Pollinator-Plant Cascade Extinction Model

A Mesa agent-based model simulating cascade extinctions in a bipartite pollinator-plant mutualistic network. Models how losing pollinators causes delayed plant collapse: a real ecological phenomenon linked to colony collapse and habitat loss.

Overview

Pollinators (bees, butterflies, etc.) and plants are connected in a bipartite network where edges only exist between the two types, never within. Each agent has continuous energy/health attributes that respond to the state of their network neighbors. When pollinators die, plants lose support and eventually collapse, potentially triggering further pollinator loss.

The key emergent behavior is the lag effect that shows plants survive for several steps after their pollinators die, then collapse rapidly. This mirrors real ecosystem dynamics where resilience masks underlying fragility.

Installation

git clone <your-repo-url>
cd pollinator_cascade
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Running the Model

Static analysis (3-panel chart):

python run.py

Interactive dashboard:

solara run app.py

Parameters

• n_pollinators (default: 20) — number of pollinator agents in the simulation
• n_plants (default: 30) — number of plant agents in the simulation
• connectivity (default: 0.35) — probability that any pollinator-plant pair is connected. Lower values produce sparser, more fragile networks
• extinction_schedule (default: [30, 60, 90]) — steps at which one random pollinator is forcibly removed, simulating external shocks like pesticide use or habitat loss
• rng (default: 42) — random seed for reproducibility

Model Details

Agents:

• Pollinator — gains energy from alive connected plants, loses energy each step. Specialists (high specialization) pay higher energy costs, making them more vulnerable to cascade failures.
• Plant — health tracks the fraction of alive pollinator partners. Resilient plants decline slower and survive longer into the cascade, staggering deaths across many steps.

Network: Built using networkx.bipartite.random_graph and wrapped in Mesa's Network discrete space. Lower connectivity produces sparser, more fragile networks where single pollinator losses have larger impacts.

Cascade mechanism: At each step agents check their network neighbors via cell.connections. Dead agents are removed from model.agents after each step. Cascade depth is tracked as a running counter on the model since dead agents are no longer queryable.

Output

run.py produces 3 panels:

• Survival rates of both species over time
• Cumulative plant deaths (cascade depth)
• Survival gap showing the lag between pollinator and plant loss

app.py produces an interactive dashboard with:

• Page 1: Live survival rate and cascade depth charts
• Page 2: Real-time energy and health distribution histograms

Project Structure

pollinator_cascade/
├── agents.py       # Pollinator and Plant agent classes
├── model.py        # Model, network setup, data collection
├── run.py          # Run model and generate static charts
├── app.py          # Interactive Solara dashboard
├── readme.md       # This file
├── requirements.txt
├── data/
└── images/         # Output charts saved here

References

• Bascompte, J. & Jordano, P. (2007). Plant-animal mutualistic networks. Annual Review of Ecology, Evolution, and Systematics.
• Memmott, J. et al. (2004). Tolerance of pollination networks to species extinctions. Proceedings of the Royal Society B.