organic matter and, therefore, more nitrogen beneath the canopies (Trivedi et al., 2019). Larger plants are also focal points for meso- and macrofaunal activity (Dean et al., 1999) and provide more shade, sequestering more carbon (Lal, 2019), thus providing more biomass, which in turn provides additional resources for other organisms that drive nutrient cycling (Bahram et al., 2018). Our three shrub species are known to live for more than half a century (Lay, 1979) and are considered important due to their large size (Yan et al., 1996), which makes them critical habitat for plants (Emmerson and Facelli, 1996) and refugia for animals during dry times (Eldridge and Rath, 2002).

In our study, plant size was skewed towards smaller plants, with very few larger plants. Only two plants M. pyramidata (low productivity, 1.7 m³) and M. aphylla (high productivity, 1.9 m³) exceeded 1.5 m³ in volume. Yet, removing these two very large shrubs did not alter our results, with larger plants still associated with greater function (model estimate = 0.32), and plant function was still stronger than either fertility or integrity functions (model estimate = 0.80). The predominance of smaller shrubs could result from grazing by goats, particularly during droughts. All three shrub species are moderately palatable to livestock (Eldridge et al., 1990), but the distribution of grazing by sheep, feral goats and kangaroos is relatively uniform across the study site, so we are confident that older shrubs are generally larger. Shrub size is a general proxy of age in woody plants (Day et al., 2002), at least during the early stages of growth, and older shrubs are likely to be larger and therefore have a greater effect on resource accumulation beneath their canopy. We used shrub size as our predictor of the fertile island effect because it is easy to measure, strongly related to age and we know that larger plants produce and trap more resources. We acknowledge that grazing complicates this relationship, so heavily grazed smaller shrubs might be relatively older than ungrazed larger shrubs. Data on woody plant age are difficult to obtain, particularly in drylands. We suggest that where possible, future studies should endeavour to sample a wider range of shrub sizes that include more larger shrubs.

Global studies indicate that canopy structure is important for promoting and maintaining organic matter and nutrients, particularly for hemispherical-shaped plants (Li et al., 2007). Plant height and crown width have been shown to be important, potentially by increasing threshold wind velocities in the vicinity of shrubs, causing aeolian material to be deposited close to and within the canopy (Yan et al., 2019), which intensifies with increasing basal cover (Field et al., 2012). Larger plants also have a greater hydrological effect on their surrounding environments, intercepting more rainfall because of their higher leaf area index (Whitford et al., 1997), and therefore potentially greater enhancement of soil moisture beneath their canopies. These effects were greater under plants with larger canopies despite the effects of larger plants infiltrating further into the interspaces (Dunkerley, 2002).

Compared with functions associated with surface integrity and fertility, the greatest effect of islands was on understorey plants and litter, and this effect was independent of shrub species identity. Our results suggest that resource enrichment is a signature of individual shrubs rather than an effect due to a specific geomorphic position that might differ in productivity. Our results showed that island effects were strongest for biotic factors (Figure 2.). Attributes associated with plants (plant cover and litter cover) far exceeded those associated with either soil fertility or soil structural stability. Furthermore, this positive effect of shrubs on plant function was independent of productivity, though there were some weak shrub size and species identity effects (Figure 3.). The strong plant size effect could also be related to the fact that larger plants tend to have more of their foliage skirting the soil surface. This trait has been identified as an important driver of function in a range of shrubs globally (Eldridge et al., 2011; Ding and Eldridge, 2024) and relates to the ability of plants to intercept aeolian material. Canopy soil contact is a key driver of processes involved in the formation of nebkhas (El-Bana et al., 2002).

Conclusions

The ability of shrubs to capture and sequester resources, improve soil fertility, support understorey plants and animals and provide forage for livestock and native herbivores during dry periods, yet survive long periods of drought stress, makes them particularly valuable in drylands (Ward et al., 2018). We show that plant size is a critically important determinant of the strength of the fertile island effect, even for plants within a narrow range of sizes. Our work complements previous work showing that the fertile island effect increases with increasing plant size across a large size gradient from small grasses to tall trees. This knowledge is important because it gives us important insights into the functional benefits of rewilding using plant species of different traits. Larger plants are likely to have a greater effect on their understorey environment, and together with their value as habitat, this makes them key species to include in any restoration programme in drylands.