If you grow coffee, run a café, or simply want to understand what Citane is and why it behaves the way it does — start here. This section explains every key idea in plain language before the full technical paper begins.
This document is a thinking tool. It is not a recipe, a processing protocol, or a guarantee of any flavour outcome. It is a framework for understanding the coffee leaf as a biochemical system — and for making sense of what happens when that system is put through different kinds of processing.
Think of it like a map of a territory that has not yet been fully explored. The map is drawn from what science already knows about similar territories — tea processing, fermented foods, plant chemistry — applied carefully and honestly to the coffee leaf. Some parts of the map are well-grounded. Others are speculative. Where we are speculating, we say so clearly.
Farmers: Understanding what is chemically active in your leaf, and when, will help you make better decisions about when to harvest, how to handle the leaf immediately after picking, and what processing choices are worth experimenting with.
Café owners: Understanding the chemistry behind different Citane processing styles will help you describe what you are serving accurately, choose products that match your menu, and ask better questions of your suppliers.
Processors: This is your core reference. The Reservoir Model in Part II is the most directly practical section for your day-to-day decisions.
The full technical paper follows this primer. It is comprehensive, detailed, and references the scientific literature. You do not need to read it all to understand the core idea — the primer covers what you need. But if you want the reasoning behind any specific claim, the full paper has it.
↑ ContentsA coffee leaf is not a simple thing. Under a microscope, it is a dense ecosystem of molecules — dozens of different chemical compounds, each stored in different parts of the leaf cell, each waiting for a specific trigger to become active.
Scientists at the University of Siena analysed Coffea arabica leaves in detail and found 39 distinct compounds. These fall into several families. Here is what each family is, in plain language:
Your leaf is full of locked potential. Most of these compounds do nothing until something activates them — heat, air, water, time, or microbes. The art of Citane processing is choosing which locks to open, in which order, and for how long.
Here is the entire framework, compressed to its essentials.
The leaf is not a finished ingredient. It is a system of potential, waiting to be directed. The compounds inside it do not have fixed flavours. They have possible flavours — depending on what you do to the leaf, and in what order.
There are three layers to understand:
The pipeline above is the skeleton of the model. But there is one more dimension: the reservoirs are not separate rooms. They interact. Sugars feed microbes. Microbes produce acids. Acids change how enzymes work. Enzymes affect how polyphenols oxidise. Polyphenols affect the colour and taste of the cup.
This web of connections — with its cascades (A triggers B triggers C) and feedback loops (C changes the rate of A) — is what makes extended processing genuinely complex, and genuinely rewarding to explore. The full paper maps these connections. The route map is how you record where your particular journey went.
This document uses four classification systems — Reservoirs (A–K), Activation Events (A–F), Process Domains (1–6), and Emergent States (α–ε). They are not four separate models. They describe the same underlying process from four different angles:
Reservoirs = what is present in the leaf · Events = what activates change · Domains = common combinations of events · States = typical destinations or outcomes.
The relationships between them are not one-to-one — a single Domain can lead to different States depending on duration and depth, and a single Event can activate several Reservoirs at once. Holding all four in mind for any single processing step is not necessary. It is enough to ask, at each step: what is being activated, by what, and roughly where it tends to lead.
The five Emergent States are the possible flavour territories your Citane can arrive at:
For the farmer: The leaf you harvest is not neutral. Its chemistry at the moment of harvest — shaped by altitude, shade, leaf age, season, and variety — determines what processing can do with it. A leaf harvested too old has degraded chlorophyll and different enzyme activity than a young flush leaf. A leaf from high altitude may have higher chlorogenic acid concentrations than one from lower ground. These differences matter.
The practical upshot: harvest timing and leaf condition are the first processing decisions, even before the leaf is touched. The framework in this document can help you think about what you are starting with — which reservoirs are full, which are depleted — before any processing begins.
For the café owner: When you buy a Citane product, you are buying the result of a set of processing decisions — activation events applied to a specific leaf. Understanding the framework gives you a vocabulary for what you taste and why. A pale, green-gold liquor with grassy notes is likely a State Alpha product — minimal processing, enzymatic domain, intact chlorophyll. A deep amber cup with body and a honeyed finish is likely a State Beta — oxidative domain, catechin transformation.
Every cup of Citane is the result of a journey through the leaf's chemistry. This document is a map of the possible journeys. No journey is inherently better than another. They lead to different places. The question is always: which place are we trying to reach — and do we know how to get back there if we find something worth repeating?
The scientific foundation. What the coffee leaf contains, how we know, and why the standard approach to beverage processing needs to be rethought for this ingredient.
Most beverage frameworks start from a known target. Tea makers know what green tea tastes like. Roasters know what a medium roast profile produces. The target precedes the process, and the process is designed to reach it reliably.
This document proposes a different starting point — one that is closer to how a chemist or plant biologist thinks about the material.
The coffee leaf is not a finished ingredient waiting to be prepared. It is a living biochemical reservoir that has been harvested at a particular moment in its metabolic life. Inside every leaf, simultaneously and in dynamic tension, are dozens of compound pools. Enzymes that have not yet been activated. Glycoside-locked aromas waiting for the right hydrolytic conditions. Oxidative cascades held in check by intact cellular structure. Microbial ecologies clinging to the leaf surface, dormant and patient.
Processing does not create these things. It redirects what is already there.
The practical consequence is significant. There is no single "coffee leaf flavour." There is a family of possible flavour territories, each accessible through a different combination of activation events — mechanical, thermal, aqueous, microbial, enzymatic, or temporal. The leaf's chemistry is the landscape. Processing choices are the paths taken through it.
The leaf contains numerous precursor pools existing simultaneously. Different processing choices alter which pools become active and which remain dormant. The role of the Citane practitioner is not to reproduce a known flavour, but to navigate the landscape and record the routes taken.
The pivot point for this model is a 2022 paper published in the journal Foods by Cangeloni, Bonechi, Leone, Consumi, Andreassi, Magnani, Rossi and Tamasi at the University of Siena. Full citation: Characterization of Extracts of Coffee Leaves (Coffea arabica L.) by Spectroscopic and Chromatographic/Spectrometric Techniques. DOI: 10.3390/foods11162495.
The paper applied two complementary analytical methods — nuclear magnetic resonance spectroscopy (NMR) and high-performance liquid chromatography coupled to tandem mass spectrometry (HPLC-MSⁿ) — to extracts of Coffea arabica L., Castillo variety, collected at approximately 1,700m altitude in the Department of Huila, Colombia. In total, 39 distinct compounds were identified and the major components quantified.
The significance for Citane is not primarily clinical or nutritional. It is architectural. The paper provides a compound inventory — a list of what is actually present in the leaf — that can serve as the raw material for a theoretical processing model. If we know what compounds are present, we can reason about what those compounds might do under various processing conditions, drawing on what is understood about their chemistry in analogous systems: tea, wine, fermented foods, plant biotransformation.
| Category | Count | Notable Members |
|---|---|---|
| Xanthones | 4 | Mangiferin, Isomangiferin, Iriflophenone-3-C-glucoside, 6-O-(p-hydroxybenzoyl)mangiferin |
| Chlorogenic Acids | 7 | 3-CGA, 4-CGA, 5-CGA (dominant), 3,4-dCQA, 3,5-dCQA, 4,5-dCQA, 5-FQA |
| Flavonoids | 7 | Rutin, Rutin glycoside, Quercetin sophoroside, Kaempferol forms, Apigenin 6,8-di-C-glucoside, Catechin/Epicatechin |
| Procyanidins | 4 | Procyanidin B, trimer A-type, tetramer B-type, Procyanidin C |
| Lignans | 2 | Cinchonain I isomers a & b (novel finding in coffee leaf) |
| Alkaloids | 2 | Caffeine (7.94 g/kg DW), Trigonelline (4.47 g/kg DW) |
| Organic acids & amino acids (NMR) | ~13 | Malic acid, Lactic acid, Leucine, Alanine, Glutamine, Aspartic acid, Cysteine, Choline + saccharides |
| Compound | g/kg Dry Weight | %RSD |
|---|---|---|
| 5-Caffeoylquinic acid (5-CGA) | 16.27 ± 1.66 | 10.2 |
| Caffeine | 7.94 ± 0.42 | 5.3 |
| Trigonelline | 4.47 ± 0.13 | 2.9 |
| Mangiferin | 4.43 ± 0.14 | 3.3 |
| 3-CGA | 1.28 ± 0.12 | 9.2 |
| 4,5-dCQA | 0.91 ± 0.05 | 5.3 |
| 4-CGA | 0.89 ± 0.07 | 8.0 |
| 3,4-dCQA | 0.63 ± 0.05 | 7.6 |
| 3,5-dCQA | 0.58 ± 0.02 | 3.4 |
| Isomangiferin | 0.52 ± 0.03 | 5.8 |
These figures come from a high-altitude Colombian Castillo variety. Thumpassery Estate's Chandragiri Arabica, grown at approximately 130m under rubber shade, will present a different phytochemical profile. The concentrations above are indicative starting points — not fixed parameters for Citane specifically. Altitude, shade, leaf age, season, and variety all influence the compound profile significantly.
More broadly: every concentration figure in this document is a reference value drawn from available published literature, not a measurement of Thumpassery leaf. Beyond the variety/altitude/shade differences noted above, leaf-to-leaf, harvest-to-harvest, and season-to-season variation within a single estate is itself substantial and is not separately modelled here. This document treats reservoir contents as fixed reference points for the purpose of explaining mechanisms — how compounds tend to behave, relative to each other — not as a prediction of the exact starting composition of any given batch of leaf. Practical guidance on leaf selection and the sources of batch-to-batch variation is addressed separately in Paper II.
The complete compound list from Cangeloni 2022. This index is the biochemical inventory from which the Reservoir Model is constructed. Compound numbers follow the original paper's numbering.
| # | Compound | Class | Rt (min) | [M−H]⁻ | g/kg DW |
|---|---|---|---|---|---|
| 18 | Quinic acid | Organic acid | 1.01 | 191 | — |
| 1 | 3-Caffeoylquinic acid (3-CGA) | Chlorogenic acid | 3.51 | 353 | 1.28±0.12 |
| 19 | Iriflophenone 3-C-glucoside | Xanthone | 5.07 | 407 | — |
| 20a | Catechin / Epicatechin | Flavonoid | 5.28 | 289 | — |
| 2 | 1-Caffeoylquinic acid | Chlorogenic acid | 6.69 | 353 | — |
| 3 | 4-Caffeoylquinic acid (4-CGA) | Chlorogenic acid | 7.06 | 353 | 0.89±0.07 |
| 4 | 5-Caffeoylquinic acid (5-CGA) | Chlorogenic acid | 7.70 | 353 | 16.27±1.66 |
| 20b | Catechin / Epicatechin (isomer) | Flavonoid | 10.38 | 289 | — |
| 21 | Procyanidin B (dimer) | Procyanidin | 11.22 | 577 | — |
| 22 | Isomangiferin | Xanthone | 11.63 | 421 | 0.52±0.03 |
| 5 | Mangiferin | Xanthone | 12.42 | 421 | 4.43±0.14 |
| 24 | Procyanidin trimer A-type | Procyanidin | 16.19 | 863 | — |
| 25 | Procyanidin tetramer B-type | Procyanidin | 17.10 | 576* | — |
| 26 | Apigenin 6,8-di-C-glucoside | Flavonoid | 17.91 | 593 | — |
| 27 | Procyanidin C (trimer) | Procyanidin | 18.95 | 865 | — |
| 28 | 5-Feruloylquinic acid (5-FQA) | Chlorogenic acid | 19.25 | 367 | — |
| 29 | Rutin glycoside | Flavonoid glycoside | 19.89 | 771 | — |
| 30 | Quercetin sophoroside | Flavonoid glycoside | 20.70 | 625 | — |
| 31a | Cinchonain I isomer a | Lignan | 21.75 | 451 | — |
| 32 | Kaempferol triglycoside | Flavonoid glycoside | 22.41 | 755 | — |
| 6 | Rutin (quercetin rutinoside) | Flavonoid | 23.36 | 609 | — |
| 33 | 3,4-Dicaffeoylquinic acid | Chlorogenic acid | 24.00 | 515 | 0.63±0.05 |
| 34 | 3,5-Dicaffeoylquinic acid | Chlorogenic acid | 24.62 | 515 | 0.58±0.02 |
| 35 | Kaempferol-3-O-rhamnoglucoside | Flavonoid glycoside | 25.50 | 593 | — |
| 36 | 4,5-Dicaffeoylquinic acid | Chlorogenic acid | 26.44 | 515 | 0.91±0.05 |
| 37 | 6-O-(p-hydroxybenzoyl)mangiferin | Xanthone (acylated) | 27.10 | 541 | — |
| 31b | Cinchonain I isomer b | Lignan | 28.52 | 451 | — |
| 38 | Trigonelline | Alkaloid† | 1.11 | — | 4.47±0.13 |
| 39 | Caffeine | Alkaloid† | 20.69 | — | 7.94±0.42 |
* Doubly deprotonated [M−2H]²⁻ † ESI positive mode NMR additionally detected: malic acid, lactic acid, leucine, alanine, glutamine, aspartic acid, cysteine, choline, saccharides
↑ ContentsThe coffee leaf is a collection of distinct chemical systems — each with its own pool of compounds, its own reactivity, and its own flavour potential. We identify eleven reservoirs (A through K). Eight are directly evidenced by Cangeloni 2022; three — Sugar (I), Lipid (J), and Volatile Precursor (K) — are proposed theoretical extensions based on known plant leaf chemistry, clearly labelled as such throughout. All eleven coexist simultaneously in every leaf. Processing determines which become active.
In plain terms: these are the most abundant compounds in your leaf. They give structure, mild bitterness, and acidity to the cup. Under heat or oxidation they transform significantly.
The dicaffeoylquinic acids (dCQAs) are structurally more complex than the monoacyl forms and may behave differently under equivalent conditions. Their specific reactivity in coffee leaf processing has not been studied. Their collective concentration (~2.1 g/kg DW) is not negligible — they may contribute to body and colour development in ways that remain to be characterised.
In plain terms: mangiferin is the compound that makes coffee leaf unusual among beverage ingredients. It is not found in meaningful quantities in tea, wine, or coffee beans. Its flavour role is not fully understood — which is precisely what makes it interesting.
Mangiferin's specific sensory contribution in a brewed Citane infusion is genuinely unknown. At 4.43 g/kg DW it is present in significant quantity. Whether it contributes a perceptible flavour — or acts as an antioxidant buffer that modifies how other compounds behave — is one of the most interesting unanswered questions in Citane chemistry. Observation in the cup is currently the only route to an answer.
In plain terms: this is the system responsible for black tea chemistry. When the leaf is damaged and exposed to air, catechins oxidise — producing amber colour, body, and rounded flavour. Coffee leaves contain the same system.
In black tea, catechin oxidation produces theaflavins (brightness, briskness) and thearubigins (body, colour, depth). Whether PPO in coffee leaf produces structurally identical or only analogous compounds has not been established. The analogy is directionally reliable; it is not a precise prediction. The presence of a large xanthone pool alongside the catechin system — absent from Camellia sinensis tea — means the overall oxidative chemistry of coffee leaf will have its own character.
In plain terms: amino acids are not flavours themselves, but they become flavours. Under heat they react with sugars to create toasted and nutty aromas. Under microbial activity they become esters, organic acids, and fruity notes. This is arguably the most reactive pool in the leaf relative to its concentration.
In coffee roasting science, amino acids are often more important than trigonelline in generating flavour complexity because they participate in vast, overlapping Maillard networks with sugars, CGAs, and each other. The amino acid pool in the coffee leaf may be small in absolute terms, but it is highly reactive. Any processing step that increases the free amino acid pool — Koji treatment, proteolytic fermentation, extended enzymatic withering — potentially amplifies every downstream thermal and fermentative pathway.
In plain terms: several flavour and colour compounds in the leaf are chemically locked to sugar molecules — like a key inside a safe. In this locked form they are flavour-inactive. Processing can break the lock and release the active compound. This is one of the least explored but potentially most rewarding reservoirs in Citane.
Aspergillus oryzae produces β-glucosidase enzymes capable of cleaving glycosidic bonds at low temperatures (28–32°C), without the thermal degradation that acidic hydrolysis at high temperature would cause. This is theoretically one of the most precise tools available for unlocking the glycoside reservoir — releasing aroma compounds without collateral damage to other systems. This remains untested on coffee leaf specifically.
In plain terms: caffeine you know. Trigonelline you may not — but it is just as important. Under roasting, trigonelline breaks down and produces the distinctive smell of roasted coffee. Coffee leaves contain significant amounts of both.
In coffee beans, trigonelline is well-studied. Its presence at 4.47 g/kg DW in the coffee leaf — and particularly its ratio to caffeine (roughly 1:1.8, compared to a much lower ratio in beans) — suggests the leaf may produce a different roast-aroma profile from the bean when taken through equivalent thermal treatments. Whether this translates into a distinctive "roasted coffee leaf" aroma distinct from roasted coffee remains an open question. The thermal domain (Section 25) is where this hypothesis can be explored.
In plain terms: chlorophyll is the green colour of your leaf. It is also a reliable indicator of what state the leaf is in. A green cup means chlorophyll is intact — minimal processing. An olive or brown cup means it has transformed. Tracking colour is a practical proxy for tracking the state of this reservoir.
In plain terms: lignans (specifically cinchonain isomers) were found in coffee leaves for the first time by Cangeloni 2022. We do not yet know what they taste like or what they do under processing. They are documented here because honest science includes documenting the unknowns.
In plain terms: sugars sit at the centre of almost everything. Microbes eat them. Heat converts them into caramel. They react with amino acids to create roasted aromas. And many of the other compounds in the leaf are locked to sugar molecules. Sugars deserve their own entry in the model.
The original Cangeloni dataset identifies sugars within glycoside structures and in the NMR saccharide region, but does not quantify free sugars as a distinct pool. The theoretical review of this framework identified that sugars function as the central energy currency connecting microbial activity, thermal chemistry, and glycoside hydrolysis — and therefore merit their own reservoir designation. The sugar quantities in coffee leaf have not been separately characterised in the available literature and represent a gap worth filling.
In plain terms: coffee leaves contain far less fat than coffee beans, but not zero. Lipids matter because their oxidation products include some of the most powerful aroma compounds in plant foods — the green, grassy, and fruity notes that define fresh leaves, and the roasted, aldehyde-rich notes that emerge under heat. Tea science takes this seriously. Coffee leaf science has not yet begun here.
Japanese green tea research has documented that the fresh, "marine" vegetal aroma of high-grade matcha is significantly influenced by lipid-derived C6 aldehyde compounds (specifically (Z)-3-hexenal) produced by lipoxygenase activity when the leaf is disrupted. The same enzyme system is present in all green leaves including coffee leaf. The grassy/vegetal character of minimally processed Citane may have a lipid-pathway dimension that has not been studied. This is an open research territory.
In plain terms: if you have ever brewed a coffee leaf infusion and noticed a faint floral or citrus note that you cannot explain from looking at the dry leaf — this reservoir may be the reason. Many plants store aroma molecules in a locked, odourless form. Processing unlocks them. The coffee leaf may contain a significant pool of these locked aromas. Nobody has mapped it yet. This is the most speculative entry in the model — and potentially one of the most important.
Reservoir K is the most theoretically grounded but least empirically supported entry in this model. Terpene glycosides have been documented in other Coffea species and in tea; their presence in Coffea arabica leaves at Thumpassery has not been confirmed. This reservoir is included because the possibility of a significant locked volatile pool — accessible via glycosidase activity — represents one of the most compelling theoretical opportunities for Citane flavour development. It warrants investigation before being dismissed.
Reservoirs do not transform spontaneously. Something must trigger them. These are the activation events — the physical, chemical, and biological interventions that unlock, redirect, or combine the leaf's compound pools. Each event has its own character, selectivity, and risks. New events can be added to this framework as they are discovered without requiring a structural rewrite.
Intact leaf cells maintain spatial separation between reactive compounds and the enzymes that would otherwise transform them. PPO (polyphenol oxidase) is compartmentalised in the chloroplast and vacuole. Its phenolic substrates are stored elsewhere. Physical disruption — rolling, wringing, crushing — breaks cell walls and membranes, bringing enzyme and substrate into contact for the first time.
This is the fundamental mechanism of oolong and black tea manufacture. A rolled leaf begins oxidising within minutes. The degree of rolling determines how much of the catechin and CGA pools become exposed to enzymatic action.
For Citane: the question is not whether to roll, but how much, when in the sequence, and for how long. Light bruising activates the oxidative domain gently. Full rolling drives it aggressively. The decision commits the leaf to a trajectory.
Water provides molecular mobility for compounds previously immobilised in dry tissue. It activates hydrolytic enzymes — glycosidases and esterases — that require aqueous conditions. It creates the medium through which microbial activity can proceed. It is the substrate for hydrolysis reactions that unlock the glycoside reservoir (E).
The temperature of water matters significantly. Hot water drives rapid extraction and hydrolysis. Cool water produces slower, more selective extraction with different selectivity. Koji water is a specific instance — it brings not only moisture but a cocktail of fungal enzymes that act on multiple reservoirs simultaneously.
Oxygen is present in every open-air processing step. The question is whether oxidation is enzymatically directed (via PPO and peroxidase, operating below approximately 70°C), chemically auto-driven (non-enzymatic, slower, less selective), or actively suppressed.
Enzymatic oxidation is structured — PPO preferentially attacks o-diphenols including catechins and CGAs, producing specific quinone products that then couple with amino acids and other phenolics in complex but partially predictable cascades. Auto-oxidation is less selective and tends to produce more random degradation products over longer timescales.
Heat operates differently at different temperatures. At low temperatures (60–80°C) it fixes enzymatic activity, denatures PPO, and stabilises the leaf in whatever oxidative state it has reached. At medium temperatures (100–150°C) it drives moisture out, concentrates flavour compounds, and begins Maillard reactions. At high temperatures (160°C+) it triggers trigonelline degradation, CGA thermal breakdown into lactones, and deeper Maillard chemistry producing pyrazines and roast character.
Heat also drives the most visible changes to the chlorophyll system — green to olive to brown — providing a practical process indicator legible to the eye.
Microorganisms do not simply consume substrate — they produce enzymes, organic acids, alcohols, esters, carbon dioxide, and secondary metabolites inaccessible through any physical or thermal process. Microbes are simultaneously catalysts and energy harvesters: consuming one reservoir and generating new compounds from another.
The coffee leaf carries a native epiphytic microbiome — bacteria and fungi living on the leaf surface in the field. Beyond this, controlled inoculation with Koji, lactic acid bacteria, or selected yeast strains opens the full range of fermentative chemistry.
Time is not a driver of chemistry in the way heat or oxygen are. But it is the medium through which every other event operates. A 4-hour wither produces a different leaf from a 24-hour wither. A 48-hour Koji contact is chemically distinct from 96 hours. Post-process rest allows microbial populations to consolidate and enzymatic work to continue at ambient pace.
Time is also the only variable that cannot be compressed. This makes it the most often overlooked — and one of the most practically significant — in a processing sequence. For Citane route mapping, time should be recorded as precisely as temperature. Subtle changes in rest periods between steps may account for large differences in final sensory profile.
An important structural insight: the framework so far has treated compounds as if they simply transform. But transformations are energy transfer events. A compound does not "choose" to become something else. An energy system pushes it across a reaction barrier, or lowers that barrier so the reaction can proceed. This layer sits beneath activation events and gives the model its physical grounding.
In chemistry, reactions do not happen unless energy is available to drive them. Some reactions are thermodynamically favourable — they would release energy if they happened — but they do not proceed because the activation energy barrier is too high. Something must either supply energy to push molecules over the barrier, or lower the barrier so the reaction can proceed at ambient conditions.
This is what energy systems do. They do not create compounds. They make reactions possible — either by supplying energy, lowering the barrier, or providing a pathway that was not otherwise available.
Understanding energy systems explains why the same compounds can produce radically different outcomes depending on the processing context. Trigonelline does not "choose" to become pyridines. Heat pushes its molecules over the thermal degradation barrier. Catechins do not "decide" to oxidise. PPO lowers the activation barrier for oxidation at ambient temperature. These distinctions matter for designing processing sequences intentionally.
↑ ContentsWhat it does: Supplies kinetic energy to molecules, pushing them over reaction barriers. More heat = more reactions accessible. Temperature determines which reactions become active.
In Citane processing: Every heat application — from pan-firing at 80°C to roasting at 180°C — is a thermal energy input. The temperature profile determines whether you are in the enzyme-inactivation range, the Maillard onset range, or the deep thermal conversion range.
What it does: Lowers the activation energy barrier for specific reactions, allowing them to proceed at ambient temperature and speed. Enzymes are highly selective — each acts on specific substrates via specific pathways.
In Citane processing: PPO (oxidation of catechins and CGAs), β-glucosidase (glycoside hydrolysis), lipoxygenase (lipid oxidation), proteases (protein breakdown). Endogenous leaf enzymes, Koji-derived enzymes, and microbial enzymes all operate in this system.
What it does: Microbes are simultaneously catalysts and energy harvesters. They consume one compound to extract chemical energy, then release new compounds as metabolic products. Unlike enzymes, they have goals — reproduction — and will continue transforming substrate until it is exhausted or conditions become hostile.
In Citane processing: Every fermentation step. Yeasts consuming sugars to produce esters. LAB consuming sugars to produce lactic acid. Native microbiota transforming the leaf surface environment before intentional inoculation.
What it does: Oxygen accepts electrons from reduced compounds, oxidising them. This releases energy and changes molecular structure. The rate of oxidation is controlled by oxygen availability, temperature, pH, and the presence of pro-oxidants or antioxidants (including mangiferin).
In Citane processing: The primary driver of colour development and flavour transformation in all open-air processing steps. Controlled by how thinly the leaf is spread, airflow, temperature, and whether enzymatic or non-enzymatic oxidation is dominant.
What it does: Water activity (aw) describes how much of the water in a system is "free" — available to support chemical reactions, enzyme activity, and microbial growth. High water activity enables all aqueous reactions. Low water activity (dry leaf) suppresses most biological activity but does not prevent thermal reactions.
In Citane processing: Controls which reactions are possible at any given stage. A dry leaf cannot support enzymatic hydrolysis of glycosides. A wet leaf in a warm environment will support rapid microbial growth. Water activity management is implicit in every drying, soaking, and resting decision.
The complete model, with Layer 0 included, reads as follows:
This layered model allows a more precise diagnosis of why a processing sequence produced a particular result — and more importantly, why it failed to produce an expected one. If the thermal energy was insufficient (temperature too low), the Maillard reactions do not proceed. If the enzymatic catalysis was interrupted (PPO denatured too early by heat), oxidation stops regardless of oxygen availability. The energy system layer is where troubleshooting begins.
Energy systems do not operate in isolation — they govern each other. Water activity is the gatekeeper: if the leaf is too dry, enzymatic catalysis and microbial metabolism cannot operate regardless of how much enzyme or how many microbes are present. Temperature determines which energy system dominates: below ~70°C, enzymatic catalysis is possible; above it, thermal energy takes over and enzymes are denatured. Oxidative potential is modulated by the antioxidant content of the leaf — a high-mangiferin leaf may have a different oxidative environment than a low-mangiferin leaf (see Reservoir B, buffering hypothesis), even under identical atmospheric conditions.
This means that a processing decision which changes one energy system almost always affects others. Drying the leaf (reducing water activity) does not just stop microbial activity — it also stops enzymatic hydrolysis of glycosides, preventing Reservoir E from unlocking further. A decision about moisture level is simultaneously a decision about which energy systems remain operative.
When activation events combine in characteristic ways, they create Process Domains — recognisable territories of chemistry with their own logic and sensory tendencies. A processing sequence can move through several domains in sequence, or commit deeply to one. The hybrid domain — where multiple domains overlap — is typically where the most complex and interesting results emerge.
Entered when the leaf is physically disrupted and exposed to air below the PPO denaturation threshold (~70°C). The classical analogue is oolong or black tea manufacture. The Oxidative Domain converts catechins and CGAs into polymeric oxidation products. Colour shifts from green toward amber and brown. Grassy notes diminish. Tea-like, honeyed, and sometimes floral notes emerge.
For coffee leaf, the xanthone pool (Reservoir B) coexists with the catechin system throughout this domain — see Reservoir B's buffering hypothesis for a possible (unconfirmed) interaction with no direct tea analogy.
The Thermal Domain begins where heat becomes the primary transformation agent, operating on the leaf's precursor pools rather than simply fixing what oxidation has done. This means temperatures above approximately 130°C, where Maillard chemistry begins to dominate and trigonelline degradation becomes significant.
The sensory products are toasted grain, roasted nut, caramel, and — at higher temperatures — pyridine and furan notes that move the profile toward coffee-adjacent territory. The trigonelline pathway in coffee leaf has not been specifically studied; its behaviour here is extrapolated from well-characterised coffee bean roasting science.
This domain uses the leaf's own enzymes as the primary transformation tools, without mechanical disruption sufficient to trigger the full oxidation cascade, and without heat or significant microbial activity. The analogue is a long, slow wither in green tea — where the leaf softens, loses moisture, and allows glycosidase and esterase activity to proceed gently.
The Enzymatic Domain is subtle and its products are the hardest to observe directly — unlocked aromatic compounds from the glycoside reservoir, mild textural changes, gradual acidity shifts. It is the domain of patience and small changes that may only become perceptible in the brewed cup.
Koji (Aspergillus oryzae) is a food-safe mould used in Japanese culture to produce sake, miso, and soy sauce. Its relevance to coffee leaf processing comes from its enzyme cocktail: proteases that break proteins into amino acids, amylases that convert starches to sugars, and β-glucosidases that cleave glycosidic bonds. Applied via Koji water or direct inoculation, these enzymes can simultaneously unlock the glycoside reservoir (E), increase the free amino acid pool (D), and create new fermentable sugar substrate (I) — all at low temperature, without thermal damage.
The Microbial Domain encompasses broader fermentative activity by multiple organism classes in a less controlled environment. This is the territory of natural fermentation — analogous to the spontaneous fermentation of coffee mucilage in wet-process coffee, or the native fermentation of pu-erh tea.
Ester and organic acid production creates profiles that can be simultaneously fruity, sour, complex, and unexpected. The risk of off-flavours — excessive acetic acid, unwanted microbial metabolites — is higher here than in any other domain. But so is the potential for sensory territory that no controlled physical or thermal process could produce.
No real-world processing sequence operates in a single domain. Every step carries forward the chemical state established by the previous step. The Hybrid Domain recognises that the most interesting Citane profiles will emerge from deliberate sequences that move through two or more domains in a designed order — each step operating on the outputs of the previous one.
A wither followed by Koji treatment followed by a gentle roast combines enzymatic aroma release, enzyme-amplified amino acid substrate, and thermal Maillard development on that enriched substrate — three domains, each feeding the next. This is the territory of State Epsilon.
The Reservoir Model presents compounds as distinct pools. But this is a simplification — the pools are connected. A sugar feeds a microbe. The microbe produces an acid. The acid shifts the pH. The shifted pH changes how an enzyme behaves. Every processing journey is a path through an interconnected network, not a sequence of isolated transformations. This section upgrades the model from a static map to a dynamic one.
The original framework treats compounds as if they merely transform — as if each reservoir is a room, and processing is the key that opens it. This is directionally useful, but it misses something important: the compounds are not in separate rooms. They are in a continuous chemical environment, and they interact.
Consider what happens in a single extended wither. Endogenous glycosidases slowly cleave glycoside bonds in Reservoir E, releasing aglycones and free sugars into the medium. The sugars feed the native microbiota on the leaf surface (Reservoir I → microbial activity). The microbiota produce organic acids that lower the pH. The lower pH changes the rate of PPO activity. PPO acts on the catechin pool (Reservoir C) at a modified rate. The modified oxidation produces different quinone products. Those quinones couple with the freed amino acids (Reservoir D) to produce new pigments and flavour compounds that would not exist in the sum of the reservoirs treated individually.
This cascade — triggered by a single activation event, propagating through multiple reservoirs in a specific sequence — is what makes coffee leaf processing genuinely complex, and genuinely interesting. The model should reflect it.
The wither example above describes a cascade: A → B → C → D, each step triggered by the previous. This is already more complex than treating reservoirs as isolated. But real biological systems also contain feedback loops — where a downstream product modifies the rate of an upstream reaction.
Example of a feedback loop in this system: Microbial organic acid production (from Reservoir I sugar consumption) lowers pH. Lower pH changes the activity of PPO. Changed PPO activity alters the rate of catechin oxidation (Reservoir C). The catechin oxidation products are different at lower pH. Those products re-enter the environment and modify the conditions in which microbes are operating. The microbes' behaviour changes as a result. The loop closes.
No single step in this loop is large or dramatic. But the accumulation of small feedback effects over hours or days is precisely what makes extended processing sequences produce outcomes that are disproportionate to what any single activation event would suggest. This is what is meant by "system dynamics" rather than "inventory."
The old question: Which reservoir are we activating?
The new question: Which pathway through the network am I initiating — and what feedback loops will it encounter along the way?
The following diagram represents the key compound nodes and the connections between them. Arrows indicate transformation pathways. The pathways are activated by specific energy systems and activation events — indicated in brackets.
[Enzymatic / Time]
GLYCOSIDES (E) ──────────────────────────> FREE AGLYCONES (quercetin, kaempferol, terpenes)
│ │
│ [Enzymatic / Microbial] │ [Microbial / Thermal]
v v
FREE SUGARS (I) ────────────────────────────────> ESTERS · ACIDS · MAILLARD PRODUCTS
│ [Microbial] │
│ │
v [Thermal > 120°C] v [Thermal > 130°C]
MAILLARD NETWORK <──── AMINO ACIDS (D) <──── PROTEINS [Protease]
│ │
v v [Microbial / Thermal]
PYRAZINES · FURANONES ESTERS · ALDEHYDES · ORGANIC ACIDS
[Mechanical / Oxygen / PPO]
CATECHINS (C) ──────────────────────────────> QUINONES
│ │
│ v [Time / Oxygen]
│ THEAFLAVIN ANALOGUES → COLOUR · BODY
│ │
│ v [Amino acids D]
│ BROWN PIGMENTS (coupled products)
│ │
│ < < [FEEDBACK LOOP] < < < < <│
│ pH drop (from microbial acids, Reservoir I)
│ modifies PPO activity → alters oxidation rate of this pool
│
CHLOROGENIC ACIDS (A) ──[PPO]──> CGA QUINONES ──> COLOUR · STRUCTURE
│
└──[Heat > 160°C]──> CGA LACTONES ──> STRUCTURED BITTERNESS
TRIGONELLINE (F) ──[Heat > 160°C]──> PYRIDINES · NICOTINIC ACID ──> ROAST NOTES
MANGIFERIN (B) ──────> [Buffering hypothesis, unconfirmed — see Reservoir B] ──> possible moderation of CGA (A) / Catechin (C) oxidation rate
│
└──[Unknown]──> POSSIBLE SENSORY CONTRIBUTION (not yet characterised)
LIPIDS (J) ──[Lipoxygenase / Mechanical]──> C6 ALDEHYDES ──> GREEN · GRASSY · FRESH NOTES
│
└──[Thermal]──> ALDEHYDES · KETONES ──> COOKED · ROASTED NOTES
VOLATILE PRECURSORS (K) ──[Glycosidase]──> FREE TERPENES ──> FLORAL · CITRUS · SPICE
│
└──[Carotenoid cleavage / Thermal]──> IONONES · DAMASCENONES ──> FLORAL · TOBACCO
CHLOROPHYLL (G) ──[Withering]──> PHEOPHYTIN ──> COLOUR SHIFT (green → olive → brown)
│
└──[Carotenoid path / Thermal]──> VOLATILE PRECURSORS (K)
Note: All pathways are theoretically proposed from published chemistry in analogous systems. Pathways marked [Unknown] have no established analogue. The network is not exhaustive — it represents current understanding of likely connections, not a complete biochemical map.
↑ ContentsA network journey is what happens when you trace a specific processing sequence through the network map. Instead of asking "what does this step do?" you ask "where in the network does this step take me — and what does the chemistry encounter next?"
This is the difference between a process description and a network journey. The process description says: "Koji treatment, then roast." The network journey says: "Koji treatment amplified the amino acid and sugar pools, which were then converted by Maillard chemistry during roasting into a richer pyrazine profile than a direct-roast pathway would produce — with a possible additional volatile layer from Koji's glycosidase activity on the terpene glycoside pool."
One describes what was done. The other describes why it produced something different.
↑ ContentsRather than targeting specific flavours, the model predicts states — broad sensory territories defined by which chemical systems are dominant at the point of brewing. A batch does not "achieve" a state; it arrives at a position in the landscape. The route maps record how it got there.
Intact chlorophyll dominates. Catechin and CGA pools are largely unreacted. No significant oxidation, no Maillard chemistry, no fermentation. Accessed by minimal-disruption, low-heat processing: steaming to kill-green, then drying at low temperature with no mechanical disruption and no fermentation.
Sensory tendency: Fresh, vegetal, herbal, bright. Possible marine or seaweed note if chlorophyll is prominent. Mild grassiness. Relatively simple profile — the leaf at its least transformed, with the highest proportion of intact compounds.
Oxidised catechin products dominate the sensory profile. This is the most "tea-like" state available to Citane. Accessed by partial to full oxidation pathways, stopped before fermentation or deep thermal conversion.
Sensory tendency: Rounded, tea-like, honeyed, possibly floral. Amber to deep amber liquor. Moderate to full body. Smooth acidity from modified CGA pool. The xanthone pool (mangiferin, Reservoir B) may contribute to this stability via the antioxidant buffering hypothesis (see Reservoir B) — a character that, if confirmed, would be specific to coffee leaf and absent from tea.
Microbial activity has had significant influence on the leaf's chemistry. Fermentation esters, organic acids, and biotransformed phenolics characterise this territory. The analogy is aged or wet-pile pu-erh tea, or naturally fermented coffee — a profile that is layered, complex, and not immediately legible to an untrained palate.
Sensory tendency: Earthy, layered, fermented complexity, possible dried fruit or vinous notes. Body can be significant. Acidity shifts from the bright CGA-acidity of green states to a softer organic-acid acidity from lactate and acetate production. This is the state where Citane can move furthest from conventional coffee leaf character.
The thermal domain, taken to sufficient depth, produces State Delta. Maillard products — pyrazines, furans, aldehydes — dominate. Trigonelline degradation adds pyridine notes. CGA lactones contribute structured bitterness distinct from raw CGA bitterness. Chlorophyll is fully transformed.
Sensory tendency: Nutty, toasted grain, caramel, roasted — coffee-adjacent but distinct from roasted coffee beans because the starting chemistry differs. The trigonelline-to-caffeine ratio in the leaf is different from the bean, and the far lower lipid content compared to the bean (far fewer diterpenes and wax esters) means the roast character will be cleaner and less heavy. The leaf's identity remains present, but the transformations are deeper and more irreversible than in any other state.
State Epsilon is not a failure of the model. It is the model's most honest entry — and arguably its most important one.
When multiple domains overlap — when enzymatic, oxidative, microbial, and thermal chemistry occur in close sequence, each acting on the outputs of the previous — the resulting compound profile cannot be predicted from individual reservoir descriptions alone. Coupling reactions between different compound classes, novel biotransformation products, and the emergence of compounds not present in the starting material: these are the chemistry of State Epsilon.
This is where an entirely new Citane identity may emerge. It cannot be designed from first principles. It must be arrived at through experimentation, recognised when it appears, and recorded precisely enough that the route back can be found.
State Epsilon is not a destination. It is a territory to be explored. Every Epsilon batch that produces something interesting becomes the seed of a new route map — and the model grows from the accumulation of those routes into an actual navigable geography of coffee leaf transformation.
Most food processing documents pretend to understand everything. Koji, oxidation, endogenous enzymes, leaf microbiome, and roasting chemistry all interacting together in a novel substrate is extraordinarily difficult to predict from first principles. State Epsilon is not a placeholder to be filled in later. It is a principled acknowledgement that complex interacting systems generate outcomes that cannot be computed from their components. This is rigour, not weakness.
A Route Map is a record of a processing sequence and its observed outcome. It is the primary tool for converting exploration into repeatable craft. Without route maps, every batch is an experiment with no memory. With them, experiments accumulate into knowledge.
After surveying the full model — eleven reservoirs, six activation events, five energy systems, six process domains, five emergent states, and the precursor network — several intersections stand out as theoretically rich and currently unexplored. The Reservoir B buffering hypothesis (mangiferin's possible role in moderating oxidation) is one. The volatile precursor pool (Reservoir K) is another. The one below is not presented as the unique conclusion the chemistry points to — rather, it is the candidate that is most practically accessible, because it sits closest to processing steps already in use.
It is not the trigonelline thermal pathway. That is well-understood in the context of coffee bean roasting and can be extrapolated with reasonable confidence.
It is not the oxidative domain. That has analogues in a century of tea processing science.
The candidate prioritised here is:
Glycosides (E) + Amino Acids (D) + Microbial Enzymes + Mild Heat
The convergence of enzymatic glycoside unlocking, microbial and Koji-driven amplification of the amino acid pool, and low-temperature thermal development of the resulting substrates.
This intersection sits directly at the centre of existing Citane experimental practice: withering, wringing, Koji water treatment, resting, drying, and gentle roasting. Each of these steps makes sense individually. What the network model reveals is that they are not independent — they are a cascade. Each step modifies the substrate that the next step works on. The Koji treatment does not simply add flavour. It rebuilds the chemical landscape that the roasting step encounters.
A direct-roast coffee leaf and a Koji-treated-then-roasted coffee leaf begin with the same starting material. They do not end in the same place — because the energy systems acting on the Koji-treated leaf operate on an enriched and partially transformed precursor network. The outcome is structurally different, not merely flavour-different.
If this hypothesis holds, it may represent a comparatively direct route toward genuinely new coffee-leaf flavour territory — flavour that is not a version of tea, not a version of roasted coffee, and not a version of fermented fruit, but something that emerges from the specific chemistry of this leaf, processed with these tools, in this sequence. Other routes, including those above, may prove equally or more significant; this one is highlighted because it is the easiest to begin testing now.
The framework is not intended to reduce coffee leaf processing to fixed pathways. It exists to make observation more precise. Every processing decision alters multiple variables simultaneously, and combinations often behave differently from individual interventions taken in isolation.
Readers are therefore encouraged to observe both isolated transformations (what does this one step do, on its own?) and combinatory transformations (what does this step do when it follows that one?) — recording not only intended outcomes but unexpected behaviour. New understanding is likely to emerge from the interaction between pathways as much as from the pathways themselves.
The traditional Engere pathway (Ethiopia) appears to occupy an interesting position within this framework, and is raised here as a question for future investigation, not a conclusion.
Engere achieves substantial extraction, sweetness, body, and sensory persistence using a preparation method that is, by the standards of this document's process domains, relatively simple — closer to a single sustained Aqueous/Thermal exposure than to a multi-step sequence moving through several domains.
This raises an open question: are some of the most distinctive outcomes in coffee leaf beverages the product of increasing process complexity (more domains, more steps, more sequence) — or of increasing extraction and transformation depth within a comparatively simple pathway (the same domain, sustained longer or more thoroughly)?
This is not intended to elevate Engere above other pathways, nor to suggest that complexity is unnecessary. It is intended to encourage comparison between isolated interventions, combinatory interventions, and traditional long-duration extraction systems — and to suggest that the framework may ultimately prove most useful when these are viewed as different points within the same landscape, rather than as competing approaches.
This document has not resolved the question of what Citane's most distinctive flavour identity is. It cannot — that question is answered in the cup, not on paper. What it has done is map the territory where that answer is most likely to be found, provide the vocabulary for describing the journey, and establish the framework for recording what is discovered along the way.
The work is in the leaf. The document is the map.
Complete compound identification data as reported in the source paper. Rt = retention time. Concentrations in g/kg dry weight where measured. All ESI negative mode unless marked.
| # | Compound Name | Chemical Class | Rt (min) | [M−H]⁻ | g/kg DW | %RSD |
|---|---|---|---|---|---|---|
| 18 | Quinic acid | Organic acid | 1.01 | 191 | — | — |
| 1 | 3-Caffeoylquinic acid (3-CGA) | Chlorogenic acid | 3.51 | 353 | 1.28±0.12 | 9.2 |
| 19 | Iriflophenone 3-C-glucoside | Xanthone | 5.07 | 407 | — | — |
| 20a | Catechin / Epicatechin | Flavonoid / Flavan-3-ol | 5.28 | 289 | — | — |
| 2 | 1-Caffeoylquinic acid | Chlorogenic acid | 6.69 | 353 | — | — |
| 3 | 4-Caffeoylquinic acid (4-CGA) | Chlorogenic acid | 7.06 | 353 | 0.89±0.07 | 8.0 |
| 4 | 5-Caffeoylquinic acid (5-CGA) | Chlorogenic acid | 7.70 | 353 | 16.27±1.66 | 10.2 |
| 20b | Catechin / Epicatechin (isomer) | Flavonoid / Flavan-3-ol | 10.38 | 289 | — | — |
| 21 | Procyanidin B (dimer) | Procyanidin | 11.22 | 577 | — | — |
| 22 | Isomangiferin | Xanthone | 11.63 | 421 | 0.52±0.03 | 5.8 |
| 5 | Mangiferin | Xanthone | 12.42 | 421 | 4.43±0.14 | 3.3 |
| 24 | Procyanidin trimer A-type | Procyanidin | 16.19 | 863 | — | — |
| 25 | Procyanidin tetramer B-type | Procyanidin | 17.10 | 576* | — | — |
| 26 | Apigenin 6,8-di-C-glucoside | Flavonoid (C-glycoside) | 17.91 | 593 | — | — |
| 27 | Procyanidin C (trimer) | Procyanidin | 18.95 | 865 | — | — |
| 28 | 5-Feruloylquinic acid (5-FQA) | Chlorogenic acid | 19.25 | 367 | — | — |
| 29 | Rutin glycoside | Flavonoid glycoside | 19.89 | 771 | — | — |
| 30 | Quercetin sophoroside | Flavonoid glycoside | 20.70 | 625 | — | — |
| 31a | Cinchonain I isomer a | Lignan | 21.75 | 451 | — | — |
| 32 | Kaempferol triglycoside | Flavonoid glycoside | 22.41 | 755 | — | — |
| 6 | Rutin (quercetin rutinoside) | Flavonoid | 23.36 | 609 | — | — |
| 33 | 3,4-Dicaffeoylquinic acid (3,4-dCQA) | Chlorogenic acid (di-) | 24.00 | 515 | 0.63±0.05 | 7.6 |
| 34 | 3,5-Dicaffeoylquinic acid (3,5-dCQA) | Chlorogenic acid (di-) | 24.62 | 515 | 0.58±0.02 | 3.4 |
| 35 | Kaempferol-3-O-rhamnoglucoside | Flavonoid glycoside | 25.50 | 593 | — | — |
| 36 | 4,5-Dicaffeoylquinic acid (4,5-dCQA) | Chlorogenic acid (di-) | 26.44 | 515 | 0.91±0.05 | 5.3 |
| 37 | 6-O-(p-hydroxybenzoyl)mangiferin | Xanthone (acylated) | 27.10 | 541 | — | — |
| 31b | Cinchonain I isomer b | Lignan | 28.52 | 451 | — | — |
| 38† | Trigonelline | Alkaloid | 1.11 | — | 4.47±0.13 | 2.9 |
| 39† | Caffeine | Alkaloid | 20.69 | — | 7.94±0.42 | 5.3 |
* Doubly deprotonated [M−2H]²⁻ † ESI positive mode [M+H]⁺ NMR additionally detected (not in table): Malic acid, Lactic acid, Leucine, Alanine, Glutamine, Aspartic acid, Cysteine, Choline, Saccharides (hexose, deoxyhexose, pentose)
↑ ContentsCangeloni, L.; Bonechi, C.; Leone, G.; Consumi, M.; Andreassi, M.; Magnani, A.; Rossi, C.; Tamasi, G. Characterization of Extracts of Coffee Leaves (Coffea arabica L.) by Spectroscopic and Chromatographic/Spectrometric Techniques. Foods 2022, 11, 2495. DOI: 10.3390/foods11162495. Published under Creative Commons Attribution (CC BY) 4.0.
All compound data, quantitative values, chromatographic parameters, and analytical findings cited in this document originate from this paper. The research belongs to the original authors. Citane / KoffyKraft make no claim of ownership over this scientific data.
This document is a conceptual model only. It is not experimentally validated. It is not a processing protocol. It is not a functional or health claim of any kind. All theoretical statements about compound behaviour under processing conditions are extrapolations from published chemistry in analogous systems (tea, wine, coffee bean roasting, plant fermentation) and have not been specifically tested on coffee leaf material from Thumpassery Estate or anywhere else.
Reservoirs I (Sugar), J (Lipid), and K (Volatile Precursor) are proposed theoretical extensions beyond the Cangeloni 2022 dataset, based on known plant leaf chemistry and documented in the relevant sections as such. They represent testable hypotheses, not established findings.
Where the model speculates, it says so. Where it does not know, it says so. This is intentional — a document that pretends to certainty it does not have is not a useful scientific tool.
KoffyKraft — Thumpassery Estate, Karavaloor, near Punalur, Kollam District, Kerala, India. Arabica Chandragiri variety, grown under rubber shade at approximately 130m altitude. The Citane beverage category is developed and documented by KoffyKraft.
Compiled and editorially structured by Citane / KoffyKraft. Source research belongs to the original scientific authors. Traditional knowledge of coffee leaf use belongs to the communities of origin who have practised it for centuries.
CITANE REACTIVE LANDSCAPE MODEL · VERSION 2.0
KoffyKraft / Thumpassery Estate · Karavaloor · Kollam · Kerala · India
Scientific pivot: Cangeloni et al., Foods 2022, 11, 2495 · CC BY 4.0
Not experimentally validated · A map, not a protocol · Speculative by design