The natural process in a closed tank

Where it is practiced

Brazil supplies nearly all the literature that describes this process with measurements. Trials on whole cherry in a closed tank have been published there since 2021. The first concerns the Caparaó region, on the border of Minas Gerais and Espírito Santo. It compares four altitudes, from 800 to 1,400 meters in steps of 200, on the cultivar Catuaí Vermelho IAC 44.

A second series works on the cultivar Arara, at the Chuá farm, in Minas Gerais, at 1,050 meters. It uses 200 liter bioreactors. In 2026, the same team publishes a trial in a 15,000 liter stainless steel tank. The process thus goes from farm trial to industrial scale, which is rare in this field.

Elsewhere, the practice has been sold since the mid-2010s by farms in several countries:

• Colombia
• Costa Rica
• Panama
• Ethiopia

Their protocols circulate through importers' sheets, not through articles. One example relayed by a reseller: the Deborah farm, in Panama, announces airtight tanks flushed with nitrogen and more than 100 hours of maceration on whole cherry. This is a commercial announcement, not a published protocol, and nothing makes it possible to verify it.

The date of the first lot sold under this name is not documented.

The principle

Whole, unpulped cherries are loaded into a tank closed with a water seal. This seal lets carbon dioxide out and keeps air from getting in. Nobody drives out the oxygen. The oxygen in the tank headspace is breathed by the microorganisms present on the fruit, and the medium becomes anoxic through depletion. The scientific literature calls this self-induced anaerobiosis fermentation, abbreviated SIAF.

The skin of the cherry, the exocarp, is a waxy layer that protects the fruit. It slows the access of microbes to the mucilage, the layer of sweet flesh stuck to the parchment. Fermentation therefore starts from the surface of the fruit and from the points where the skin is open. This barrier is the fundamental difference from a closed tank of pulped coffee, where the sugar is offered right away.

Three metabolisms follow one another and overlap. Lactic acid bacteria turn sugars into lactic acid. Yeasts turn them into ethanol and carbon dioxide, through alcoholic fermentation. Acetic bacteria oxidize this ethanol into acetic acid, but this work requires oxygen, so it dies out as the tank is depleted. Ethanol and acetic acid then combine into ethyl acetate, an ester with a fruity smell.

The starting population is not the one that finishes. On whole Arara cherry, the genus Methylobacterium occupies 64.3 percent of the bacterial community at entry into the tank, then recedes. Lactobacillus, among the bacteria, and the yeast Kazachstania, among the fungi, dominate afterward. Kazachstania reaches 61.8 percent of the fungal community at 72 hours. Saccharomyces appears only in inoculated tanks, which shows its natural absence from the cherries.

These acids, alcohols and esters cross the pericarp and reach the seed. The fruit then stays on the bean throughout drying, which prolongs these exchanges for several days.

The sequence

1. Harvest of ripe cherries. Available sugar is measured with a refractometer: 18 to 19.3 degrees Brix on Catuaí Vermelho IAC 44, in the Caparaó series. Brix tells how much raw material the microorganisms will have to work with.

2. Washing, or not. In the Arara series, the cherries are washed with drinking water from a farm well before loading. Washing lowers the starting microbial load, hence reduces the share of chance in the population that will take over. The Caparaó series, for its part, loads the fruit without washing or disinfection. The step is therefore not mandatory, and both practices are published.

3. Loading into a closed tank fitted with a water seal. The published volumes span three orders of magnitude:

• the 20 liter polypropylene bucket
• the 200 liter high-density polyethylene bioreactor, loaded with about 124 kilos of cherries
• the 15,000 liter stainless steel tank

The water seal does the essential: it evacuates the carbon dioxide produced without letting air in.

4. Choice between solid phase and immersion. In solid phase, nothing is added. In immersion, water is added, 60 liters in the 200 liter bioreactor, which the authors note as 30 percent by volume. Water homogenizes temperature and exchanges, but it dilutes the sugars and changes the population that dominates.

5. Yeast added or not. A commercial strain of Saccharomyces cerevisiae, RX60, was inoculated at 10 percent by volume, at 10 to the power 9 cells per milliliter. Inoculation serves to impose a known population instead of putting up with the one found on the fruit.

6. Duration. The published trials on whole cherry compare 24, 48 and 72 hours. At 48 hours, no butyric or propionic acid is detected. The authors conclude that this duration is safe, these two acids being the markers of fermentation off-flavors. Science goes well beyond this, however. One SIAF trial follows the tank up to 180 hours. It measures a rise in lactic acid there, at 3.57 grams per kilo on average. Farm protocols also announce long durations, more than 100 hours in the Panamanian case cited.

7. Temperature. In an unregulated tank, the coffee mass is recorded between 18 and 25 degrees, for an ambient air of 8 to 23.1 degrees. At 48 hours precisely, it ranges from 21.6 to 25 degrees depending on altitude. The Arara series compares this free handling to a chamber held at 27 degrees. Temperature governs speed: the hotter it is, the faster the acids accumulate, and the shorter the margin before over-fermentation.

8. Acidification. On pulped coffee in a closed tank under water, pH falls from 5.8 to 4.4 in 72 hours. In another closed tank trial, pH stops moving after 36 hours, whatever the total duration chosen. pH is the simplest witness of progress, because it directly follows the accumulation of acids.

9. Emptying. In the Arara series, the cherries are rewashed with drinking water on leaving the tank, before going to dry. This rinse is not systematic elsewhere, but it is in the protocol of the best-described trial.

10. Drying as a natural, whole fruit included, on a patio, down to 11 to 12 percent moisture. This is the second half of the process, and the longest.

11. Drying management. A coffee in the dry process needs about 12 days of drying, against 7 in the washed process, because the water has to cross the pulp. Drying first goes through six to eight days at a water activity close to 0.95, where harmless bacteria dominate. Then comes a phase of about six days around 0.90. The lot ends at 12 percent moisture, that is a water activity of 0.60.

12. Resting and hulling. In the Arara series, lots rest at room temperature and are hulled only after 30 days.

What it gives in the cup

What is measured, first.

On whole Caparaó cherry, at 48 hours, acetic acid is the most abundant of the acids assayed: 2.74 to 5.49 grams per kilo. Lactic acid stays between 0.54 and 1.38, and citric acid between 0.91 and 3.07. But this ranking does not hold from one series to the other. In the 15,000 liter industrial tank, also on whole cherry, it is lactic acid that dominates. The two published trials on whole fruit therefore do not give the same winner.

On pulped coffee in a closed tank, lactic clearly dominates: 5.06 grams per liter at 72 hours, against 1.00 for acetic. The idea that the cherry skin would tip the profile toward acetic is consistent with the Caparaó series, but the industrial trial contradicts it. The point is not settled.

On Brazilian coffees run by SIAF, the fruity attribute stands out strengthened compared with conventional handling. The measurement comes from a temporal dominance of sensations analysis, a method in which the taster signals at each moment the attribute that dominates.

On the 15,000 liter industrial tank, a 48 hour fermentation in solid phase with Saccharomyces cerevisiae improved the sensory score by up to 2.7 points. The number of volatile compounds identified rises by 46 percent. Sixty-one volatiles are identified in all, of which 55 percent are ketones, aldehydes and pyrazines. All the treatments in this trial give specialty coffees.

On the Arara series in 200 liter bioreactors, the median score of all treatments is 83.75 points. The authors use it as a threshold to separate the lots into two groups of equal size. They do not publish the score of each treatment.

A useful point of comparison, but obtained on pulped coffee: in a closed tank, the score rises with duration. It goes from 80.96 at 12 hours to 84.60 at 96 hours, in regular steps. This result does not carry over as is to whole cherry, for lack of a published equivalent trial.

What is claimed, next.

Lots go on sale with descriptors such as rum, candied fruit or cinnamon. No published comparison supports them. This would require setting an anaerobic natural lot against the same coffee run as a plain natural. The trials read compare tank durations with one another, or a closed tank with conventional handling. The pair that interests the buyer, anaerobic natural against plain natural, same harvest and same plot, has not been published.

What goes wrong

Over-fermentation in the tank. This is the defect named fermented or vinegary (fermento or vinagre, in Spanish). Cenicafé has published a reference scale, established on washed coffee, pulped and under water. It describes a succession of smells that appear in this order:

• sour taste and vinegar smell
• winey character
• smell of decomposed pineapple
• smell of onion or rancid
• discoloration of the beans

The hourly thresholds associated with this scale could not be verified at the source. They would not transfer in any case to whole cherry, which the skin protects. The succession of smells, however, remains the right guide for sensing where a tank stands.

Acetic drift. In the Caparaó series, acetic acid is the acid that rises highest at 48 hours. Now acetic bacteria need oxygen to make this acid from ethanol. A leaking tank, a tank opened and then closed again, cherries loaded warm: each of these actions lets air back in. The cup then turns to vinegar.

Heterogeneity of the tank. On pulped coffee, at 72 hours, lactic acid measures 5.06 grams per liter at the bottom against 2.18 at the surface, and acetic acid 1.00 against 0.40. On whole cherry, the 15,000 liter tank trial tested this question directly. Only the solid phase fermentations proved uniform from one sampling point to another. A single tank can therefore contain several coffees, and immersion does not remedy this.

Drying, which is the second half of the risk and the longest. The fruit stays on the bean, and the lot takes about 12 days to dry instead of 7. It goes through a phase of about six days around a water activity of 0.90. It is during this phase that ochratoxigenic Aspergillus can produce ochratoxin A. A layer that is too thick or a bed that is not turned creates damp pockets. From these come mold and stinkers, those isolated beans smelling of rot or vinegar that are enough to condemn a cup.

A laboratory result goes the other way, and deserves to be stated with its limits. Whole cherries were fermented for 36 hours in an anaerobic jar, at 20 then at 37 degrees. The fermentation liquid at 20 degrees contained more lactic acid: 0.279 milligram per milliliter, against 0.107 at 37 degrees. Put in contact with Aspergillus niger spores on a Petri dish, this liquid slowed their germination. It is an in vitro trial on 100 grams of cherries, with an oxygen-absorbing sachet. It does not demonstrate that a real lot is protected, only that tank metabolism produces acids hostile to the fungus.

The seed itself. Tank metabolites migrate to the embryo, and the activity of germination enzymes is modified as a result. The trial that measured this nevertheless concludes that the dry process preserves a higher proportion of viable seeds than the pulped process. The loss of viability is therefore not a marker specific to whole cherry in a closed tank. The link between this loss and cup quality is not established.

The economic risk, finally, which is not a technical detail. The more critical steps a process has, the more the probability of a failure rises, and over-fermentation is easy to obtain when the variables are not tracked. The market for these lots is narrow and saturates. A professional buying guide recommends that the buyer commit financially before the process is launched, so that failure does not rest on the producer alone.

Benchmarks

• Durations compared on whole cherry24, 48 and 72 hours
• Longest duration followed by a SIAF trial180 hours
• Acids at 48 hours on whole cherry, Caparaóacetic 2.74 to 5.49 g/kg, lactic 0.54 to 1.38 g/kg
• Dominant acid in the 15,000 liter industrial tank, whole cherrylactic
• Sugar of the cherries at entry18 to 19.3 degrees Brix
• Mass temperature in an unregulated tank18 to 25 degrees
• Drying in the dry processabout 12 days, against 7 in the washed process
• First drying phasesix to eight days at a water activity close to 0.95
• Ochratoxin window during dryingabout 6 days around a water activity of 0.90
• Target final moisture11 to 12 percent
• Rest before hulling in the Arara series30 days