Skip to main content

Research

Human impacts on biodiversity and the effectiveness of conservation measures

My research examines human impacts on biodiversity, the effectiveness of conservation measures, and approaches to monitoring biodiversity. I combine large-scale spatial data, quasi-experimental and statistical methods, and acoustic and computational tools.

My work has informed policy assessments by the European Environment Agency (EEA), the European Commission, the Food and Agriculture Organization of the United Nations (FAO), and the United Nations Environment Programme, including the EEA’s State of Nature in the EU report, the European Commission’s Fitness Check of the Birds and Habitats Directives, and FAO’s Mainstreaming Biodiversity in Forestry.

Asian elephants feeding in dense vegetation

Protected areas and landscape ecology

Protected areas are central to global conservation policy, yet their effectiveness in reducing human pressures varies widely. My research focuses on quantifying these pressures and uses quasi-experimental designs to assess how effectively protected areas, including the EU’s Natura 2000 network, mitigate them.

Show studiesHide studies

Riparian zones across the European Union (EU) face continued threats from agriculture and urbanization, although pressures are lower within Natura 2000 sites.

Agriculture and urban areas covered 43% of the EU’s riparian zones outside Natura 2000 sites, compared with 17% inside, with Natura 2000 protection more effective at limiting urbanization than agricultural encroachment.

Biological Conservation, 2026

Many recently established protected areas in Southeast Asia have not curbed deforestation.

We compared 80 protected areas with synthetic controls derived from unprotected forests and found that only 36 (45%) reduced deforestation. An annual investment of at least US$17 million to improve management of the remaining 44 could enhance their effectiveness in curbing forest loss and protecting carbon stocks, with carbon markets offering a potential funding source.

Current Biology, 2024

Remote assessments of human pressure on biodiversity may miss important human threats.

We evaluated three widely used indices of human pressure, including the Human Modification Index, against threats recorded in Natura 2000 sites and found that they failed to reflect threats such as hunting and invasive species. These findings highlight the need for researchers and practitioners to complement remote assessments with field-based information whenever possible.

Integrative Conservation, 2022

Soundscape ecology

Passive acoustic monitoring allows vocalizing animals to be surveyed across large spatial and temporal scales. My work in this area focuses on evaluating and improving acoustic methods, including machine-learning approaches, for biodiversity monitoring.

Show studiesHide studies

A large ecoacoustic dataset from the Mediterranean Basin biodiversity hotspot.

We made more than 5,200 hours of recordings from 61 sites across Cyprus openly available through Hugging Face and the ARBIMON platform. Such large-scale datasets from biodiverse areas are essential for improving the accuracy of passive acoustic monitoring methods.

Scientific Data, 2025

Acoustic indices are only moderately correlated with biodiversity.

We analyzed 49 acoustic studies and found a moderate positive correlation between acoustic indices and biodiversity (r = 0.32), which was stronger at higher latitudes. This suggests that acoustic indices may underestimate biodiversity in the tropics, highlighting the need to investigate further the mechanisms underlying this latitudinal pattern.

Ecological Indicators, 2024

Non-focal sounds confound the relationships between acoustic indices and bird species richness.

We tested seven commonly used acoustic indices in southern China and found that sounds such as insects, wind, and human noise influenced them more strongly than bird species richness. The indices also reflected the amount of bird sound rather than the number of species, limiting their usefulness for monitoring bird diversity.

Global Ecology and Conservation, 2024

Computational methods in ecology

I am interested in exploring how various computational tools, including large language models and conformal prediction, can advance ecological research, from synthesizing scientific literature to quantifying uncertainty in machine-learning predictions.

Show studiesHide studies

Large language models have potentially transformative implications for ecology and conservation science.

We synthesized emerging applications of large language models, described the associated challenges, and outlined potential ways to address them. Applied thoughtfully, these models can be a valuable addition to the ecologist’s toolkit, but they require careful implementation to limit their risks.

Conservation Biology, 2026

Conformal prediction can quantify the uncertainty of machine learning models used to monitor animal communities.

We demonstrated how conformal prediction provides guaranteed-coverage prediction intervals for machine learning estimates of bird diversity from acoustic data, improving the reliability of biodiversity estimates by quantifying their uncertainty.

Environmental and Sustainability Indicators, 2026

Large language models can facilitate interdisciplinary research, but carry important risks.

Large language models can facilitate interdisciplinary research by helping ecologists adopt methods from other fields, but their use in complex tasks can generate inaccurate information. Building capacity through formal training in academic programs is therefore essential.

Methods in Ecology and Evolution, 2024

Policy and society

My research also explores the broader social and policy context of conservation, including how the media represents biodiversity, how the European Union allocates conservation funding, and which species benefit most from its efforts.

Show studiesHide studies

Media coverage of biodiversity falls short compared to climate change and popular culture.

We analyzed 227,940 media stories worldwide and found that COP27 received nearly seven times more coverage than COP15, the summit at which the Kunming-Montreal Global Biodiversity Framework was adopted, which was also overshadowed by popular culture. This limited media attention represents a missed opportunity to communicate the urgency of biodiversity action to the public and decision-makers.

npj Biodiversity, 2025

EU conservation funding is unevenly distributed among habitats and member states.

We analyzed LIFE projects funded since 1992 and found that a quarter of habitats had never received funding and that transnational cooperation was limited. The EU should prioritize projects in underfunded member states, particularly those targeting habitats with an unfavorable conservation status, and encourage greater transnational cooperation.

Environmental Conservation, 2020

Many of the EU’s threatened species are overlooked by its conservation efforts.

An analysis of EU conservation legislation and LIFE Programme funding revealed a taxonomic bias, with conservation efforts focusing largely on species listed in the Birds and Habitats Directives, which over-represent vertebrates, particularly birds.

Biodiversity and Conservation, 2019