INTEROP.md

★ N.I.C. ★

INTEROP — feeding the world's existing systems

Design-stage concept — nothing built. The figures are verified against the parts' sheets and the parts before a board is made.

The claim, up front: NIC does not compete with the global scientific networks — it feeds them. Most of the world's geophysical infrastructure is not a closed club; it is a set of aggregators that live on incoming data (SuperMAG, Madrigal, MADIS, the IOC sea-level facility, OpenAQ…), plus a few strict observatory clubs (INTERMAGNET, IGS core) whose job — absolute reference — we deliberately leave to them. The path is proven: Raspberry Shake got an FDSN network code (AM), its ~2k community stations flow into EarthScope, and the ISC cites their picks. NIC walks the same path with ten quantities instead of one.

Why it is cheap for us: the architecture already assumes it. The HMC archive is raw and self-describing (derive-from-raw, archive/HMC.md), so an export format is just another reader — no change to the station, the bus, or the protocol. How each one is written is archive/EXPORTERS.md. And the trade goes both ways: we give the aggregators the density they don't have (see ../gaia/ — India, Turkey, Indonesia…), and they give us credibility and calibration for free (the variometer-vs-observatory regression in ../gauss/CONSTRUCTION.md is exactly this, and FDSN locations validate our picks).

The map — system by system

domain	system	takes us?	format / channel	from the station
seismology	FDSN / EarthScope, SeisComP, ObsPy	yes — Raspberry Shake precedent	miniSEED + StationXML	the miniSEED exporter on the HMC reader (Steim-2 + StationXML); the network code is a registration (the gate below)
	ISC (bulletin)	yes — takes picks/data from registered networks	via FDSN	paperwork only
geomagnetics	INTERMAGNET	no — needs an absolute observatory: weekly DI-flux measurements by a human on a fixed pier (see below)	—	out of scope by design
	SuperMAG	yes — aggregates ~600 variometers, strips baselines itself	IAGA-2002, Data Type: variation	the IAGA-2002 exporter on the HMC reader (calibrated nT out of the archive)
GNSS: TEC	Madrigal (MIT Haystack)	yes — takes any dual-frequency RINEX	RINEX	the RINEX exporter on the HMC reader, from Sputnik's Tier B
GNSS: geodesy	Nevada Geodetic Lab, EarthScope	yes — NGL processes any submitted RINEX (20k+ stations) → PPP positions, crustal motion	RINEX upload	the same RINEX export — pairs beautifully with Quake
	CORS / IGS densification	yes — the UM980 is survey/geodetic class	RINEX + site info	needs a decent antenna
	IGS core	per-station — needs a calibrated antenna (ANTEX) + monument + site log; the receiver is NOT the blocker	RINEX	achievable where a station earns it
tsunami / sea level	IOC Sea Level Station Monitoring Facility (VLIZ)	yes — aggregates ~1000 radar gauges worldwide	station registration + HTTP push	the IOC exporter on the server; Pascal is a bottom-pressure gauge, a class the facility also takes
	magnetometer tsunami array	research-grade → science collaborations, not operations	IAGA-2002 / raw	doctrine in ../gauss/ARRAY.md
weather	CWOP → NOAA MADIS	yes — citizen stations feed forecast assimilation	CWOP protocol (trivial)	the CWOP exporter on the server
	WIS2 / GBON — a national service's surface network	per national service	BUFR 307096 hourly, 307092 ten-minute	the server encodes a mode-B Palatine's HOUR and TEN rows (../palatine/WMO.md)
air quality	sensor.community, OpenAQ	yes, takes anything	HTTP push	the air-quality exporter on the server, from Chinook's units
lightning	Blitzortung	it IS a community network	TOA receiver protocol	Tesla could join as a receiver, by agreement on their protocol
cosmic ray	NMDB	case-by-case	—	science collaboration

Where calibration lives — the one rule to keep straight

Two exporters, two deliberate conventions, one raw archive:

• miniSEED carries raw counts; calibration belongs in StationXML metadata (that is seismology's own split — data vs. response). Both are written by the miniSEED exporter.
• IAGA-2002 carries physical nT; the exporter applies the scale of the field's layout in the HMC header (physical = (raw + add) · mantissa · 10^exp10, mantissa 0 ≡ 1) on the way out.

The archive itself never stores corrected values (derive-from-raw, archive/HMC.md); an export is a view, re-derivable forever with a better model.

No field calibration — precision by selection, zero by reference

The station stores no calibration curves and no instrument-response tables; the one reference it ever establishes is zero. The sensors are chosen so a field or laboratory calibration buys nothing:

• A factory-calibrated digital sensor is its own reference. Calibrating it in the field needs a reference better than the sensor, which a station does not carry. Where a known local offset appears and the unit exposes a writable offset register, the offset is written over the bus in place; where the element itself has degraded, the unit is replaced.
• The rain gauge is precise by construction, not by a fitted curve: a ratiometric 24-bit converter and a cell worked at a few percent of its span. What it measures is the change from a zero re-established at every drain; acceptance is a go/no-go check with known weights. A damaged cell is replaced — its error is path-dependent, and no static table corrects it.
• The seismic MEMS are flat across the band: one factory sensitivity, no poles or zeros to fit.

So the StationXML response is flat — InstrumentSensitivity only — which is correct for this sensor set and is what the miniSEED exporter emits. A sensor that one day needs a response is given a write-once calibration block per unit, carried to the head as one out-of-band packet and into the HMC header's RESPONSE section, from which StationXML takes poles and zeros.

Why INTERMAGNET is a "no" — and why that costs nothing

An INTERMAGNET observatory guarantees absolute field values for decades: a non-magnetic pier in its own hut (arcsecond-stable orientation, an azimuth reference mark), a DI-flux (fluxgate on a non-magnetic theodolite) that a trained human works weekly to measure true declination/inclination, a proton/Overhauser magnetometer for absolute intensity, and the resulting baseline that anchors the continuous record to single-nT/year stability. That is a building, an instrument worth thousands, and a decade of weekly ritual — the exact opposite of an unattended potted sonde.

And it does not matter: everything the network hunts — storms, pulsations, SSC, tsunami, dB/dt, GIC — is a variation, and the absolute baseline cancels in every variation product. The world has run on a handful of absolute observatories + a sea of variometers for 150 years. We are the sea; SuperMAG is where the sea reports. (The variometer doctrine ①–④: ../gauss/CONSTRUCTION.md, Calibration without rotating the site.)

Registrations — and the gate on them

The registrations and channels (FDSN network code · SuperMAG station registration · IOC SLSMF registration · CWOP ID + push · NGL / Madrigal upload jobs) are paperwork + the server's exporters, no station change — they live on the server, never on the node.

The gate: no registrations until the hardware is validated. A network is registered when it produces real data. The exporters exist so the door is ready; it opens the day a real station has run and its output has been checked. The station never changes for interop — the archive is raw, and the exporters are readers.