Survey data → LERC and Natural England templates (England)
What you are seeing — the survey is done, and every receiving body wants the same records in a different shape — the county centre has its spreadsheet, Natural England wants its own CSVs, and each names its own columns.
There is no single "records centre format". A submission conforms to one receiver's own template — a Lincolnshire species spreadsheet is not a London one, and Natural England's habitat CSVs are their own three shapes again. Every conversion between them is a chance to lose rows, truncate names, or ship coordinates under the wrong datum, and the receiving body will usually discover none of it.
The damage happens on the way out of the GIS, not at the receiver's door. A shapefile export caps column names at 10 characters, the .prj beside it is the only surviving record of the datum, and a grid-reference column retyped from coordinates — or coordinates retyped from grid references — can drift without anything looking wrong.
So this corridor is measured rather than promised. We ran two Natural England-published datasets — 26,870 rows, both under the Open Government Licence v3 — through 7 receiver targets across 6 receiving templates: a LERC species spreadsheet and shapefile, a second LERC's records sheet, a baseline-habitat shapefile, and Natural England's area, hedgerow and watercourse CSVs. Every output arrived with row parity — zero rows lost — and the shapefile outputs were checked structurally: record counts in the .dbf header, byte sizes, and the .prj byte-identical to the receiver's spec. The build refuses to publish this page if any of that stops being true of the artifact behind it.
The same run audits what publishers themselves ship: of 23,570 OS grid references compared against their own geometry, 99.996% agreed — and 1 reference was impossible outright, a square lettered SI, which no British National Grid square carries. A drifted grid reference looks completely normal; only computing it back from the coordinates can see it.
One honest limit: those runs are on published open data, end to end, because that is what we can show receipts for. Your file goes through the same corridor — columns confirmed from values, refusals coded and named — and the receiver templates on offer are gated by your file's own coordinates, so a survey outside a receiver's jurisdiction is never offered its sheet.
.prj and .cpg with the shapefile. The .prj is the only record of what your coordinates mean, and the two failure modes read very differently — metres fail loudly, a wrong datum fails silently.Upload the table or the zipped shapefile. Columns are confirmed from their values, refusals carry coded reasons, and the receiver targets offered are the ones whose jurisdiction your file's own coordinates sit inside — measured the same way as the 26,870-row corridor above. Free until 1 September 2026, 00:00 UTC · no account · your rows are discarded after the check.
Check the file you have →Measured on a file GBIF had already accepted and served — their validator asks will this load, and it had. Every figure is read from ops/data.json, never typed here. We issue no rating and no seal — each finding names the rule it applied.