|
|
Hokkaido University Collection of Scholarly and Academic Papers >
Theses >
博士 （情報科学） >
Information Dynamics for Complex Ecosystem Prediction and Design
|
Files in This Item:
|
|
|
Related Items in HUSCAP:
|
| Title: | Information Dynamics for Complex Ecosystem Prediction and Design |
| Other Titles: | 複雑生態系の予測・設計に関する情報ダイナミクス |
| Authors: | 李, 杰 Browse this author |
| Issue Date: | 30-Jun-2021 |
| Publisher: | Hokkaido University |
| Abstract: | An ecosystem is a complex assembly of an uncountable number of living organisms,physical components of the environment and all interrelationships in a particular unit of space. Healthy ecosystems are “balanced” systems in which interactions among components contribute to a certain stable state of ecosystems, ensuring steady requisite ecological services for living organisms. Nevertheless, ecosystems are always exposed to variable disturbances such as the fluctuations of environmental factors, alien species invasion and internal diseases and disorders, that may influence the structure and function of ecosystems and result in ecosystem degradation and biodiversity loss. Ecosystems are highly dynamical and nonlinear complex systems, making it challenging to monitor, understand and regulate adequately. In recent decades, data-driven network approaches and mathematical analyses have been increasingly used in ecosystems research thanks to their visualization, simplicity and analyzability. Complex ecosystems are therefore abstracted as a set of nodes representing individual species and environmental factors and a set of links characterizing biotic and abiotic interactions among these living and nonliving components, forming the formalism of graph with particular structure and function. Therefore, methodological frameworks in graph theory can be well exploited to investigate the dynamics and stability of ecosystems, and recognize species-specific features and collective behavior. In this research, complex network models are used to disentangle the complexity of ecosystems, and study the information dynamics and dissemination among components. Information-theoretic variables including transfer entropy, mutual information and Shannon entropy are incorporated into complex networks, formulating an integrated Optimal Information Flow (OIF) model. When inferring complex networks for ecosystems, the detection of interrelationships between components is one of the fundamental work for ecosystem modeling and graphical representation. The proposed information-theoretic OIF model quantifies these interrelationships by measuring causal interactions that can be perceived as information fluxes. The performance of OIF in inferring causal interactions is validated on a mathematically simulated predator-prey model, a real-world sardine-anchovy-temperature system and a multispecies fish community by comparing to the well-documented Convergent Cross Mapping (CCM) model. Results from the validation work demonstrate that OIF outperforms CCM since it provides a larger gradient defining causal interactions at higher resolution, smaller fluctuations, more accurate prediction for ecological indicators and no requirement for convergence. Thus, the proposed OIF can be used as a robust model to infer causal interactions and networks in ecosystems. The information-theoretic causal interactions should be considered here as nonlinear predictability of ecological information about species communities. This research also explores OIF’s applications in two real-world ecosystems:gut microbes and a marine fish community. The gut-associated microbiome is an extremely complex ecosystem considering the large number of bacteria and their interactions. In this case study, to untangle the complexity of human microbiome for the Irritable Bowel Syndrome (IBS), OIF is used to infer species interaction networks for healthy, transitory and unhealthy groups. It is observed that healthy networks are characterized by a neutral patterns of species interactions and scale-free topology versus random unhealthy networks. The top ten interacting species are the least relatively abundant for the healthy micriobiome and the most detrimental. These results are useful for public health and disease diagnosis and etiognosis, as well as the personalized design treatments and microbiome engineering. In the case study of the marine fish community, to study the biological responses of the ecosystem to global ocean warming caused by climate change, OIF with Kernel estimator is employed to infer species interaction networks for the fish ecosystem considering five temperature ranges: 10 C, 10-15 C, 15-20 C, 20-25 C, 25 C. OIF-inferred networks present different patterns in structure and function for each temperature range that indicate the evolution of system dynamics with the change of sea temperature. Network-based species-specific analysis is also performed to identify critical species that have more impacts on the fish community, and species more sensitive to the fluctuations of sea temperature. This work provides a data-driven tool for analyzing and monitoring fish ecosystems under the pressure of ocean warming and is valuable to formulate accurate fishery policy to maintain fish ecosystems stable and sustainable. |
| Conffering University: | 北海道大学 |
| Degree Report Number: | 甲第14627号 |
| Degree Level: | 博士 |
| Degree Discipline: | 情報科学 |
| Examination Committee Members: | (主査) 教授 大鐘 武雄, 教授 齊藤 晋聖, 教授 坂本 雄児, Convertino Matteo(元・本学大学院情報科学研究院准教授) |
| Degree Affiliation: | 情報科学研究科(メディアネットワーク専攻) |
| Type: | theses (doctoral) |
| URI: | http://hdl.handle.net/2115/82444 |
| Appears in Collections: | 課程博士 (Doctorate by way of Advanced Course) > 情報科学院(Graduate School of Information Science and Technology)
学位論文 (Theses) > 博士 （情報科学）
|
|
