Arctic Tracker / What changed
What this rebuild shows differently
This site shows less than the previous one in three places. Each reduction is a correction, and each is stated below with its numbers.
Reference lists were identical across species. The same class of error was then found in the seizure table, where it was present and unreported. The three entries below follow from that.
How this page was produced curated
Each delta is a named finding. Counts that appear in the deposit manifest are read from it. Counts that describe the previous live site or the source-CSV rejection pass are stated as verified at analysis time; the application role cannot re-query those paths.
No single “Total Quantity Traded” derived
The published figure summed a quantity column whose unit varies row by row, adding kilograms to individual animals. All fifteen species in that figure had a total mixing mass with counts.
Separately, 226,046 records (45.9% of 492,095) sit inside a shipment coordinate reported from both the importer and the exporter. A total spanning both sides counts those shipments twice.
The coordinate is eight fields: year, taxon, term, unit, importer, exporter, purpose and source. It leaves out origin, which the two sides themselves report differently, so fixing it splits apart pairs that are one shipment. Fixing it anyway gives 213,838 records (43.5%); also fixing appendix gives 213,763 (43.4%). The overlap is 43.4 to 45.9% of the corpus depending on the key, and no key in that range supports a combined total.
Trade is now reported per unit dimension, per traded term, and per reporter side. Those axes are not additive.
Illegal trade: retained rows only derived
The source file contains 919 seizure rows. Of those, 225 are kept because the reported taxon matches the species they were filed under. The remaining 694 named a different species and are not attributed here.
225 + 694 = 919. Each retained record shows the taxon name the source reported, so the attribution can be checked rather than trusted.
Quantities in the seizure source are unitless, so no total is offered. Records are the only additive axis.
Finding document: seizure taxon attribution
The previous version of this site showed a live table of 742 seizure rows, produced by a different construction path that did not apply the taxon-match rule above. Comparing that earlier live count to the retained set is a comparison of two pipelines. It is not an account of how many source rows failed the taxon check.
No machine-drafted prose or proposed references withheld
343 narrative sections and their reference bindings exist and are withheld, because no subject expert has reviewed and signed them. The count is disclosed on every affected species page.
Reference lists were identical across species. There are now no reference lists until they are signed: withholding unsigned material is defensible; serving it while calling it repaired is not.