Ekodama

Integrity//How we work

Scientific integrity

The standards this practice holds itself to when it publishes and advises, written down so you can check the work, and check us against them.

Why // This page

Standards worth writing down

Most of what Ekodama publishes appears outside peer review, and says so plainly: the articles, the catalogue references, the reports. Working outside the journals does not lower the bar; it removes the referee. The discipline has to come from the practice itself.

This page states that discipline. Each commitment names the mechanism behind it, so a sceptical reader can see how the work is made and where to press on it.

None of it is exceptional. It is the ordinary deontology of science, carried into consulting work: name your sources, state your limits, correct your errors. That holds for everything published under the Ekodama name.

Commitments//Five

What we hold ourselves to

  1. 01/05

    Source everything

    Claims of fact rest on named sources: peer-reviewed studies where they exist; public research, institutional reports, and open datasets where those are the best available. A reader should never have to take a number on trust.

    In practice

    • Articles cite numbered, DOI-linked references, so a claim can be followed to the publisher's version of record.
    • Catalogue titles trace every number and equation to the paper it came from.

  2. 02/05

    Say what we don't know

    Honest work separates what the data shows from what it implies. Uncertainty is stated, not smoothed over; where the evidence is thin or contested, the text says so.

    In practice

    • Detection limits, error budgets, and model assumptions are treated as subject matter. Several of our articles exist to work through them.
    • Published references state their limits alongside their results.

  3. 03/05

    Stay independent

    Clients pay for the analysis, not the answer. We hold no stake in the credits or projects we assess, and no fee of ours depends on how a result comes out. If an interest could colour a conclusion, we disclose it.

    In practice

    • No carbon credits are sold, brokered, or marketed by this practice.
    • We do sell reference publications; they are openly priced, and no engagement depends on buying them. No vendor sponsors this practice.

  4. 04/05

    Verify before publishing

    References are checked against the published record, including whether a cited paper has since been corrected or retracted. Where that checking has reached so far is spelled out below.

    In practice

    • Cited DOIs are resolved against their Crossref records, correction and retraction notices included.
    • We are building a claim-by-claim evidence ledger, article by article, comparing each sentence against its source text.

  5. 05/05

    Correct in the open

    Substantive errors get fixed in place once found, then logged with the date and what changed: wrong numbers, wrong citations, claims that overreach. Typos are just fixed; anything that could have misled a reader gets a line in the record.

    In practice

    • The corrections log lists each substantive fix with its date, the affected article, and what changed.
    • Anyone can flag a suspected error; the log ends with the address for exactly that.
    See the corrections log

Method//Verification

How verification works

The chain a published reference passes through, from citation to standing record.

  1. 01/04

    Cite at the source

    Every reference carries an identifier, a DOI wherever one exists, pointing at the publisher's version of record rather than a summary of it.

  2. 02/04

    Check the record

    Each DOI is resolved against its Crossref record, including the update notices that flag corrections, errata, and retractions. A citation to a retracted paper is a defect; this step is where it gets caught.

  3. 03/04

    Trace the claim

    Claims are compared, sentence by sentence, against the text of their sources and given a verdict: supported, or source unavailable. Where we could not obtain a source, the claim says so rather than getting the benefit of the doubt.

  4. 04/04

    Version and revise

    Catalogue references carry dated editions and are revised as the literature moves, so you always know which version you are citing and when it last changed.

Where this stands

Sourced citations and versioned editions are standing practice. The registry check and the claim ledger are newer: five articles have been through them in full so far.

Adopted August 2026, extended article by article. This note is updated as coverage grows.

Transparency//AI

What AI touches, and what it does not

This chapter covers everything published under the Ekodama name: this website, the insight articles, the catalogue references. It also covers how we use AI tools inside client engagements. First written in August 2026, and reviewed at least once a year.

The legal frame is Regulation (EU) 2024/1689, the EU AI Act, whose transparency obligations (Article 50) have applied since 2 August 2026. Those rules exempt AI-assisted text that a person with relevant expertise has substantively reviewed, under named editorial responsibility. Everything Ekodama publishes goes through that review, so none of it legally needs an AI label. We disclose anyway, because how the work is made is part of what you are assessing.

Used for

  • Literature and source searches, alongside the usual scholarly databases.
  • Help with structuring and drafting articles, reports, and site copy.
  • Analysis support: writing code, checking data handling, stress-testing calculations.
  • Translating the site into the languages we do not write natively.
  • Generating some of the imagery. Every page carrying a generated image says so at the foot, and generated figures inside articles are credited at the figure.
  • Building and maintaining this website.

Never used for

  • Publishing without review. A person reads every text before it goes out.
  • Scientific judgment. Method choices, interpretations, and recommendations in client work are the founder's own.
  • Passing generated images off as field photography. Where an image was generated rather than taken, the page says so.
  • Training models on client material. Nothing a client sends us feeds a model.

Review // Responsibility

A person answers for every page

Dr. Thomas Fungenzi reads every Ekodama text before it goes out: the substance of the claims, the numbers, and the references, which he traces back to their original sources. Editorial responsibility for everything under the Ekodama name is his.

That review is what the Article 50 exemption rests on. The editorial note at the foot of each article goes beyond it: an article that used AI somewhere along the way says so on its own page, instead of leaving it to the small print here.

Checking is a practice, not a guarantee. Errors still slip through, and we correct them once we find them, then log them in the corrections record below.

Client work // Confidentiality

Client data and independence

Client documents and data stay under the confidentiality terms of each engagement. AI tools touch client material only where an engagement's terms allow it, and then under settings that exclude use for model training. Identifying or sensitive material goes in only with the client's explicit agreement; otherwise it stays out, and the work uses anonymised or non-identifying excerpts instead. Where confidentiality and convenience pull against each other, confidentiality wins.

Ekodama has no sponsorship, affiliation, or commercial relationship with any AI vendor. We pay for the tools like any other instrument, and no vendor had a word in this statement. AI changes how a draft comes together. It does not change whose judgment you are buying.

Last reviewed

August 2026

Reviewed at least once a year, and whenever our practices or the law change materially.

Corrections//The record

What we got wrong

Substantive errors in published work are corrected in place and recorded here, dated, newest first. The bar for an entry: anything that could have misled a careful reader.

  1. 2026-10-04

    Correction

    The Lab page (then called Ecosystem) described the Map of Soil Science as covering about 85 fields and 300 sub-topics. Those figures were early estimates that matched nothing in the map's data. The published map has 10 domains, 150 entries, and 601 sub-branches, and the page now says so. The page also listed Solum as live. Solum is in a closed beta, and is now labelled as a private beta.

  2. 2026-09-07

    Correction

    How Large Should Your Soil Carbon Sampling Campaign Be?

    A verification pass read every retrievable source cited in this article and checked all twenty-seven citations against them. Eighteen of the twenty-three checkable claims needed work. The economics of change detection were stated backwards: the cited paper finds measure-and-remeasure viable above roughly five thousand fields, not across ten to thirty farms. The claim that compositing destroys the variance structure needed for inference is rejected by name in the textbook cited for it. A stratification finding was credited to a paper that never tested stratification, and the saving it reported was 17 percent on average rather than 15 to 35. Nine further passages claimed more than their sources support, among them the GHG Protocol's data requirement, which permits calibrated models alongside direct measurement. Three reference titles had been truncated or paraphrased and have been restored. Four claims rest on two sources that could not be obtained, and are now marked as unverified rather than presented as checked.

  3. 2026-09-07

    Correction

    Is Soil Inorganic Carbon Critical to Include in Your Soil Carbon Accounting?

    A source-by-source reconciliation read the twenty sources held for this article and corrected what did not survive the check. The clearest error reversed a ranking: lime-induced CO₂ release was called the second-largest pathway returning soil inorganic carbon to the atmosphere, where the cited review makes it the largest, at 273 against 7.5 Tg C/yr. Three claims about market and regulatory frameworks were absent from the documents cited for them: an ICVCM endorsement of Verra's VM0042, a set of ICVCM approval and rejection counts, and an eligibility rule barring projects on calcareous land. The GHG Protocol definition of the soil carbon pool had acquired a list of minerals the standard does not contain, including gypsum, which carries no carbon at all. A Chinese cropland ratio was divided into the wrong quantity, three figures that apply only to the upper 40 cm lost that qualifier, and a concentration range for calcareous soils had no source and contradicted the stock figure in its own sentence. Smaller corrections reached a field transect that had been called a synthesis, an analytical method presented as scheduled to replace another, and four reference entries with the wrong volume, issue or year. Two claims are now marked unverified because their sources could not be obtained.

  4. 2026-09-05

    Correction

    France on the Edge of a Climate Precipice: Droughts, Heatwaves, and the Collapse Thresholds of Ecosystems

    A third pass applied the remaining corrections the source-by-source review had queued: thirty-three passages that were directionally right but claimed more than the source supports. The largest concerns the article's own argument. Its section on the 2030-2040 window described transitions at +2.7 °C, a level the TRACC trajectory places at 2050; the section is now keyed to the +2 °C France reaches around 2030, and the water finding it cited is explicitly dated to 2050. The per-species forest thresholds are no longer presented as documented, because no cited source publishes one, and the article now says so. Several figures gained the scenario conditions or horizons their sources attach to them, three station temperatures and a Paris record were deleted because no cited source contains them, and a claim about heatwave days, 50 °C peaks and generalised fire risk was removed for the same reason. The macroeconomic modelling is now correctly attributed to ADEME economists and dated to 2100.

  5. 2026-09-04

    Correction

    France on the Edge of a Climate Precipice: Droughts, Heatwaves, and the Collapse Thresholds of Ecosystems

    A follow-up pass checked every reference entry against the page it resolves to, and corrected six that misdescribed their source. The paper on the macroeconomic cost of climate damage was credited to INSEE; it is by two ADEME economists using ADEME's ThreeME model, published in an INSEE series, and its results are stated for 2100, not the 2040s. The Haut Conseil pour le Climat report was dated 2023 from its filename rather than 2024 from its cover, and carried a title the report does not use. Two study titles had been shortened in ways that changed their scope: one drops "in China", the other drops "heat and" from a paper about heat and drought together. The G20 Climate Risk Atlas entry now names its publisher and date and distinguishes its 2050 figures from its 2100 ones. Three links had rotted and now point to pages that still carry the cited material.

  6. 2026-08-27

    Correction

    France on the Edge of a Climate Precipice: Droughts, Heatwaves, and the Collapse Thresholds of Ecosystems

    A verification pass read all forty-five sources cited in this article against the live pages and corrected what did not survive the check. The most consequential: the article placed +2 °C of global warming around 2030, where the TRACC trajectory it cites places it at 2050; the JRC PESETA IV projections for France were read off the wrong macro-region, turning projected wheat gains into losses of up to 49%; the beech study was attributed to the wrong journal; a quotation from the Haut Conseil pour le Climat spliced two separate sentences into one that appears in neither the French report nor its English summary; INRAE's heatwave-and-drought frequency multipliers were attached to harvest years instead; and a drought-restriction timeline was dated a year late. Three citation markers pointed at sources that do not contain the claims beside them and were removed. The per-ecosystem collapse thresholds are no longer presented as documented findings, because no cited source reports them in that form.

  7. 2026-08-10

    Correction

    How Do You Model Soil Carbon Change Over Time?

    A verification pass read all nineteen sources cited in this article and corrected what did not survive the check: the AMG model's origin story was stated backwards (first tested on Argentine pampas trials, rebuilt on sixty French field trials at twenty sites), a conclusion attributed to a 26-model ensemble study is not drawn by that paper, a quantified initialization claim was absent from its cited source, and three further sentences claimed more than their sources support. The GHG Protocol citation now points to the final 2026 guidance.

  8. 2026-08-04

    Correction

    Representative of What? When a New Sample Falls Outside Your Model’s Domain

    The interactive domain explorer scaled its dissimilarity index by the wrong denominator, so every value read about three times too large, and the verdict shown beside it came from a different test than its label implied. Both were corrected, along with three passages that promised more than the article delivered.

Spotted something?

If a number, a citation, or a claim looks wrong, say so. Reports go to the founder and get checked against the source; where the reader is right, the fix lands here.

Write to contact@ekodama.pro
Ekodama // Field practiceRef 2026-10

Work with us

Hold us to it

Every engagement runs under the standards on this page, and the scientific judgment behind it stays with the founder. If that is what your project needs, the conversation starts here.