Predictability and transferability of local biodiversity environment relationships

Area of Science:

• Ecology
• Remote Sensing
• Biodiversity Science

Background:

• Biodiversity varies spatially and temporally, often linked to environmental heterogeneity.
• Local biodiversity measures (e.g., species richness) are key indicators.
• Remote sensing data now allows global characterization of environmental heterogeneity.

Purpose of the Study:

• To quantitatively evaluate the accuracy of predicting local biodiversity measures using environmental heterogeneity.
• To assess the predictability within and transferability across biodiversity surveys.

Main Methods:

• Combined terrestrial local biodiversity survey data with global remote sensing data on environmental heterogeneity.
• Employed a cross-validation framework to assess prediction accuracy.

Main Results:

• Prediction errors for local biodiversity measures can be substantial.
• Error magnitudes vary by biodiversity measure, taxonomic group, sampling technique, and environmental characterization.
• While within-survey predictability may be low-error, across-survey transferability is questionable.

Conclusions:

• Current methods may overestimate our ability to accurately predict and project local biodiversity.
• Future biodiversity predictions must be rigorously evaluated for accuracy and uncertainty.
• Ecological theories should be tested against the actual predictive power derived from local biodiversity data.