The temperature has always been a major variable in extraction. It's one of the first taught to trainee baristas. It's also very instinctive
However, it's more complicated when we try to understand its impact on the flavour profile isolated from the extraction. We've all heard someone saying too hot water can burn coffee - whatever that actually means. So what's the deal with the temperature?
This article is a summary of a new study by the UC Davis Coffee Center. It's free to access so I encourage everyone to have a look:
It's a bold title since it has only tested 3-litre batch brewing within a limited range of temperatures. But I'm getting ahead of myself.
Before dissecting the paper I'm going to outline the basic theory on temperature for context. In the end, you will find my opinion on the narrative and methods of the research.
Temperature as a variable
Temperature influences the speed of extraction (how quickly compounds dissolve in water). It affects what is being extracted - for example, some compounds are not soluble in water at room temperature - think the less acidic flavour profile of cold brew (Cordoba, 2019).
As we move into very high temperatures, the polarity of compounds comes into play. Water is polar and thus easily extracts other polar compounds. The majority of coffee solubles are polar. But roasty, smoky, acrid flavours are generally non-polar. They need much more energy to be extracted. You get more of them as the temperature gets closer to 100°C (Lopez, 2016). That's why darker roasts taste better with a lower brew temperature.
A slurry temperature in a batch brew
The temperature of the water is only one part. What is actually important is the slurry temperature and its stability. And that is only partially affected by the brewing water. To name just a few other factors:
- the temperature of the ground coffee,
- how well preheated is the brewing basket,
- the material and design of the brewer,
- cumulative mass of slurry,
- pulses setting.
Batch brewing places lower importance on the temperature of brewing water. The mass of brewing water in the batch brew is much greater than single-cup V60 - it will lose its hear more slowly. It will also be more resilient to ambient temperature and draughts. The length of the brewing is much longer too, thus the speed of extraction matters less.
With that context, we can now look into the study.
The goal of the research
This study is part of a bigger project looking into TDS and extraction percentage (PE). To quote from a bio of one of the authors, Mackenzie Batali:
"Her work focused on using descriptive sensory techniques to expand and update the classic Coffee Brewing Control Chart for the coffee industry. The goal is to quantify how flavor profiles change with different variables in different brewing methods, to create a more specific and scientifically-based Coffee Brewing Control Chart" [edited for clarity]
Previously the UC Davis Coffee Center published other studies focusing on the factors of sensory profile on coffee. Such as effect of basket geometry (Frost, 2019), brew development over time (Batali, 2020) and effects of strength, yield and roast level (Frost, 2020). The last one is very similar in the design to this study. In summary it tested different combinations of TDS and PE, using the same Curtis G4 brewer with 3075ml batch at 90.5°C. The critical variable was roast level (light, medium, dark). It might explain why having established the effects of roast this paper doesn't vary roast level. Although I suspect combining roast and brewing temperature as variables might affect the sensory profile to a great extent.
Study design
The strength and extraction percentage were isolated. To achieve that, different temperature setpoints were offset with other variables.
Coffee was brewed at each zone of the Coffee Control Chart for each temperature (87°C, 90°C and 93°C). That produced 27 type samples.
The total of 270 batches was measured and plotted on this graph:
As you can see there is a wide range of extraction levels from over 15% to nearly 27%. TDS varied from 0.9% to 1.7%.
31 attributes - 3 aromas, 22 flavours, 2 mouthfeels, and 4 basic tastes were evaluated by a team of 12 trained panellists.
The statistically significant attributes were plotted on this spider graph:
Results
The strength (TDS) has the biggest impact on the flavour profile, followed by Extraction Percentage. Across combinations of TDS and PE, the temperature didn't have a significant impact on the sensory profile.
A very interesting aspect of this study is correlating attributes with TDS and PE. This provides a scientific basis for extraction theory.
The Response Surface Methodology produced contour plots which are intuitive and easy to apply in real-life. It identified sensory extraction markers. That wouldn't be possible without advanced data science methods.
The results provide a strong scientific basis for understanding how the sensory profile is impacted by TDS and PE. It's certainly a better model than Lockhart's Brewing Chart.
Methods and limitations of the study - my opinion
Now that we've covered the major points of the study, let's have a closer look at technical details.
The recipe. All samples are detailed here. However, just to give you an idea what we're working with I've extracted the ranges for each variable:
Target TDS: 1% - 1.5 %
Target PE: 16% - 24%
Temperature: 87° - 93°C
Dose: 119 - 265 g
Brewing water: 3075 ml
Brewing ratio: 11.7 - 25.8
Grind size (Mahlkonig Guatemala): 3 - 5
Brew duration: 6:12 - 11 min
Brew pulses: 2 - 10
Let's discuss some of them.
First, the temperature range. 87°-93°C. I would barely call that a range for reasons I explained at the very beginning of this article. In batch brewing, even light roasts extract well around 90°C if the brewing cycle is long enough, without dramatic effects on acidity levels. I suppose the 3-litre batch could manage the 87°C, especially given the development of the roast. Therefore I can't imagine this range to produce vastly different flavour profiles.
They could go lower, but I'm particularly interested in the temperatures above 93°C. That could change the key findings on temperature. The authors might be aware of it since it's been mentioned in the introduction of the paper. But it would be super interesting.
Unfortunately, the brewer - Curtis G4 was partly at fault. To quote Professor Ristenpart: "93°C was the highest we could reproducibly achieve for the large volumes of coffee needed for the trained panel assessment".
The equipment was donated by Wilbur Curtis. Thus the choice of the brewer was not elicited by the study design.
Breville Corporation, which provided support for this research had an interest. As a producer of home brewers it benefits from the study. It's easier to manufacture equipment operating at lower temperatures. The use of energy and carbon footprint are certainly a concern as well. But I think the SCA Gold Cup certification standard - a minimum of 92°C is the main concern. Thus, I can understand why other equipment allowing higher temperatures wasn't sought.
I've never used Curtis G4, but I heard it has poor water dispersion and basket design. That is known to cause astringency and low flavour clarity. It is purely my speculation, but if that's true it could undermine the results.
Roasting. "Green coffee was mixed then roasted in a Loring S35 Kestral in 29 kg batches for an 11:29 (minutes:seconds) total roast time, with first crack recorded at 9:00 and a 2:30 development time". There's nothing unusual about it. I can't speculate about the development since the water details are confusing. Though 2:30 seems to be a fairly long development time for filter coffee, but I understand it's based on the US market. Testing different roasting styles would certainly benefit the study.
The roasting curve below (I've added conversions into Celsius):
Water specification is published in great detail, but I'm not sure if it's correct. I've emailed the authors about it, but at the time of writing, there was no reply.
"Brewing water was prepared by dissolving 1.16 g CaSO4·2H2O, 4.97 g MgSO4, 3.26 g NaHCO3, and 2.57 g KHCO3 per litre of deionized water" - that seems very hard, more like a composition of a concentrate you would then dilute with deionised water to make up brewing water.
Using this calculator, I came up with these values:
I wouldn't want to put that water in any water heating device, let alone drink coffee brewed with this. I will update this part of the article as soon as I get a response.
It would be great to have different water compositions used for this study.
Grinding. Mahlkonig Guatemala grinder was used (it has a dial numbered 1-7 for context).
3 grind settings were used:
- "3", average particle grind size 770 ± 10 µm
- "4", average particle grind size 928 ± 3 µm
- "5", average particle grind size 1115 ± 74 µm
3075 ml batch requires a very coarse grind setting for any reasonable results. I would certainly try to stick to the upper 20% of the range (5.5-7). The 770-micron particle size for 3075 ml proves poor extractability in this setting. I wonder whether the data provided is simplified or they have actually used such big steps (from 3 to 4) without any intermediate settings.
Using an effect of basket geometry study (Frost, 2019) for context, with the same grinder, the same range of settings was used. Although in that study the batch size was much smaller. 1200 ml of brewing water at 95°C for 66 g of coffee. Even though Breville Precision brewer was used, I can't imagine how 3075 and 1200 ml batch could successfully use the same grind size.
The coffee used was a washed Honduras from Marcala, La Paz. No other coffee was used for the testing.
Conclusion
I don't want to attack the paper or the authors. It's a good study. This research is needed. I'm impressed with the use of Principal Component Analysis and Response Surface Methodology. These techniques simplify the complexity of multiple brew variables and allow to show them in an intuitive way. It's an advanced data science applied to coffee allowing anyone to understand it, i.e. the contour plots. That's exactly what most coffee folks can't do themselves.
The sensory correlations with TDS and PE lead to a better scientific understanding of dialling-in. It's much better than the original Coffee Brewing Chart. I can totally see myself using a version of the contour plots in a sensory training session.
My main concern is the clickbait-like title of the study. The temperature findings apply to a very limited range of scenarios. It has limited implications for the industry.
I believe the narrative around the study is a bit too bold. Using an example from an interview with Ristenpart: "If those metrics [TDS and PE] are the same in two different brews, it doesn’t matter what brew temperatures were used to achieve them because tasters can’t differentiate them: they taste the same” - I would urge to be more careful with that statement - based on this research I agree with this sentence only for batch brewing (over 3 l) in a very narrow range of temperatures (87°-93°C).
I'm not sure if the interest of supporting companies (Breville, Curtis) had any influence on that. But even if, I'd be the first to defend that.
This study is free access - that means significant cost researchers have to pay if they want their paper published in a reputable journal. Any support makes that easier. I'd argue democratising access to scientific research is more important than the narrative.
References:
Images:
Title Photo by earlybird coffee on Unsplash
0 comments:
Post a Comment