Monday, November 21, 2022 - Friday, November 25, 2022 Metz, France

Think big: agent-based modelling meets data science

24 November 2022
S18 10:00 > 12:00 Think big: agent-based modelling meets data science Room 05

Volker Grimm, Helmholtz Center for Environmental Research-UFZ, Leipzig,

Uta Berger, Technische Universität Dresden, Dresden,

Session description
Ecological phenomena result from the behavior of their constituent agents and their interaction with their biotic and abiotic environment. Agent-based modeling aims to capture these processes and is thus an important complement to more aggregate or statistical modeling approaches. However, the amount of data required to parameterize and test agent-based models (ABMs) often limits them to small areas or specific species and systems. Advances in monitoring (e.g., genetic markers, motion tracking, remote sensing), powerful workflows including machine learning methods to exploit large data sets, and the increasing availability of high-performance computing systems have now enabled agent-based modeling to "go big," i.e., cover large regions and target more general applications and theories. In addition, advances in transferability, scaling, and first-principles theory development through ABM are enabling a mechanistic explanation of the processes behind patterns detected by statistical tools in large data sets. So-called "digital twins" for biodiversity, as envisioned in a recent European Commission funding initiative, appear to be becoming possible: They would allow us to continuously incorporate new data to reduce uncertainty in models and use them for new regions, making them a tool for scenario assessment and policy development. This symposium reviews advances in monitoring, data science, and agent-based modeling, and presents a vision of "big" ABMs that can take ecological application and theory to a new level.

INT64 Combining the forces of data science and agent-based modeling and their perspectives on ecology and environmental management > U. Uta BERGER
Content : BERGER, Uta, Technische Universität Dresden, Dresden, & GRIMM, Volker, Helmholtz Centre for Environmental Research, Leipzig,

We will introduce into the themes of the symposium and provide an overview of the talks to follow.
INT65 Improving the realism of predictive systems models with field data-based calibrations: a large-scale study case on honey bees > F. Fabrice REQUIER
Content : REQUIER, Fabrice, Université Paris-Saclay, CNRS, IRD, UMR Évolution, Génomes, Comportement et Écologie,

Predictive system models are used in ecology to tackle problems that are too complex to be investigated in the field. However, model predictions are simplifications of real processes. This talk aims at proposing a rigorous method of implementing empirical parameters in models as a preliminary calibration step. This calibration step would help at improving the likelihood of the predictions, but also to better consider model outputs for applied purposes. As an example, I am using data from the large-scale ECOBEE monitoring of 250 honey bee colonies and the model BEEHAVE. This model has the ability to address multiple stress interactions that can affect colony resilience and explain the current global decline of bees.
INT66 Energy budget models as tools for understanding population dynamics and predicting disturbance impacts: the harbor porpoise in the North Sea > C. Cara GALLAGHER
Content : GALLAGHER, Cara, University of Potsdam, Plant Ecology and Nature Conservation,

Animals must balance their foraging rates with their energy expenditure to successfully grow, reproduce, and survive. However, changes in environmental conditions, such as those resulting from anthropogenic disturbances and climate change, can alter individual behavior and energy balance, ultimately leading to impacts on population vital rates. I will discuss the use of energy budget models as tools in both theoretical and applied investigations of the impacts of environmental changes on population dynamics. As an example, I will present an agent-based energy budget model developed for harbor porpoises (Phocoena phocoena) which was thoroughly calibrated and tested using empirical data for the species following a pattern-oriented modelling approach. I will discuss two applications of the model to assessing the implications of 1) acoustic disturbance in the North Sea and 2) climate change-induced alterations in prey structure on the porpoise population.
INT67 Next generation forest simulation models in sustainable forest management: an integration of remote sensing, machine learning and individual-tree-based modelling > Y. Yue LIN
Content : LIN, Yue, Timberlands Ltd., Rotorua 3010, New Zealand,

Forest managers need new forest simulation models to evaluate and predict the impacts of environmental changes (such as climate, land use, disease, wind disturbance, and forest harvesting) on the sustainability of forest ecosystems. I will present a new forest simulation model platform, which (1) is based on diverse modelling approaches, including empirical statistical models, machine learning, and mechanistic individual-based modelling, and (2) utilises massive data collected by modern remote sensing techniques (such as satellite images, UAV aerial photography, and LiDAR). It can effectively transfer model results, from individual tree level to forest system level, to forest managers. This new modelling approach can precisely predict growth and yield and reduce uncertainties under climate change.
INT68 Predicting desert locust behaviors and outbreaks with agent-based models: from individuals to continents > C. Cyril PIOU
Content : PIOU, Cyril, CIRAD - UMR CBGP, Montferrier sur Lez, , Lucile Marescot, Camille Vernier, Fanny Herbillon, Pierre-Emmanuel Gay, Sory Cissé, Ahmed Salem Benahi, Maeva Sorel, Christine Meynard

Desert locusts can build cohesive swarms of billions of individuals that travel hundreds of kilometres. The impacts on agriculture can be disastrous. The preventive management strategy aims in localizing the habitats that favour the concentration of solitarious individuals toward the plastic change of becoming gregarious and swarming. Two ecological modelling approaches exist to better understand and forecast this threat: (1) Agent-based models (ABM) are used to understand the role of vegetation structure in triggering outbreaks, the collective marches of the gregarious hoppers, the migration of swarms in relation to wind, temperature and vegetation and the management system that sometimes misses the onset of outbreaks. (2) Remote sensing and machine learning are used to help in the forecasting of areas of potential outbreaks. In recent works, we attempt to merge both approaches with a meta-agent based model to simulate population dynamics over continental scales and bring more mechanisms into the forecasting tools of desert locust managers.
INT69 Agent-based models for predicting range dynamics under global change > D. Damaris ZURELL
Content : Talk 6. ZURELL, Damaris, University of Potsdam, Ecology/Macroecology,

Researchers called for more process-based models in biodiversity science and global change ecology for decades, yet progress has been slow due to data and technical challenges. I will present recent advances in spatially explicit, agent-based modelling platforms for predicting biodiversity dynamics. By comparing these developments to the success story of static distribution models, I derive key recommendations for facilitating wider usage in decision support and conservation.
Copyright © key4events - All rights reserved