Impact statement
Although woody plants and herbivores are widely acknowledged as major dryland ecosystem engineers, we still know relatively little about their contribution to shaping understory vegetation structure and soil condition. Using a field survey supporting a nearly orthogonal design that effectively isolates herbivores from woody encroachment effects, we show that herbivores are a dominant player in dryland biodiversity and functioning compared with woody plant cover. We also show that most of the negative effects are due to pressure from livestock rather than from native herbivores, and that those effects, as well as those of woody species, are dampened under productive ecosystems. We show that dryland engineering effects are only partly driven by their effects on heterogeneity, and that mitigation by increasing the diversity of these engineers, as previously suggested, has limited effects. Our results suggest that focusing on livestock management rather than on woody removal can be a more effective way to sustain the structure and function of drylands.
Introduction
Professor Walter (Walt) Whitford had a fascination for the contribution of animals – from small soil-digging biota to livestock grazing – as ecosystem engineers, influencing a large range of ecosystem attributes including plant and insect diversity and a number of soil attributes (De Soyza et al., Reference De Soyza, Whitford, Turner, Van Zee and Johnson2000; Whitford, Reference Whitford2002). Livestock grazing is a major land-use type in drylands (ILRI, 2021; Briske et al., Reference Briske, Niamir-Fuller, Davies, Waters-Bayer, Hutchinson and Samuels2026), and both livestock and wild herbivores play important roles by dispersing seeds (Janišová et al., Reference Janišová, Škodová, Magnes, Iuga, Biro, Ivașcu, Ďuricová and Buzhdygan2025), altering soil surface heterogeneity through trampling, grazing and dung deposition and may reduce the abundance of the more competitive plants (Milchunas et al., Reference Milchunas, Sala and Lauenroth1988; De Bello et al., Reference De Bello, Leps and Sebastià2007; Price et al., Reference Price, Sitters, Ohlert, Tognetti, Brown, Seabloom, Borer, Prober, Bakker, MacDougall, Yahdjian, Gruner, Olde Venterink, Barrio, Graff, Bagchi, Arnillas, Bakker, Blumenthal, Boughton, Brudvig, Bugalho, Cadotte, Caldeira, Dickman, Donohue, Grégory, Hautier, Jónsdóttir, Lannes, McCulley, Moore, Power, Risch, Schütz, Standish, Stevens, Veen, Virtanen and Wardle2022). Whitford identified generally negative effects of overgrazing by large European ungulates in drylands, though mostly through their impacts around watering points and dryland condition assessment (e.g., Whitford et al., Reference Whitford, De Soyza, Van Zee, Herrick and Havstad1998; Kerley and Whitford, Reference Kerley and Whitford2000). He also recognized the ecosystem engineering roles played by woody plants in drylands (Whitford et al., Reference Whitford, Anderson and Rice1997; Eldridge et al., Reference Eldridge, Bowker, Maestre, Roger, Reynolds and Whitford2011). For example, Whitford investigated how desert shrubs such as Larrea tridentata generate fertility islands, altering their physical and chemical environment (De Soyza et al., Reference De Soyza, Whitford, Martinez-Meza and Van Zee1997; Whitford et al., Reference Whitford, Anderson and Rice1997). Initially, Whitford and colleagues drew attention to the role of woody plants encroachment as being symptomatic of desertified environments (Schlesinger et al., Reference Schlesinger, Reynolds, Cunningham, Huenneke, Jarrell, Virginia and Whitford1990). However, the relationship between woody encroachment and desertification has since been further refined, with the functional traits of the encroaching woody plants, the densities they attain or environmental factors ranging from climate to the intensity of livestock grazing modulating this relationship (e.g., Maestre et al., Reference Maestre, Bowker, Puche, Belén Hinojosa, Martínez, García-Palacios, Castillo, Soliveres, Luzuriaga, Sánchez, Carreira, Gallardo and Escudero2009; Soliveres et al., Reference Soliveres, Maestre, Eldridge, Delgado-Baquerizo, Quero, Bowker and Gallardo2014a; Stanton et al., Reference Stanton, Boone, Soto‐Shoender, Fletcher, Blaum and McCleery2018). Due to the important role of both herbivores and woody plants as dryland ecosystem engineers, a better understanding of their ecological effects and context-dependencies is critical for advancing our understanding on how to sustainably manage drylands.
The ecological effects of woody plants and herbivores may depend on their complex interactions and influence on spatial heterogeneity, as well as the environmental context in which they operate. A number of factors can modulate the sign and intensity of these dryland engineering effects. First, the abundance of both woody plants and herbivores is highly influential. Though beneficial at low abundances, both high grazing loads and woody encroachment levels often impair ecosystem diversity and function, with ecological attributes including bird and plant diversity or soil fertility showing hump-shaped relationships with the abundance of these dryland engineers (Sirami et al., Reference Sirami, Seymour, Midgley and Barnard2009; Soliveres et al., Reference Soliveres, Maestre, Eldridge, Delgado-Baquerizo, Quero, Bowker and Gallardo2014a; Chillo et al., Reference Chillo, Ojeda, Capmourteres and Anand2017; Maestre et al., Reference Maestre, le Bagousse-Pinguet, Delgado-Baquerizo, Eldridge, Saiz, Berdugo, Gozalo, Ochoa, Guirado, García-Gómez, Valencia, Gaitán, Asensio, Mendoza, Plaza, Díaz-Martínez, Rey, Hu, He, Wang, Lehmann, Rillig, Cesarz, Eisenhauer, Martínez-Valderrama, Moreno-Jiménez, Sala, Abedi, Ahmadian, Alados, Aramayo, Amghar, Arredondo, Ahumada, Bahalkeh, Ben Salem, Blaum, Boldgiv, Bowker, Bran, Bu, Canessa, Castillo-Monroy, Castro, Castro, Castro-Quezada, Chibani, Conceição, Currier, Darrouzet-Nardi, Deák, Donoso, Dougill, Durán, Erdenetsetseg, Espinosa, Fajardo, Farzam, Ferrante, Frank, Fraser, Gherardi, Greenville, Guerra, Gusmán-Montalvan, Hernández-Hernández, Hölzel, Huber-Sannwald, Hughes, Jadán-Maza, Jeltsch, Jentsch, Kaseke, Köbel, Koopman, Leder, Linstädter, le Roux, Li, Liancourt, Liu, Louw, Maggs-Kölling, Makhalanyane, Issa, Manzaneda, Marais, Mora, Moreno, Munson, Nunes, Oliva, Oñatibia, Peter, Pivari, Pueyo, Quiroga, Rahmanian, Reed, Rey, Richard, Rodríguez, Rolo, Rubalcaba, Ruppert, Salah, Schuchardt, Spann, Stavi, Stephens, Swemmer, Teixido, Thomas, Throop, Tielbörger, Travers, Val, Valkó, van den Brink, Ayuso, Velbert, Wamiti, Wang, Wang, Wardle, Yahdjian, Zaady, Zhang, Zhou, Singh and Gross2022). Second, the diversity of both encroaching woody species and herbivores may influence their effect. Mixed assemblages of grazers and browsers can have more positive effects on ecosystem functioning as well as on plant, microbial and animal diversity than mono-specific assemblages (Prins and Fritz, Reference Prins, Fritz, Gordon and Prins2008; Wang et al., Reference Wang, Delgado-Baquerizo, Wang, Isbell, Liu, Feng, Liu, Zhong, Zhu, Yuan, Chang and Liu2019; Andreoni et al., Reference Andreoni, Bestelmeyer, Lightfoot and Schooley2024). Woody assemblages with diverse traits (e.g., nitrogen cf. non-nitrogen-fixing plants; canopy shape) have contrasting effects on ecosystems (De Soyza et al., Reference De Soyza, Whitford, Martinez-Meza and Van Zee1997; Eldridge et al., Reference Eldridge, Bowker, Maestre, Roger, Reynolds and Whitford2011). Therefore, diverse arrays of encroaching woody species could show complementary effects and enhance the positive response of ecosystem diversity and functioning to shrub encroachment; something still poorly explored. Moreover, the effects of woody plants and herbivores on vegetation structure and function can be context dependent, and especially responsive to ecosystem productivity. More productive ecosystems are often less heterogeneous. Thus, increasing heterogeneity, for example, by generating open spaces or concentrating nutrients and water, could be more beneficial under productive than unproductive environments (Proulx and Mazumder, Reference Proulx and Mazumder1998; Maestre et al., Reference Maestre, le Bagousse-Pinguet, Delgado-Baquerizo, Eldridge, Saiz, Berdugo, Gozalo, Ochoa, Guirado, García-Gómez, Valencia, Gaitán, Asensio, Mendoza, Plaza, Díaz-Martínez, Rey, Hu, He, Wang, Lehmann, Rillig, Cesarz, Eisenhauer, Martínez-Valderrama, Moreno-Jiménez, Sala, Abedi, Ahmadian, Alados, Aramayo, Amghar, Arredondo, Ahumada, Bahalkeh, Ben Salem, Blaum, Boldgiv, Bowker, Bran, Bu, Canessa, Castillo-Monroy, Castro, Castro, Castro-Quezada, Chibani, Conceição, Currier, Darrouzet-Nardi, Deák, Donoso, Dougill, Durán, Erdenetsetseg, Espinosa, Fajardo, Farzam, Ferrante, Frank, Fraser, Gherardi, Greenville, Guerra, Gusmán-Montalvan, Hernández-Hernández, Hölzel, Huber-Sannwald, Hughes, Jadán-Maza, Jeltsch, Jentsch, Kaseke, Köbel, Koopman, Leder, Linstädter, le Roux, Li, Liancourt, Liu, Louw, Maggs-Kölling, Makhalanyane, Issa, Manzaneda, Marais, Mora, Moreno, Munson, Nunes, Oliva, Oñatibia, Peter, Pivari, Pueyo, Quiroga, Rahmanian, Reed, Rey, Richard, Rodríguez, Rolo, Rubalcaba, Ruppert, Salah, Schuchardt, Spann, Stavi, Stephens, Swemmer, Teixido, Thomas, Throop, Tielbörger, Travers, Val, Valkó, van den Brink, Ayuso, Velbert, Wamiti, Wang, Wang, Wardle, Yahdjian, Zaady, Zhang, Zhou, Singh and Gross2022).
In addition to the context-dependent effects of woody plants and herbivores, their engineering effects are not independent. Grazing enhances woody encroachment, which in turn may reduce pastoral value of drylands (Kerley and Whitford, Reference Kerley and Whitford2000; Whitford and Steinberger, Reference Whitford, Steinberger, Dubinsky and Seckbach2010; Anadón et al., Reference Anadón, Sala, Turner and Bennett2014). These two engineers also interact to drive ecosystem diversity and functioning, with large grazing loads impairing the otherwise positive effects of shrubs (Soliveres and Eldridge, Reference Soliveres and Eldridge2014; Stanton et al., Reference Stanton, Boone, Soto‐Shoender, Fletcher, Blaum and McCleery2018) and woody encroachment can affect the intensity of herbivory via ‘landscapes of fear’ (Laundré et al., Reference Laundré, Hernández and Ripple2010) or reduced accessibility (Riginos and Grace, Reference Riginos and Grace2008). Their likely interactions and mutual influences make the role of herbivores and woody encroachment difficult to tease apart. Both receive substantial research attention separately, and both have profound implications for dryland sustainability. Yet, an important research gap involves the relative importance of encroaching woody plants and herbivores as engineers of dryland biodiversity and functioning. A greater importance of woody encroachment would require prioritizing woody plant management, whereas a larger importance of grazing would imply that reducing livestock loads should be the main management target. Both perspectives have received conflicting support in previous research (e.g., Schlesinger et al., Reference Schlesinger, Reynolds, Cunningham, Huenneke, Jarrell, Virginia and Whitford1990; Anadón et al., Reference Anadón, Sala, Turner and Bennett2014; Eldridge and Soliveres, Reference Eldridge and Soliveres2015).
Here, we present an observational study with a nearly orthogonal design for both woody plants and herbivore abundance (independent of each other) as well as for their abundances compared with their diversities. Our study includes 451 sites in eastern Australia. This database has been used previously to evaluate the effects of grazing on ecosystem functions, soils and biota (e.g., Eldridge et al., Reference Eldridge, Delgado-Baquerizo, Travers, Val, Oliver, Dorrough and Soliveres2017, Reference Eldridge, Delgado-Baquerizo, Travers, Val and Oliver2018a; Val et al., Reference Val, Eldridge, Travers and Oliver2018; Travers et al., Reference Travers, Eldridge, Val and Oliver2019; Eldrige et al., Reference Eldridge, Oliver, Val and Travers2020). Here, we advanced previous work to evaluate the relative contribution of both dryland engineers to different aspects of vegetation composition and structure (plant richness and cover and biocrusts), and soil attributes (soil nutrients, biological activity and erosion). We teased apart the effects of woody diversity from their abundance, as well as those of herbivores, considering also the interactions between both engineer effects as well as their dependency on ecosystem productivity and the role of environmental heterogeneity. We hypothesized that: i) since grazing is a major driver of woody encroachment, we expect the effects of grazing to exceed those of woody encroachment, ii) the response to woody encroachment and grazing will differ among the different response variables assessed. We expect, for example, plant attributes to be more responsive than soils to both engineers, soils to be more positively affected by woody plants than herbivores due to their larger and more continuous organic inputs and biocrusts to be more negatively influenced by herbivores than woody plants due to their sensitivity to trampling, iii) effects of woody plants and herbivores will become more positive with higher diversity within these groups (positive abundance × diversity interactions) as well as with higher productivity (positive productivity × abundance interactions) and iv) most effects of woody plants and herbivores will be related to changes in heterogeneity (i.e., effect on variation related to effect on the average of different ecosystem attributes within a given site).
Methods
Our study includes 451 sites distributed across eastern Australia (900 km longitudinal gradient; and area of ~0.4 million km2; described in Eldridge et al., Reference Eldridge, Delgado-Baquerizo, Travers, Val, Oliver, Dorrough and Soliveres2017; Figure 1). Average annual rainfall ranges from 385 to 460 mm, and average annual temperature is ~18°C in all cases. Soil textures range from sandy clay loams to silty loams. Our 451 sites were equally distributed among three different communities (black box [Eucalyptus largiflorens], white cypress pine [Callitris glaucophylla] and river red gum [Eucalyptus camaldulensis]), with contrasting ecosystem productivity (Eldridge et al., Reference Eldridge, Delgado-Baquerizo, Travers, Val and Oliver2018a; Figure 1).
Location of the 451 study sites in southeastern Australia. Different land-uses and community types are shown. Inset shows location within Australia.
Figure 1. Long description
The layout is divided into three main sections.
On the far left are two legends. The Land use legend includes Conservation (light green), Cropland (yellow), Forestry (purple), Rangelands (orange), Other (grey), and Water (light blue). The Community legend identifies Blackbox (grey squares), Cypress pine (green triangles), and Red gum (pink circles).
In the center is a large map of southeastern Australia spanning 32 degrees South to 36 degrees South and 142 degrees East to 148 degrees East. A 200 kilometer scale bar is at the bottom right. An inset map at the top left shows the study area as a blue rectangle within the state of New South Wales, Australia. The main map is dominated by orange Rangelands and yellow Cropland. The 451 study sites are clustered in the center. Green triangles (Cypress pine) are concentrated in the North and East. Grey squares (Blackbox) are primarily in the West and center. Pink circles (Red gum) are clustered in the South and center.
On the far right are three vertical reference photos. The top photo shows a Blackbox community with low, scrubby bushes. The middle photo shows a Cypress pine community with upright, needle-leaf trees. The bottom photo shows a Red gum community with tall, white-barked eucalyptus trees.
The study originally focused on herbivory, establishing herbivory intensity gradients based upon distance to water points and dung counts, but also including grazing enclosures and low grazing conservation reserves (Eldridge et al., Reference Eldridge, Delgado-Baquerizo, Travers, Val, Oliver, Dorrough and Soliveres2017). However, this gradient implicitly accounted for a nearly orthogonal gradient in woody plant abundance (grazing cf. woody abundance (Spearman’s correlation = −0.05; 0.03 for their richness; see Supplementary Table S1 for results within each community), known to independently affect different aspects of plant species composition (Eldridge et al., Reference Eldridge, Delgado-Baquerizo, Travers, Val and Oliver2018a; Travers et al., Reference Travers, Eldridge, Val and Oliver2019), including the abundance of exotic species and also soil attributes (Eldridge et al., Reference Eldridge, Delgado-Baquerizo, Travers, Val, Oliver, Dorrough and Soliveres2017, Reference Eldridge, Oliver, Val and Travers2020). Here we build upon this previous knowledge to better understand: i) the relative importance of woody plants and herbivores, and their potential interactions, as dryland engineers, ii) the influence of their abundance, diversity as well as ecosystem productivity, in the effects of these engineers and iii) the contribution of heterogeneity in mediating the observed effects.
At each site, we established a 200-m-long transect, along which we placed five 5 m × 5 m quadrats every 50 m. Within each quadrat, we recorded plant species composition (cover by species) and embedded a smaller (0.5 m × 0.5 m subplot) to assess the soil surface condition and sample the soils. These quadrats included a range of open, timbered and shrubby microsites, depending on whatever microsite occurred in the randomly selected location of each quadrat. This soil assessment included biocrust cover, as well as different indicators related to indices of nutrient cycling, resistance to erosion (stability) and water infiltration (Tongway and Hindley, Reference Tongway and Hindley2001; see further details in Eldridge et al., Reference Eldridge, Delgado-Baquerizo, Travers, Val, Oliver, Dorrough and Soliveres2017). We sampled soils (0–5 cm) to measure available nutrients (labile carbon, available nitrogen and phosphorus) and measured soil bulk density as an indicator of soil compaction (higher bulk density = greater soil compaction). From each of the 451 sites we calculated the average across the subplots for soil attributes, including biocrust cover. We also obtained their standard deviation as an estimate of heterogeneity (assuming that larger deviations = more heterogeneous). Adding up the five quadrats, we obtained an overall estimate of native plant cover and richness and exotic plant cover.
We measured woody species abundance by measuring all woody plants, by species, within a 10 m by 200 m plot along the transect. Under conditions of very high abundance, we measured two or three smaller sections (e.g., 20 m by 10 m) of the larger 2,000 m2 plot. Woody plants were separated into trees and shrubs and N-fixers and non-fixers. Woody species abundances ranged from 25 to ~56,000 individuals ha−1. Woody richness, that is, presence of trees and shrubs, N-fixing and non-fixing, ranged from 1 to 4. To estimate grazing intensity, we identified and counted the dung or pellets of all herbivores within the large (cattle, sheep/goat, kangaroo) and small (kangaroo, rabbit, sheep/goat) quadrats. These dung included a mixture of fresh and older dung, but there was no evidence that dung integrity varied among the three communities, that is, in relation to productivity. We then counted, collected, dried and weighed the dung collected from 10 large random quadrats to obtain a relationship between dung counts and dry mass for each herbivore type. This relationship was then used to calculate the total oven-dried mass of dung per hectare per herbivore as our measure of recent grazing intensity. This dung mass provided a good proxy of recent grazing over the past 2–3 years (Eldridge et al., Reference Eldridge, Delgado-Baquerizo, Travers, Val and Oliver2018a, Reference Eldridge, Sáez-Sandino, Maestre, Ding, Guirado and Delgado-Baquerizo2025). Dung mass ranged from 0 to 1,159 kg ha−1 and herbivore richness from 0 (no herbivores present) to 4 (cattle, sheep/goat, rabbit and kangaroo present). Woody plant richness and abundance were not strongly correlated (Spearman’s rank correlation = 0.33), whereas herbivore abundance and richness were (Spearman’s rank correlation = 0.76).
We obtained net primary productivity (NPP) for each site using MODIS satellite imagery data (https://modis.gsfc.nasa.gov/data/) with a spatial resolution of 1 km to equate that of climatic data. Average NPP values were 0.64 g C m−2 day−1 at Black box, 0.75 g C m−2 day−1 at White cypress pine and 1.03 g C m−2 day−1 at River redgum communities (Eldridge et al., Reference Eldridge, Delgado-Baquerizo, Travers, Val, Oliver, Dorrough and Soliveres2017). Although communities differed in their average productivity, there was also substantial variation within them (Supplementary Figure S2). Thus, we used NPP instead of community type as a predictor in our statistical analyses, as it is a continuous, more informative and more generalizable predictor.
Statistical analyses
We performed linear models using the same model structure for all response variables (species richness of native plants, understory plant cover, proportion of exotic plants by cover, biocrust cover, labile C concentration, available N and P, surface stability, infiltration and stability indices and soil bulk density). The model structure included NPP, woody abundance (individuals ha−1) and richness (1–4; N-fixing and non-fixing shrubs and trees) and grazing abundance (dung mass kg ha−1) and richness (0–4; cattle, sheep/goat, kangaroo and rabbit). To evaluate context-dependencies on the effect of dryland engineers, we included pairwise interactions between our predictors (NPP, woody plant abundance, woody plant species richness, herbivore abundance and herbivore richness, as well as three-way interactions between richness and abundance of woody plants or herbivores with NPP). Our models always included latitude and longitude to account for the spatial structure in our data, as well as precipitation and temperature seasonality to account for climatic variation that could generate additional heterogeneity across our sites not captured by NPP. From the latter, only precipitation seasonality was retained in our models, as latitude, longitude and temperature seasonality generated multicollinearity (VIF 5.1, 8.6 and 13.4, respectively). The same VIF scores showed that, despite their strong correlation, herbivore richness and abundance could be used in the same model without inducing multicollinearity (VIF < 1.5); hence, we retained both as they were directly linked to our hypotheses. In addition to removing those predictors inducing multicollinearity, we removed those predictors from our models that did not contribute to the explanatory power of any of the response variables tested (based on both AIC and adjusted R 2). These included the three-way interactions (richness × abundance of woody plants or herbivores × NPP), as well as the interactions among richness × abundance of herbivores, richness × abundance of woody plants and grazing × woody plant abundance; see Results below). We retained the rest of terms in the models regardless of the parsimony or their specific effects on each response variable, in order to obtain a common model structure for all response variables that would allow us to compare results between them.
Australian soils are poor in phosphorus (Kooyman et al., Reference Kooyman, Laffan and Westoby2017) and N-fixing plants could cause detrimental effects by altering nutrient stoichiometry (N-to-P ratios), as observed elsewhere (Delgado-Baquerizo et al., Reference Delgado-Baquerizo, Maestre, Gallardo, Bowker, Wallenstein, Quero, Ochoa, Gozalo, García-Gómez, Soliveres, García-Palacios, Berdugo, Valencia, Escolar, Arredondo, Barraza-Zepeda, Bran, Carreira, Chaieb, Conceição, Derak, Eldridge, Escudero, Espinosa, Gaitán, Gatica, Gómez-González, Guzman, Gutiérrez, Florentino, Hepper, Hernández, Huber-Sannwald, Jankju, Liu, Mau, Miriti, Monerris, Naseri, Noumi, Polo, Prina, Pucheta, Ramírez, Ramírez-Collantes, Romão, Tighe, Torres, Torres-Díaz, Ungar, Val, Wamiti, Wang and Zaady2013). In turn, the relatively short evolutionary history of grazing by European livestock on the Australian continent could generate more detrimental effects of hooved animals than other herbivores (Milchunas et al., Reference Milchunas, Sala and Lauenroth1988; Eldridge et al., Reference Eldridge, Delgado-Baquerizo, Travers, Val, Oliver, Dorrough and Soliveres2017; Price et al., Reference Price, Sitters, Ohlert, Tognetti, Brown, Seabloom, Borer, Prober, Bakker, MacDougall, Yahdjian, Gruner, Olde Venterink, Barrio, Graff, Bagchi, Arnillas, Bakker, Blumenthal, Boughton, Brudvig, Bugalho, Cadotte, Caldeira, Dickman, Donohue, Grégory, Hautier, Jónsdóttir, Lannes, McCulley, Moore, Power, Risch, Schütz, Standish, Stevens, Veen, Virtanen and Wardle2022). Thus, to disentangle these potential effects, we duplicated our models using the abundance of N-fixing woody species rather than overall woody abundance and richness and the abundance of livestock, instead of overall herbivore abundance and richness.
The same modeling procedure was performed separately for both average and standard deviation (across the five quadrats per site) of all our response variables, and the results compared to determine those effects on the average that showed a similar response to those on their heterogeneity. Analyses using heterogeneity (standard deviation) were only considered for the soil attributes determined from samples collected in the 0.5 m × 0.5 m subplots. For native plant species richness, cover and proportion of exotic species, we obtained an overall measure using all sampling quadrats for a given site. In all cases, we obtained model outputs as implemented in the MASS and lm.beta libraries in R (Ripley and Venables, Reference Ripley and Venables2002; Behrendt, Reference Behrendt2025).
Results
Stronger influence of herbivores than woody plants
Herbivores showed negative associations with native plant richness and cover (all natives including woody and herbaceous) and had varying relationships (positive for abundance, negative for richness) with the dominance of exotic plants (Figure 2). The larger importance of herbivores as dryland engineers was also apparent for soil attributes, with lower infiltration and stability indices under high herbivore pressure and a slight, though nonsignificant, decline in nutrient cycling and increase in soil compaction (higher bulk density; Figure 2). Most of the negative associations were found for herbivore abundance (with the exception of those on native plant species richness) and linked to the abundance of livestock (cattle, sheep/goat). Indeed, when limiting herbivore abundance to livestock only, stronger negative and positive effects were apparent for nutrient cycling and bulk density, respectively (Figure 2).
Heatmap showing the linear model results on plant- and soil-related attributes (rows). We show the standardized regression coefficients of the different predictors (columns) evaluated, with bold font and asterisks showing statistically significant (P < 0.05) effects. Predictors are organized by productivity (NPP) and the two dryland engineers (woody plants, herbivores), as well as the interactions between NPP and the effect of dryland engineers. For all variables, the adjusted R 2 (Model adj. R 2) is shown in the first column, as well as a comparison between overall dryland engineers’ effects (abundance [abund.], richness) versus those solely related to N-fixing woody plants and livestock abundance (N-fixers, livestock). Native plants = native species plant richness; Cover = understory plant cover, Exotic prop = proportion of understory cover pertaining to exotic species and Biocrust cover = cover of biological soil crusts (lichens, mosses and cyanobacteria associated to soil surfaces). Precipitation seasonality was included in all models, although the effects are not shown for simplicity. N = 451 for all models. Figure improved using AI (Claude).
Figure 2. Long description
A heatmap grid displays standardized regression coefficients. A vertical color scale on the right ranges from dark blue (negative 0.6) to dark red (positive 0.6).
Rows are divided into two main sections:
1. Plant-related attributes: Native plants, Cover, Exotic prop, and Biocrust cover.
2. Soil-related variables: Labile carbon, Available nitrogen, Available phosphorus, Stability index, Infiltration index, Nutrient index, and Soil compaction.
Columns are organized into four main predictor groups:
1. Model adj. R squared: Values range from 0.03 (Stability index) to 0.33 (Soil compaction).
2. Main Effects: Includes N P P, Woody plants (abund., rich., N-fixers), and Herbivores (Herb. abund., Herb. rich., Livestock).
3. Interactions between N P P and: Woody plants (abund., rich., N-fixers) and Herbivores (Herb. abund., Herb. rich., Livestock).
Significant effects (marked with asterisks and bold text) include:
- N P P has a strong negative effect on Native plants (-0.47) and positive effects on Labile carbon (0.49) and Available nitrogen (0.44).
- Woody plant richness has a strong positive effect on Biocrust cover (0.67).
- Livestock has strong negative effects on Cover (-0.65) and Stability index (-0.57), but a positive effect on Soil compaction (0.56).
- Interactions show Livestock and N P P together have a strong positive effect on Cover (0.66) and Stability index (0.50).
Woody species, however, showed generally weaker and less frequently significant correlations with vegetation structure and soil condition than herbivores. We found a positive relationship between woody richness and both abundance and biocrust cover (Figure 2) and a reduction in both nitrogen and phosphorus with increasing richness of woody plants (combination of N-fixers and non-fixers, shrubs and trees). Water infiltration also declined with increasing woody abundance (Figure 2). Surprisingly, woody plant effects related to biocrusts and infiltration were linked to the abundance of N-fixers, whereas those on soil nitrogen and phosphorus were not.
In general, soil attributes (with the exception of the soil stability and infiltration indices) were more responsive to the environmental predictors considered here (NPP, woody plants and herbivores; adjusted R 2: 0.19–0.33) than diversity-related variables (native and exotic plants, biocrusts; adjusted R 2: 0.05–0.25; Figure 2). Labile carbon was insensitive to both woody plants and herbivores, responding strongly and positively to NPP. Understory species cover, exotic plant cover and soil stability and infiltration indices were highly responsive to our dryland engineers, though their influence explained only a very small portion of the overall variation (adjusted R 2: 0.03–0.08).
Productivity dampens positive and negative associations among woody plants and herbivores with vegetation structure and soil conditions
Perhaps one of the more consistent results is the role of NPP in buffering the impacts of dryland engineers (Figure 3). In most cases, we found strong interactions between NPP and the abundance or richness of woody plants and herbivores, and these interactions always had the opposite sign of the main effect, thus dampening both the positive and negative effects of dryland engineers on ecosystems. This was true for the negative effect of herbivores on native plant richness and cover, or soil stability, as well as for the positive effects of herbivores and woody plants on exotic plant and biocrust abundance, respectively (Figures 2 and 3). The soil water infiltration index also became more resistant to woody plant abundance under productive (high NPP) environments (Figure 2). These buffering interactions remained consistent when herbivore richness and abundance were restricted to livestock only and when restricting woody plants to N-fixers only. Interestingly, we found substantial interaction effects with NPP for both dryland engineers when using mean values, but not when analyzing their standard deviation values (Figure 4).
Buffering interactions between net primary production (NPP; in g C m−2 day−1) and dryland engineers´ effects. We illustrate the predicted response curves (from the linear model results) of three example variables (biocrust cover [panels above], soil stability index [below left] and native plant richness [below right]) regarding (i) their response to woody plants (above) and herbivore abundance (below, left) and richness (below, right) and (ii) how this response changes depending on NPP. Woody abundance = individuals Ha−1, woody richness = range 1–4; presence of N-fixing shrubs (+1) and trees (+1), and non-fixing shrub (+1) and trees (+1). Herbivore abundance = kg of dung ha−1; herbivore richness = range 0–4; presence of cattle (+1), sheep/goats (+1), rabbits (+1) and kangaroos (+1).
Figure 3. Long description
A multi-panel set of four contour plots. All panels share a common y-axis representing Net Primary Production (N P P) ranging from 0 to 1.5.
* Top-Left Panel: X-axis is Woody abundance (0 to 40,000). The dependent variable is Biocrust cover (predicted), shown on a color scale from white (10) to dark green (30). The plot shows that biocrust cover is highest (dark green) when N P P is low and woody abundance is high, with contour lines curving toward the bottom-right.
* Top-Right Panel: X-axis is Woody richness (1 to 4). The dependent variable is Biocrust cover (predicted). Similar to the first panel, the highest values are concentrated at low N P P and high woody richness, creating a green gradient in the bottom-right corner.
* Bottom-Left Panel: X-axis is Herbivore abundance (0 to 900). The dependent variable is Stability index (predicted), on a scale from 40 (white) to 90 (dark green). The highest stability is found at high N P P (top of the graph) across all herbivore abundances, while a secondary peak appears at low N P P and low herbivore abundance.
* Bottom-Right Panel: X-axis is Herbivore richness (0 to 4). The dependent variable is Native plant richness (predicted), on a scale from 25 (white) to 45 (dark green). The plot shows a complex interaction where richness is highest at the extremes of the y-axis (very low or very high N P P) when herbivore richness is also high, indicated by green zones in the top-right and bottom-right corners.
Heatmap illustrating the comparison of linear model results on the mean (M) and standard deviation (SD; our measure of heterogeneity) of soil-related variables (rows). We show the standardized regression coefficients of the different predictors (columns) evaluated, with bold font showing statistically significant (P < 0.05) effects. Predictors are organized by the two dryland engineers (woody plants, herbivores), productivity (NPP), and the interactions between NPP and the effect of dryland engineers. For all variables, the adjusted R 2 (Model adj. R 2) is shown in the first column, highlighting in grey the most substantial changes. To ease interpretation, asterisks and circles show when results are consistent between mean and standard deviation, and when they are not. Biocrust_cover = cover of biological soil crusts (lichens, mosses and cyanobacteria associated to soil surfaces); Stability index = soil erosion resistance index; Infiltration index = soil water infiltration index; Nutrient index = soil nutrient cycling index. Soil compaction = soil bulk density. Precipitation seasonality was included in all models, although the effects are not shown for brevity. N = 451 for all models. Figure improved using AI (Claude).
Figure 4. Long description
A heatmap grid with 8 main soil variables on the Y-axis, each split into Mean (M) and Standard Deviation (S D) rows. The X-axis contains columns for Model adjusted R squared, N P P, Woody plants (abundance and richness), Herbivores (abundance and richness), and their interactions with N P P.
* The Model adjusted R squared column shows values ranging from 0.01 to 0.33, with the highest values for Soil compaction M (0.33) and Labile carbon M (0.32).
* The N P P column shows strong positive effects on Labile carbon M (0.49) and Available nitrogen M (0.44), and negative effects on Soil compaction M (-0.35) and S D (-0.23).
* Woody plants show the strongest positive effects on Biocrust cover, with coefficients of 0.57 and 0.67 for abundance and richness respectively.
* Herbivore abundance has a significant negative effect on Stability index M (-0.42) and Infiltration index M (-0.42).
* The interaction section shows that Woody richness and N P P interaction has a strong positive effect on Stability index S D (0.50).
Color scale on the right indicates red for positive coefficients (up to 0.6) and blue for negative coefficients (down to -0.6). Bold text with asterisks or circles denotes statistical significance (P < 0.05).
In contrast, the three-way interactions (NPP x diversity x abundance of woody plants and herbivores) as well as interactions between richness and abundance of herbivores, richness and abundance of woody plants and herbivores and woody plant abundance, were negligible. Akaike Information Criterion (AIC) values exceeded 7 units when considering these interactions in our models, which only accounted for an increase in R 2 < 0.05 and were nonsignificant in all cases. Thus, our results do not support the notion that increasing richness of dryland herbivores buffers the impacts of their abundance. The only exceptions to this general pattern were nutrient cycling and soil bulk density. The latter responded significantly to livestock abundance, but these effects dampened when including a larger variety of herbivores (rabbit, kangaroos).
Heterogeneity-related effects: Only part of the story behind dryland engineering
Heterogeneity-related contributions showed similar responses to those found using mean values for biocrust cover, soil resistance to erosion (stability index) and soil compaction (bulk density) only (Figure 4). In contrast, the results for the effect of 1) woody richness on available soil nitrogen or phosphorus and 2) herbivore effects on the soil infiltration index were only apparent when using mean values, but not on their standard deviation. A similar pattern was found for interactions involving NPP, which were only significant when using mean data, but not their standard deviation. Indeed, heterogeneity was substantially less responsive to our environmental predictors than the mean, with adjusted R2 values generally lower, particularly for labile carbon (mean cf. deviation: 0.32 vs 0.09), available nitrogen (0.29 cf. 0.06), nutrient cycling index (0.29 cf. 0.01) and soil compaction (0.33 cf. 0.01).
Discussion
Prioritize livestock management over woody removal to promote dryland function
Although woody plants and herbivores are widely known dryland engineers due to their multiple effects on ecosystem attributes (Wilby et al., Reference Wilby, Shachak and Boeken2001; Paz-Kagan et al., Reference Paz-Kagan, Zaady, Shachak and Karnieli2016), their relative importance is still poorly understood, perhaps due to their strong interdependencies and potential interactions. Indeed, both grazing intensification and abandonment are linked to shrub encroachment (Kerley and Whitford, Reference Kerley and Whitford2000; Archer et al., Reference Archer, Andersen, Predick, Schwinning, Steidl, Woods and Briske2017; Lopez-Diaz et al., Reference Lopez-Diaz, Rolo, Benitez and Moreno2015) and their effects are known to interact (Riginos and Grace, Reference Riginos and Grace2008; Eldridge et al., Reference Eldridge, Soliveres, Bowker and Val2013; Stanton et al., Reference Stanton, Boone, Soto‐Shoender, Fletcher, Blaum and McCleery2018; Grinath et al., Reference Grinath, Larios, Prugh, Brasjares and Suding2019). In the case of ecosystems from eastern Australia, we demonstrated the generally stronger (i.e., more ecosystem attributes affected) and more negative contribution of herbivores, particularly livestock, than woody plants on dryland biodiversity and functioning. Shrub removal is proposed as a restoration tool to improve dryland biodiversity and functioning (reviewed in Ding et al., Reference Ding, Travers, Delgado-Baquerizo and Eldridge2020), yet our results suggest that it is grazing (predominantly by livestock) management that should be prioritized, at least in our study area (see also Eldridge et al., Reference Eldridge, Soliveres, Bowker and Val2013; Eldridge and Soliveres, Reference Eldridge and Soliveres2015). This does not necessarily mean the removal of livestock, but more appropriate management that ensures the maintenance of productive and functional dryland ecosystems. Our results, supported by previous research, suggest that it is possible to enhance the positive, or reduce the negative, impacts of these dryland engineers. We suggest that this could occur by: i) diversifying livestock herds/flocks (Prins and Fritz, Reference Prins, Fritz, Gordon and Prins2008; Wang et al., Reference Wang, Delgado-Baquerizo, Wang, Isbell, Liu, Feng, Liu, Zhong, Zhu, Yuan, Chang and Liu2019) as well as combining the consumption of wild and domestic herbivores (Eldridge et al., Reference Eldridge, Delgado-Baquerizo, Travers, Val, Oliver, Dorrough and Soliveres2017), both supported by our results comparing livestock vs overall herbivore effects, and ii) focusing rangelands management on more productive environments, given the strong NPP × herbivore interactions (see also Maestre et al., Reference Maestre, le Bagousse-Pinguet, Delgado-Baquerizo, Eldridge, Saiz, Berdugo, Gozalo, Ochoa, Guirado, García-Gómez, Valencia, Gaitán, Asensio, Mendoza, Plaza, Díaz-Martínez, Rey, Hu, He, Wang, Lehmann, Rillig, Cesarz, Eisenhauer, Martínez-Valderrama, Moreno-Jiménez, Sala, Abedi, Ahmadian, Alados, Aramayo, Amghar, Arredondo, Ahumada, Bahalkeh, Ben Salem, Blaum, Boldgiv, Bowker, Bran, Bu, Canessa, Castillo-Monroy, Castro, Castro, Castro-Quezada, Chibani, Conceição, Currier, Darrouzet-Nardi, Deák, Donoso, Dougill, Durán, Erdenetsetseg, Espinosa, Fajardo, Farzam, Ferrante, Frank, Fraser, Gherardi, Greenville, Guerra, Gusmán-Montalvan, Hernández-Hernández, Hölzel, Huber-Sannwald, Hughes, Jadán-Maza, Jeltsch, Jentsch, Kaseke, Köbel, Koopman, Leder, Linstädter, le Roux, Li, Liancourt, Liu, Louw, Maggs-Kölling, Makhalanyane, Issa, Manzaneda, Marais, Mora, Moreno, Munson, Nunes, Oliva, Oñatibia, Peter, Pivari, Pueyo, Quiroga, Rahmanian, Reed, Rey, Richard, Rodríguez, Rolo, Rubalcaba, Ruppert, Salah, Schuchardt, Spann, Stavi, Stephens, Swemmer, Teixido, Thomas, Throop, Tielbörger, Travers, Val, Valkó, van den Brink, Ayuso, Velbert, Wamiti, Wang, Wang, Wardle, Yahdjian, Zaady, Zhang, Zhou, Singh and Gross2022).
Consistent with previous findings (e.g., Eldridge et al., Reference Eldridge, Poore, Ruiz-Colmenero, Letnic and Soliveres2016), we did not find any specific level of herbivore abundance at which the effects of grazing shift their sign or substantially alter their effect size. Indeed, such abundance-dependent effects should be linked to the effects of herbivores on environmental heterogeneity, which we found to be particularly weak in our study (Figure 4). Irrespectively, the more numerous and predominantly negative effects of herbivores in our study could be driven by the short evolutionary history of domestic livestock grazing on native Australian plants and soils (Eldridge et al., Reference Eldridge, Maestre, Koen and Delgado‐Baquerizo2018b; Price et al., Reference Price, Sitters, Ohlert, Tognetti, Brown, Seabloom, Borer, Prober, Bakker, MacDougall, Yahdjian, Gruner, Olde Venterink, Barrio, Graff, Bagchi, Arnillas, Bakker, Blumenthal, Boughton, Brudvig, Bugalho, Cadotte, Caldeira, Dickman, Donohue, Grégory, Hautier, Jónsdóttir, Lannes, McCulley, Moore, Power, Risch, Schütz, Standish, Stevens, Veen, Virtanen and Wardle2022), or potentially, the relatively unproductive and nutrient poor environmental conditions (Maestre et al., Reference Maestre, le Bagousse-Pinguet, Delgado-Baquerizo, Eldridge, Saiz, Berdugo, Gozalo, Ochoa, Guirado, García-Gómez, Valencia, Gaitán, Asensio, Mendoza, Plaza, Díaz-Martínez, Rey, Hu, He, Wang, Lehmann, Rillig, Cesarz, Eisenhauer, Martínez-Valderrama, Moreno-Jiménez, Sala, Abedi, Ahmadian, Alados, Aramayo, Amghar, Arredondo, Ahumada, Bahalkeh, Ben Salem, Blaum, Boldgiv, Bowker, Bran, Bu, Canessa, Castillo-Monroy, Castro, Castro, Castro-Quezada, Chibani, Conceição, Currier, Darrouzet-Nardi, Deák, Donoso, Dougill, Durán, Erdenetsetseg, Espinosa, Fajardo, Farzam, Ferrante, Frank, Fraser, Gherardi, Greenville, Guerra, Gusmán-Montalvan, Hernández-Hernández, Hölzel, Huber-Sannwald, Hughes, Jadán-Maza, Jeltsch, Jentsch, Kaseke, Köbel, Koopman, Leder, Linstädter, le Roux, Li, Liancourt, Liu, Louw, Maggs-Kölling, Makhalanyane, Issa, Manzaneda, Marais, Mora, Moreno, Munson, Nunes, Oliva, Oñatibia, Peter, Pivari, Pueyo, Quiroga, Rahmanian, Reed, Rey, Richard, Rodríguez, Rolo, Rubalcaba, Ruppert, Salah, Schuchardt, Spann, Stavi, Stephens, Swemmer, Teixido, Thomas, Throop, Tielbörger, Travers, Val, Valkó, van den Brink, Ayuso, Velbert, Wamiti, Wang, Wang, Wardle, Yahdjian, Zaady, Zhang, Zhou, Singh and Gross2022; Price et al., Reference Price, Sitters, Ohlert, Tognetti, Brown, Seabloom, Borer, Prober, Bakker, MacDougall, Yahdjian, Gruner, Olde Venterink, Barrio, Graff, Bagchi, Arnillas, Bakker, Blumenthal, Boughton, Brudvig, Bugalho, Cadotte, Caldeira, Dickman, Donohue, Grégory, Hautier, Jónsdóttir, Lannes, McCulley, Moore, Power, Risch, Schütz, Standish, Stevens, Veen, Virtanen and Wardle2022).
Trade-offs in woody encroachment
For woody encroachment, we found trade-offs between the positive influences of increasing abundance and richness on biocrust cover and with the negative responses registered for soil nutrients. Global analyses have shown a generally strong suppression of woody encroachment on forage productivity (Anadón et al., Reference Anadón, Sala, Turner and Bennett2014; Ding and Eldridge, Reference Ding and Eldridge2024), perhaps by shading, or via more efficient competition for soil nutrients (reduced nitrogen and phosphorus beneath more diverse woody assemblages in our study). The latter is partly supported by the (nonsignificant) negative effects of woody abundance on understory cover (Figure 2) and could expose suitable bare soil for biocrust colonization; hence the more extensive biocrust cover at encroached sites. Nonlinear woody encroachment-dryland diversity and functioning relationships have been demonstrated (Sirami et al., Reference Sirami, Seymour, Midgley and Barnard2009; Peng et al., Reference Peng, Li, Li, Zhang, Zhang, Li, Zhao, Jiang and Ma2013; Soliveres et al., Reference Soliveres, Maestre, Eldridge, Delgado-Baquerizo, Quero, Bowker and Gallardo2014a; Bestelmeyer et al., Reference Bestelmeyer, Peters, Archer, Browning, Okin, Schooley and Webb2018). Nonlinear relationships often show a peak (both on diversity-related and functioning-related variables) at moderate levels of woody cover. These peaks typically decline at low (homogeneous grasslands, limited vegetation structure) or high (strong shading, resource concentration or woody competition) woody cover. In our case, however, we failed to find such nonlinear relationships, with steady and sustained changes across a large woody abundance gradient (from 25 to >55,000 individuals ha−1; 0 to 135% woody cover). Indeed, woody plant effects were relatively weak and infrequent in our analyses, with no detected effects on vascular plants and many soil variables including labile carbon. Considering that the variables we studied are commonly responsive to woody plant encroachment (e.g., Eldridge et al., Reference Eldridge, Bowker, Maestre, Roger, Reynolds and Whitford2011), and that our woody plant gradient was very large, these weak effects could be explained by the wide grazing intensity gradient that our dataset covers; with high grazing levels dampening the effects of woody plants on ecosystems (Eldridge et al., Reference Eldridge, Soliveres, Bowker and Val2013; Stanton et al., Reference Stanton, Boone, Soto‐Shoender, Fletcher, Blaum and McCleery2018).
Importantly, our study also shows that enhancing the diversity of encroaching woody species is not particularly beneficial, a counter-intuitive result we did not expect, but which was consistent for soil nitrogen, phosphorus and soil compaction (Figure 2). Diverse plant assemblages may be able to occupy a diverse range of niches (Deng et al., Reference Deng, Schmid, Bruelheide, Chen, Li, Li, Morsdorf, Ray, Schuman, Tang and Von Oheimb2025) and have more efficient nutrient uptake (Tilman and Downing, Reference Tilman and Downing1994). Thus, a more efficient soil nutrient uptake by woody plants could deplete the already limited soil nutrients for understory vegetation. More likely, however, is that interspecific variation in rooting depths of a more diverse woody community allows soil water to be exploited in the upper layers (Breshears et al., Reference Breshears, Myers and Barnes2009), leading to reductions in faunal habitat and soil porosity (Decaëns et al., Reference Decaëns, Jiménez, Gioia, Measey and Lavelle2006). Contrary to findings from biodiversity-functioning studies (Temperton et al., Reference Temperton, Mwangi, Scherer-Lorenzen, Schmid and Buchmann2007), and previous results from drylands (Riginos et al., Reference Riginos, Grace, Augustine and Young2009), our results suggest that N-fixing woody species are unrelated to observed soil chemistry. The very low nutrient levels often observed in Australian soils (Kooyman et al., Reference Kooyman, Laffan and Westoby2017; Eldridge et al., Reference Eldridge, Maestre, Koen and Delgado‐Baquerizo2018b) could have reduced the potentially negative effect of high abundances of N-fixing woody species on nutrient stoichiometry and P availability (Riginos et al., Reference Riginos, Grace, Augustine and Young2009; Delgado-Baquerizo et al., Reference Delgado-Baquerizo, Maestre, Gallardo, Bowker, Wallenstein, Quero, Ochoa, Gozalo, García-Gómez, Soliveres, García-Palacios, Berdugo, Valencia, Escolar, Arredondo, Barraza-Zepeda, Bran, Carreira, Chaieb, Conceição, Derak, Eldridge, Escudero, Espinosa, Gaitán, Gatica, Gómez-González, Guzman, Gutiérrez, Florentino, Hepper, Hernández, Huber-Sannwald, Jankju, Liu, Mau, Miriti, Monerris, Naseri, Noumi, Polo, Prina, Pucheta, Ramírez, Ramírez-Collantes, Romão, Tighe, Torres, Torres-Díaz, Ungar, Val, Wamiti, Wang and Zaady2013). In addition, complementarity could also be prevalent among woody species, irrespective of their capacity to fix nitrogen, as both show facilitative effects (Soliveres et al., Reference Soliveres, Maestre, Bowker, Torices, Quero, García-Gómez, Cabrera, Cea, Coaguila, Eldridge, Espinosa, Hemmings, Monerris, Tighe, Delgado-Baquerizo, Escolar, García-Palacios, Gozalo, Ochoa, Blones, Derak, Ghiloufi, Gutiérrez, Hernández and Noumi2014b), contrasting canopy and root patterns (Martinez-Meza and Whitford, Reference Martinez-Meza and Whitford1996; Eldridge et al., Reference Eldridge, Bowker, Maestre, Roger, Reynolds and Whitford2011), invoked mechanisms to enhance complementarity in resource use.
Productivity and heterogeneity partly modulate the effects of dryland engineers
We found that more productive environments dampened both negative and positive effects of dryland engineers on ecosystems, and that this modulation was independent of heterogeneity-driven patterns, that is, there were no significant NPP × herbivore or NPP × woody species interactions in the models using standard deviation (Figure 4). These results suggest that managing shrub encroachment and grazing would be more efficient on the most unproductive areas (see Proulx and Mazumder, Reference Proulx and Mazumder1998; Soliveres et al., Reference Soliveres, Maestre, Eldridge, Delgado-Baquerizo, Quero, Bowker and Gallardo2014a; Maestre et al., Reference Maestre, le Bagousse-Pinguet, Delgado-Baquerizo, Eldridge, Saiz, Berdugo, Gozalo, Ochoa, Guirado, García-Gómez, Valencia, Gaitán, Asensio, Mendoza, Plaza, Díaz-Martínez, Rey, Hu, He, Wang, Lehmann, Rillig, Cesarz, Eisenhauer, Martínez-Valderrama, Moreno-Jiménez, Sala, Abedi, Ahmadian, Alados, Aramayo, Amghar, Arredondo, Ahumada, Bahalkeh, Ben Salem, Blaum, Boldgiv, Bowker, Bran, Bu, Canessa, Castillo-Monroy, Castro, Castro, Castro-Quezada, Chibani, Conceição, Currier, Darrouzet-Nardi, Deák, Donoso, Dougill, Durán, Erdenetsetseg, Espinosa, Fajardo, Farzam, Ferrante, Frank, Fraser, Gherardi, Greenville, Guerra, Gusmán-Montalvan, Hernández-Hernández, Hölzel, Huber-Sannwald, Hughes, Jadán-Maza, Jeltsch, Jentsch, Kaseke, Köbel, Koopman, Leder, Linstädter, le Roux, Li, Liancourt, Liu, Louw, Maggs-Kölling, Makhalanyane, Issa, Manzaneda, Marais, Mora, Moreno, Munson, Nunes, Oliva, Oñatibia, Peter, Pivari, Pueyo, Quiroga, Rahmanian, Reed, Rey, Richard, Rodríguez, Rolo, Rubalcaba, Ruppert, Salah, Schuchardt, Spann, Stavi, Stephens, Swemmer, Teixido, Thomas, Throop, Tielbörger, Travers, Val, Valkó, van den Brink, Ayuso, Velbert, Wamiti, Wang, Wang, Wardle, Yahdjian, Zaady, Zhang, Zhou, Singh and Gross2022). Another interesting speculation related to Whitford’s work on soil disturbing animals (Whitford and Kay, Reference Whitford and Kay1999) is a potential combination of an additional dryland engineer acting in concert with herbivores and woody plants. Soil disturbing animals enhance plant productivity and soil heterogeneity in drylands (Whitford, Reference Whitford1993; Whitford, Reference Whitford2002; Eldridge and Soliveres, Reference Eldridge and Soliveres2023); effects sometimes reported to be even larger than those of woody plants per se (Eldridge and Whitford, Reference Eldridge and Whitford2014). Thus, considering the combination of this third dryland engineer could minimize effects of both woody plant encroachment and herbivores when these are detrimental, something that deserves further exploration. Irrespective of the potential role(s) or additional engineers, our results show that ecosystem productivity produces the most consistent, commonly dampening dryland engineering effect over and above the effects of either the abundance or diversity of engineers.
Conclusion
The importance of herbivores and woody plants as dryland engineers is widely known due to their multiple effects on the diversity, structure and functioning of these ecosystems. However, which is their relative importance, how do they interact, how their diversity and abundance influence their effects and which are the roles of productivity and heterogeneity as modulators of the effects of these dryland engineers is less well understood. Here we took advantage of a nearly orthogonal observational design to evaluate the relative contribution of the diversity and abundance of woody plants and herbivores on plants (native and exotic), biocrusts and multiple soil attributes. We found more frequent and consistently negative effects of herbivores than woody plants on drylands, mainly attributable to livestock abundance, which suggests that focusing on livestock management rather than woody removal may more effectively improve dryland condition. Productivity was the strongest modulator of dryland engineering effects, as shown by the interactions between the effects of these dryland engineers and productivity (always dampening both positive and negative effects under more productive conditions). These modulating interactions were stronger and more frequent than results related with heterogeneity or the interplay between engineers´ diversity and their abundance.
Open peer review
For open peer review materials, please visit https://doi.org/10.1017/dry.2026.10052.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/dry.2026.10052.
Data availability statement
Data associated to this manuscript are available in: https://doi.org/10.5061/dryad.851d0
Acknowledgments
We thank James Val, Samantha Travers and Marta Ruiz-Colmenero who undertook field work and data entry, and Ian Oliver, who managed the NSW Grazing Study, from which these data are derived.
Author contributions
All authors contributed to research conceptualization; S.S.: data analyses and writing of the first draft, D.J.E.: funding acquisition, field- and lab-work, data gathering and curation and text revision; M.D.-B.: text revision.
Declaration of interest
The authors declare no competing interest.
Funding statement
S.S. and D.J.E. were supported by the Hermon Slade Foundation (HS210474). M.D.-B. acknowledges support from GRASS4FUN (Biodiversa+2022)/MCIU/AEI/10.13039/501100011033/ Unión Europea.
Open access
Comments
June 17, 2026
Dear Editor,
We wish to submit an original research article entitled “Engineering effects of herbivores surpass those of woody plants” for consideration by Cambridge Prisms: Drylands. This is an invited submission to the Special Issue in honour of the late Professor Walter Whitford.
In our work, we highlight the work of Prof. Whitford regarding the role of dryland engineers to then present our results on a nearly orthogonal design which effectively isolate the effect of two major engineers: large herbivores and woody plant encroachment. We evaluate the influence of the density of dryland engineers (abundance of herbivores and woody plants), their diversities, and overall ecosystem productivity as modulators of dryland engineering effects on plant richness and cover, dominance of exotic plants, biological soil crusts cover and soil attributes (nutrient availability and cycling, resistance of erosion and water infiltration). We also evaluated changes in heterogeneity as the main driver of dryland engineering effects. Our study, which includes 451 sites spanning ~0.4 million km2 in Southeastern Australia, show that herbivores played a larger role as dryland engineers than woody species, which showed more mixed effects. Most of the negative effects found for herbivore abundance were linked to livestock, only partly explained by heterogeneity, and substantially dampened with ecosystem productivity. The latter was an overall modulator of dryland engineering effects, as it softened both positive and negative effects of both herbivores and woody plants. We show that dryland engineering effects are only partly driven by their effects on heterogeneity, and that mitigation through increasing the diversity of these engineers, as previously suggested, has limited effects. Our study contributes to a better understanding on the relative importance and context-dependency of two major dryland ecosystem engineers, showing that prioritizing livestock management over woody removal, diversifying herds and locating them in more productive areas may aid conserving drylands´ biodiversity and functioning.
We confirm that this article is original and not submitted anywhere else, although its data has been used elsewhere (e.g., Eldridge et al. 2017, cited in the text).
Thank you for considering our manuscript for publication in Cambridge Prisms: Drylands.
Sincerely,
Santiago Soliveres, on behalf of all authors.