Island Safe Havens from Climate Change (Jones et al., 2020)

📖 21 min read~3767 words

This is an original editorial summary prepared by the Komodo Guide team pointing readers to the primary source: Jones, A.R., Jessop, T.S., Ariefiandy, A., Brook, B.W., Brown, S.C., Ciofi, C., Benu, Y.J., Purwandana, D., Sitorus, T., Wigley, T.M.L., & Fordham, D.A. (2020). "Identifying island safe havens to prevent the extinction of the World's largest lizard from global warming." Ecology and Evolution, 10(19), 10492–10507. https://doi.org/10.1002/ece3.6705. We encourage all researchers and conservation practitioners to consult the full open-access paper directly.

Table of Contents

• Quick Facts
• Paper Overview
• Modelling Approach
• Projected Island Outcomes
• Two Compounding Threats: Warming & Sea-Level Rise
• Link to the 2021 IUCN Endangered Uplisting
• Conservation & Management Implications
• Myths vs Facts
• Key Takeaways
• Frequently Asked Questions
• Sources & Further Reading

Quick Facts

Item	Detail
Full citation	Jones et al. (2020), Ecology and Evolution 10(19): 10492–10507
DOI	10.1002/ece3.6705
Publication date	15 September 2020 (open access)
Focal species	Varanus komodoensis (Komodo dragon)
Wild population at time of study	~4,000 individuals
Scenarios modelled	RCP 2.6 & RCP 8.5, combined with three climate-sensitivity parameterisations (six combinations)
Projected habitat loss by 2050	8%–87% reduction across range, depending on scenario
Projected population decline by 2050	Patch occupancy loss 25%–97%; abundance decline 27%–99%
Identified refugia	Komodo Island; Rinca Island
Most at-risk areas	Nusa Kode, Gili Motang, Flores (projected extirpated under all scenarios)
Key policy outcome	Findings contributed to IUCN uplisting of V. komodoensis from Vulnerable to Endangered (2021)

Paper Overview

Before 2020, no published study had attempted to forecast how climate change would reshape the distribution and abundance of Varanus komodoensis, the world's largest living lizard. Jones and colleagues filled that gap with rigorous spatial population modelling — an approach that combines projected climate envelopes with detailed demographic data to simulate how populations actually grow, decline, and go locally extinct across fragmented landscapes.

The team's starting point was simple but sobering: the Komodo dragon is already one of the most spatially constrained large vertebrates on Earth. It occupies a handful of small Indonesian islands spread across roughly 1,300 km² of suitable habitat. That extreme insularity, which has defined the species for hundreds of thousands of years, becomes a liability under climate change — islands cannot be migrated away from, and refugia are only as safe as their topography and ecological integrity allow. The paper's central question was therefore not whether climate change would affect Komodo dragons, but which islands, if any, could absorb the coming pressure and sustain breeding populations into the next century.

The answer that emerged from over 1.2 million model iterations was both geographically precise and policy-relevant: Komodo Island and Rinca Island — both already within Komodo National Park — are projected to retain suitable habitat under even moderately pessimistic emissions trajectories, while the smaller and lower-lying islands of Nusa Kode and Gili Motang, and the populations on Flores, face substantially higher extinction risk across all six scenarios tested.

Modelling Approach

The methodological architecture of this study is what distinguishes it from simpler species-distribution forecasts. Rather than projecting habitat suitability alone and leaving the reader to infer population consequences, Jones and colleagues explicitly coupled two model types: an ensemble ecological niche model (ENM) and a spatially explicit demographic simulation. Each layer handled a different biological question, and the two were integrated so that projected habitat changes fed directly into population dynamics.

Ecological Niche Modelling

The ENM was constructed using four widely validated algorithms — generalized linear models (GLM), generalized additive models (GAM), gradient boosting machines (GBM), and MaxEnt — run in ensemble to reduce the sensitivity of any single algorithmic choice. Predictor variables included temperature seasonality, topographic position, and proximity to the coast, all of which are known to influence dragon activity budgets and prey availability. Model performance was strong: the ensemble achieved a true skill statistic (TSS) of 0.583 and an area under the curve (AUC) of 0.842, both comfortably above thresholds considered acceptable for conservation-relevant projections.

Six future climate scenarios were derived from two Representative Concentration Pathways — RCP 2.6, representing a stringent mitigation trajectory sometimes described informally as the "policy" scenario, and RCP 8.5, the high-emissions "business as usual" trajectory — combined with three climate-sensitivity parameterisations corresponding to equilibrium warming of approximately 1.5°C, 3°C, and 6°C above pre-industrial levels. This structure forced the models to confront a wide range of plausible futures rather than a single projection, giving decision-makers a cleaner picture of which outcomes are robust across uncertainty and which depend heavily on emissions choices.

Population Modelling

The demographic component used RAMAS GIS v5, a platform designed for spatially structured population viability analysis. Dragons were modelled in a stage-structured, female-only framework with density dependence governed by a Ricker logistic function. Vital rates — hatchling survival, juvenile survival, adult survival, and fecundity — were estimated from a 10-year capture-mark-recapture dataset spanning 2003–2013 in which more than 1,000 individuals were tracked across the park. This demographic foundation is unusually solid for a reptile PVA and substantially reduces the uncertainty that plagues models built on literature estimates alone.

Sensitivity analysis identified the maximum intrinsic rate of population increase (Rmax = 1.18) and the habitat-derived carrying capacity as the parameters to which projections were most sensitive — a finding with a direct management implication: protecting the habitat that sets carrying capacity is at least as important as preventing direct mortality.

In total, the study ran more than 1.2 million model iterations (200 parameter sets × 6 climate scenarios × 1,000 stochastic replicates), providing robust confidence intervals around every projection.

Projected Island Outcomes

The most policy-critical output of the paper is its island-by-island risk ranking. The Komodo dragon's range comprises five principal land masses: Komodo Island, Rinca Island, Nusa Kode, Gili Motang, and the far larger but human-modified Flores Island. The models produced sharply differentiated prognoses for each.

Island	Status in Komodo NP	Projected Climate Fate	Rationale
Komodo	Core zone	Refugium — retained habitat under all six scenarios	Larger area, elevation gradient buffering temperature extremes, existing legal protection
Rinca	Core zone	Refugium — retained habitat under all six scenarios	Similar size advantage; topographic variation provides thermal microhabitat diversity
Nusa Kode	Within NP boundary	High risk — increasingly unsuitable habitat across scenarios	Small area, low topographic relief, high vulnerability to sea-level inundation of lowlands
Gili Motang	Within NP boundary	High risk — increasingly unsuitable habitat across scenarios	Small area, low elevation, comparable vulnerability to Nusa Kode
Flores	Three reserves outside core NP	Extirpated under all scenarios modelled	Already heavily degraded by human land use; climate pressure compounds existing threats; populations isolated from park-protected refugia

The range-wide projections are striking in their span. Under the most optimistic scenario (RCP 2.6 with low climate sensitivity), the models still project a 15%–45% decline in population abundance by 2050. Under the most pessimistic scenario (RCP 8.5 with high climate sensitivity), the projected decline reaches 95%–99%, with the minimum expected female abundance across the range falling to approximately 96 individuals. Patch occupancy — the fraction of currently inhabited habitat cells sustaining viable subpopulations — is forecast to fall by 25%–97% depending on scenario. These numbers do not represent guaranteed outcomes; they represent the probabilistic space within which the species is expected to land if current emissions trajectories continue or are reversed.

Two Compounding Threats: Warming and Sea-Level Rise

Jones et al. modelled both thermal warming and sea-level rise as distinct drivers with different timescales of impact. This dual-threat framing is one of the paper's important conceptual contributions, and it is worth unpacking separately from the population numbers.

Temperature increase is the dominant driver of projected habitat loss through the 2050 modelling horizon. Rising mean temperatures and increasingly severe dry seasons — both expected across the Lesser Sunda region under all RCP scenarios — reduce the suitability of lower-elevation terrain where dragons currently concentrate. Because V. komodoensis is ectothermic, its activity windows, foraging success, and reproduction are tightly coupled to ambient thermal conditions. Higher temperatures do not simply make the landscape uncomfortable; they can shift the energetic balance of large-bodied lizards past viability thresholds for reproducing females.

Sea-level rise acts more slowly in the near term but is projected to become progressively more important after 2050. The low-lying coastal valleys on Komodo and Rinca, and the entire footprint of the smaller islands, currently support some of the highest dragon densities recorded anywhere in the range. These valleys are precisely where sea-level inundation would strike first. The paper notes that even moderate inundation scenarios would permanently eliminate significant portions of the highest-density habitat — not temporarily flood it, but convert it to subtidal or intertidal zones beyond any plausible recovery. This asymmetry between thermal stress (potentially reversible with aggressive emissions cuts) and sea-level inundation (effectively irreversible on management timescales) has direct implications for how conservation resources should be prioritised.

The interaction between the two stressors is also nonlinear: as warming shifts suitable thermal habitat toward higher elevations, sea-level rise simultaneously narrows the lower boundary of the habitable zone from below. On small, low-relief islands like Nusa Kode and Gili Motang, this "elevation squeeze" leaves virtually no thermally tolerable, dry-land terrain between the encroaching sea and the intolerably hot lowland fringe.

Link to the 2021 IUCN Endangered Uplisting

In September 2021, the IUCN Red List reclassified Varanus komodoensis from Vulnerable to Endangered — a significant uplisting that reflected new evidence of how climate change threatens the species' future, not merely its current abundance. Jones et al. (2020) was among the peer-reviewed studies informing that reassessment.

The logic connecting the paper's modelling to the IUCN outcome is direct. The Red List criteria for Endangered status include quantitative thresholds for projected population decline over three generations. The Jones et al. models demonstrated, across multiple climate scenarios, that declines of sufficient magnitude to satisfy those thresholds are plausible even under moderate warming trajectories — and become near-certain under business-as-usual emissions. The finding that three of the five island populations face extirpation under all scenarios, while the remaining two show substantial but not catastrophic decline, gave assessors the spatial and quantitative granularity to make a defensible threat categorisation.

It is worth emphasising what the IUCN uplisting does and does not mean. It does not mean extinction is imminent or inevitable. It means that the combination of restricted range, small total population, and credible climate projections places V. komodoensis in the category where urgent, targeted conservation action is both scientifically justified and practically necessary. The Jones et al. paper did not simply provide alarming numbers — it identified exactly which geographic areas warrant that action.

Conservation & Management Implications

Jones et al. (2020) explicitly identified Komodo Island and Rinca Island as likely long-term climate refugia — locations where habitat conditions are projected to remain suitable for Varanus komodoensis under most climate scenarios through 2050. This spatial prioritisation shapes a concrete conservation agenda: these two islands should receive the strongest protection, active habitat management, and the highest monitoring investment.

The explicit identification of Komodo Island and Rinca Island as likely long-term refugia shapes a concrete spatial conservation agenda. Jones et al. argue that these two islands should be the priority targets for strengthened protection, habitat restoration, and active management — not because the other islands are unimportant, but because they are the locations where conservation investment is most likely to yield a species-level benefit under any realistic climate trajectory.

Several management conclusions follow from the modelling results:

Maintaining habitat quality in the refugia. The sensitivity analysis showed that carrying capacity — set by habitat extent and food availability — is among the most influential parameters in the population model. Any degradation of savanna, forest edge, or prey base on Komodo or Rinca will compress carrying capacity and amplify the vulnerability already imposed by warming. Invasive species management, control of illegal hunting of prey animals (deer, wild pigs, water buffalo), and limiting development footprint within the park become direct climate-adaptation measures.

Rethinking the status of at-risk island populations. The projected extirpation of Flores populations under all climate scenarios raises a difficult question: should conservation investment in currently marginal and heavily human-modified populations be redirected toward reinforcing the refugia? The authors do not prescribe a single answer, but they make explicit that scenario-conditioned planning — deciding how to allocate finite resources given a range of possible futures — is more defensible than assuming the worst-case scenario or ignoring quantified risk.

Emissions reduction as a conservation tool. The difference between the optimistic and pessimistic scenario endpoints — 15%–45% decline versus 95%–99% decline — is determined entirely by global greenhouse gas emissions trajectories, not by anything that park managers can control locally. The paper implicitly frames climate mitigation as the single most consequential intervention available to Komodo dragon conservation. Local management can shift a population from the lower end to the higher end of a given scenario's probability distribution, but it cannot substitute for the difference between RCP 2.6 and RCP 8.5.

Monitoring priorities. Because the sensitivity analysis identified Rmax and carrying capacity as the most influential model parameters, and because Rmax depends on adult and juvenile survival rates derived from mark-recapture work, continued long-term demographic monitoring on Komodo and Rinca is directly conservation-relevant. The Komodo Survival Program's multi-decade mark-recapture effort is not just scientific record-keeping — it is the empirical foundation that makes population viability projections credible.

Myths vs Facts

Common Misconception	What Jones et al. (2020) Actually Show
"Komodo dragons have survived past climate swings, so they'll be fine."	Past climatic resilience occurred over millennia without concurrent human habitat modification. The projected rate of 21st-century warming is unprecedented in the species' recorded history, and current populations are also smaller and more fragmented than in prior interglacial periods.
"The whole national park is protected, so climate threats are managed."	Legal protection prevents direct habitat destruction but does not constrain temperature rise or sea-level inundation. Three of five island populations are projected to decline toward extirpation even within park boundaries under most scenarios.
"Only worst-case emissions lead to serious declines."	Even the most optimistic RCP 2.6 scenario with low climate sensitivity projects a 15%–45% abundance decline by 2050. No scenario tested produced a stable or increasing population range-wide.
"Sea-level rise is the primary threat."	Temperature increase drives the majority of projected habitat loss through 2050. Sea-level rise becomes progressively more important later in the century, particularly for smaller, lower-relief islands.
"All five island habitats are equally at risk."	The study shows clearly differentiated outcomes: Komodo and Rinca retain suitable habitat across all scenarios; Nusa Kode and Gili Motang face high risk; Flores populations are projected to be extirpated under every scenario tested.
"There is nothing conservation managers can do about climate impacts."	The paper identifies spatial refugia where local management investment maximises long-term survival outcomes, and emphasises that reducing habitat degradation in refugia is a direct climate-adaptation strategy.

Key Takeaways

• The citation is confirmed. Jones, A.R., Jessop, T.S., Ariefiandy, A., Brook, B.W., Brown, S.C., Ciofi, C., Benu, Y.J., Purwandana, D., Sitorus, T., Wigley, T.M.L., & Fordham, D.A. (2020). Ecology and Evolution, 10(19), 10492–10507. DOI: 10.1002/ece3.6705.
• The study is the first spatially explicit population viability analysis for V. komodoensis under climate change. Over 1.2 million model iterations across six scenarios give its conclusions unusual statistical credibility.
• Komodo and Rinca Islands are the identified refugia. Both retain projected habitat suitability under all six scenario combinations and should receive priority conservation investment.
• Flores is projected to lose its dragon populations under every scenario. Nusa Kode and Gili Motang face high risk, driven by small area, low topographic relief, and susceptibility to thermal stress and sea-level inundation.
• The scale of projected decline spans an enormous range depending on emissions choices. A 15%–45% decline under strong mitigation versus a 95%–99% decline under business-as-usual. The difference is determined by global policy, not local management.
• The paper contributed to the 2021 IUCN uplisting of V. komodoensis to Endangered. That reclassification increased international visibility and legal obligations surrounding species protection.
• Habitat carrying capacity and Rmax are the most management-relevant model parameters. Protecting prey populations and native vegetation on refugia islands translates directly into population resilience under warming.

Frequently Asked Questions

What exactly is a "safe haven" or "refugium" in this context?

A refugium, in the climate-change conservation literature, is a location where a species can persist through a period of environmental stress that eliminates it elsewhere. Jones et al. use "safe haven" to mean an island that retains projected habitat suitability — defined by temperature, topography, and productivity conditions within the dragons' modelled niche — under all climate scenarios tested. Komodo and Rinca Islands meet this criterion; the smaller islands do not.

Why are Komodo and Rinca Islands better positioned than the others?

Both are substantially larger than Nusa Kode and Gili Motang, giving them greater topographic heterogeneity. Elevation gradients buffer against heat extremes: as lowland areas become thermally marginal, populations can shift upslope and still remain within suitable thermal envelopes. The smaller islands lack this elevational buffer. Flores, though large, carries heavily human-modified habitat that compounds climate pressure rather than absorbing it.

What are RCP 2.6 and RCP 8.5, and why do they matter so much?

Representative Concentration Pathways (RCPs) are standardised greenhouse gas concentration trajectories used by the Intergovernmental Panel on Climate Change to bracket a range of possible futures. RCP 2.6 assumes aggressive near-term emissions cuts consistent with ambitious international climate agreements. RCP 8.5 assumes continued high fossil-fuel use through the century. The gap between the two scenarios in this study — a difference of roughly 50–80 percentage points in projected population decline — illustrates in concrete species-level terms why global emissions trajectories are not an abstract policy question but a direct determinant of biodiversity outcomes.

Does the 2021 IUCN Endangered listing change what can be done inside Komodo National Park?

The IUCN Red List categorisation is not itself a legal instrument — it does not automatically impose new regulations on park management. Its primary effects are reputational and indirect: an Endangered listing increases pressure on donor governments, conservation organisations, and international bodies (such as UNESCO, which designates Komodo National Park as a World Heritage Site) to prioritise and fund protective measures. It also strengthens the political argument for maintaining or expanding the park's protection zone and for resisting development proposals that would degrade refugium habitat.

Could Varanus komodoensis be translocated to new islands if existing habitat becomes unsuitable?

Translocation has been discussed as a speculative long-term option. It would face substantial ecological, logistical, and regulatory hurdles: recipient islands would need appropriate prey assemblages, absence of competing predators, suitable thermal and topographic conditions, and legal protection. Dragons are also highly site-fidelal and socially structured in ways that complicate reintroduction success. Jones et al. do not model translocation as a scenario, focusing instead on in-situ refugium management as the more immediately actionable priority.

What role did the 10-year mark-recapture dataset play in the model?

The demographic parameters — stage-specific survival and fecundity rates — are the biological engine of any population viability analysis. Estimating these rates accurately requires repeated observation of individually identifiable animals over many years. The Komodo Survival Program's dataset of more than 1,000 individually marked dragons followed from 2003 to 2013 provided the vital rate estimates that anchor the Jones et al. population model. Without that long-term field investment, the demographic uncertainty in the projections would have been far larger and the policy conclusions correspondingly weaker.

Are the modelling results likely to have changed since 2020?

The core findings — refugia identification and the broad scenario envelope — are likely to remain robust. Updated climate projections from IPCC's Sixth Assessment Report (AR6, 2021) generally affirm the range of outcomes considered by Jones et al. However, new demographic data, updated land-cover information on Flores, and revised sea-level rise estimates for the Lesser Sunda region could refine the projections. This paper represents the best available quantified forecast as of its publication; it should be read as a living model awaiting iterative refinement rather than a final word.

How does this study relate to the genetics and genomics research on Komodo dragons?

Jones et al. (2020) is a demographic and spatial modelling study; it does not address genetic diversity or inbreeding risk. Complementary genomics work — including research on the adaptive potential of small, isolated island populations — would be needed to assess whether the identified refugium populations possess sufficient genetic variation to respond to rapid environmental change. The two lines of evidence together would give a more complete picture of long-term viability than either provides alone.

Sources & Further Reading

1. Jones, A.R., Jessop, T.S., Ariefiandy, A., Brook, B.W., Brown, S.C., Ciofi, C., Benu, Y.J., Purwandana, D., Sitorus, T., Wigley, T.M.L., & Fordham, D.A. (2020). Identifying island safe havens to prevent the extinction of the World's largest lizard from global warming. Ecology and Evolution, 10(19), 10492–10507. https://doi.org/10.1002/ece3.6705 (open access)
2. IUCN SSC Monitor Lizard Specialist Group (2021). Varanus komodoensis (amended version of 2019 assessment). The IUCN Red List of Threatened Species 2021. https://doi.org/10.2305/IUCN.UK.2021-3.RLTS.T22884A192523447.en
3. Ariefiandy, A., Purwandana, D., Jessop, T.S., et al. (2015). Longevity, growth and reproduction of Komodo dragons at Rinca Island. Biotropia, 22(2), 114–126. (Source of the mark-recapture demographic data underlying the Jones et al. population model.)
4. Purwandana, D., Ariefiandy, A., Jessop, T.S., et al. (2014). Ecological allometries and niche use dynamics across Komodo dragon ontogeny. Science of the Total Environment, 466–467, 39–48. https://doi.org/10.1016/j.scitotenv.2013.06.065
5. Fordham, D.A., Resit Akcakaya, H., Araújo, M.B., et al. (2012). Plant extinction risk under climate change: are forecast methods overly optimistic? Global Change Biology, 18(5), 1704–1714. (Methodological background on coupled niche-population models used in Jones et al.)
6. University of Adelaide press release (17 September 2020). Climate change may eliminate world's largest lizard. https://set.adelaide.edu.au/news/list/2020/09/17/climate-change-may-eliminate-worlds-largest-lizard
7. Komodo Survival Program. (2023). Publications. https://komododragon.org/publications/ (Repository of long-term field research underpinning the demographic parameters used in Jones et al.)