Abstract:
Biodiversity is under threat across our planet from climate change, deforestation, and overexploitation. My lab focuses on mathematical and computational approaches to understanding biodiversity. In this talk, I will first discuss a recent project applying modelling tools to historical biodiversity data in Singapore to estimate how many species have gone extinct in the highly developed tropical island nation over the last 200 years. I will then present work exploring the fundamental question of what mechanisms maintain biodiversity in nature using a combination of mathematical modelling, observational data, and experiments. Lastly, I will discuss potential applications of high-performance computational mechanistic biodiversity models to open-world games and virtual reality.
Bio:
Associate Professor Ryan Chisholm is a theoretical ecologist at the National University of Singapore. He completed a BSc in Mathematics and Statistics and a BA in German at the University of Melbourne, a PhD in Ecology and Evolutionary Biology at Princeton, and a post-doctoral fellowship at the Smithsonian Tropical Research Institute in Panama. His primary research interests are tropical biodiversity and extinctions. He uses a range of mathematical, computational and statistical tools to explore these topics.
Summary:
Extinctions in species in Singapore
Many species have gone extinct over the past 200 years (e.g. tigers in 1930s)
Study used historical records (e.g. photos of species over the centuries)
3k species
50k records
Approach: MODGEE (matrix of detection gives detection estimates)
Given matrix of whether species i was observed in time period t, detectability of a species and the amount of effort put into each species’ detection
Estimates the likelihood that each species has gone extinct
~37% of species have gone extinct over the past 200 years
Estimating that ~20% of species will go extinct over Southeast asia over the coming century
What mechanisms structure biodiversity patterns?
Why do some places have more/fewer species?
Hypotheses
Niches maintain diversity?
E.g. polar bears and beavers have different environments and thus don’t compete
Lotka-Volterra competition model: infer effective interference between pairs of species
Hard to operationalize (e.g. estimate size of niches)
Works well mostly in species-poor environments with few interactions
Dispersal maintain diversity?
Species move/are moved from place to place
Start competing with species that already exist in new environment
Neutral models:
Neutral: all species treated the same, so natural selection and species diversity is not modeled
Grid of cells with individuals from different species
Each time step individual dies randomly, replaced with descendant of another individual
Either nearby or arbitrary parent, depending on details of the physical dispersal process
More generally, differential equation that evolves the probability distribution of different species
Pros:
Easy to simulate
Analytically tractable
Extensible to new dynamics
Have fit the stochastic spatial model to real data to get good accuracy
Works better for steady state prediction, more poorly for timing
What maintains diversity?
Niches maintain diversity?
Evidence from observations and experiments in species-poor systems
Dispersal maintains diversity?
Evidence from fits of neutral models and related models to broad statistical patterns
Hybrid niche-neutral model
Create a set of niches, assign each species a niche
Only allow intra-niche dispersal
Model reproduces the real observation that there’s a lower bound on the number of species in a real ecosystem
Experimental diversity data from island archipelago:
Diversity scales log-linearly with area of island
But there’s a lower bound for small islands
Weaknesses:
Data is observational
Island area is a confounder
How to separate immigration/dispersal from island area?
Experiment:
Boxes on seawalls that allow crustaceans to come inside and live inside the box
Each box is an “island”
Boxes have holes of different size to control immigration
Most communities have few niches and beyond these dispersal dominates
Biodiversity models and gaming
MicroVerse for Unity: https://assetstore.unity.com/packages/tools/terrain/microverse-core-collection-232976
No Man’s Sky generates ecosystems procedurally
Real patterns:
Spatial clustering
Niche associations
Spatial turnover
Most species are common, most are rate
Congruence with geographical barriers (e.g. species on opposite sides of Panama are related but 3.5m years of apart, which is the age of that land separation)
Diversity gradients: latitudinal / altitudinal
Games only succeed in the first 2 dimensions
Want to add realistic ecologies in games
Technical requirements
Arbitrarily large: on-demand runtime generation (must be fast)
Self-consistency
Arbitrary landscape geometry (e.g. flat, spherical)
Can leverage individual-based biodiversity models for this task
To improve performance can run models backwards in time: Coalescence Algorithms
Given sample area, we don’t know the species at each spot
Run it backwards to see where each plot’s individual came from
When the history of different plots merges, they coalesce: must have same species, can model them together
Has been used to simulate distribution of species in different landscapes with different immigration patterns, even inter-planetary
Possible extensions:
Non-equilibrium dynamics (landscape/immigration changes over time)
Multi-gene
Applications:
Large-scale open-world games
Design aid for games
Educational software: biodiversity discovery, bio-geographic puzzle solving (infer geographical/ecological history from evidence), effect of climate on biodiversity