Forecasting species and landscape change with ecological individual-based models | 9am PT Feb 13

[Grigory Bronevetsky]

Modeling Talks

Forecasting species and landscape change with ecological individual-based models Julie Heinrichs, Colorado State University Tuesday, Feb 13 | 9am 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/forecasting-species-and-landscape-change-with-ecological-individual-based-m Abstract: Many landscapes are undergoing ecological transformations, with potentially surprising implications for species at risk of decline. Spatially explicit individual-based models (IBMs) are increasingly used to understand complex ecological systems and anticipate future change. They are well-suited to integrate disparate datasets, represent mechanisms that can be robust to change, and quantify uncertainty. Ecological individual-based models are representing nuanced relationships between environments and species responses and yielding ecologically-nuanced and actionable results. This talk will outline the data integration, model mechanics, and emergent insights provided by spatial IBMs, using examples constructed in the HexSim modeling platform. The presentation will emphasize progress in using IBMs to replicate patterns in empirical wildlife data and forecast population dynamics through space and time. It will provide examples of scenario-driven experimentation to assess the value of information, implications of (in)action, and the ecological return on conservation investments. The presentation will describe how the co-development of ecological individual-based models can support decision-makers in planning current and future landscapes that include wildlife. Bio: Julie Heinrichs is a landscape ecologist and conservation biologist at Colorado State University, Colorado, and the Director and Chief Scientist at Computational Ecology Group Inc. in Canmore, Canada. Dr. Heinrichs develops landscape-populations models to simulate the implications of ecological change for species and landscapes at risk of decline. She builds spatially explicit individual-based models that emulate complex environments and develops scenario-based forecasts of species abundance and distribution. She co-builds models with partners to help them understand the impacts and benefits of environmental change and develop spatial strategies to manage and conserve sensitive species and places. More information on previous and future talks: https://sites.google.com/modelingtalks.org/entry/home

[Grigory Bronevetsky]

Video Recording: https://youtu.be/YSwibbgJ5wg

Slides: https://drive.google.com/file/d/1yPNhYQlDRhnr-5ABwP50BhHW5M9t6txz/view?usp=sharing

Summary:

• Focus: complex ecological relationships in wildlife systems (environment, animals plants)

  • Climate change is pushing these systems into new behaviors

  • Change varies significantly across space and time

  • E.g. pocket gopher’s distribution has been changing significantly

• Cascading changes

  • Individual system components pushing each other

  • Climate -> vegetation -> animal habitat -> populations …

• Challenge: multiple interacting factors/pathways

  • Large-scale

  • Long-term

  • Decision paralysis by many stakeholders

  • Compounding complexity

• Ecological simulation modeling

  • Focus: individual-based models (mechanistic with statistical and spatial elements)

  • Environment <-> behavior <-> demography

  • E.g., Fire -> habitat change (fire breaks) -> behavior change (animals can’t cross break) -> species population change (changes in spatial distribution, higher mortality)

  • Individual agents 

    • Differ from each other via history, traits, etc.

    • Individual decisions

    • Emergent behavior at population level

    • Spatially explicit

  • Configuration

    • General species information (optional): distribution, abundance, demography, life history data, movement data (e.g., summarized from GPS tracklog)

• HexSim modeling platform: https://www.hexsim.net

  • Workflow

    • Define problem

    • Collect resources

    • Conceptual model

    • Build IBM

    • Verification

    • Build scenarios

    • Forecast scenarios

  • Data-finding: data sets, literature, expert interviews

  • Verify against real observations (as available)

  • Process: <1week for simple models to years for complex questions with many stakeholders

  • Scenario run times: few minutes of compute time for simple dynamics to days for many agents with complex dynamics, intensive analyses, and large landscapes

  • Variables: individual traits, maps (interactions), distributions, dynamical equation parameters

  • Inherent stochasticity in individual runs (optional); aggregate many runs into a probability distribution

• Applications: Characterize, assess, plan ecological systems

• Application 1: Modeling historical occupancy of red tree vole

  • Motivation: decline of old growth forests via land use change and wildfire

  • Many species depend on old trees

  • Red tree voles live high in trees, travel between them in canopies, limited dispersal (<200m)

  • Hard to find, track (must climb trees)

  • Habitat change model captures forest structure, composition, climate; 1096 vole nest locations (maximum entropy model). Applied to landsat to create 36-year time series of habitat maps, disturbed by fire and timber harvest

  • Dynamics: Life history, population, animal movement, life cycle

  • Model shows changes in occupancy over time, and predicts reductions in occupancy area

  • Observations:

    • Fire strongly affects their spatial footprint (where voles could occupy habitat)

    • Timber extraction in patches of forest creates disconnections among populations in different forest patches

  • Can try to mitigate damage by creating stable refuges of old growth forest

  • Publication: Red tree voles 

• Application 2: simulating connectivity for Humboldt marten

  • Population isolation is a threat to biodiversity, long-term species viability

  • Mobile species (15-130km)

  • Cannot cross fenced highways, wide rivers; crossing some roads, rivers can pose mortality risk

  • What landscape features limit connectivity and could be targets for study and mitigation?

  • Animal movement modeled based on observed data

  • Runs: ~700 martens, 100 years, 50 replicates

  • Finding: A limited number of roads are constraining modeled connectivity. These features could be the target of movement data collection; sites evaluated for mitigation opportunities.

• Application 3: greater sage-grouse

  • Declining population across much of North America

  • Species use large areas that vary across seasons

  • Change in vegetation due to climate change and oil/gas development is a threat

  • Goal: quantify species’ ability to withstand change and identify the ability of populations to withstand major threats in next 50 years

  • Evaluate different economic intensity scenarios (E.g. model new oil/gas wells; update by removing vegetation on site and re-applying spatial-statistical model)

  • Complex wildlife behavior model (return to old spots, avoidance of infrastructure)

  • Model predicts 

    • Population declines around areas with high development (more decline where more development); less decline in areas planned for sage-grouse conservation.

    • Risk of spatial disconnection among populations

    • Species is potentially vulnerable over the coming 50 years

  • Publication: Sage-grouse

• Ecological IBM opportunities

  • Increased use of IBMs

  • More integration with other modeling tools, ML, data platforms

  • New Tools: calibration, automation of some tasks, processes, tech transfer to non-modelers