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.
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