Choosing a sensor

Summary

Every H2 sensor measures a proxy (heat of combustion, thermal conductivity, a surface reaction on palladium, a faint Raman line; see Why hydrogen is hard to sense), and each proxy fails somewhere that matters. The exploration problem spans eleven decades, from the 0.53 ppm atmospheric background to the 4 vol% (40,000 ppm) LEL, and no single technology covers it. Selection means picking the device whose errors you can tolerate for a specific job, running two or three complementary instruments, and logging enough covariates to tell a real anomaly from an artifact.

The firm facts:

• Electrochemical cells are the ppm-capable field workhorse; they drift, cross-talk with CO, and live 1–3 years.
• GC-RGD/PDHID is the only class that resolves the atmospheric background; slow and lab-bound.
• Nothing is field-proven downhole. The one plausible route is passive Pd-coated fiber, unvalidated above room conditions.
• Mud-gas logging screens the whole wellbore while drilling; the legacy FID base is H2-blind and raw readings are artifact-dominated.
• Mass spectrometry is the reference for composition and isotopes and never leaves the surface.

Context numbers used throughout

Atmospheric H2 background ≈ 0.53 ppm (531 ± 6 ppb). Exploration anomalies of interest span ≈ 1 ppm to percent level. LEL = 4 vol% = 40,000 ppm; 10% LEL (4,000 ppm) is the usual portable-monitor alarm point.

Start from the use case

Buying an instrument for its headline LOD, then discovering it was designed for a different regime, is the standard procurement failure. Pick the scenario first.

(a) ppm-level soil-gas survey (roving)

The target is high-tens-to-hundreds-of-ppm anomalies against a noisy few-ppm floor, tens of readings per day, from a probe at ~1 m depth. That rules out capital spectroscopy (too slow, too heavy) and catalytic/TC families (LOD hundreds to thousands of ppm). Amperometric electrochemistry is the only mature, portable, ppm-capable option; metal-oxide is a cheap densifying layer.

Hardware:

• Electrochemical multi-gas handheld (primary). Electrochemical cells at 0.5–2 ppm LOD, ~1 ppm field precision, seconds-to-90 s response. Instruments named in the published natural-H2 survey literature: Geotech/QED GA5000 (0–1,000 ppm H2 cell, dual-wavelength IR CH4/CO2), Industrial Scientific iBrid MX6 (0–2,000 ppm), Dräger X-am 8000 (safety class). OEM cells: Alphasense H2-AF/H2-B4, Membrapor H2/M-1000, Winsen ME3-H2.
• MOX chemiresistor array (optional pre-screen). MOX sensors (Figaro TGS2600/TGS821, Bosch BME688) cost $2–50 and reach sub-ppm in the lab, but are non-selective and humidity-dominated. Use them to decide where to point the electrochemical instrument, never as the quantitative record.
• Portable GC for confirmation (scenario e). A borderline reading is confirmed only when a GC measures it.

Log alongside: soil temperature at probe depth, volumetric soil moisture, barometric pressure and its rate of change, gas-line RH (Environmental covariance); a co-measured proxy gas (CO2, CH4, ideally He) for interferents and a He/H2 ratio; purge volume (3–5 dead-volume exchanges for an equilibrated hole, 8–12 fresh) and settling/repeat-read times.

Classic mistakes:

1. Drilling once and reading once. A systematic Yilgarn/Perth Basin study found drilling-induced hydrogen artifacts at 85% of locations (up to 464 ppm), only 15% naturally occurring: a single post-drill reading measures drill-bit chemistry about six times out of seven.
2. Using a catalytic LEL monitor (0–100% LEL) for ppm work: wrong regime by three orders of magnitude.
3. Reading during a pump transient or depressurization, which biases the cell high (the GA5000 over-reads ~13% at 500 mbar).
4. Storing samples in untreated butyl-septum Exetainers, which can inflate H2 ~3× in 30 days from septum off-gassing (scenario e).

Details in Soil-gas surveys.

(b) fixed continuous monitoring station

The tradeoffs invert: power budget, calibration-free lifetime, and drift immunity outrank absolute LOD, plus the station must catch transients. Documented real signals are sporadic multi-day pulses and ~6-hour diurnal emissions with spikes to ~598 ppm, invisible to one-time visits and under-sampled by a multi-minute GC cycle. Electrochemical is again the only technology demonstrated at ppm resolution and packaged for unattended operation.

Hardware:

• Buried diffusive electrochemical probe. The Geo Numeric/ISTerre MONHyTOR (0–1,000 ppm, 0.1 ppm resolution, months of autonomy, non-pumping) is buried at ~1 m to damp diurnal temperature swings; CO is its only stated cross-sensitivity.
• Autonomous pumped electrochemical logger. The Axiom Sensing WHALI/EVA (0–2,000 ppm, solar-capable, ~300 mL/min duty-cycled) delivered the first autonomous H2 microseepage monitoring in the Perth Basin.
• TC MEMS for wellhead safety only. TC/acoustic sensors (Infineon XENSIV TCI, Posifa PGS1000, Sensirion STC31) are poison-immune, low-drift, low-power, but the ~100–2,000 ppm floor and He/CH4/CO2 confounds make them a %-level safety instrument, not a ppm anomaly tracker.
• Plasmonic/optical: watchlist. Plasmonic sensors (Insplorion NPS-P2) are spark-free and reach ppb in the lab, but the sole shipping product covers 0.4–4 vol% and its vendor was divesting the business as of early 2026.

Log the same four covariates as (a), continuously, for at least several weeks and ideally a full season; only a record that long lets multivariate regression separate a geologic pulse from barometric pumping, humidity shock, and drilling relaxation. The daily ambient 0.53 ppm zero check is a free drift test.

Classic mistakes:

1. Expecting to buy a station off a shelf. Every rigorously interpreted station in the literature (MONHyTOR, WHALI/EVA, PARHyS) is a bespoke research build.
2. Using MOX or TC as the primary channel and drowning in humidity/He artifacts.
3. Deploying without covariate logging, leaving anomalies indefensible.
4. Assuming a multi-minute GC station can catch fast pulses.

(c) downhole / borehole

Almost nothing survives here. Borehole temperatures reach 150–175 °C onshore and ~200 °C offshore; downhole-rated silicon electronics stop near ~175 °C; and Hestia prefers no downhole electronics at all. Every active chemical family fails: catalytic needs O2 that may be absent, electrochemical electrolytes are neither temperature-rated nor long-lived, Pd-resistive sensors lose sensitivity as the α↔β transition shifts with T and are H2S-vulnerable, and solid-electrolyte devices need 300–900 °C with no validated 100–300 °C bridge.

No field-proven downhole sensor

There is currently no commercial, field-proven, purpose-built downhole H2 sensor of any kind.

Hardware (today = sampling; tomorrow = fiber):

• Physical sample to surface + surface analysis (the field-ready answer). A Leutert geothermal positive-displacement sampler (−20…250 °C, titanium, gas-tight) or a fully mechanical Kuster-type PVT bottle captures pressure-compensated fluid for surface GC/MS. This is the workflow of the only published direct downhole natural-H2 campaign (Llanos Basin, Colombia).
• Passive fiber-optic (the R&D bet). Fiber-optic sensors (Pd-FBG, gasochromic WO3) are the only route with zero downhole electronics and can share a well with existing DTS/DAS fiber. Demonstrated performance is 1–4 vol% at room temperature and pressure. A Raman immersion probe (Kaiser/Endress+Hauser RXN2 + WetHead) has run in a flooded borehole for CO2/CH4 but is not calibrated for trace H2.
• High-T work-function/SiC (research only). Work-function devices (Pt-AlGaN/GaN, Pd/SiC Schottky) are demonstrated at 500–800 °C but uncharacterized at downhole pressure or for H2S.

Log alongside: fluid temperature at the extraction point; total pressure (Pd transduction follows Sieverts’ √p, so doubling total pressure at fixed mole fraction multiplies the reading ~1.41×); co-gases (CH4, CO2, He, N2, H2S); DBM tracers (CO, ethene). For fiber, baseline and track hydrogen darkening (H2 diffuses into the glass and attenuates signal). Seals: Viton/FKM has the lowest H2 permeability of common elastomers; EPDM is ~5× worse, silicone up to ~75–450× worse.

Classic mistakes:

1. Scoping a program around a “downhole H2 sensor” product that does not exist.
2. Lowering an electrochemical/MOX/pellistor probe into a hot, anoxic well.
3. Applying a surface (1 ATM, room-T) Pd calibration downhole without Sieverts and temperature corrections.
4. Forgetting fiber darkening, or sealing with EPDM/silicone.

See Downhole deployment.

(d) drilling / mud-gas

Mud-gas logging is a during-drilling measurement chain (gas extractor + transport lag + surface analyzer), not a discrete sensor. It observes the entire open-hole interval in real time during the one operation every well gets. The detector in the logging cabin determines whether the data is usable.

Hardware:

• Dedicated H2 fluid-logging service. SLB FlairFlex (dual inbound/outbound extraction, GC-MS, now marketing H2/He/CO2) and the SLB DQ1000 wellsite quadrupole MS (~50 ppm–15/20 vol% H2 once QC’d, ~90 s cycle). Halliburton GasFACTS and Excellence Logging are positioning around H2.
• Detector chemistry to verify. Trust only a GC-TCD on Ar/N2 carrier (not He, below), a GC-MS, or a calibrated quadrupole MS with 3-point H2 calibration.

Log alongside: lag time (minutes shallow to hours deep), mud temperature at extraction (Henry’s-law partitioning), mud type (OBM vs WBM fail differently), drilling mechanics (ROP, WOB, RPM, bit type/vibration), and DBM tracers CO and ethene/propene. Track He/H2 and watch for H2 that persists through shut-in (corrosion signature, not DBM).

Classic mistakes:

1. Assuming “the rig has a mud logger” means H2 capability. The legacy FID installed base has zero response to H2 at any concentration; an absent gas show proves nothing.
2. Trusting a He-carrier GC-TCD, which produces an inverted, non-monotonic “W-peak” above ~8–10% H2.
3. Reading raw numbers as formation values. Colombia saw raw mud-gas H2 >40 vol%, much of it later attributed to drilling artifact.
4. Contracting a “mud-logging” line item without requiring raw and DBM-corrected H2 curves, disclosed carrier gas/detector type, and tracer co-logs.

See Mud-gas logging.

(e) lab-grade truth & isotopes

To confirm a borderline reading against the 0.53 ppm background, or to fingerprint a source (δD-H2, δ13C-CH4, He R/Ra), only chromatography and mass spectrometry qualify, at minutes-per-sample and bench/skid scale.

Hardware:

• GC with reduction-gas or pulsed-discharge detector (the quantitative gold standard). Gas chromatography with RGD or PDHID is the only class that resolves sub-background H2: demonstrated LODs of 800 ppt–25 ppb (SRI 8610C-class RGD; AMETEK ta3000R/ta5000R; Peak Performer 1 RCP). Micro-GC (Agilent 990, INFICON Micro GC Fusion) reaches ~1 ppm in a field case. Plain GC-TCD is limited to ~10–100 ppm by the He-carrier artifact.
• Mass spectrometry (reference composition and isotopes). Mass spectrometry: bench RGA/QMS (SRS RGA, Hiden QGA), membrane-inlet MS for dissolved gas (Hiden HPR-40 DSA), and IRMS/noble-gas MS for δD-H2 and He R/Ra.

Log alongside: container type and hold time, where samples degrade. Untreated butyl-septum Exetainers gain H2 to ~16 ppm in 9 days; use silicone-plug/steel-modified Exetainers (stable >60 days) or Tedlar bags. Note carrier-gas purity and HgO reagent-bed status for RGD.

Classic mistakes:

1. Trusting an N2-calibrated MS sensitivity, which under-reports H2 by ~2×.
2. Ignoring He interference (truth 7 below).
3. Using Summa steel canisters, whose wall chemistry can generate H2 in situ.
4. Treating aluminum-cylinder ppm-level H2 standards as stable past ~1 year.

See Calibration & QA/QC.

The decision tree

The same logic as the bullet tree below, as a picture. The three dashed gates at the bottom modify any surface outcome.

• Is this a during-drilling measurement on an active rig?
  • Yes → Mud-gas logging (page). Verify detector chemistry and contract terms per scenario (d) above; cross-check before any resource decision.
  • No → continue.
• Does the sensor have to sit inside the borehole, above ~150 °C?
  • Yes → no field-proven downhole H2 sensor exists. In priority order: pressure-compensated sample to surface for GC/MS; passive Pd-FBG or Raman-immersion fiber as R&D (<~200 °C); neutron/NMR wireline as corroboration only. Details in scenario (c).
  • No (surface / near-surface) → continue.
• Do you need to resolve the 0.53 ppm background, or confirm a borderline few-ppm reading?
  • Yes → GC-RGD / GC-PDHID (page); add GC-IRMS / noble-gas MS (page) for fingerprinting. Slow, confirmatory, surface-only.
  • No, the target is hundreds-of-ppm-to-percent anomalies → continue.
• Roving spatial survey, or fixed continuous station?
  • Roving → electrochemical multi-gas handheld (GA5000-class) as the primary channel (page); optional MOX array for dense pre-screening. Covariates per scenario (a).
  • Fixed → buried diffusive EC probe (MONHyTOR-class) or autonomous EC logger (WHALI). Expect to build it (scenario b). Reserve TC for %-level wellhead safety.
• Cross-cutting gates (any surface choice):
  • O2 absent (sealed headspace, N2 purge, reducing pocket)? Avoid pellistors and standard O2-dependent EC cells; prefer O2-independent families (Pd-resistive, work-function, TC, optical).
  • H2S present? Protect or avoid every Pd/Pt-surface family (pellistor, Pd-resistive, work-function, plasmonic, fiber, FET); favor poison-proof TC or lab GC/MS; add an H2S scrubber where a Pd sensor must be used.
  • Percent-level wellhead gas (up to ~98 vol% at a discovery well)? An easy regime (TCD/GC, catalytic LEL monitor) and a different measurement from ppm prospecting. Do not conflate the two.

Master comparison matrix

The same trade space as a picture. All four synthesis charts live on the charts page.

One row per technology family, populated from the modality datasheets in src/data/modalities/ and the underlying dossiers.

How to read this table

Values marked ”~” or “unver.” are order-of-magnitude or vendor-unverified; cost is a class band (legend below), and many families publish no list price. LOD/range in ppm unless noted. This is a synthesis table; follow the page link for fully sourced numbers, and see Reading the specs before any cell drives a procurement decision.

Family	LOD (ppm)	Range (ppm)	Response t90	Selectivity	T range (°C)	O2 needed?	Drift	Key poisons	Power	Cost class	TRL	Field verdict	Page
Catalytic pellistor	40–2,000	0–40,000	3–30 s	none	−40…75	yes (>10–12%)	high	silicones, H2S, lead	5 mW–0.4 W	A–B	9	Cheap %LEL safety backstop; too coarse, O2-bound	link
Thermal conductivity / acoustic	~1–2,000	500–10⁶	1–60 s	low	−40…90	no	low	none	0.05 mW–5 W	A–C	9	Poison-proof, low-drift; He/CH4/CO2 fool it, not ppm	link
MOX chemiresistor	0.001–50	1–10,000	1–120 s	low	−30…85	yes (>18%)	high	siloxanes, H2S	15 mW–0.9 W	A	9	Cheap dense pre-screen; humidity-dominated, non-selective	link
Electrochemical cell	0.5–2	0–50,000	10–90 s	medium	−30…50	yes (typ.)	medium	H2S, siloxane	1 mW–0.5 W	A–B	9	The soil-gas field workhorse; drift, CO, 1–3 yr life	link
Solid-electrolyte / high-T EC	0.2–10,000	0.2–980,000	9–900 s	low	200…900	yes	medium	H2S, Pt sintering	1–20 W	B (unver.)	4	Needs 300–900 °C; no 100–300 °C bridge; unproven	link
Palladium resistive / nanomaterial	0.04–5,000	0.04–10⁶	0.02–90 s	high	−40…150	no	medium	H2S, high CO, siloxanes	0.1 µW–10 W	A–C (unver.)	9 (%-level)	O2-free, commercial at %-level; H2S-poisoned, ppm not field-hard	link
Work-function (MIS / Schottky / GasFET)	0.005–40	0.005–50,000	1.5–172 s	low	−40…800	no	high	CO, H2S, SO2, siloxanes	1 mW–0.5 W	A–C (unver.)	7	Only branch demoed to 500–800 °C; H2S, drift	link
2D-material / FET (graphene, CNT, MXene)	0.005–20,000	0.005–40,000	3–73 s	low	0…255	no	high	H2S, humidity/condensation	1 µW–0.3 W	A (unver.)	3	Low-power lab promise; no product, humidity-wrecked	link
SAW / QCM / MEMS mechanical	2–7,700	2–10⁶	1–300 s	medium	−40…350	no	high	H2S, CO, In2O3 reduction	0.1 nW–0.4 W	A–C (unver.)	4	Passive-wireless niche only; severe T-drift, no product	link
Raman (spontaneous)	190–2,100	190–10⁶	1–900 s	high	−20…150	no	low	none (optical fouling)	0.15–300 W	D	9 (process)	Multi-gas fingerprint incl. H2; ~1000× too coarse for ppm	link
Enhanced Raman (CERS / FERS)	0.06–400	0.06–10⁶	0.3–250 s	high	5…400 (physics)	no	medium	SERS-oxidation, Pd/H2S	0.01–240 W	D	5	Cavity/fiber routes reach ppb in lab; not fielded	link
IR absorption (proxy gases)	0.0003–50	0–10⁶	1–120 s	high	−25…200	no	low	dust/condensate, H2S corrosion	1–300 W	C–D	9	Cannot see H2; senses CH4/CO2/C2H6 proxies only	link
UV/EUV & fluorescence	n/a	n/a	n/a	none (field)	n/a	no*	high	LiF solarization	0.5–5 kW	C–E	2	Air is opaque in VUV; lab-only	link
Fiber-optic / chemochromic	3–5,000	3–50,000	0.34 s–200 min	high	−54…232	no	medium	H2S, SO2	0 W head (0–15 W system)	B–D	4	Only zero-downhole-electronics path; unproven >4 vol%, room T/P	link
Plasmonic / metamaterial (LSPR)	0.25–5,000	0.25–40,000	0.4 s–40 min	medium	−30…130	no	medium	CO, H2S/SOx	unver.	unver.	5	Spark-free ppb lab demos; sole vendor divesting, %-level	link
Exotic optical (LIBS / interferometry / flame-UV)	~20 (LIBS)	up to 10⁶	1 ms–60 s	none	−55…200	no*	medium	window fouling	0.5–500 W	C–D	9 as safety / 2–4 as H2	Wellsite safety/leak-imaging role; not H2-selective	link
Gas chromatography (RGD / PDHID / TCD)	0.0008–100	0.0008–10⁶	15 s–30 min	high	−10…45 (instrument)	no	low	HgO-bed depletion, column	15 W–1 kW	C–D	9	Only tech resolving sub-background H2; slow, surface-only	link
Mass spectrometry (RGA / QMS / MIMS / IRMS)	0.1–25	0.1–10⁶	1 s–20 min	high	−10…70	no	medium	siloxane, H2S, pump oil	35 W–1.5 kW	C–D	8	Reference truth + isotopes; m/z 2–4 crowded, surface-only	link

Cost class: A = <$100 (bare element) · B = $100–1k · C = $1k–20k · D = $20k–200k (capital instrument) · E = >$200k / day-rate service. Bare-element vs. full-instrument pricing differs by family; most OEMs publish no list price. * on O2: UV/EUV and flame-UV do not need O2 to sense H2, but flame detectors only fire after ignition, and ambient O2 makes air opaque in the VUV.

Mud-gas logging is deliberately not a matrix row: it is a wellsite measurement chain built on the GC/MS/TCD families above, treated as scenario (d) and on its own page.

Cross-cutting truths

Seven theses recur across every dossier and matter as much as any single spec.

1. Only GC-RGD/PDHID resolves sub-background today. Demonstrated LODs of 800 ppt–25 ppb make it the only class that separates the 0.53 ppm background from a modest anomaly. Every other family’s realistic field floor sits at or above background (the electrochemical workhorse’s ±2.5% FS accuracy is tens of ppm), so their honest job description is “detect a high-tens-to-thousands-of-ppm anomaly against a noisy few-ppm floor.”
2. Electrochemical cells are the workhorse, with artifact caveats. Uniquely ppm-capable, portable, low-power, and cheap; also drifting, CO-cross-sensitive, O2-dependent at the counter electrode, high-biased under depressurization, and 1–3 years to end of life. Fit-for-purpose because its covariates are better characterized than the alternatives’, not because it is clean.
3. H2S poisoning threatens everything Pd/catalytic. Pellistors, Pd-resistive, work-function, plasmonic, Pd-fiber, and Pd-FET sensors all need a clean noble-metal surface, and H2S sulfides those sites. Anchor quantitation on poison-proof TC or lab GC/MS; run Pd/catalytic sensors as scrubber-protected secondaries.
4. Nothing survives downhole T yet except fiber concepts. Every active family is ruled out by temperature, the no-electronics preference, or its own failure mode; passive Pd-fiber fits the constraints and is unvalidated above room T/P or ~4 vol%. Today’s downhole program is sampling plus surface analysis.
5. Humidity is the universal confounder. Dominant for MOX and 2D-FET (one FET lost 51% of response between 50–80% RH); sign-flipping bias for TC (peaking near 40–60% RH); slows Pd and fiber response; swells or dries electrochemical electrolyte. No raw reading is interpretable without co-logged RH.
6. FID is H2-blind. Structurally zero H2 response at any concentration (it detects carbon). “The rig has a mud logger” is not evidence of H2 capability.
7. He needs its own path. Raman and IR cannot see it; TC reads 1 vol% He as ~+7,700 ppm-equivalent of false H2; it is isobaric with D2/HT at m/z 4 and co-elutes near H2 on molecular-sieve GC columns. Yet a stable He/H2 ratio is one of the best natural-vs-artifact diagnostics, and noble-gas MS (R/Ra) does source attribution. Plan a deliberate helium channel.

What we would actually deploy today

Opinionated, costed straw-man kits. Cost bands use the matrix legend; unverified prices are flagged. The reasoning is constant: a fast, cheap screening layer decides where and when to look; a slow, quantitative confirmation layer establishes what the number really is.

The through-line

Buy two layers, not one instrument: a continuous/portable screener (electrochemical, or a mud-gas chain) plus a GC/MS confirmatory backstop. No single device spans the eleven-decade problem.

(a) Roving soil-gas survey, ~$15–40k of hardware (indicative; handheld list prices undisclosed). Primary: an electrochemical multi-gas handheld (GA5000-class, cost class C) with an H2 cell plus NDIR CH4/CO2 and O2/H2S channels. Optional densifier: MOX/BME688 boards (class A) between quantitative reads. Confirmation: grab samples in silicone/steel-modified Exetainers or Tedlar bags for GC-RGD. Covariate pack: soil-moisture/temperature probe, logging barometer, gas-line RH sensor.

(b) Fixed continuous station: a build, not a purchase, ~$10–30k per node. Core: a buried diffusive EC probe (MONHyTOR-class) or autonomous pumped EC logger (WHALI-class), class B–C, solar + battery + LoRa. Add a TC MEMS channel (class A–B) only for wellhead %-level safety. Full covariate telemetry, run for at least a season (scenario b).

(c) Downhole: sampling today, fiber as an R&D bet, ~$5–30k per run. A pressure-compensated mechanical sampler (Kuster PVT bottle or Leutert PDS, class B–C, zero downhole electronics) analyzed at surface by GC/MS; corroborate with reinterpreted neutron/NMR wireline. R&D line item: a passive Pd-FBG or Raman-immersion fiber on an existing DTS/DAS string for targets <~200 °C; a 3–7 year bet, not a capability. Seals and darkening per scenario (c).

(d) Drilling / mud-gas: a service contract, not a box, ~$0.5–2M/well day-rate. Contract an H2-branded fluid-logging service (SLB FlairFlex or a DQ1000-equipped unit; class E) with the QC written into the contract: detector must be GC-MS, QMS with 3-point H2 calibration, or GC-TCD on Ar/N2 carrier, never FID-only or He-carrier TCD; raw and DBM-corrected H2 curves plus CO/ethene co-logs; cross-check against discrete can samples at an offsite lab.

(e) Lab-grade truth & isotopes, ~$20–150k capital or send-out. Anchor: a GC-RGD/PDHID (SRI 8610C-class or online AMETEK ta3000R, class C–D). Add access to GC-IRMS and noble-gas MS for δD-H2, δ13C-CH4, and He R/Ra, the evidence that turns a high reading into a defensible resource claim; and a membrane-inlet MS (Hiden HPR-40) if dissolved-H2 groundwater work is in scope. Guard the sample chain per scenario (e).

Related pages

• Why hydrogen is hard to sense: the physics that constrains every choice above
• Reading the specs: how to read the datasheet numbers behind the comparison matrix
• Geologic context · Soil-gas surveys · Downhole deployment · Environmental covariance · Calibration & QA/QC
• Electrochemical cells · Gas chromatography · Mass spectrometry · Mud-gas logging · Fiber-optic sensors