Known Limitations and Assumptions
2022 is a real gap, not a bug
ANP has no published files at all for 2022 — confirmed by exhaustively probing every month and both filename-separator conventions ANP has used, not assumed from a missing download. A hole in the chart for 2022 reflects a gap in ANP's own publication record, not a failure of this pipeline.
Two different granularities, one aggregation trap
The source publishes every variable at shipper/contract granularity, but not every variable is a genuine per-shipper split:
- Volume Solicitado / Programado / Realizado and Allocation (%) are real per-shipper figures and are summed across shippers/contracts to get a point's total — this is a correct sum.
- Average Pressure and every pipeline-system ledger variable (GUS,
unaccounted-for gas, losses, imbalance, linepack) are a single physical
or system-wide reading, broadcast identically to every shipper row active
that day — confirmed by inspecting raw rows directly (the same value
repeated across every shipper reporting for a pipeline on a given date).
Summing these would multiply the true value by however many shippers
happened to be active that day. Both are aggregated with
median()instead ofsum()specifically to avoid that.
Shipper- and contract-level detail exists in the source but is deliberately collapsed at build time to keep the dashboard's payload a reasonable size — POC Contracts already covers shipper/contract-level detail for held capacity; this dashboard is about physical flow totals. Average Pressure is computed and stored in the pipeline's own data store but left out of the dashboard's embedded payload entirely (it was roughly 30% of the payload's size on its own).
TSO overlay takes priority over ANP when both exist
TAG, TBG, and NTS each publish their own programmed/actual meter volumes, often sooner than ANP. Where the same point and date exist in both sources, the TSO's own figure is used in preference to ANP's. Ledger, pressure, allocation, and requested-volume series stay ANP-only (the TSOs don't publish those in a comparable form).
Data lag and revisions
ANP typically publishes files with a lag of several weeks after month-end, and revises recently-published months in place — a figure for last month quoted today may not match the same figure quoted next month, once ANP's own revision settles.