The problem centers on a question of sampling. When oil flows from a reservoir to a platform, it arrives as a multiphase mixture of crude oil, natural gas, formation water, and suspended solids. Onboard, the aqueous portion is separated off as produced water, and this water has long served as the convenient proxy for the reservoir microbiome. The logic seemed sound: subsurface reservoirs are anoxic, and microbes are thought to dwell in oil-water transition zones or in tiny water droplets suspended within the oil, so some reservoir organisms should end up in the produced stream. Extracting DNA from water is also vastly easier than wrestling it out of sticky, flammable crude oil, which requires non-polar solvents and surfactants and triggers strict safety rules for offshore handling and helicopter transport. But whether produced water faithfully mirrors the community living in the reservoir had, remarkably, never been directly tested with paired samples from the same wells.

That gap is precisely what a research team set out to close. Working with material provided by the Norwegian operator Aker BP, they collected paired reservoir fluid and produced water samples directly at the wellheads of production wells across four North Sea oil fields: Alvheim, Vilje, Volund, and Bøyla, all located 200 to 300 kilometers offshore of Stavanger and tied back to the Alvheim floating production vessel. Sampling used standard eight-liter miniseparators connected straight to the production wells, flushed for roughly two minutes at 20 bar to eliminate carryover contamination, then drained and refilled before the final sample was taken. In-situ reservoir temperatures ranged from 63 to 66 degrees Celsius, and critically, none of the fields had received seawater injection, so the main external water source was the underlying aquifer used for pressure maintenance, along with produced water re-injection at some sites.

The analytical workflow was exhaustive. Samples were sealed into sterile glass bottles with PTFE-lined caps and no headspace to preserve anoxic conditions, and all subdivision was performed in an anoxic glove box. Reservoir fluid samples, which contained between 0.1 and 10 percent aqueous phase, were mixed with phosphate-buffered saline containing Tween 80, incubated overnight at 55 degrees, and filtered, with added sodium dodecyl sulphate boosting DNA yields from the oil-rich matrix. Both 16S rRNA gene amplicon sequencing on an Illumina MiSeq and deep shotgun metagenomic sequencing on a NovaSeq 6000 were applied, the latter enabling reconstruction of metagenome-assembled genomes, profiling of the ribosomal protein S3 marker gene, and quantification of metabolic marker genes. Statistical rigor came through PERMANOVA models, indicator species analysis, and compositional differential-abundance testing with careful corrections for multiple comparisons.

The results were striking. Reservoir fluids harbored the thermophilic, halophilic, and anaerobic taxa expected of a hot deep subsurface environment, including genera such as Caminicella, Petrotoga, Flexistipes, Halanaerobium, and Thermovirga. Produced water told a completely different story, dominated by marine-associated, aerobic genera like Marinobacter and Malaciobacter, organisms typical of seawater and aquifers. In one reservoir, Marinobacter pelagius reached up to 98 percent of the inferred relative abundance in produced water, while Caminicella accounted for 86 percent of the reservoir fluid from the same field but a mere 2 percent of its produced water. The genetic overlap was astonishingly small: only 41 percent of genera and just 21 percent of amplicon sequence variants detected in reservoir fluids also appeared in the paired produced water, which additionally carried 67 genera and 928 sequence variants found nowhere in the reservoir fluids.

The statistical architecture of the study made clear that this was not random variation. When reservoir identity and sample type were modeled together, sample type explained more of the community variation than the reservoir itself, a remarkable result given that the four fields are geologically distinct. The separation held on both abundance-based and presence-absence data, meaning it reflects genuine differences in community membership, not merely in how many cells of each type were present. Reservoir fluid samples were also significantly more dispersed and more diverse, with Shannon diversity indices of 1.9 versus 0.9 for produced water, while produced water samples from different fields clustered tightly together, consistent with their shared imprint of aquifer-derived water. Notably, 25 of 28 indicator species flagged by the analysis pointed to reservoir fluid, and every single indicator for produced water belonged to one genus, Marinobacter.

The functional dimension reinforced the taxonomic one. Across all seven reservoir-relevant metabolic pathways tested, the abundance of curated metabolic marker genes differed significantly between sample types, with sample type explaining 67 percent of the variation in functional composition compared to a non-significant contribution from reservoir identity. Metagenome-assembled genomes added strain-level resolution: strains of anaerobic halophiles and thermophiles were detected in both sample types but were on average between roughly 10 and 18 times more abundant in reservoir fluids, while 16 strains, primarily from the phyla Bacillota_A, Deferribacterota, and Synergistota, were found exclusively in reservoir fluids. Interestingly, genome quality was higher for produced water assemblies, likely because the water matrix interferes less with DNA extraction than crude oil does, a technical reminder that easier processing does not equal better biology.

The ecological explanation is elegantly simple. Produced water represents the bulk aqueous phase of the reservoir, and that water mixes readily with aquifer water drawn in to maintain pressure, picking up external microorganisms along the way. The oil phase, being immiscible, is buffered from this exchange, so the cells associated with it, dwelling in transition zones, biofilms on suspended particles, and water droplets within the oil, retain the reservoir signal. When the reservoir fluid reaches the platform, those oil-associated populations ride along concentrated in the hydrocarbon, while the bulk produced water dilutes them and swamps them with aqueous outsiders. This also explains why the findings contradict an earlier study that reported higher diversity in produced water: bulk formation water vastly outweighs the small volumes of droplet and transition-zone water where the native community thrives.

Perhaps most remarkable is what survived. These fields have been producing for years, some since 2008, and Bøyla only since 2015, yet the reservoir fluids still yielded the thermophilic, halophilic, and anaerobic taxa characteristic of the deep subsurface, even where produced water had been re-injected. After decades of production, the putative native community remains detectable in the oil phase, suggesting the reservoir microbiome is more persistent, and more physically protected, than the aqueous sampling paradigm implied. The authors acknowledge limitations, including a single sampling time point, the absence of sequenced references for the external water sources, and one reservoir that yielded too little DNA from its fluid for metagenomic sequencing, but the paired design itself controls for temporal variation.

The practical implications reach well beyond petroleum geology. For operators monitoring souring and microbiologically influenced corrosion, a survey built on produced water alone will disproportionately reflect aquifer and injection-derived organisms while under-sampling the anaerobic specialists actually driving hydrogen sulfide production in the reservoir. The study’s recommendation is direct: wherever safety and logistics permit, reservoir fluids should be sampled to characterize the in-situ microbiome, and where only produced water is available, its directional biases, missing the reservoir-native community and overweighting water-associated taxa, must be stated plainly. More broadly, the work is a cautionary tale for subsurface microbiology everywhere: the easiest sample to collect is not necessarily the one that tells the truth about the ecosystem being studied, and only paired, controlled comparisons can reveal when a convenient proxy has quietly become a misleading one.

Subject of Research: Microbial community comparison between reservoir fluids and produced water in North Sea oil fields

Article Title: Reservoir fluids versus produced water as proxies for oil reservoir microbiomes

Article References: Alibrandi, A., Plewka, J., Vuillemin, A., Bartholomäus, A., di Primio, R., Probst, A. J., & Kallmeyer, J. (2026). Reservoir fluids versus produced water as proxies for oil reservoir microbiomes. Metabarcoding and Metagenomics, 10, Article e194096. https://doi.org/10.3897/mbmg.10.194096

Image Credits: AI Generated

DOI: 10.3897/mbmg.10.194096

Keywords: oil reservoir microbiome, produced water, reservoir fluids, metagenomics, 16S rRNA, North Sea, hydrocarbon degradation, reservoir souring, sulfate-reducing bacteria, thermophilic archaea, microbially influenced corrosion, subsurface microbiology