Agri-inputs & biocontrol
Soil microbiome: what is actually measured
What sits behind a soil microbiome test: spatial variability, the compositional nature of sequencing, relic DNA, the gap between taxonomy and function — and why an introduced strain rarely establishes.
Soil microbiome diagnostics rest on one measurement operation: total DNA is extracted from a few grams of soil (direct extraction is standardised as ISO 11063), a region of the 16S rRNA gene is amplified for bacteria and archaea or ITS for fungi, the amplicons are sequenced, and reads are matched against a reference database. Everything afterwards called “the state of the microbiome” is an interpretation of that table. It is worth being clear about what the table can and cannot say.
Four limits inside the measurement itself
Space. A gram of soil holds thousands of taxa, and variability within a single field often exceeds the difference between treatments. Rhizosphere and bulk soil differ by orders of magnitude. A composite of many cores is a spatial average, and repeatability is set by the sampling scheme and depth rather than by the laboratory.
Proportions, not quantities. Sequencing returns composition: the share of reads assigned to a taxon. A rise in one organism’s share can mean it multiplied or that others declined. Absolute abundance needs a separate measurement — quantitative PCR, or a spiked internal standard.
Relic DNA. DNA from dead cells adsorbs to clay minerals and organic matter and persists, so an appreciable fraction of extracted DNA does not belong to the living community. Detecting a sequence is not the same as an active population being present; that requires RNA or an activity assay.
Amplification bias. The 16S rRNA operon is present in roughly one to fifteen copies depending on the genome, primers cover groups unevenly, and ITS length varies between fungi. Read share is therefore not proportional to cell share — and a large fraction of soil fungi has no confident match in reference databases at all.
Taxonomy is not function
Even a flawless taxon list does not answer the agronomic question. The 16S rRNA gene resolves to genus at best, while nitrogen fixation, phosphate solubilisation and antagonism towards pathogens differ between strains of one species. Shotgun metagenomics shows that a gene is present, not that it is expressed. Composite “soil health scores” are therefore built on models trained on a provider’s own dataset, and results from different laboratories are not comparable: primers, clustering thresholds and pipelines all differ.
Why prescribing a microbe is easier than installing one
The intervention side has the mirror problem. An introduced strain arrives in a medium already holding on the order of a billion cells per gram, with carbon and niches allocated. Inoculum abundance typically falls by orders of magnitude within weeks, and the effect lasts as long as the population does. Disease-suppressive soil is a property of a community and of the supply of organic carbon rather than of any single strain, which is why rotation, residue and cover crops shift it more reproducibly than a single inoculation. The measurement earns its place mainly as diagnosis — a quantitative PCR for a specific soilborne pathogen before planting supports a decision that a survey-level community profile does not.