Reconstructing large interaction networks from empirical time series data

Chang, Chun‐Wei; Miki, Takeshi; Ushio, Masayuki; Ke, Po‐Ju; Lu, Hai‐Lin; Shiah, Fuh‐Kwo; Hsieh, Chih‐hao

doi:10.1111/ele.13897

Cited by 45 publications

(70 citation statements)

References 53 publications

(79 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Furthermore, our analyses based on CCM cannot access the sign of feedbacks (i.e., positive or negative), although it is known that the sign is important in determining the response of feedbacks to external perturbations (e.g., amplified or dampened). Although methods to estimate the sign of interactions were proposed (e.g., S-map 22 , 57 ), the robustness of these methods has not been thoroughly examined 58 . Lastly, due to limitations of data availability, our analysis only quantified causal strength across systems at a consensus monthly scale, acknowledging that state-space reconstruction methods (e.g., CCM) are scale-dependent 59 , e.g., one causal driver dominated monthly might not necessarily dominate at other time scales.…”

Section: Resultsmentioning

confidence: 99%

Causal networks of phytoplankton diversity and biomass are modulated by environmental context

Chang

Miki

et al. 2022

Nat Commun

Self Cite

View full text Add to dashboard Cite

Untangling causal links and feedbacks among biodiversity, ecosystem functioning, and environmental factors is challenging due to their complex and context-dependent interactions (e.g., a nutrient-dependent relationship between diversity and biomass). Consequently, studies that only consider separable, unidirectional effects can produce divergent conclusions and equivocal ecological implications. To address this complexity, we use empirical dynamic modeling to assemble causal networks for 19 natural aquatic ecosystems (N24◦~N58◦) and quantified strengths of feedbacks among phytoplankton diversity, phytoplankton biomass, and environmental factors. Through a cross-system comparison, we identify macroecological patterns; in more diverse, oligotrophic ecosystems, biodiversity effects are more important than environmental effects (nutrients and temperature) as drivers of biomass. Furthermore, feedback strengths vary with productivity. In warm, productive systems, strong nitrate-mediated feedbacks usually prevail, whereas there are strong, phosphate-mediated feedbacks in cold, less productive systems. Our findings, based on recovered feedbacks, highlight the importance of a network view in future ecosystem management.

show abstract

Section: Resultsmentioning

confidence: 99%

Causal networks of phytoplankton diversity and biomass are modulated by environmental context

Chang

Miki

et al. 2022

Nat Commun

Self Cite

View full text Add to dashboard Cite

show abstract

“…Although ESNs have been applied to causal analysis [Huang et al 2020, Duggento et al 2021, Wang et al 2022], we developed for the first time a reliable method of causal network inference by integrating adaptive online ESNs [Jaeger 2002] into an ensemble machine learning framework. In addition to its applicability to nonlinear dynamics, our approach is not affected by the reliability of nonlinear prediction methods on arbitrary underlying dynamics, which differs from previous approaches based on nonlinear time series analysis [Sugihara et al 2012, Suzuki et al 2017, Ushio et al 2018, Cenci et al 2019, Chang et al 2021]. As we have confirmed (Fig.…”

Section: Discussionsupporting

confidence: 50%

Decomposing predictability to identify dominant causal drivers in complex ecosystems

Suzuki

Matsuzaki

Masuya

2022

Preprint

View full text Add to dashboard Cite

Ecosystems are complex systems of various physical, biological, and chemical processes. Since ecosystem datasets are composed of a mixture of different levels of stochasticity and nonlinearity, handling these data is a challenge for existing methods of time-series based causal inferences. Here we show that, by harnessing contemporary machine learning approaches, the concept of Granger causality can be effectively extended to the analysis of complex ecosystem time series and bridge the gap between dynamical and statistical approaches. The central idea is to use an ensemble of fast and highly predictive artificial neural networks to select a minimal set of variables that maximizes the prediction of a given variable. It enables decomposition of the relationship among variables through quantifying the contribution of an individual variable to the overall predictive performance. We show how our approach, EcohNet, can improve interaction network inference for a mesocosm experiment and simulated ecosystems. The application of the method to a long-term lake monitoring dataset yielded new but interpretable results on the drivers causing cyanobacteria blooms, which is a serious threat to ecological integrity and ecosystem services. Since performance of EcohNet is enhanced by its predictive capabilities, it also provides an optimized forecasting of overall components in complex ecosystems. EcohNet could be used to analyze complex and hybrid multivariate time series in many scientific areas not limited to ecosystems.

show abstract

“…the number of data points is more than or roughly equal to the square of the dimensions of reconstructed state space, which is required for robust estimations of the S-map coefficients). Note that, in the case that the number of dimensions far exceeds the number of data points, a recently proposed S-map method would be more suitable [ 25 ].…”

Section: Resultsmentioning

confidence: 99%

Interaction capacity as a potential driver of community diversity

Ushio

2022

Proc. R. Soc. B.

Self Cite

View full text Add to dashboard Cite

How patterns in community diversity emerge is a long-standing question in ecology. Studies suggested that community diversity and interspecific interactions are interdependent. However, evidence from high-diversity ecological communities is lacking because of practical challenges in characterizing speciose communities and their interactions. Here, I analysed time-varying interaction networks that were reconstructed using 1197 species, DNA-based ecological time series taken from experimental rice plots and empirical dynamic modelling, and introduced ‘interaction capacity', namely, the sum of interaction strength that a single species gives and receives, as a potential driver of community diversity. As community diversity increases, the number of interactions increases exponentially but the mean interaction capacity of a community becomes saturated, weakening interspecific interactions. These patterns are modelled with simple mathematical equations, based on which I propose the ‘interaction capacity hypothesis': that interaction capacity and network connectance can be two fundamental properties that influence community diversity. Furthermore, I show that total DNA abundance and temperature influence interaction capacity and connectance nonlinearly, explaining a large proportion of diversity patterns observed in various systems. The interaction capacity hypothesis enables mechanistic explanations of community diversity. Therefore, analysing ecological community data from the viewpoint of interaction capacity would provide new insight into community diversity.

show abstract

Reconstructing large interaction networks from empirical time series data

Cited by 45 publications

References 53 publications

Causal networks of phytoplankton diversity and biomass are modulated by environmental context

Causal networks of phytoplankton diversity and biomass are modulated by environmental context

Decomposing predictability to identify dominant causal drivers in complex ecosystems

Interaction capacity as a potential driver of community diversity

Contact Info

Product

Resources

About