Using multivariate cross correlations, Granger causality and graphical models to quantify spatiotemporal synchronization and causality between pest populations

摘要: This work combines multivariate time series analysis and graph theory to detect synchronization causality among certain ecological variables represent significant correlations via network projections. Four different statistical tools (cross-correlations, partial cross-correlations, Granger causality) utilized quantify correlation strength biological entities. These indices correspond ways estimate the relationships between construct networks using as nodes edges. Specifically, introduce rules that define associations (links) (nodes). approach is used for first analyze of moth populations well temperature relative humidity in order spatiotemporal over an agricultural study area illustrate interactions graphical models. The resulting from approaches are trimmed show how configurations affected by each construction technique. provide a simple test determine whether one (population) caused another (i.e. environmental variable or other population) even when they not correlated. In most cases, related models, revealed intra-specific links, fact may be linked similarities pest population life cycles synchronizations. Graph theoretic landscape projections reveal subject characteristics. Populations great distances through physical features such rivers only at adjacent locations which more likely appear. some incidental connections, with no explanation, were also observed; however, this was expected because methods non trivial connections cannot interpreted phenomenologically. Incorporating causal probabilistic sense comes closer reality than doing per se binary constructs former conceptually incorporate dynamics all kinds within network. advantage have dynamic easy way examine relations multiple time-series variables. constructed intuitive, advantageous representation populations’ can realized agro-ecosystem. due cycle synchronizations, exposure shared climate complicated moving behavior, dispersal patterns host allocation. Moreover, useful drawing inferences regarding their spatial management. Extending these models including should allow exploration intra interspecies larger systems, identification specific traits might constrain structures areas.