How Do I Know That’s Really Kona Coffee?

shawnBy Shawn Steiman, PhD

There are times in life when you
want confirmation that the contents of a package really match what is advertised on the outside of the package. Is the olive oil really from Italy? Is the sparkling wine really from Champagne? 
Is it truly Manuka honey? The reason we want to know these things is because these products are almost always more expensive than their alternatives.
 Thus, if we’re going to pay more
for them, we want to be sure we’re
getting exactly what we pay for
 (the issue of whether they taste
as good as they’re supposed to
is a topic for another section in
the book). How do we prove the
product is what it claims to be? 
Is the coffee really from Kona,
 Hawaii?

In a perfect world, rare, special, or expensive coffees would taste so different that we’d be able to verify their origins upon tasting them. But, being able to taste with that level of precision is difficult and it requires extensive knowledge of coffees from all over the world. Moreover, every coffee grown within a particular place must have a shared and globally unique taste. Well, these prerequisites are never all met simultaneously, so, using taste to confirm the origin of a coffee will never work.

Alternatively, a government can establish rules and laws for packaging and labeling and expect its citizens to follow them. Most governments do this and they do their best to enforce them with the limited resources available to them. However, there are always clever miscreants and a government’s power doesn’t exist past its borders.

What is needed is an objective, product-based method for determining where a coffee was grown. All one has to do is discover the right chemical or combination of chemicals that will fingerprint a growing location. If every fingerprint is unique, then one just has to analyze any sample, match it to a fingerprint, and voilà!

Sounds easy, right? The actual lab work is usually fairly easy but discovering a fingerprint is incredibly tricky. Many scientists, including this author, have worked on this problem. Nobody has figured it out yet. There are two big hurdles to this problem. One is settling on the right fingerprint and the other is being able to properly analyze the data to ensure everything works correctly.

Scientists have tried all kinds of different analytical techniques and markers to build the fingerprint: near-infrared spectroscopy (NIRS), fourier transform infrared spectroscopy (FTIR), high performance liquid chromatography
(HPLC), solid phase microextraction—gas chromatography—time of flight mass spectrometry (SPME-GC-TOF-MS), brewed coffee volatiles, stable isotopes, elemental content, molecular compounds, and who knows what else! The aim has been to find a very quick, cheap, reliable method that can detect the right markers.

Most of these methods and chemical markers suit this purpose well and much of the data is very promising. The data is promising because many of these methods allow the detection of many signals or markers rather than a small handful. They can be 2,000 reflectances of light at different wavelengths, hundreds of volatile compounds, or dozens of molecules. The more markers one has to create a fingerprint, the more likely that fingerprint will be unique. Moreover, the current state of computer power and statistical software packages allows for adequate analysis of all the data, so building a fingerprint and testing its efficacy is relatively simple.

So, where’s the problem? The problem is twofold. One, there are never enough samples in a dataset to build a truly robust fingerprint. Two, any given bean is, well, complicated!

Large datasets are important for statistical power and simply being able to paint the right picture. The statistical analysis used in origin discrimination work requires many samples for the analysis to work well. Many studies do the analysis with too few samples and the numbers crunch well, too well, really. The end result is 
too perfect because so many markers are being used to describe a small set of samples. The data is overfit. Painting the right picture is just as important. If you want to be able to tell a Hawaii coffee from a Costa Rican coffee from
a Rwandan coffee, you need many samples from each location to capture the variation from that location. Now, with eighty-plus countries in the world growing coffee and each country having many individual regions, acquiring enough samples to paint the big picture is daunting.

As for coffee being complicated, there are just so many things that influence coffee’s chemical composition. These include, but are not limited 
to, year of production, the genetic makeup, the climate in which it grew, the nutritional health of the plant, the fertilizer regime, ripeness at harvest, cherry processing method, storage of green coffee,
 age of the green coffee, roasting, blending, and freshness. In order for a geographic fingerprint to work, it must be able to account for all these compositional influences every year across many locations!

I believe we have the knowledge and capability to build a geographic indicator system. It may never be perfect but it probably could be effective a very high percentage of the time. All we need are time, manpower, and adequate resources.

In the meantime, how do we know where the coffee in our cups is actually from? Trust. Trust in all the people whose hands touched that coffee and belief that they acted with integrity.

Screen Shot 2015-10-29 at 11.45.50 AMThis text is an excerpt from The Little Coffee Know-It- All: A Miscellany for Growing, Roasting, and Brewing, Uncompromising and Unapologetic, a forthcoming book written by Shawn Steiman, PhD, and published by Quarry Books. The book is laid out in an FAQ format where each question is addressed using the most current and accurate scientific data.