Broder Breckling (with Ulrike Middelhoff)

Individual-based models as tools for ecological theory and application

Understanding the emergence of organisational properties in ecological systems

Individual-based models open a structurally unique (and unifying) approach to ecological applications. Model results also provide an important input to ecological theory. The approach operates on the lowest organisational level considered in ecology. Simulating actions of single organisms allows to study how the properties of higher level ecological entities like swarms, populations and trophic networks emerge. Unlike other approaches working on a higher abstraction level, individual-based models can represent structural-functional relationships in a close qualitative and quantitative relation to the form and content of ecological knowledge. To demonstrate the application range of the approach for the advancement of ecology this presentation takes four steps:

First, a generic model structure for individual-based models operating on the basis of object oriented programming is explained. It allows to capture a large variety of different ecological interactions and thus a coherent representation of ecological knowledge during model development.

In a second step application examples from various fields of ecology will be demonstrated. Plants and animals, active in terrestrial or aquatic environments, exhibit interaction types, which lead to self-organised structural-functional networks resulting from single organismic interactions. Spatial relations, dispersal, bio-energetics, plasticity of growth and form are issues which can be successfully dealt with in individual-based models. The wide range of qualitatively different interactions is responsible for the importance the approach has gained in ecology.

In a third step we show, how the approach is used in a current research project on an advanced, level-integrating study anticipating properties and coexistence implications of genetically modified plants in agriculture. In a holistic approach, an individual-based model is used to simulate dispersal effects of genetically modified organisms (GMO). Oilseed rape (Brassica napus) serves as an example. The model represents cultivation, feral populations, cultivation practice and environmental characteristics. It is run for various environmental conditions to allow an up-scaling from single plots to the landscape level. Physiological data, climate information and remote sensing data were used as model input. The results allow estimations of the presence of transgenic material outside cultivated areas as well as unintended dispersal between cultivated fields.

The fourth and final step discusses epistemological implications of individual-based models. It is concluded, that a successful application of the approach requires detailed biological information about the represented species. This makes a very close involvement of field ecologists essential. On the other side, it also opens a theoretical access how to connect quantitative and qualitative aspects of cause-effect chains in ecology in a conceptual way.

Joseph C. Coughlan

An overview of ecological modeling and machine learning research within the U.S. National Aeronautics and Space Administration

In the early 1980’s NASA began research to understand global habitability and quantify the processes and fluxes between the Earth’s vegetation and the biosphere. This effort evolved into the Earth Observing System Program which current encompasses 18 platforms and 80 sensors. During this time, the global environmental research community has evolved from a data poor to a data rich research area and is challenged to provide timely use of these new data. This talk will outline some of the data mining research NASA has funded in support for the environmental sciences in the Intelligent Systems project and will give a specific example in ecological forecasting, predicting the land surface properties given nowcasts and weather forecasts, using the Terrestrial Observation and Prediction System (TOPS).

Cesare Furlanello

GIS-based predictive models for ecology

Joint work with S. Merler, S. Menegon, M. Neteler, S. Fontanari, R. Blazek, A. Rizzoli, and C. Chemini

This talk will discuss how machine learning methods may be integrated within a Geographical Information System (GIS) for the development of new approaches in ecology research. As needed in tasks in landscape epidemiology and wildlife management, it is now possible to develop unified environments in which methods of statistical learning and spatial statistics are combined and applied to feature vectors derived from GIS analysis of digital maps, or from relational databases with embedded GIS capabilities. For example, multitemporal predictive maps may be obtained by modeling with classification trees or with Breiman's Random Forest in R, analyzing geodata and digital maps in the GRASS GIS, and managing biodata samples and climatic data in PostgreSQL. Overall, connecting a working data notification and management system to the predictive models is of crucial importance for a practical use of machine learning on ecological data.

We first describe a risk mapping system for tick-borne diseases, applied to create a multitemporal risk model of exposure to Lyme borreliosis and TBE in Trentino, Italian Alps. The system input features include vegetation data derived from the Forest register and multitemporal climatic data, also from remote sensing. As a second example, a predictive risk model for deer-vehicle collisions will be presented, considering variables as distance from urban areas and from waters, wildlife population density maps, vector line analysis (road curvatures) and traffic data. Features for the model include a multiscale accident site characterization based on the integrated use of orthophoto landuse classification and morphometrical analysis of the digital elevation model. The methodology has been applied at mesoscale (6200 km^2) for the predictive modeling of deer-vehicle collisions, a project for the Wildlife Management and Road Transportation Services of Trentino. We will present methods for variable importance analysis, classification with combined models and the resulting roe deer-vehicle accident risk maps.

Bai-Lian Li

Modeling ecological complexity: Challenges and opportunities

Ecological complexity is an emerging and rapidly growing interdisciplinary field in ecology. It focuses on how and why complex ecological systems emerge from the nonlinear interactions of living entities at all levels and spatiotemporal scales and with all facets of their external environment including the human dimension. The field is based on a complexity theoretical framework for solving real world environmental problems. It has been recognized as the most important and exciting frontier of the 21st century ecology.

Modeling has played the most significant role in studying complex ecological systems and has provided the mathematical tools indispensable for studying their dynamics. In this talk, I will start with a brief overview of current research and paradigm shift in modeling ecological complexity, then outline computational, mathematical and statistical challenges from complex ecology to ecological modelers, along with several examples from our own recent studies (e.g., emergent properties of scaling and power law, spatiotemporal complexity and chaos of ecological pattern formations, ecological phase transitions, assessment of sustainability, etc.). I will also show how we could take those challenges as a great opportunity for seeking a truly quantitative and integrative approach towards a better understanding of the complex, nonlinear interactions (behavioral, biological, chemical, ecological, environmental, physical, social, cultural) that affect, sustain, or are influenced by all living systems, including humans, which may enable us to explain and ultimately predict the outcome of such interactions.

Jacqueline McGlade

Spatial assessments of Europe's environment

Integrated environmental and ecosystem health assessments rely on combining information from local and global attributes derived from surveys and case studies. In order to properly examine issues such as the impact of climate change, loss of biodiversity, environmental threats to human health or the long-term effects of infrastructure development on Europe’s landscapes, the European Environment Agency (EEA) needs to be able to analyse changes across a range of scales and media (water, air, soil etc.). But despite extensive monitoring and research, the current situation in Europe is that we cannot meet the challenge of supporting consistent environmental and sectoral policies at a European, national and regional levels.

The knowledge needed will not be obtained solely through the accumulation of observations on individual systems but, will require such in situ data to be integrated within overall frameworks of models and data analysis to generalise their information content. In relation to the demands of understanding changes in Europe’s environments, using spatially distributed data and information on ecosystems and human activities is a key factor, as they:

• can help identify where conflicts in use of the territory take place, and under which type of pressure;
• contribute to the stratification of data and knowledge from existing monitoring networks and research programmes;
• help in designing efficient sampling schemes for new monitoring networks as well as targeting research programmes to priority needs;
• provide important input to modelling, in particular when very heterogeneous information from the bio-physical, social and economic realms need to be integrated and
• can be up- and downscaled to the appropriate levels of decision making of the various public and private bodies.

In this context, land accounts for Europe are being implemented by the EEA. The purpose of land accounts is to observe, qualify and quantify the cover of land resulting from ecosystem and land use. Stocks of land cover are described as well as their change. A first set of land cover accounts is under construction using CORINE land cover data from 1990 and 2000. Within this accounting framework, assessments of ecosystem condition has been produced; for example, in the case of European wetlands, spatial data on the change in extension, fragmentation, connectivity and neighbourhoods can provide insights into the possible destruction and stress. These first variables are being supplemented by data on flora and fauna and by quantitative and qualitative data on water. Other spatial data to be included, are land use in agriculture, urban development and transport infrastructure which will help to identify the sources of stress. These spatial data will be supplemented, at a more aggregated level, with social and economic statistics, from the perspective of the development of land use accounts to show how social and economic activities influence our environment.

The results of land accounts will be fully made available on the EEA website with the aim of facilitating access to these data and approaches to a range of users, including researchers and the wider public.