Are trait-growth models transferable? Predicting multi-species growth trajectories between ecosystems using plant...

Area of Science:

• Ecology
• Plant Biology
• Trait-based Ecology

Background:

• Plant functional traits are increasingly used to generalize ecological patterns across species.
• However, evaluating the predictive power of trait-based models in novel species or environments is limited.
• Understanding trait-environment interactions is crucial for ecological forecasting.

Purpose of the Study:

• To assess the predictability of plant growth trajectories using functional traits across different ecosystems.
• To investigate the influence of specific leaf area (SLA), woody density, seed size, and leaf nitrogen on plant growth.
• To test out-of-sample prediction capabilities of trait-based growth models between distinct environmental settings.

Main Methods:

• Utilized three independent datasets from south-eastern Australia, including non-resprouting species height data across time-since-fire chronosequences.
• Quantified relationships between functional traits (SLA, woody density, seed size, leaf nitrogen) and plant growth metrics (maximum relative growth rate, age at maximum growth, asymptotic height).
• Performed out-of-sample predictions of growth trajectories between ecosystems to evaluate model transferability.

Main Results:

• Strong trait-growth relationships were observed in one dataset: low SLA correlated with greater asymptotic height, high leaf nitrogen with faster growth rates, and low seed mass with earlier peak growth.
• These specific growth-trait relationships did not hold consistently across the other two datasets.
• Accurate out-of-sample prediction of growth trajectories between ecosystems was unachievable.

Conclusions:

• Plant growth trajectories may fundamentally differ between ecosystems.
• Trait-growth relationships are not universally transferable and can change across environmental gradients.
• The predictive power of functional traits for plant growth is context-dependent, limiting broad generalizations.