Mathematical and computational modelling of biodiversity | 3pm PT Tues June 4, 2024

[Grigory Bronevetsky]

Modeling Talks

Mathematical and computational modelling of biodiversity Ryan Chisholm, National University of Singapore Tues, June 4 | 3pm PT Meet | Youtube Stream Hi all, The presentation will be via Meet and all questions will be addressed there. If you cannot attend live, the event will be recorded and can be found afterward at https://sites.google.com/modelingtalks.org/entry/mathematical-and-computational-modelling-of-biodiversity 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. More information on previous and future talks: https://sites.google.com/modelingtalks.org/entry/home

[Grigory Bronevetsky]

Video Recording

Slides

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