Periodic Reporting for period 1 - IslandLife (Determinants of island ecological complexity in the context of global change)
Período documentado: 2022-11-01 hasta 2025-04-30
Biodiversity is declining globally at an unprecedented rate, with islands —hosting c. 30% of biodiversity hotspots— being particularly vulnerable. Island species account for 80% of reported extinctions, yet much of their unique biodiversity, including intricate species interactions, remains unknown. It is urgent to document these interactions and predict the effects of their loss to mitigate further declines. Detecting ecological interactions is a major challenge, but emerging theories and tools like critical transition theory and complex network analysis offer hope. As relatively simple systems with clear boundaries, islands present a unique opportunity to deepen our understanding of ecological complexity.
IslandLife will provide the most comprehensive and quantitatively sophisticated study of multilayer networks to date in any terrestrial ecosystem. We are focusing on six archipelagos encompassing five oceans and a wide latitudinal gradient, comparing for the first time the food web structure of ‘pristine’ (little-disturbed) islands (of a few km2) with areas of similar size in nearby disturbed (human-inhabited) islands. The overarching goal is to unveil the unique biodiversity of these ecosystems, understand their complexity, and evaluate their fragility to global change drivers, such as biological invasions. The specific objectives are four: (1) To assess and compare the ecological complexity of pristine and disturbed islands, (2) To identify commonalities in the structure of multilayer networks across archipelagos located at different latitudes, spanning from the tropics to the Arctic and Antarctic zones, (3) To evaluate the effect of alien invasive species on the structure of multilayer networks and their role as potential drivers of ecosystem collapse, and (4) To test our ability to predict network reassembly during ecosystem restoration.
We are combining direct observations during intense fieldwork, automated video monitoring and deep-learning, cutting-edge molecular techniques, and newly developed coextinction models to predict persistence and resilience of island biota to disturbances. The project will represent a major breakthrough towards understanding the effects of global change on these valuable ecosystems, of great relevance to both theoretical ecologists and applied conservationists.
IslandLife will provide the most comprehensive and quantitatively sophisticated study of multilayer networks to date in any terrestrial ecosystem. We are focusing on six archipelagos encompassing five oceans and a wide latitudinal gradient, comparing for the first time the food web structure of ‘pristine’ (little-disturbed) islands (of a few km2) with areas of similar size in nearby disturbed (human-inhabited) islands. The overarching goal is to unveil the unique biodiversity of these ecosystems, understand their complexity, and evaluate their fragility to global change drivers, such as biological invasions. The specific objectives are four: (1) To assess and compare the ecological complexity of pristine and disturbed islands, (2) To identify commonalities in the structure of multilayer networks across archipelagos located at different latitudes, spanning from the tropics to the Arctic and Antarctic zones, (3) To evaluate the effect of alien invasive species on the structure of multilayer networks and their role as potential drivers of ecosystem collapse, and (4) To test our ability to predict network reassembly during ecosystem restoration.
We are combining direct observations during intense fieldwork, automated video monitoring and deep-learning, cutting-edge molecular techniques, and newly developed coextinction models to predict persistence and resilience of island biota to disturbances. The project will represent a major breakthrough towards understanding the effects of global change on these valuable ecosystems, of great relevance to both theoretical ecologists and applied conservationists.
Data collection has been completed across five study archipelagos. Most taxa from the Balearic and Canary Islands and the Galapagos have been identified, while taxa from the remaining archipelagos are still being processed. Interaction networks have been constructed for functions where taxa are identified. Fungal networks from uninhabited islands are under analysis, with networks for inhabited islands awaiting molecular identifications.
The first multilayer network we obtained is that from Na Redona (Balearic Islands) and includes six functions: pollination, herbivory, seed dispersal, decomposition, nutrient uptake, and fungal pathogenicity. Using this dataset, we developed a new theoretical framework for constructing Multifunctional Ecological Networks (MFEN) which allows identifying a ranking of species and functions in their importance contributing to network cohesiveness. This dual perspective, integrating species-level and functional approaches, provides a deeper understanding of ecosystem complexity and enables more precise evaluation of how multifunctionality shapes ecosystem functioning and biodiversity. We have published this new framework in Nature Communications (2024) 15:8910.
The second multilayer network we constructed, from Montaña Clara (Canary Islands), revealed a simpler structure with higher versatility and modularity compared to the continental island Na Redona. These patterns align with ecological and evolutionary predictions from island biogeography theory and are attributed to the stronger effects of disharmonic biodiversity on oceanic islands, particularly smaller ones. These findings are in press in Ecology.
Our project has also introduced an innovative methodology for monitoring plant-pollinator interactions. Recent advancements have focused on tracking plant-insect interactions for specific plant species, improving our capacity to document insect presence, behavior, and interactions. However, these efforts have yet to address the accuracy of automated systems across plants with varying heights and structural complexities at a community level. Moreover, direct comparisons between automated methods and traditional manual observations remain scarce. Closing these gaps is essential for developing a more reliable automated system that can enhance our understanding of pollinator interactions across diverse ecosystems. We have submitted a paper that is in second review in the journal Methods in Ecology and Evolution describing this novel methodology.
The first multilayer network we obtained is that from Na Redona (Balearic Islands) and includes six functions: pollination, herbivory, seed dispersal, decomposition, nutrient uptake, and fungal pathogenicity. Using this dataset, we developed a new theoretical framework for constructing Multifunctional Ecological Networks (MFEN) which allows identifying a ranking of species and functions in their importance contributing to network cohesiveness. This dual perspective, integrating species-level and functional approaches, provides a deeper understanding of ecosystem complexity and enables more precise evaluation of how multifunctionality shapes ecosystem functioning and biodiversity. We have published this new framework in Nature Communications (2024) 15:8910.
The second multilayer network we constructed, from Montaña Clara (Canary Islands), revealed a simpler structure with higher versatility and modularity compared to the continental island Na Redona. These patterns align with ecological and evolutionary predictions from island biogeography theory and are attributed to the stronger effects of disharmonic biodiversity on oceanic islands, particularly smaller ones. These findings are in press in Ecology.
Our project has also introduced an innovative methodology for monitoring plant-pollinator interactions. Recent advancements have focused on tracking plant-insect interactions for specific plant species, improving our capacity to document insect presence, behavior, and interactions. However, these efforts have yet to address the accuracy of automated systems across plants with varying heights and structural complexities at a community level. Moreover, direct comparisons between automated methods and traditional manual observations remain scarce. Closing these gaps is essential for developing a more reliable automated system that can enhance our understanding of pollinator interactions across diverse ecosystems. We have submitted a paper that is in second review in the journal Methods in Ecology and Evolution describing this novel methodology.
We have advanced biodiversity studies by expanding beyond traditional approaches, which often focus on single-species interactions that mediate specific ecological functions like pollination or herbivory. While these studies provide valuable insights, they overlook the multifaceted roles species play across interconnected ecological functions. Our recent work addresses this gap by integrating various species interactions into a unified mathematical framework. This model provides a holistic view of biodiversity dynamics, illuminating how species contribute to ecosystem functionality and resilience. The framework also offers potential applications beyond ecology, such as in genetics and economics, demonstrating its versatility and value for interdisciplinary research. This innovative approach has garnered significant interest and positive feedback from the scientific community.
In tandem, our novel methodology for monitoring plant-pollinator interactions represents a groundbreaking advance in field research. Using Automatic Camera Systems (ACS) powered by Raspberry Pi 4 computers and deep learning algorithms, we automated the collection and analysis of plant-pollinator data. This method overcomes traditional limitations by accommodating plants of varying heights and structures, enabling large-scale, community-level monitoring. By leveraging convolutional neural networks (CNNs), ACS effectively detects pollinator presence and behavior, while enhancing manual observations to improve plant-pollinator network resolution. The system’s open-source design ensures it is replicable and accessible, offering a transformative tool for researchers and environmental managers to deepen their understanding of pollination dynamics and support ecosystem management efforts.
In tandem, our novel methodology for monitoring plant-pollinator interactions represents a groundbreaking advance in field research. Using Automatic Camera Systems (ACS) powered by Raspberry Pi 4 computers and deep learning algorithms, we automated the collection and analysis of plant-pollinator data. This method overcomes traditional limitations by accommodating plants of varying heights and structures, enabling large-scale, community-level monitoring. By leveraging convolutional neural networks (CNNs), ACS effectively detects pollinator presence and behavior, while enhancing manual observations to improve plant-pollinator network resolution. The system’s open-source design ensures it is replicable and accessible, offering a transformative tool for researchers and environmental managers to deepen their understanding of pollination dynamics and support ecosystem management efforts.