Data leakage and measurement error inflate the apparent predictability of overyielding from plant traits
Kopp, E. B.; Koenig, N.; Vonmetz, L.; Wuest, S. E.; Niklaus, P. A.
Show abstract
Predicting plant mixture overyielding from functional traits is central to understanding how biodiversity influences ecosystem functioning. Combining empirical data from a large mixture experiment (764 mixtures) and two trait-measurement experiments containing 90 soybean genotypes, with complementary simulations where trait-function relationships and noise levels were defined a priori, we show that apparent model performance depends critically on how predictive models are validated. When training and testing data share genotypes, predictive ability is strongly inflated because shared monoculture and trait measurements create data leakage. Measurement errors further propagate through these shared components, inducing spurious correlations and amplifying noise. When validation uses completely independent genotypes, the predictive power of both linear and machine-learning models declines sharply, revealing limited but genuine predictability. These results show how data structure and measurement error can produce misleading model performance and underscore the need for rigorous validation to achieve robust ecological prediction.
Matching journals
The top 8 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- Plant defense synergies and antagonisms affect performance of specialist herbivores of common milkweed 94%
- Plant neighbors differentially alter a focal species' biotic interactions through changes to resource allocation 93%
- Imaging spectroscopy reveals topographic variability effects on grassland functional traits and drought responses 92%
Similar papers in this journal
- Direct and indirect fitness effects of plant metabolites and genetic constraints limit evolution of allelopathy in an invading plant 93%
- Phenological mismatch between trees and wildflowers: Reconciling divergent findings in two recent analyses 93%
- Earlier snowmelt increases the strength of the carbon sink in montane meadows unequally across the growing season 92%
Similar papers in this journal
- pnetr: An R package for the PnET family of forest ecosystem models 93%
- Forest and Biodiversity 2: a tree diversity experiment to understand the consequences of multiple dimensions of diversity and composition for long-term ecosystem function and resilience 92%
- Model-based ordination for species with unequal niche widths 92%
Similar papers in this journal
- Capturing spatiotemporal variation in salt marsh belowground biomass, a key resilience metric, through geoinformatics 92%
- Change in liana density over 30 years in a Bornean rain forest supports the escape hypothesis 91%
- In Silico Evaluation of Plant Nitrification Suppression Effects on Agroecosystem Nitrogen Loss 91%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.