Home

Background for project:

Our goal is to test for distance decay in structure between insular communities using spatial modularity and disentangle the factors that generate this pattern. In multilayer networks, a module can span several local communities. Therefore, distance decay in structure is a pattern in which the farther apart two local communities are, the less they share modules. We use data from local plant-pollinator networks in the Canary Islands, which we represent as a spatial multilayer network. This data set is particularly suitable because a previous study suggests signs of distance decay in species and interactions between islands in this system. Distance decay in species composition should strongly affect spatial modularity because if two islands are remote and do not share species, they are unlikely to share modules. We hypothesized that species turnover ($H_1$) and interaction rewiring across islands ($H_2$) drive distance decay in structure at the regional scale because both are processes that contribute to shape community structure. Finally, we further hypothesized that factors occurring within each island (i.e., local factors) could affect structure distance decay ($H_3$) because they influence processes that favor modular structure, such as resource partitioning and coevolution. To test these alternative, yet not mutually exclusive, hypotheses we developed a set of three null models, which alter different components of the spatial multilayer network. These models allowed us to disentangle local vs. regional drivers in spatial modularity.

Folder breakdown:

• R- All used code.
• csvs_nuevo- Every notable output saved as a csv.
• graphs- Every graph created as a part of this project.
• Extra_analysis- Previous analysis not used in the main manuscript.

R (utility):

functions

This portion of the code contains functions used in more than one analysis.

• The functions are organized by their main goal.
• Before every function there is an explanation of the intention of the function or what it calculates.

R (pipeline):

Empirical

This code turns the data into a multilayer network with 5 layers (islands as layers) and tests distance decay in module composition of the empirical network.

• Gets an interactions csv organized by Noa Fridman based on the interactions recorded and documented in the original research. Also gets a csv of the distance between islands (layers).
• Creates interlayer and intralayer edges data and normalizes the weights between 0 and 1.
• Interlayers and intralayers edges, nodes, and layers are combined into a multilayer network.
• Modularity analysis is conducted. It also contains basic exploratory analysis regarding modules such as the number of shared modules between islands.
• Tests distance decay in module composition in the empirical network.

---

Null_M1

This portion of the code runs the null model M₁, in which we shuffled species between islands (layers) in one of 3 versions:

1. shuffling plants among themselves
2. shuffling pollinators among themselves
3. shuffling plants among themselves and then pollinators among themselves

The shuffled networks are then compared to the empirical network to determine whether species turnover is influencing the pattern found.

---

Null_M2

This portion of the code runs the null model M₂, in which we shuffled interactions between islands (layers).

• We randomly shuffled interactions of each pair of species between all the layers in which they co-occur. For example, if a plant and a pollinator co-occurred in layers 1 and 3 but interacted only in layer 1, they would still co-occur in the same layers but may interact in layer 3 after shuffling.

The shuffled networks are then compared to the empirical network to determine whether interaction rewiring is influencing the pattern found.

---

Null_M3

This portion of the code runs the null model M₃, in which we shuffled interactions within islands (layers).

The shuffled networks are then compared to the empirical network to determine whether local factors influence the pattern found.