top of page

The ECL's Australian projects

Tropical rainforests are the most diverse terrestrial systems with innumerable species interactions. Humans are disturbing these ecosystems in a multitude of ways, often simultaneously (e.g. one forest can suffer from deforestation, predator loss, hunting and invasive species). Research to date has focused on how individual threats affect particular species, but we have a poor understanding of how multiple threats interact to reshape wildlife communities (diversity and food-web metrics). Given that most remaining forests face multiple threats, this enormous knowledge gap limits our ability to effectively manage protected areas (PAs).

​

Urgent problem - Nowhere are these issues more important than the tropical rainforest landscapes of Australia and Southeast Asia, which are home to a billion people and four global biodiversity conservation hotspots. In Australia, fragmented rainforest cover <1% the land area – a >70% decline from historic levels – but contain >50% of plants and 35-40% of mammal and bird species. Meanwhile, Southeast Asia faces the highest global deforestation rates with fragmented PAs surrounded by agriculture and large human populations.

​

A pressing conservation issue is the decline of wildlife inside PAs. The key threats facing wildlife are habitat fragmentation, hunting, and invasive species, and these vary substantially among PAs. Consequently, ostensibly similar fragmented PAs experience drastically different outcomes, including being defaunated or overrun with pests or invasive species (Fig 2 & 3). This PA-specificity suggests contingent, interacting, and/or non-linear processes. Crucially, multiple threats routinely precipitate rapid losses in native mammals. If we are going to prevent extinctions, we need to assess the combined effects of threats on wildlife communities.

​

Australian rainforests have seen >20 recent mammal extinctions and most are linked to multiple threats. However, scientists have only begun to examine how threats interact to accelerate extinctions, in particular how fire elevates invasive predator impacts and forest fragmentation induces native predator losses, which in turn magnifies impacts from smaller invasive predators. Another pressing issue is the combined effects from multiple invasive species that plague the same PA (feral cats, foxes, pigs, and deer). Addressing such 3-way interactions in Australia requires big data that spans entire wildlife communities in many landscapes.

​

Efficiency – Many millions of dollars are spent annually implementing wildlife surveys, and the newest and richest method for mammals is camera traps (CT) that passively document all terrestrial species >1 kg. However, single surveys in isolation offer little guidance on dynamics and are rarely tied to management decisions. To address this, new approaches are needed for cross-site analyses of CT data, which can provide far greater insights than any single study. This new approach would also leverage troves of existing wildlife data that is underutilised.

​

Aims – The ECL develops a novel cross-site analytical approach to study the interactive effects from multiple threats on rainforest wildlife communities. We will apply this approach in Australia using new data collected in the Wet Tropics of Queensland World Heritage Sites, including pre-existing data from Australian collaborators. We will also compare and contrast key drivers and interaction strengths among Australian, Southeast Asian, and global datasets. In this way, we can address fundamental questions in wildlife ecology and also produce applied management guidance.

​

Innovation - Untangling impacts of interacting threats on mammal communities requires a comparative framework and ‘big data’ (many landscape-scale datasets of mammal communities). Camera trap (CT) surveys are ideal for this but have not been systematically assembled for cross-site comparisons within Australia or Southeast Asia, with the exception of the ECLs work. The ECL addresses this in two ways: first by developing a novel meta-regression approach to compare wildlife communities using CT data from different sites (in collaboration with a USA statistician). Second, by developing a meta-structural equation modelling (metaSEM) approach for CT data, in close collaboration with the lead statistician authoring the metaSEM R-package (based in Singapore, see Feasibility).

​

Benefit – The results will provide scientists and managers with clear guidance about the state of wildlife and mammal communities, the threats driving specific changes, and the importance of interacting effects. The analyses will be used to model the benefit of alternative conservation strategies, e.g. should we deal with (i) hunting, (ii) invasive predators, or (iii) invasive herbivores? And when managers should focus on reducing the biggest single threat, versus if reducing a combination of threats at the same time can produce outsized improvements.

​

Objectives:

  • (i) Advance fundamental knowledge on ecological impacts from multiple threats;

  • (ii) innovate new statistical approaches for multi-site analyses of wildlife communities,

  • (iii) boost the value of expensive wildlife surveys, (iv) produce the largest wildlife community dataset in Australian rainforests,

  • (v) build a collaborative network to share data across Australian institutions,

  • (vi) communicate research outcomes to government, NGOs, and PAs managers to improve threatened species conservation and train these groups to use new approaches.

 

Conceptual framework - Wildlife ecology and ecological cascades (EC) 

​

Conserving tropical forests requires understanding how they function. Mammal community food webs are maintained through a delicate balance of connections between predators, prey, competitors, and symbionts. Disturbances can disassemble food webs by altering diversity, species population sizes, and species interactions. Ecological cascades (ECs) research tracks how changes to one species produce indirect, often cryptic, ‘knock-on’ effects on other species and the ecosystem. The secondary and tertiary consequences of losing a single species (e.g. apex predator) can outweigh the initial direct effects. For example, a 50% reduction of predators may allow herbivore densities to double, which in turn strongly suppress plants. ECs can also reshape ecosystem processes such as carbon sequestration, fire dynamics, and even disease dynamics. Even in diverse tropical rainforests, changes to a single keystone species can cause food webs to collapse, reducing wildlife diversity and altering tree communities (Fig 1).

​

ECL research has extended ECs to situations when forest-dwelling pest animals crop-raid (e.g. pigs foraging in farms) and increase their densities within forests (Luskin et al. 2017b in Nature Comms3). Food-subsidized animals then overconsume native plants and degrade the ecosystem (Fig 1 & 2; Jia+ et al & Luskin+ 2018 in PNAS).

​

There are diverse types of ECs, for example hunting herbivorous versus omnivorous game species (e.g. pigs, deer) produces the opposite effect as predators declines (Fig 1). ECs also depend on the ecology of wildlife species involved. For example, hunting seed-dispersing animals (e.g. primates, birds) produces different ECs than hunting seed-predating animals (e.g. pigs or rodents). The removal or addition of keystone species, such as native or feral pigs that disturb soils and vegetation, can produce magnified ECs relative to changes in more benign species (e.g. civets; Dunn et al 2022, Dehaudt et al 2022). 

​

A key lesson in Australia is that negative effects from ECs are magnified by invasive species. There is a preponderance of work showing that invasive predators in Australia cause trophic cascades. Likewise, feral pigs (Sus scofa, hereafter ‘pigs’) in Australian rainforests (Fig 3C) are a particularly egregious example, costing billions in economic losses in Australia alone36, a topic we have studied in depth (Luskin et al 2014, 2017b, 2019). ECL work shows that pigs even alter rainforest tree diversity and negative density dependence in tree recruitment (Figs 2, 3 & 4).

​

Australian rainforests have dramatically contracted in the Pleistocene while Southeast Asia rainforests have been comparatively stable. Work by Betts et al (2019, Science) shows that sites with Pleistocene disturbances like glaciation filtered out sensitive species from communities so that contemporary fragmentation has a comparatively lower impact on diversity relative to more climatically-stable environments. The resulting hypothesis is that Australian mammal communities are more resilient to contemporary fragmentation than those in Southeast Asia or globally.

​

Taken together, fragmentation, predator loss, crop-raiding, hunting/culling, and invasive species each produce distinct changes to wildlife communities, and these outcomes differ in Australia versus other regions.

fig1.png

Fig 1: Types of ecological cascades and interactions. Greyscale animals indicate their densities dramatically decline and grey arrows show lost regulation. Red arrows show unnatural regulation. (A) Trophic cascades from the loss of area-demanding apex predators (commonly in fragments), releases prey species populations. (B) Farming provides food subsidies to crop-raiding native wildlife (pigs eating oil palm fruits19 ) which then disturb forest vegetation. (C) Hunting commonly extirpates all large animals. (D) Fragmentation reduces apex predators (dingos), which releases invasive mesopredators (cats, foxes) and omnivores (pigs). Invasives thrive in edges where can forage in nearby farmlands, and they predate and compete with native animals.

​

Knowledge gap:   The majority of remaining tropical forest PAs face multiple threats13 and a key challenge for ecologists is discerning the trajectories of wildlife populations. A telling statistic is that 70% of remaining forests globally now lie <1 km from an edge42. In Australia and Asia, most forests are also bordered by agriculture and nearby to high human populations that exploit wildlife, e.g. through hunting, and have invasive or pest species42.

​

Tropical forest food-webs are notoriously difficult to study because they are statistically ‘noisy’ with many indirect effects and feedbacks. Furthermore, the large spatial and temporal scales over which ECs operate has made controlled experiments difficult. ‘Natural experiments’ can be used when there is ecological data from before and after disturbances, but these are exceedingly rare24. As a result of these difficulties, most research focuses on single threats in observational settings. Multiple landscapes have rarely been included in a quantitative framework to assess relative and interactive effects. This has limited robust inferences about realistic landscapes where multiple threats are the norm. This DECRA outlines an approach to disentangling the ECs from multiple threats in tropical forests throughout Australia, Southeast Asia, and globally.

​

Domestic problem:   Australian rainforests are well protected, but they are fragmented and are home to many invasive species. Pigs and invasive predators like cats and foxes often thrive in forest edges and degrade diversity (Fig 3). This suggests an interaction between fragmentation (more edge habitat) and invasive species (Fig 3). We utilise both new and existing data to explore how such interacting threats affect Australia’s wildlife and compare this globally to sites sharing threats (e.g. North and South America also face invasive pig problems).

​

Regional problem:   A key issue in Southeast Asia is the rapid expansion of oil palm. For example, Malaysia has ~7 Mha of oil palm and ~15 Mha of forest. This has led to a majority of remaining forests also being fragmented and in close proximity to plantations (within 5km). ECL work has shown that remaining forest patches often lose predators and gain pigs, which crop-raid in nearby plantations. ECL work has also highlighted a strong interaction between hunting and oil palm expansion, as the immigrant workers in plantations often bring new hunting practices. New ECL work is comparing forests facing different threats using a global dataset.

​

Feral pigs, a keystone species in their native and invasive range:   As humans spread across the planet, they often hunted native large mammals to extinction and then replaced this protein by introducing game species. Pigs (Sus scrofa) are particularly successful in both their native and introduced range due to crop-raiding (Fig 3C).

​

Pigs are ‘ecosystem engineers’ that negatively impact soils, plants, and animal communities, as well as crops, livestock, and property. Pigs impact vertebrates through direct interactions, such as monopolising a discrete food source and indirectly by altering the habitat and consuming available food (‘apparent competition’). In addition to this, due to their rapid reproductive rate, any competitive advantage is magnified quickly over the course of a few years. Our new work highlights that pigs thrive in forest edges and fragments (Fig 2).

Recently there are reports showing that habitat fragments with invasive pigs had 22-26% lower vertebrate richness in the United States41 and Brazil (Fig 2B). Multiple reviews highlight that pigs’ negative impacts extend throughout their native range (Eurasia and Southeast Asia) and introduced range (e.g. Australia, North and South America). It is also clear that invasive species have numerous direct and indirect impacts on Australian food webs and ecosystems. What remains unclear is how pigs interact with other pressures (fragmentation, invasive predators) to magnify negative impacts on wildlife communities.

Fig2.png

Fig 2: (A) Forest pig densities are higher near edges and farmland. (B) Invasive pigs cause strong negative impacts on mammal species richness. (C) Relative abundances of different species in 109 camera trap surveys from Asian forests (prelim DECRA data). Smaller forests (yellow and red points) are dominated by pigs while larger forests have higher relative abundances of herbivores and predators.

​

Research Questions and Hypotheses:      An advantage to the dataset and analyses will be used to address a variety of basic ecology and applied conservation questions. The dataset also offers opportunities for honours, masters, PhD students to be involved, as well as future postdocs, with the project having the potential to comfortably result in >10 peer-reviewed publications. Here we outline three ‘big-picture’ research questions and hypotheses. For each question, we will assess how multiple aspects of the mammal community respond to threats, where the term ‘mammal community’ includes changes to species (e.g. occupancy of quolls), diversity (species richness and evenness), and food webs (trophic levels and web-generality; Table 1). For simplicity, we also use the term ‘fragmentation’ to refer to forest size and isolation, although the actual analyses will differentiate these effects.

 

Q1: Does fragmentation mediate invasive predators and pest species impacts on mammal communities?

Globally, the effect of native predators suppressing herbivores and mesopredators is a positive ‘top-down’ influence on mammal communities (higher species richness and evenness). However, invasive predators can produce the opposite consequences, especially in Australia Likewise, pest species like pigs have been negatively associated with diversity (Fig 3 & 5C). What remains unclear is how fragmentation mediates these relationships. This section focuses on 2-way interactions between fragmentation*predators and fragmentation*invasive species.

 

Fragmentation*Predators: We hypothesise mammal communities are positively and negatively associated with native and invasive predators, respectively (both have been previously reported). In Australia, invasive predators (cats, foxes) thrive in edges while some native predators (e.g. quolls) avoid edges. Building on this, we hypothesise stronger negative impacts from invasive predators in smaller forests.Fragmentation*Pests: We hypothesise that the negative effects from pest species and invasive herbivores (e.g. native and feral pigs) vary with forest area, with worse impacts in smaller forests where other mesopredators and competitors have been lost, so the niche breadth of pigs may expand. We also hypothesise stronger interaction strength in Australia (feral pigs) versus Southeast Asia where pigs are native. Fragmentation*Region: Based on forest contraction during the Pleistocene that filtered out sensitive species (described earlier), we hypothesise mammal communities in Australia’s small fragments suffer less (smaller percentage change in species richness and evenness) relative to similar-sized forests in Southeast Asia.  

​

Q2: How do native and invasive predators and omnivores interact to shape mammal communities?

Predators can limit edge-adapted generalist herbivores/omnivores (pigs, deer) from crop-raiding via behavioural responses (e.g. landscapes of fear). Further, top-down predation control maybe more important where resource subsidies have removed bottom-up control of herbivores. What remains unclear is the influence from both predators and invasive herbivores on mammal communities, and their 3-way interaction with forest fragmentation.

 

Native Predator*Pest*Fragmentation: In Southeast Asia, we hypothesise that the beneficial effect of native predators (tigers, leopards) in reducing pest herbivores (e.g. crop-raiding pigs) is stronger nearby farmlands. Invasive predator*Invasive Pest: In Australia, we hypothesise negative effect of invasive predators (foxes, cats) on mammal communities and a negative effect of invasive herbivores (deer, pigs) via overconsuming the vegetation (apparent competition) and disturbing the habitat, and a negative interaction when both invasives are abundant that produces dramatic declines in mammal community metrics. The interaction could arise because pig-disturbed habitats may increase rates of predation on native mammals.Invasive predator*Invasive Pest*Fragmentation: In Australia, both invasive predators and invasive herbivores thrive in edges, so we hypothesise fragmented forests with more edges may suffer most from negative impacts. 

​

Q3: What are the indirect effects from human-imposed threats like fragmentation and hunting?

The ECL is quantifying the relative importance direct and indirect of threats on mammal communities using structural equation modelling (SEM) – and for the first time – will incorporate both within and between landscape variation in species occurrence using meta-SEM. SEM can be used to quantify the strength of indirect pathways such as fragmentation altering predators or pests which in turn will contribute to altered diversity.

 

Previous work has used SEM to show that predators’ top-down role regulating mammal communities was less important than direct human influences like hunting. We hypothesise a similar pattern in Australia, with beneficial impacts of native predators (dingos, quolls) on mammal communities being relatively small compared to other variables like fragmentation and presence of invasive species (Fig 5). Pests thrive in small forests near farmland4,6,36(Fig 2 & 5A). We hypothesise this is driven by the combination of direct effects from beneficial habitat and crop-raiding, and indirect effect of predator declines. We hypothesise the impact of fragmentation on large native herbivores is mediated indirectly by increased hunting in Southeast Asia (Fig 5), but not in Australia where hunting is highly regulated. These recent advances now enable multiple CT surveys to be compared.

​

Research to date has also focused on single species. By studying whole wildlife communities, we can capture the secondary effects from species interactions, which enables much richer scientific and conservation inferences.

pic3.png

Fig 3: Camera traps (CT) have been used to compare mammal communities globally (A) but not in Australia (C). Panel (A) shows results from 16 sites (“TEAM Network”) but these have poor coverage in Australia and Southeast Asia. Panel (B) shows the preliminary Southeast Asia database that includes >50 sites (see Fig 5 and Feasibility). We will focus on new data collation and fieldwork in Australia. Panel (C – left) shows heat map of feral pigs in Australia, highlighting that the most biodiverse areas (Queensland rainforests) also have the worst pig problems. The inset in (C) shows Zach's fieldwork for new landscape-scale CT surveys.

 

Methods for data collection – The ECL's key data collection task is to (a) collate 20 previous Australian CT surveys through local partnerships, (b) conduct 10 new CT surveys in Queensland tropical rainforests. The 30 Australian CT surveys data will be collected with two HDR students and will be used for Zach and Zoe's PhD thesis. The results from Australia will be compared with >100 CT surveys from Southeast Asia and globally, which we have collated in partnerships with NGOs and dozens of scientists over the last year (Fig 4 & 5).

​

Methods for quantifying species responses to anthropogenic threats

Each threat has a direct impact on each species (e.g. agriculture --> pigs +50%). We use multi-species multi-site occupancy modelling (MSMS OccMod) to assess relationships between species and threats. These models provide detectability-corrected occupancy for all species as well as site-specific relationships between species occupancy and threat covariates. Fig 4A & 4B show MSMS OccMod results from the Asia CT surveys.

​

Site-specific results are useful for local managers; however, to make generalisations across the region, and compare Australia to Southeast Asia and globally, we use meta-regressions to assess if there are consistent relationships. Here, the response variable is the MSMS OccMods regression coefficients for a particular species and the explanatory variables are the threats (e.g. forest size) with observations weighted by the standard error. Multiple threats may interact to magnify outcomes for species, leading to superadditive impacts (e.g. agriculture + predator loss --> pigs +200%). These interactions will be tested using meta-regressions.

fig4.png

Fig 4: Panels A and B show wildlife occupancy is shaped by human threats (A and B). This data comes from recent CT work in Southeast Asia (in prep) and highlight how generalist pigs thrive in smaller forests and where landscape human populations are high. Panels C and D show communities are also shaped internal food-web effects from species present (pigs and carnivores). This data comes from 113 surveys in Asia and globally (Luskin in prep).

 

Methods for quantifying community responses to anthropogenic threats

The net effect of multiple threats on entire communities can be measured by diversity and food-web metrics. We assess this using generalised linear mixed models (GLMM) wherein the response variable is a community characteristic (species evenness) and the explanatory variables are the threats (e.g. Table 1). For proof-of-concept, Fig 4C & 4D show preliminary results from GLMMs using the Southeast Asian database (109 surveys, does not include any Australian sites).

 

Methods to quantifying indirect pathways shaping mammal communities (metaSEM)

SEM can be used to compare the relative importance of different pathways through which threats affect mammal communities. For example, SEM can clarify the following question: is the positive relationship between fragmentation and pigs mediated by the presence of edge habitat or the loss of apex predators, and which pathway is more important? SEM uses the correlations among species and threats and in many different sites to establish directional relationships. For example, Fig 5 shows that hunting affects predators via both a direct negative effect and an indirect negative effect mediated by reduced herbivores prey base.

fig5.png

Fig 5: Direct and indirect pathways threats affect wildlife, assessed using meta- structural equation modelling (metaSEM). Blue arrows show positive relationships, red show negative relationships, solid arrows indicate significant results (P < 0.05) and dashed grey arrows are non-significant. Arrow weights are drawn relative to their effect size (data from 109 CT surveys in Asia). For simplicity, this SEM grouped species into 3 main feeding guilds. Crop-raiding and hunting are not observed variables but inferred from two measured proxies each.

bottom of page