BioTRaCE

Simons Collaboration on
Biodiversity Trajectories and Causal understanding of Ecoevolutionary dynamics

I live in one of the most biodiverse regions of the world. My house is a stone’s throw away from Table Mountain National Park, which is home to almost 3000 vascular plant species.

Unfortunately, it is changing.

In 2017, with members of this team, I published the first documented evidence of climate-change driven diversity loss in this region.

Since then, we’ve experienced our worst megadrought on record, multiple heatwaves, extreme heavy rainfall and flooding events, and aseasonal fires are becoming increasingly common.

This year, we’ve had our driest rainfall season on record.

This photo I took near my house a week or two ago. We’ve seen roughly 90% mortality in these endemic large shrubs across more than a third of its range.

What’s happening to the thousands of less conspicuous species we simply don’t know.

Neither do we know the implications for Nature’s Contributions to People like carbon storage, water provision or future fire regimes.

Sadly, this uncertain state and trajectory of biodiversity and associated contributions to people is not unique to my home. It’s a global crisis.

We can observe change. We can identify candidate drivers. But we struggle to reliably determine mechanisms, or predict how biodiversity will respond to change at the scale of proximal drivers or interventions.

That is the scientific problem BioTRaCE is designed to address.

Forecasting Biodiversity Trajectories

Current State

Predicting biodiversity dynamics is hampered by fragmented, inconsistent or context-dependent theory

Desired State

Coherent integrative theory allows prediction of biodiversity dynamics and iterative learning from new observations

We need to be able to rapidly detect and forecast change and its drivers in order to be agile in response to threats to nature and its contributions to people.

Unfortunately, this is rarely feasible for most ecosystems at scales at which management interventions can be applied, largely because existing theory is fragmented, inconsistent or context dependent.

-Slide-

We need coherent integrative theory that facilitates prediction of biodiversity dynamics and allows iterative learning as new observations are made.

Three Empirical Challenges

Complexity of Biodiversity

Incomplete Observation

Limited Causal Inference

We haven’t had the tools needed to test and integrate theory across the scales required for prediction of biodiversity!

Existing efforts have been hampered by three empirical challenges…

The first is conceptual: How do we characterize the complexity of biodiversity in a way that allows us to explore mechanistic processes and linkages?

• For example, the Cape Floristic Region is home to >9000 vascular plant species. We can’t hope to directly model every single one of them.

The second is data-limitation: How do we overcome the challenges associated with patchy biodiversity observations in space and time?

• GBIF visualization of species locality records
• To forecast trajectories with confidence we need good, high-frequency data

The third is methodological. How do we test hypotheses and infer causation at the scales relevant to theory and biodiversity prediction?

• Historically, our only tool has been randomised control trials. These are constrained in space and time, and even some of the largest and longest running, such as the Park Grass Experiment shown here, are only up to the size of a small farm.

In short, we simply haven’t had the tools to build a coherent body of integrative theory for understanding and forecasting biodiversity and its contributions to people.

Three Solutions

Complexity of Biodiversity

Trait-Based Ecology

Mechanistic links between scales from organism to ecosystem and Nature’s Contributions to People (NCP)

Fortunately, the past few decades and recent developments are allowing us to overcome these challenges.

Firstly, trait-based ecology characterizes biodiversity in a manner that allows us to explore mechanistic links and processes that scale from individual organisms to ecosystems and Nature’s Contributions to People, such as carbon sequestration, provision of clean water, or even human well-being.

This makes quantifying and modelling biodiversity tractable.

Three Solutions

Incomplete Observation

Satellite Remote Sensing

Provides repeated observations at scales from metres to the globe

Secondly, satellite remote sensing provides repeat observations at scales from metres to the globe, which when validated with ground observations can tell us a lot about the state and trajectory of various facets of biodiversity.

This makes biodiversity observable.

Three Solutions

Limited Causal Inference

Modern Causal Statistics

Allows us to test causation at the scale of observations

Third, new methods in modern causal statistics are freeing us from the constraints of randomised control trials and allowing us to infer causation at scales that match our observations and application.

This makes theory and the drivers of biodiversity change testable across scales.

Three Solutions

Trait-Based Ecology

Remote Sensing

Causal Statistics

Characterizable

Observable

Testable

---

Combining these solutions will be transformative, because it makes biodiversity theory testable and continuously falsifiable across spatial and temporal scales.

Combining these solutions will be transformative, because it makes biodiversity characterizable and observable, and the drivers and theory testable and continuously falsifiable across spatial and temporal scales.

• Slide -

This is our offering.

BioTraCE

• Remote Sensing of Biodiversity
• Theoretical Ecology
• Trait-based Ecology
• Statistical Ecology
• Hydroclimatology
• People and Nature
• Conservation Planning
• Science and Policy
• and more…

BioTraCE brings together a team with the skills, experience and tools required to combine these solutions to create something transformational for biodiversity theory, observation and forecasting.

We plan to start in the MTEs, where we have some of the best data for this work, and may expand if there is need or opportunity.

Combining the solutions

Mapping traits with hyperspectral remote sensing

We have a roadmap for combining the three solutions.

Firstly, PI Townsend’s lab have been leading efforts to map plant traits from imaging spectroscopy (hyperspectral remote sensing), groundtruthed with field data - combining solutions 1 and 2.

• Here we see false colour composites of 18 NEON landscapes in the US showing three traits, Phenolics, Chlorophyll and Leaf Mass per unit Area mapped to red, green and blue. They show clear variation in these traits within and between landscapes, revealing the legacies of ecological histories and land-use.

Combining the solutions

Mapping traits with hyperspectral remote sensing

We can also do this with spaceborne hyperspectral imagery, but these are currently patchy in space and time.

Several of our team members have been involved in the development of hyperspectral satellite missions that will provide repeated global coverage, such as NASA’s EAGLE-VSWIR and ESA’s CHIME, that are scheduled to be launched in 2028 - during the lifetime of this proposed collaboration.

Combining the solutions

Trait time-series from multispectral satellites

We need to be ready to harness these data. Satellite missions typically only have a planned operational lifetime of 3-7 years, so waiting for time-series would be critical time lost.

We will use field data and hyperspectral imagery from airborne and existing space-borne platforms to calibrate and develop trait time-series from multidecadal multispectral satellite records, such as Landsat and Sentinel.

This allows us to leverage the hyperspectral data the minute the satellites are operational.

• slide-

This figure shows preliminary work from PI Townsend’s lab developing an annual timeseries of traitmaps for Yosemite and surrounds in the Sierra Nevada from 2000 to 2024. We plan to extend these as far back as the early 1980s and at a higher temporal frequency, and to increase to 30 or so traits.

This provides biodiversity datasets at spatial and temporal extents and resolutions that have never been available to science before.

Not only does this present new opportunities to test existing theory, it is very likely to reveal phenomena that we’ve never seen or even thought to look for.

Combining the solutions

Causal analysis of theory and drivers of traits and NCP

The third leg of the solution is modern causal statistics, which will allow us to test the predictions of theory and the drivers of traits and NCP at the scales at which they are most relevant for observation, prediction, or societal application.

These two figures are simplistic representations of exploring biodiversity responses or effects:

• one where we intervene on drivers like fire, to explore their effects on traits (representing biodiversity responses)
• the other where we intervene on traits, to explore their effects on ecosystem functions related to Nature’s Contributions to People (representing biodiversity effects on ecosystem function)

Combining the solutions

Iterative forecasting of traits and NCP under different scenarios

Together this facilitates forecasting of traits and NCP that iteratively update as new imagery are acquired and processed.

Each update confronts our predictionc with new data, testing understanding and model specification, allowing us to learn and refine or rebuild as we go.

This allows us to leverage decades of existing data, and then refine as the higher quality hyperspectral data become available in 2028.

A framework to harmonize theory

Describing eco-evolutionary dynamics by harmonizing existing theory in a common trait-based framework

Synthesizing theory across physiology, population biology, evolutionary biology, community ecology, ecosystem ecology, and global ecology.

We’ll also apply an approach that leverages the trait time series and knowledge gained from the causal models in a Trait Driver Framework.

Trait Driver Theory borrows from quantitative genetics to recast ecological theories in the common currency of traits, linking dynamics to selection, dispersal, drift, and speciation.

This makes competing theories directly comparable on the same data - allowing us to partition the contribution of different processes - or test them in a causal framework.

BioTraCE Success

A new foundation for biodiversity science, where new global observations enable iterative testing and refining of theory to explain and forecast biodiversity change.

BioTraCE Success

A better future for biodiversity and Nature’s Contributions to People

And hopefully, a better future for biodiversity and Nature’s Contributions to People

BioTraCE: A Unified Framework for Understanding and Predicting Biodiversity Change

[Closing slide]

We’ve structured the project around 4 working groups, each with their own set of work packages, questions and responsibilities, but together they form an iterative learning cycle that allows us to learn and share across groups.

The Theory and Synthesis group involves the entire team and others and is responsible for both guiding the other groups and iteratively reviewing, synthesizing and redistributing knowledge gained within the project, while also engaging with the needs of biodiversity management and policy decision-makers.

Working group 2 will focus on developing rich biodiversity and environmental time-series from satellite and airborne imagery, field data and expert elicitation. This is the fuel for working group 3 and 4.

Working group 3 is focused on developing iterative ecological forecasting models and causal analysis to detect and attribute drivers of biodiversity change and test theory.

Working group 4 will use the data and findings from WG 2 and 3 in a Trait Driver Framework to reframe existing theory in terms of traits and test their utility for biodiversity prediction (again Cory - Help?).