Hydroscape-Group

1. What the figure is

Hydro-Vericount returns a minimum confirmed population for an area over a period, derived from reports made by independent observers who may be looking at the same birds.

Anyone quoting the figure should quote it with that qualification. Hydroscape does so wherever it appears.

2. The method in one sentence

Divide the area into fixed 2 km cells, take the highest single count recorded in each cell in any calendar month, add those together for that month, and report the best month in the period.

3. The three stages

The order is the method.

Stage 1. Deduplicate within cell and window

Within one cell in one calendar month, take the single highest report and discard the rest. Six anglers watching one flock of 40 yields 40, not 240. This is the stage that removes observer duplication, which is the dominant source of inflation in voluntary reporting.

Stage 2. Sum across cells

Add the cell figures for that month. Separate waters are treated as separate birds.

Stage 3. Peak across windows

Across all months in the period, take the highest monthly total. That figure is the result.

4. Grid definition

The grid is anchored at a fixed origin so the same water always falls in the same cell regardless of which subset of data is passed in. Without a fixed origin, cell boundaries shift with the dataset and results are not reproducible between runs.

Longitude scaling is taken at a fixed reference latitude rather than each point's own latitude. Using the point's latitude made the grid a fan: the same longitude produced column 275 at 50°N and column 245 at 55°N, roughly 60 km of drift across Britain, so cells were not a fixed partition of space and a catchment spanning latitudes received skewed cells. With a fixed reference, cells are true 2 km at 54°N and vary by about 10% at the extremes of the British Isles, which is inside the resolution the figure claims.

Dates are handled in UTC. Reading local calendar components made the figure depend on the viewer's clock: the same two reports returned 75 in the UK and 40 in New Zealand, because a day boundary moved. An evidence figure cannot change with who opens the page.

5. Why a calendar month

A daily window is more conservative and is the right choice on a single water, where the boundary is one cell and the answer is the peak count either way.

Nationally, and on any catchment, a day only sums the waters that happened to be reported that day. Anglers do not report every water every day, so a daily figure measures reporting coincidence rather than population. Measured against 12,157 reports on 13 September 2026, the busiest single day summed 278 cells for 4,692 birds, while 2,071 cells carried reports across that month. The monthly window summed those cells to 19,961.

A calendar month is the unit the Wetland Bird Survey uses for the same reason. WeBS, run by BTO, RSPB and JNCC, has counted non-breeding waterbirds including cormorants since 1993 using predefined count sectors, monthly counts and an annual maximum. Hydro-Vericount v2 follows that structure: fixed area, monthly window, peak reported.

The cost is that a flock moving between waters within a month can be counted at each. WeBS accepts this and so do we. It pushes the figure up, which is stated in the limitations below.

6. Why 2 km

Cell size is a resolution choice and no single value is right for every water. Two cases pull in opposite directions. Both were measured on 8 September 2026.

2 km is the default because the figure is published as a minimum. Splitting one flock across cells inflates it, which would make the claim false. Merging two waters deflates it, which keeps the claim true. Where the method has to be wrong, it should be wrong downwards.

The consequence, which belongs wherever the figure appears: a fishery with several small waters close together will read low.

7. What v1 did and why it was replaced

Version 1 used single-linkage spatial clustering. Any two reports within 5 km joined, and joins were transitive. Measured against constructed cases it returned 34 for a 120 km river holding 885 reported birds, because every stretch was within 5 km of the next and the whole river chained into a single cluster.

It also took the peak across space rather than time, so two lakes 4 km apart each holding 40 returned 40. And it never read a date, so a whole season collapsed to its best day with no way to tell that was what had happened.

v2 fixes the area and takes the peak across time. Publishing this is deliberate: a method that has been wrong should say so, and say how.

8. Worked example, live data

Run against the full Hydroscape cormorant dataset on 18 September 2026.

The raw cumulative figure of 71,250 and the Vericount figure of 26,363 are both true statements about the same dataset. The first counts reports, the second counts birds confirmed present at one time. Quoting the first as a population figure would be wrong, which is why the platform publishes both side by side.

9. Undated reports

A report without a usable date cannot enter the windowing stage. Such reports are deduplicated to the highest value per cell and held in a separate bucket, reported alongside the result as undatedContribution. They are never silently dropped and never silently added. On the 18 September 2026 run this bucket was empty.

10. Limitations

All of these should be stated wherever the figure is quoted.

Cell size is the resolution. Two waters inside one cell cannot be separated and the larger count is what carries.

A flock moving between cells within the window is counted in each. Cormorants commonly forage within 10 km of a roost and can range 20 to 25 km, so this is real and it pushes the figure up.

Counts are observer estimates made in the field. No observer training, no detection probability correction and no confidence interval are implied.

Reporting effort is not uniform. Cells with no reports contribute nothing, so the figure reflects where people fish and report, not where birds are. It is a floor for the reported area, not for the country.

The method does not distinguish age, sex or subspecies, and does not separate Phalacrocorax carbo carbo from P. c. sinensis.

11. Reproducibility

The implementation is a single self-contained JavaScript module with no dependencies, served publicly at hydro_vericount.js. It runs in a browser or under Node. The same input returns the same output on any machine, because the grid origin, the reference latitude and the date basis are all fixed constants rather than derived from the data.

Every result object carries a method block recording version, cell size, window, grid origin, reference latitude and date basis, so any figure can be traced to the parameters that produced it.

Anyone wishing to verify the figure against their own data, or to run the method with different parameters, is welcome to. Ask and we will supply the dataset in whatever format is useful.

12. Citation