For the last several weeks, members of the Catracha Quality Project have been presenting results of a data survey conducted during the 2016 harvest season in Santa Elena, Honduras. To go back and read some of that work — which includes detailed process maps and an analysis of correlation between cherry selection and cup score — I suggest beginning here.

Catracha Community producers collected data pertaining to the harvest, pulping, fermentation, and drying of their coffee. Our hopes within the Quality Project are to dissect and analyze the collected data with the goal of creating a baseline dataset against which we can compare future harvests, as well as potentially directing further, more focused inquiry.

img_7373

With the analysis of the fermentation data provided by Catracha Community producers, I hope to shed some light on any potential correlations between the recorded fermentation process and the overall quality of each coffee. We have cast a wide net and collected a large variety of data. Without a rigid experimental structure, it would be wrong to attempt to formulate any causal relationships.

My analysis focuses on information collected surrounding fermentation–specifically the length of fermentation, the beginning and end pH of the slurry, and the average cup score of each sample. The hope is to identify general trends in the relationship between variables in the fermentation process and coffee quality, potentially allowing us to improve quality in the future by informing best practices and to guide future experimentation.

After the harvest, while farmers processed their coffee, the length of fermentation was recorded, beginning and end pH were logged, and samples maintained traceability through the cupping table. Only the coffees from Atanacio Nolasco, Luis Nolasco, Jose Antonio Nolasco and Fidelina Perez are shown below – these farmers share processing facilities. I felt their data could be looked at as a group for this reason. Producers made these observations while maintaining an otherwise regular process.

In the figure below, the measured pH at the end of the fermentation period is represented on the y-axis, cup score on the x-axis, and the relative size of each data point represents the total length of fermentation (larger circles indicate longer fermentation times).

fermentation

Firstly, notice how tight the distributions are – most fermentation times hover around 36 hours, final pH is generally less than 3.4 but greater than 3, and most samples were scored between 84 and 85. Given this view of the data, I decided to further examine the trials which resulted in cup scores significantly higher than average.

The sample with the highest average cup score was Atanacio’s first experimental pick of his coffee, which happened on January 18th. Although Atanacio demonstrates in these experiments a very high level of regularity in his processing – all fermentation times are about 34.5 hours – this sample showed a relatively fast change in pH, dropping to the lowest recorded final pH value.

Fidelina’s first experimental pick happened on January 15th, and this coffee was the next highest scoring. The fermentation numbers for this run are odd – although fermentation lasted 38 hours, the pH only dropped from 4.5 to 3.7, representing both a relatively more acidic beginning value and less acidic end value than the averages.

In the next figure, the size of each data point represents the overall quality of the respective coffee sample: the larger the circle, the higher that experiment’s average cup score. Here, we have plotted the change in pH on the y-axis versus the length of the fermentation on the x-axis. The scores seem to be consistently higher with fermentation times just over 35 hours and pH changes of at least two.
change

More obviously, the consistency in fermentation seems to lead to consistency in overall cup quality: if we assume that each producer’s coffee has a baseline potential quality, that level of quality is achieved more regularly when the fermentation process varies less.

Ultimately, more data is needed. At this point, it seems cherries that ripen early in the season scored better, and this cannot be discounted. One thought: perhaps the bouts of cold which moved in through the month of February negatively affected the quality of the coffee still on the plants.

Since metabolic activity of the collective “fermenters” is likely temperature dependent, I propose the addition of a temperature probe that can safely rest in the fermentation tank and passively record the temperature of the slurry. Fermentation, being a biological process, produces heat as a byproduct of metabolic activity – as such, simply measuring ambient temperature will be insufficient in a focused study of fermentation in Santa Elena.

Within a second series of experiments, training producers on data collection techniques will be paramount. This will help us reduce any variability due to errors and inconsistencies in the collection of data. Additionally, we could have the opportunity to control variables such as final pH or the length of fermentation. This might allow us to arrive at a more cause-and-effect analysis of the relationship between fermentation and quality.

img_7392