Research profile

My academic path has not been a typical one. The key choices in my career have been driven by what I find most important and intellectually compelling, not what will help me climb a university ladder toward tenure. This also means my research interests have constantly changed: I am less interested in what I have already done and more intersted in what I have not yet done. My tendency to think outside the box and across disciplinary boundaries has compelled me to secure external funding for my research, totaling some ten million euros in competitive grants since my habilitation. My work has been published in Science, Nature, PNAS among others.

1 Population models and viability analyses

What are population models? When one thinks that smaller or declining wildlife populations have a higher risk of disappearing, one has started to become a modeller. Models are simplified description of the world. Conceptually linking population size with extinction risk is a qualitative model. In ecological sciences, we use more quantitative models where, for example, we would describe the dynamic of a population with equations to be able to give estimates of its risk of extinction. This is how I started my career: calculating the extinction risk of small populations of large carnivores, such as wolves and bears.

Quantifying extinction risk

How do we do that? When quantifying extinction risk, models – or their use – are often termed a Population Viability Analysis (PVA) and rely on Monte Carlo simulations. These simulations are repeated stochastic (i.e. random) trajectories where the model starts each trajectory from the same initial conditions and parameters. Extinction risk is simply the ratio of extinct trajectories divided by the total number of run trajectories. My first paper (published in 2003 in Comptes Rendus Biologie 10.1016/S1631-0691(03)00148-3) estimated the extinction risk of wolves especially when some individuals are killed after they attacked livestock.
There are however other approaches to estimate extinction risk. One such approach is to estimate the probability density of the population growth rate: if the distribution contains a substantial area smaller than 1 (i.e. stable growth rate), this indicate a non-negligible extinction risk. We published in the Proceedings of the National Academy of Sciences (PNAS) the first rigorous quantitative evaluation of the conservation status of lions, showing with co-authors that lion populations are rapidly declining, except in intensively managed areas (10.1073/pnas.1500664112). I have also used Bayesian models to investigate the demography of populations of large carnivores hunted at low population size (10.1002/eap.2063).

Hierarchical models

As their names tell, Bayesian hierarchical state-space models are on Bayesian statistics where everything, data, model, hypotheses and parameters is a probability. Such models allow the mechanistic investigation of ecological processes by combining multiple sources of data in a statistically coherent way. This is very useful to estimate unobserved quantities (such as poaching 10.1098/rspb.2011.1275) and to quantify both process and data uncertainties (such as error in population estimates) when estimating parameters (10.1890/11-1309.1).

Markov Decision Processes

Numerous challenges in environmental conservation involve making decisions about the best choice among a set of competing actions. For example, if you could buy every year some areas each hosting threatened species and convert them into reserves, which areas would you buy considering that you cannot buy all areas at once and areas you did not buy a year will be developed the next year? Finding the sequences of actions to maximise your outcome (the utility function) requires solving a Markov Decision Process (MDP). Our R package ‘MDPtoolbox’ (10.1111/ecog.00888) proposes functions related to the resolution of discrete-time MDPs: finite horizon, value iteration, policy iteration, linear programming algorithms with some variants and also proposes some functions related to artificial intelligence (Reinforcement Learning).

Approximate Bayesian Computation

Model-based inference generally relies on estimating the likelihood function, which is the probability of data given a model and parameters. For example hierarchical state-space models typically require writing the likelihood. However, for complex models (such as IBM, where individuals are explicitly described as the functional unit of the model and the population trend is the emerging population level consequence of individual events), the likelihood function is unknown and cannot be written. Monte Carlo Markov Chain based methods can therefore not be used. Enters Approximate Bayesian Computation (ABC), a class of computational Bayesian methods that bypasses the direct evaluation of the likelihood function by approximating it and allowing IBM to be fitted to data (10.1016/j.ecolmodel.2016.08.012).

Quantifying illegal mortality

Large carnivores commonly face illegal mortality, and I have focused on quantifying poaching. This is not straightforward because there are strong incentives for poachers to conceal evidence of their illegal activity. Radio-marked animals typically disappearing with no tracks. I brought hierarchical state-space Bayesian models to the rescue, making use of their unique ability to merge multiple sources of data in a coherent statistical framework. In the Proceedings of the Royal Society B (10.1098/rspb.2011.1275), I quantified the impact of poaching on wolves in Scandinavia and demonstrated that half their mortality was poaching, two-thirds of which remained undetected. I investigated further how poaching reacts to policy signals, assessing the hypothesis that allowing legal culling would reduce illegal killing, as often propounded by governments. I developed a Bayesian model, separating the policy signal (the allowance of culling) from its implementation (actual culling), to show that—contrary to the intended conservation outcomes—allowing culling was more likely to be associated with an increase in poaching. These findings were also published in the Proceedings of the Royal Society B (10.1098/rspb.2015.2939) and generated three replies. I made a video with Playmobil figurines to explain our results to the public. Also on the topic of poaching, we have proposed a SES framework to better understand it (10.1007/s13280-016-0852-z), documented its impact on the demography of lynx in Switzerland (10.3389/fcosc.2021.665000) and lions in Uganda with an agent-based model (10.1016/j.biocon.2023.110147).

R packages for population simulations

population: Models for Simulating Population: Run population simulations using an Individual-Based Model (IBM) compiled in C.
pop.wolf: Models for Simulating Wolf Populations: Simulate the dynamic of wolf populations using a specific Individual-Based Model (IBM) compiled in C.
pop.lion: Models for Simulating Lion Populations: Simulate the dynamic of lion populations using a specific Individual-Based Model (IBM) compiled in C.
MDPtoolbox: Markov Decision Processes Toolbox: The Markov Decision Processes (MDP) toolbox proposes functions related to the resolution of discrete-time Markov Decision Processes: finite horizon, value iteration, policy iteration, linear programming algorithms with some variants and also proposes some functions related to Reinforcement Learning.

2 Conservation of European megafauna

Return of wolves, bears and lynx in Europe

The contribution for which I am most often known is my paper documenting the recovery of large carnivores in Europe published in Science with >70 authors from most European countries (10.1126/science.1257553). We reported that brown bears, Eurasian lynx, grey wolves, and wolverines are now recovering across large portions of Europe, even in densely populated and heavily modified landscapes. This revealed a significant conservation success that challenges prevailing assumptions about the relationship between humans and large carnivores. We have recently published an update about the recovery of wolves in Europe (10.1371/journal.pstr.0000158). We report that wolves are continuing to make a remarkable comeback across Europe, with their population growing to over 21,500 individuals by 2022–a 58% increase in a decade. Annually, wolves kill around 56,000 domestic animals in the EU, leading to annual costs of about 17 million EUR for damage compensation. Coexisting with newly established wolf populations in Europe entails managing impacts on human activities, including livestock depredation, competition for game, and fear of attacks on humans, amidst varying social and political views on wolf recovery. Sustainable coexistence continues to operate in evolving and complex social, economic, and political landscapes, often characterized by intense debates regarding wolf policies.

Towards a theory of coexistence with large carnivores

We have attempted to theorize the conservation of large carnivores as securing the maintenance of a community of species with one of them being hyper-predatory humans (Trends in Ecology & Evolution 10.1016/j.tree.2016.06.003). The resource these species mainly compete for is space. Ecologists have shown that communities of competing species will converge to single species communities when one species displays strong competitive abilities and has a limited niche differentiation with others. In contrast, species can coexist when they show moderate competitive abilities and large niche differentiation. The question of how to conserve species in the Anthropocene therefore requires understanding how a hyper-predator –humans– may avoid displacing other competing species by deliberately developing i) a higher niche differentiation and ii) a lesser competitive ability. In our conceptual framework, we illustrate how coexistence between large carnivores and people emerges from differences in competitive abilities and ecological niches: A) Extinction: unregulated killing increases human competitive abilities and inadequate human practices maintain strong niche overlap. B) Weak conservation: lack of effective protection maintains high human competitive abilities, but protected areas increase niche differentiation. C) Weak conservation: high tolerance to predators reduces human competitive abilities but inadequate human practices maintain strong niche overlap. D) Strong conservation: environmental rule of law reduces human competitive abilities and protected areas increase niche differentiation.

What works in conservation?

Carnivore predation on livestock often leads people to retaliate and persecution by humans has contributed strongly to global endangerment of carnivores. It is therefore likely that preventing livestock losses would help to achieve three goals common to many human societies: preserve nature, protect animal welfare, and safeguard human livelihoods. Between 2016 and 2018, four independent reviews (including ours) evaluated >40 years of research on lethal and nonlethal interventions for reducing predation on livestock. When pooling these 114 studies reviewed, we found a striking conclusion: scarce quantitative comparisons of interventions and scarce comparisons against experimental controls preclude strong inference about the effectiveness of methods (PLOS Biology 10.1371/journal.pbio.2005577). Immense resources are spent globally each year to protect livestock from carnivores, but too often without science-based evidence that the methods work. Evidence of the effectiveness of a deterrent should be a mandatory prerequisite to large-scale funding, policy-making and implementation. An appropriate evidence base is needed, and we recommend a coalition of scientists and managers be formed to establish and encourage use of consistent standards in future experimental evaluations.

3 Legal conservation

Making law and ecology talking to (and not past) each other

Just over three decades ago, the scientific community created a new discipline –conservation biology– to address the loss of biodiversity. That discipline has since succeeded in delivering answers to many ecological questions to stop the loss of biodiversity. However, it has been argued that further contributions from this discipline alone may somewhat be limited, with additional ecological knowledge likely to bring only marginal gains for understanding how to conserve biodiversity. More substantial gains can certainly be gained through interdisciplinary research. One discipline with an untapped potential is law. However, conservation scientists appear in general to be ignorant of nature protection laws, while environmental lawyers generally have a limited understanding of ecological systems. By grounding law in ecological knowledge and more systematically understanding law when planning and enacting conservation policies, better policy outcomes may be achieved. We believe the time has come to break walls between these two disciplines – epistemologically distinct but nevertheless interacting on a daily basis – and to develop what we termed legal conservation.

Favorable Conservation Status

One of the most important concept used in the EU Habitats Directive is the one of favorable conservation status (FCS). There are however many contested aspects when interpreting FCS that have not yet been conclusively settled. In a paper published in Conservation Letters (10.1111/conl.12200), we have provided legal-ecological clarifications of the most contested aspects: at what level should FCS be measured, what it means for a species to be a “viable component of its natural habitat”, what is a “long‐term basis”, what does it mean for a species to “maintain itself”, whether FCS should be measured from extinction or carrying capacity and whether FCS requires that a population approaches historical levels.

When is hunting legal?

The European Union (EU) bans the killing of strictly protected animals through the Habitats Directive. This law allows exceptions in special circumstances when doing so would not be detrimental to the conservation status of species' populations. Some decisions to kill animals have triggered litigation regarding how broadly the provision on exceptions can be interpreted. In Conservation Science and Practice (10.1111/csp2.18) we review several contested aspects of the law to conclude that it would be very difficult for countries within the EU to allow hunting of strictly protected animals because of the restrictive interpretations supported by prior decisions of the EU Court of Justice and other sources of law.

Derogations to strict protection

Whether or under what circumstances the hunting of species listed as strictly protected in the Habitats Directive's Annex IV can be allowed has been the subject of extensive controversy and litigation in several EU Member States. In fall 2017, Finland has asked the Court of Justice of the European Union for a preliminary ruling on several related questions, which we have addressed in a paper published in European Energy and Environmental Law Review (10.54648/eelr2018009). Specifically, we focus on the permissibility of the management hunting of strictly protected species, the permissibility of allowing hunting with the goal of preventing poaching, and at what scale the "favourable conservation status'' of species populations should be considered.

Obligations to reintroduce

The obligations of Member States when a population of species protected in the Habitats Directive has almost gone extinct also remain unclear. In a paper published in Biological Conservation (10.1016/j.biocon.2018.09.027), we use the case of the quasi-extinct wolf population in the Sierra Morena region in Spain to provide legal-ecological clarifications on the obligations by Member States. We show that Articles 6 and 12 of the Directive require Member States to restore populations, and the complete extinction of the species does not exonerate Member States from their obligations regarding the conservation of the species in Natura 2000 sites.

A database of species protection litigation in Europe

As a quantitative ecologist, I did not want this interdisciplinary tack to preclude a quantitative approach in legal conservation, so with the help of a jurist research assistant we assembled a team of international legal experts and collected all the court rulings from all the administrative or constitutional courts of all EU Member States relevant to derogations to the protection of species listed in the Birds or Habitats Directive. This massive effort has led to the compilation of a database with > 6,800 court rulings (www.clawsandlaws.info), each of them coded to facilitate further analysis. This dataset has made possible to document patterns of litigation in France during two decades of wolf recovery (preprint, 10.1101/2022.10.11.511781). We are also now using Large Language Models to extract legal arguments from the rulings and to carry a deeper cross-European analysis of species protection litigation.

4 Politics and values in nature conservation

Current political controversies

Law and litigation cannot be dissociated from politics. Since my doctorate I have published more than twenty letters in Science and Nature, critically commenting on current affairs in conservation. These letters provide insights about recent events in conservation, highlighting challenges and opportunities at the science-policy interface. Some have generated media coverage and debates. I have written about proposed and significant legal changes to biodiversity protection (10.1126/science.adk7686, 10.1126/science.abe6191, 10.1126/science.aam6200, 10.1126/science.aay8053), recent court rulings from the CJEU and what they imply for biodiversity policies (10.1126/science.aaz8424, 10.1126/science.abj9226), general trends in conservation in the EU (10.1126/science.adp1306, 10.1126/science.adf2714) or in conservation more broadly (10.1126/science.320.5872.47a, 10.1038/d41586-018-02064-4, 10.1126/science.343.6176.1199-b, 10.1126/science.322.5904.1049b), the importance of science and scientists in conserving biodiversity (10.1126/science.314.5806.1682c, 10.1038/442627a, 10.1038/451127b) and the risk that science is abused for political purpose (10.1126/science.339.6127.1521-a, 10.1038/516289a).

What are the values of conservation?

Politics is about who gets what, when and how, and this naturally applies to conservation too. I believe that it is beneficial to reflect more on the values and politics in conservation. Writing in Conservation Biology (10.1111/cobi.13065), I introduced the concept of “political populations”, which are populations with ecological attributes, such as size or abundance, that are constructed to serve political interests by misrepresenting scientific evidence. Again in Conservation Biology (10.1111/cobi.13485), I explained that some of the intellectual premises in the field of conservation—by blaming conservation for creating social conflicts—may prevent such efforts from truly addressing the accelerating conservation crisis. Also in Conservation Biology (10.1111/cobi.13541), I explained the need for more ‘ecocentric’ perspectives. I led a perspective in Science (10.1126/science.aav5601) where we discussed at large the emerging idea of rights of nature, although my work on this question has subsequently convinced me this idea carries risks for human rights abuse. Finally, I wrote a satire in Trends in Ecology and Evolution (10.1016/j.tree.2017.12.010), issuing a “A Final Warning to Planet Earth” whereby I provocatively used irony to deliver a serious message (explained further in 10.1016/j.tree.2018.04.017), mocking the relentless pursuit of economic growth and technological advancement at the expense of ecological stability.

Do Europeans still support the recovery of large carnivores?

To answert this question, we ran a questionnaire that reached 10,807 respondents in all EU member states that host breeding populations of wolves, lynx or brown bears. In every country surveyed, we found that support for large carnivore recovery outweighed opposition, often by a substantial margin. The strongest support was found in southern and eastern Europe and the presence of abundant large carnivore populations did not seem to diminish public support. However, this support for recovery does not necessarily translate into support for population increases: Europeans are generally not keen on further growth in the numbers of large carnivores. Paradoxically, Europeans also tend to oppose hunting these animals; opposition to hunting was strongest in southern Europe, with more than 50% opposing it in many countries. Noticeably, we found little difference between the opinions of rural and urban residents, and support for large carnivore recovery was substantial even in rural areas. Although a relative majority of Europeans support large carnivore recovery, at least one‐third of the population remains neutral; even on the issue of large carnivore hunting, which is often thought of as highly contentious, we found that many Europeans neither agree nor disagree. Our assessment of public opinion suggests that Europe displays far less polarization than is often thought. Published in Nature Ecology & Evolution (10.1038/s41559-025-02914-1 and 10.1038/s41559-025-02915-0)

5 How new technologies may improve policies

Trusted environmental policies through crypto-governance

I have not restricted my interdisciplinary interests to law but have also looked for technological innovations in conservation and sustainability. I explained in Nature (10.1038/545403a) that traditional modes of governance are failing to achieve sustainability, largely due to corruption and the inability to foster trust among diverse stakeholders. I argued that the introduction of blockchain technology, originally developed for cryptocurrencies, may offer a radical solution by decentralizing trust and eliminating the need for intermediaries such as governments or banks. This technology could potentially ensure the integrity of environmental data, enhance transparency, and enforce conservation efforts through “smart contracts,” which execute actions automatically when specific conditions are met. By securing and tracing resources like land titles, wildlife products, or carbon credits, blockchain could revolutionize how environmental governance is conducted, making it more resilient to fraud and political interference.

Artificial intelligence in strategic decision making

I have pursed additional efforts to become even more interdisciplinary by linking artificial intelligence (AI) and strategy. In my dissertation for my Master in International Relations at King’s College London (“Towards deep strategic theory”, 2023, 10.5281/zenodo.15094987), I raised the question of whether AI can master strategic thinking, examining the most recent AI advances and discussing their implications for strategy. I reviewed AI achievements in games like Go or more complex games involving negotiation and betrayal such as Diplomacy. I highlighted key traits that make strategy challenging even for AI, connected strategic thinking to deep learning algorithms and integrated Clausewitz’s treaty On War with complexity theory. I concluded by discussing challenges arising from the psychological and unpredictable aspects of strategy.

6 Generative modeling of complex systems

Can we predict the outcomes of policies (before they fail)?

Rather than relying on past evidence to determine “what works”, I propose using advanced simulations to instead explore the question “will it work?” In a new project (2024-02029_Formas), we will use Large Language Models as agents impersonating key stakeholders in Social-Ecological Systems models and run realistic simulations that capture the complexities of real-world dynamics. These simulations, structured as sophisticated serious games, will model the dynamic of conservation policies in real-world contexts, accounting for both supportive and opposing stakeholders. We will apply this approach to case studies in several European countries, focusing on ongoing policy changes related to large carnivore conservation, an often controversial and therefore well-documented issue. After validating and calibrating our models through Turing tests, we will conduct Monte Carlo simulations to generate distributions of policy outcomes. Our approach is general and will provide policymakers with an open, transparent, and accessible foresight tool to anticipate the success or failure of conservation policies, helping them design more effective policies.

Generative wargaming of socio-ecological warfare

This new project (2025-05075_VR) pioneers a novel methodology to enhance Europe's societal preparedness by simulating adversarial disruptions using AI-powered wargames. Building on wargaming traditions, I will integrate Large Language Models (LLMs) into agent-based models to simulate complex, adaptive interactions between societal actors facing strategic threats. The project introduces the concept of socio-ecological warfare—a previously overlooked form of “hybrid" conflict in which adversaries destabilize societies by exploiting or engineering ecological crises. Case studies include adversarial manipulation of opinion polarization in environmental policies, deliberate introduction of African Swine Fever, and the use of genetic engineering to amplify tick-borne diseases. Agents are modeled using nested LLM architectures, enabling them to mimic real-world institutions and stakeholders with context-aware reasoning, persuasion, and adaptation. Each wargame is run at scale using Monte Carlo simulations and followed by a pre-mortem analysis to identify weaknesses in our society. The project advances scientific understanding of complex threats and provides a new tool for policy testing under uncertainty, supporting society's long-term resilience planning through a rigorously interdisciplinary and scalable simulation framework.

Website of Guillaume Chapron © 2016-2026.
The views and opinions expressed in this website are mine only and do not whatsoever reflect the official policy or position of any other agency, organization, employer or company I am or have been associated with.
Contact: gchapron@carnivoreconservation.org or guillaume.chapron@slu.se