Abstract

Sustainable management of the environment is a major challenge facing not only Slovenia, but also the entire mankind. Large ecosystems, especially forests, play a major role when addressing this task, having critical impact on the quality of life and obvious social-economic benefits for the society. Systematic and complete monitoring of the evolution of such ecosystems is extremely difficult, up to now even impossible, due to their vast geographic scales and huge amounts of their miniature basic elements. Only recent advances in remote sensing technologies that have revolutionized the area of Earth observations provide us with possible insights into the dynamics of such ecosystems. Sophisticated satellite observation systems from the Copernicus program and state-of-the-art laser scanning technologies like LiDAR, allow for periodical monitoring of large geographical areas with high enough resolution and precision to distinguish the smallest basic elements of ecosystems, such as trees, undergrowth, and shrubs. However, the huge amounts of heterogeneous and complex data they acquire remains a major challenge for the future as contemporary software solutions are incapable to deliver data analytics in a systematic, organized manner. Before a holistic information space for efficient management of large ecosystems can be developed, major issues have to be addressed, regarding integration of heterogeneous Earth observations data, implementation of relevant analytic tools for their processing and, finally, relevant models of their dynamics.

The proposed project meets these challenges by introducing a new paradigm for data integration based on the decomposition of heterogeneous Earth observations into the contained basic semantic elements, their fusing and enrichment with complementary information from within different data types, and their inter-linking into a complex network. Through advanced concepts of mathematical morphology, formalizing arithmetic of shapes for sophisticated pattern analysis, the decomposition of specific data types and the recognition of the basic ecosystems’ elements will be achieved. Their geometric features will be used to determine their social status, and consequently the likelihood of their mutual influence. These will be represented with a complex network, enabling us to develop a wide range of new algorithms based on up to now unexploited mathematical and analytical methods at such large scale. This new type of data analytics will be derived primarily from methodological studies of partially ordered sets based on lattice theory and statistical-topological features based on the theory of complex networks. Such fundamental shift in the design of the pattern recognition algorithms will provide the thoughtfully required capabilities for the development of new approaches to recognition of complex structures, composed of multiple basic elements, while comparison of complex networks will allow for systematic monitoring of their evolution. Hence, the foundation for recognizing interactions between the basic elements will be established, giving us the framework for modelling dynamics of large ecosystems.

While in-situ measurements will be used to validate these algorithms, a study of forest dynamics due to the competition of trees for accessing resources and leaving space will provide the proof of concept. All the developed methods will be implemented in the form of weakly coupled service for this purpose and integrated into an existing platform for geographic data management and processing. A user-friendly environment for services orchestration and execution of analytic scenarios on-demand will be provided to experts in a form of end-user application.