The problem
Microbiome data — and many other relative-abundance datasets — are compositional: they carry only relative information and live on the simplex, where the parts sum to a constant. Analyzing them with ordinary statistics that ignore this constraint can manufacture spurious correlations and misleading conclusions.
This work builds a coherent CoDA program with two sides that reinforce each other: the methodological foundations, and the applied ecological questions about river microbiomes that motivate them.
The approach
Methodology — a literature review
A Statistical Science–style review surveys the tools of CoDA: log-ratio transformations (CLR / ILR), the geometry of the simplex, and methods for correlation, regression, and ordination that respect compositional constraints rather than fighting them.
Application — river microbiomes
On the applied side, the program analyzes bacterial community composition across riverine ecosystems — identifying environmental drivers, linking watershed land use to community structure, and surfacing microbial taxa that act as bioindicators of water quality and ecosystem health.
Martinez, Bergen & Gareis (2024) compare bacterial communities of the Yamuna (India) and Mississippi (USA) rivers, finding notably greater diversity below the Yamunotri Glacier. Read the paper ↗
Why it matters
Treating compositional data correctly changes the conclusions — and in this setting it turns microbial surveys into a practical read on ecosystem health. The bioindicator framing gives ecologists and water managers an accessible signal of how landscapes shape the life in their rivers. A follow-up paper, “Bacterial community fingerprints as indicators of watershed land use,” is currently under review.