Seagrass Dynamic Bayesian Network.

<p>A Bayesian Network (BN) is a class of graphical models representing the probabilistic relationships among a set of variables (or nodes) with directed arcs between them. These variables form a directed acyclic graph (DAG), and each node in the DAG has a state that varies depending on the states of other nodes [<a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0298853#pone.0298853.ref147" target="_blank">147</a>]. Information about those states is transmitted throughout the DAG, and as a result, inferences may be made by adding new data or evidence to the network. While BNs are suitable approaches for inferring static processes, the Dynamic Bayesian Network (DBN), a BN extension in which nodes represent variables at specific time slices, may be used to model temporally varying processes in which the state of a variable may change over time [<a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0298853#pone.0298853.ref148" target="_blank">148</a>]. When adding a dynamic component to a BN and creating a DBN, the characteristic processes of complex systems, including cumulative effects and feedback processes, can be captured. The DBN approach allows us to predict the resilience of a system given the temporal dynamics of the components of the ecosystem and their interactions with natural and anthropogenic stressors [<a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0298853#pone.0298853.ref101" target="_blank">101</a>, <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0298853#pone.0298853.ref107" target="_blank">107</a>]. The overall DBN network structure is depicted with ovals representing factors (or nodes) and arrows indicating causal parent-child relationships. In this structure, a parent node (e.g., Meadow Type) exerts an influence on a child node (e.g., Location Type). Conversely, the absence of a link between nodes signifies conditional independence. Rounded rectangles denote subnetworks [<a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0298853#pone.0298853.ref077" target="_blank">77</a>]. Note the presence of complex interdependencies and feedback loops in the seagrass ecosystem. Node colors denote different categories: site condition (white), recovery (purple), resistance (green), environmental (blue), and population (yellow). The following symbology is used in this figure: A node with a check mark: A node is ticked when inference has been successfully executed. A curved arrow back onto itself: Denotes a link to the node’s next time slice (i.e., <i>t</i> + 1). A double-headed arrow: Indicates that arcs are heading both ways between two subnetworks. For example, a double-headed arrow between a node N and a subnetwork S means that there is at least one node in S that depends on N and that there is at least one node in S that influences N. An arrow labeled with a [<a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0298853#pone.0298853.ref001" target="_blank">1</a>]: Indicates a connection to a subsequent time slice (<i>t</i> + 1).</p> <p>(PDF)</p>