Eight Japan reports, one missing distance, and the method that refused to invent it.
There are now eight Japan Dairy Field reports.
Hokkaido. Tohoku. Kanto. Chubu. Kinki. Chugoku. Shikoku. Kyushu and Okinawa.
Together they contain all 47 Japanese prefectures.
They show raw milk, dairy cattle, holdings, regional shares, internal distribution and source notes.
They do not show the number the project was originally built to find.
They do not say how close people in Japan live to the dairy field.
That number is unavailable.
This is not a missing page.
It is part of the result.
With the approved public sources and current method, a defensible Japan proximity estimate is not yet available.
A Field Made From Evidence
A dairy field is not a photograph of every farm.
It is a modeled geographic layer.
It takes an official measure of dairy activity, places that activity inside an audited geography, and asks what the resulting distribution can tell us.
In the European reports, the next step is to compare that dairy layer with where people live. Population cells are measured against a continuous regional dairy surface. The output is a proximity reading: not the distance to a named farm, but the population-weighted relationship between residents and the country's modeled dairy geography.
The first public series contains 27 EU country reports.
They are Level 1 regional estimates.
They do not map processors. They do not map cheese identity. They do not score animal welfare, climate performance or agricultural virtue. They do not know the location of every herd.
They make one spatial relationship more legible.
That is useful.
It is also limited.
Japan Is Not Europe With Different Borders
The Japan edition could not be produced by changing the country code and running the European template again.
The source geography is different. The reporting geography is different. The available statistics are different. The hidden parts of the data are different.
The source layer begins with 47 prefectures.
For public reading, those prefectures are grouped into eight conventional regions. JP8 is a Hedegreen Research reporting frame. It is not an official MAFF agricultural-region classification. Mie is placed in Kinki. Okinawa is grouped with Kyushu. Those choices remain visible because an editorial geography should not disguise itself as an official one.
The eight reports keep the prefectures underneath.
Nothing disappears merely because the reader is given a larger frame.
The Ground Had Moved
The first obstacle was not a missing dairy table.
It was the ground underneath the table.
The agricultural-settlement boundaries used in the research declare JGD2000. The accepted population grid uses JGD2011.
Those names are not interchangeable labels for approximately the same map.
The Geospatial Information Authority of Japan introduced JGD2011 after the wide-area deformation caused by the 2011 Tohoku earthquake. Across the affected part of northern Honshu, older coordinates must be corrected with an official displacement grid before they can be compared with newer geography.
The report factory reproduced the official transformation locally and matched four GSI control coordinates to sub-millimetre agreement.
For the 20 affected prefectures, the method uses the official PatchJGD parameter grid. A point is transformed only when all four surrounding official grid nodes exist. For the other 27 prefectures, the method uses the published zero-translation operation.
The system does not accept an automatic identity result inside the affected area.
Japan had moved.
The model was not allowed to pretend that it had not.
Most of the settlement geometry could be corrected. Not all of it.
Some official grid gaps cover places where the post-earthquake deformation was too complex to represent reliably. The first full audit found 327 settlement features touching a gap. A narrower representative-point method reduced the unresolved set to 217 features. Most of those carried source-defined zero dairy heads. Thirty-four carried suppressed, unknown values.
That was a real boundary.
It was not the largest one.
Later sensitivity work showed that those 34 grid-blocked rows represented less than 0.024 percent of the included raw milk under either tested allocation. The wider suppression problem was much larger. Two allocation methods compatible with the source data moved the equivalent of 19.27 percent of included dairy heads between settlements.
The earthquake had made coordinate correction unavoidable.
Suppression made the final field unstable.
Together they closed the route to a defensible fine-grained proximity result.
They did not close the route to a regional report.
What Can Be Said
Japan's official statistics support several strong regional readings.
The reports show:
- raw-milk production;
- dairy-cattle holdings;
- dairy-cattle head count;
- share of national raw-milk production;
- cattle per holding.
The raw-milk values reconcile across all 47 prefectures, the eight regions and the national total.
The cattle and holdings totals carry small source-rounding differences. Those differences are shown, not redistributed until the table looks perfect.
Each reading remains separate.
A region with more milk is not therefore more efficient.
A region with more cattle is not therefore better.
A region with more cattle per holding is not therefore more sustainable.
There is no composite winner because the data does not contain one.
What Cannot Yet Be Said
The difficult part begins below the prefecture.
Japan publishes detailed agricultural-settlement material, but many local values are suppressed. Suppression protects information that should not be reconstructed carelessly. It also leaves a spatial question.
The larger total may be known.
Its exact distribution inside the smaller geography may not be.
A model can allocate that hidden amount in several plausible ways. It can spread it across eligible settlements. It can follow one hierarchy or another. It can smooth the field. It can choose representative points.
Each choice can still reconcile to the same total.
That does not make the resulting maps equivalent.
In the Japan research, two allocation approaches compatible with the source data produced materially different spatial fields.
Arithmetic could prove that the totals balanced.
It could not prove which field was true.
So the project stopped.
No proximity value was calculated.
No proximity order was published.
The missing distance was not replaced with zero.
Unavailable is not zero.
Why Population Cannot Repair Production
There is an easy-looking shortcut.
Place more of the hidden dairy activity where more people live.
That would make a smoother map. It would also make the method partly circular.
The project would first use population to guess where dairy activity is, then use population again to measure how close residents are to that guessed dairy activity.
The answer would begin inside the assumption.
That route was rejected.
A Report Factory, Not a Dashboard
The public output is a set of fixed PDF records.
That choice is deliberate.
A dashboard usually makes the current state convenient. A record system must also preserve what the current state replaced.
In Camelot, a final PDF is not silently overwritten. A correction becomes a new issue or version. The old record remains part of the history. Draft and final objects can be distinguished. A reader can see that a method changed instead of being asked to trust that the newest screen was always correct.
The eight Japan reports therefore belong to a larger method:
source, calculation, review, fixed record, correction.
The PDF is not the end of the research.
It is the point where one reading becomes stable enough to cite.
One Series, Different Contracts
The European and Japanese reports belong to the same Dairy Field series, but they do not currently make identical claims.
The 2026 EU release asks a proximity question across 27 countries using a harmonised regional source ladder.
The Japan release publishes eight regional statistical readings while keeping proximity unavailable.
A national Japan overview was planned as part of the product family. It has not been published.
That difference matters.
A series becomes weaker when every edition is forced into the same template.
It becomes stronger when the common question survives while the evidence contract changes visibly.
The Next European Field
The 2026 EU reports are frozen as an immutable reader release.
That does not mean the method is finished.
Later audits found country cases that require different care: missing regional rows, fallback metrics, island scope, incomplete harmonised production data, and national sources that may be more detailed but less comparable across the Union.
The next EU direction is therefore comparability first.
A better national dataset is not automatically a better European ranking input. It must also have the right definition, geography, year, licence, privacy boundary and reproducible parser.
A stronger layer for one country may be a useful diagnostic. It should not silently reorder Europe while the other countries remain on a harmonised regional baseline.
The next public EU comparison should rebuild all 27 countries as one set under one approved source ladder. Country improvements should remain visible without changing the rules for everyone else.
Old PDFs should remain.
New conclusions should receive new versions.
Correction
The phrase “dairy field” can sound more physical than it is.
There is no green layer waiting in nature to be downloaded.
The field is a model produced from a metric, a geography, a population layer and a set of decisions about missing or suppressed evidence.
That does not make it fictional.
It makes the source contract part of the result.
The map is useful when the reader can still see the contract.
It becomes misleading when the model is mistaken for a farm.
Sources and Method
The Japan readings use official aggregate statistics and public geography. Their core inputs are:
- MAFF/e-Stat 2024 Milk and Dairy Products Statistics, table 2 and 2025 Statistics on Livestock, table 1-1;
- MAFF's 2020 agricultural-settlement boundary data and the Statistics Bureau's 2020 JGD2011 population grid;
- GSI's JGD2011 explanation and official PatchJGD parameter archive.
The European series uses Eurostat's regional animal-production metadata, GISCO NUTS 2024 geography and GISCO population grids.
Hedegreen Research transformed, audited, reconciled and grouped these sources. JP8 remains an editorial reporting frame. Suppressed values were not published or reconstructed as observed local facts.
The Missing Number
The Japan reports are not complete in the sense that every possible metric is present.
They are complete in a more important sense.
They say what the evidence supports.
They preserve what it does not.
A blank can be a failure to finish.
It can also be the place where a system refused to lie.