Code from: Spatial resource heterogeneity stabilizes local and regional predator-prey dynamics in ecologically-realistic networks

Published Aug 05, 2025 on Dryad. https://doi.org/10.5061/dryad.c59zw3rgn

Data files

Aug 05, 2025 version files 1.58 MB

all.data.for.supplement.csv

1.42 MB
final.df.csv

158.06 KB
README.md

4.01 KB

Abstract

Spatial resource heterogeneity (SRH; the variable spatial distribution of resources) is a surprisingly understudied component of oscillatory consumer-resource dynamics. SRH may be particularly important in large, ecologically-realistic networks where different patterns of resource distribution can manifest, which has important implications for spatial synchrony. Here we explore how SRH in large spatial networks influences both local and regional consumer-resource stability. To do so, we employ a spatially-explicit Rosenzweig-MacArthur model and vary resource distribution accordingly: homogeneously-distributed resources of low, medium and high productivity, and heterogeneously-distributed resources. The latter includes networks with SRH of random variability in productivity (“random networks”) or a spatial productivity gradient (“gradient networks”). We analyze the effects of local patch factors (i.e., productivity and connectivity) and regional factors (i.e., productivity distribution and structure). First, we find that SRH, regardless of productivity distribution type, stabilizes regional dynamics via statistical stabilization of asynchronous oscillations and local dynamics by reducing the amplitude of oscillations and bounding them further from zero. Second, the local-level stabilizing effect increases with both productivity and connectivity. Lower productivity patches are subsequently destabilized in return, albeit minimally. Lastly, random variability in productivity provides the greatest local-level stabilization effect, because high productivity patches are often (highly) connected to lower ones in a way not possible in gradient networks. We conclude that SRH is a particularly strong driver of consumer-resource stability in that it provides local-level stability in a way that other forms of heterogeneity do not. To promote consumer-resource stability, management decisions should prioritize 1) variable resource distribution patterns in large spatial networks, and 2) high connectivity between patches of different productivity levels.

Dataset DOI: 10.5061/dryad.c59zw3rgn

Description of the data and file structure

We simulated predator-prey dynamics in metacommunities using a spatially explicit version of the well-known Rosenzweig-MacArthur model (Rosenzweig and MacArthur, 1963). We have submitted Mathematica code (“SRM_Manuscript_2024.nb”) for model creation and numerical simulation, and two output datasets, one containing the results for the main manuscript (final.df.csv”) and one containing the results for the supplementary information (“all.data.for.supplement.csv”).

Files and variables

Description of code and file structure

The Mathematica code (“SRM_Manuscript_2024.nb”) simulates the spatial predator-prey dynamics and produces output for predator-prey stability metrics (amplitude and minimum of oscillations) as a function of network structure (dendritic vs. ring lattice), network size (7, 15, 31, and 63 patches), and spatial resource heterogeneity (homogeneous low productivity, homogeneous medium productivity, homogeneous high productivity, random heterogeneous distribution, gradient heterogeneous distribution). The first section (“Initialize Program”) defines parameters, creates empty tables, and defines the type of network structure to simulate. The second section (“Start Loop”) includes analysis for the single patches and the networks. The post-processing stores the data, along with visualizing it. Specifics are described in-code through comments.

Description of data and file structure

There are two datasets we created from running the model under various scenarios. “final.df.csv” is the resulting table from running the Mathematica code for the main manuscript findings, which focuses on 63-patch dendritic (tree) network structures across productivity level (low, medium, high) and productivity distribution (homogeneous, heterogeneous [random, gradient]). “all.data.for.supplement.csv” involves the additional results presented in the supplement, which includes networks of all sizes (7, 15, 31, 63 patches), structures (dendritic, ring lattice), as well as productivity level and productivity distribution patterns. Each data file has the same variable column headings, with variable descriptions and differences in the variable levels between the datasets described below.

Run.Number: Number of simulations (50). Consistent between each data file.
structure: Network structure, including tree (dendritic) or ring (ring lattice) in all.data.for.supplement.csv. Includes only tree in final.df.csv.
patch: Number of patches in a network. In final.df.csv, includes only 63. In all.data.for.supplement.csv, includes 7, 15, 31, and 63.
K: Refers to the distribution of productivity or carrying capacity. Both datasets include constant, random, and slope. Constant pertains to a homogeneous productivity distribution. Random pertains to a random productivity distribution. Slope pertains to a slope-like or gradient distribution.
productivity: Refers to the average productivity level. Both datasets include low, mid (medium) and high.
treatment2: For figure purposes. Combines K, productivity, and network vs. single patch.
variable: Describes variable type, whether single patch or network, patch or regional scale results, and minimum or amplitude metric.
value: the resulting minimum or amplitude value.
prod.dist: For figure purposes. Refers to productivity distribution, whether constant, random, or slope and with an average productivity level of mid (medium), low, or high.
log_value: the log of “value”.

Code/software

Mathematica version 14.1.0.0 was used to run the script and generate the datasets. Annotations are provided throughout the script for 1) parameter definitions, 2) model creation, 3) processing and analysis, and 4) figure creation.