Reviewed August 2026 against MSHA, the USACE National Levee Database and Copernicus Sentinel-1 mission documentation.

Jump to the calculator →

IoT in Mining: Sensors, Embankments & the Cloud

IoT in mining means putting networked instruments — piezometers, inclinometers, dust and gas heads, vibration and equipment-health sensors — on assets that were previously checked by eye, then routing those readings somewhere they can be trended and alarmed. The two jobs it does best are embankment surveillance (levees, tailings dams, haul-road fills) and equipment condition monitoring. The hard part is almost never the sensor: it is connectivity, calibration upkeep, and deciding which decisions stay at the edge and which belong in the cloud.

This page covers all four in order, with figures you can re-check yourself, and it deals with two things most overviews skip: how cloud cover measured in oktas quietly decides whether your optical satellite layer works on any given day, and what the maintenance-repair-and-operations (MRO) tail of an environmental sensor network actually costs.

Table of Contents

Start with the scale of the monitoring gap on the civil side, because it sets realistic expectations for what instrumentation coverage looks like even in a wealthy jurisdiction. The US Army Corps of Engineers’ National Levee Database catalogues about 22,000 miles of levee across 6,131 systems, with an average age of 62 years, protecting roughly 23 million residents, 7 million buildings and 5 million acres of farmland across 2,288 communities. The American Society of Civil Engineers’ Infrastructure Report Card entry for levees (consulted August 2026) grades them D+, puts the repair backlog at $70 billion on ASCE’s 2021 estimate, and notes that nearly two-thirds have never been risk-assessed at all.

The four layers of a mine IoT stack

Vendors sell “IoT in mining” as one product. Operationally it is four layers with different owners, different failure modes and different regulators. Separating them is what stops a monitoring programme from becoming a dashboard nobody trusts.

The layers earn their keep together. Ground instruments are precise but sparse; satellite deformation is spatially complete but slower and coarser. A crest movement seen by both is a real movement.

Monitoring embankments: levees, tailings dams and haul-road fills

“Embankment” covers three quite different assets that share one physics problem: water inside the fill. Rising pore pressure reduces effective stress, and the failure is usually announced by pressure and displacement long before anything is visible from a truck window.

The three asset classes and their governing frameworks

• Flood levees (United States). The National Levee Database is the authoritative register; ownership of more than a third of catalogued levees is unknown, which is itself a monitoring problem — you cannot instrument what nobody admits to owning.
• Tailings storage facilities (global). The Global Industry Standard on Tailings Management is structured as six topic areas, 15 principles and 77 auditable requirements. Oversight now sits with the Global Tailings Management Institute, established 21 January 2025, which accredits auditors and reviews audit reports; its Technical Committee was set up in February 2026 and it ran a public consultation on GISTM implementation from 30 March to 30 April 2026. For disclosure baselines, the Global Tailings Portal holds more than 1,800 facilities from more than 100 companies against 20 disclosure questions, launched January 2020 with roughly 60 companies having verified their submissions at that point.
• Slope and road fills (Italy and alpine Europe). Italy’s IFFI inventory, maintained by ISPRA, records 636,496 landslides across 25,138 km² — 8.3% of national territory — with data updated to 16 October 2025 and a reference period running from 1116 to 2025. ISPRA notes this is about two-thirds of all mapped landslides in Europe. Check the live figure at the ISPRA environmental indicators portal, which is revised as regions submit new mapping.
• Jump to the calculator

The instrument set, and what each one is actually for

This list does not expire when prices change. It is the durable core of any embankment programme:

1. Vibrating-wire piezometers at two or three depths per section — the primary indicator. Trend the rate of pore-pressure rise, not the absolute value.
2. In-place inclinometer strings across the suspected shear surface — converts “the crest looks fine” into millimetres per week.
3. Crest prisms or GNSS monuments — an independent displacement check that does not share a failure mode with the buried instruments.
4. Seepage weirs with flow and turbidity at toe drains — clear flow rising is bad; cloudy flow is an emergency, because it means the embankment is losing material.
5. Rain gauges on the structure itself — regional forecasts are not the loading your fill received.
6. Redundant pairs at the two highest-consequence sections so a single dead sensor does not blind the trigger-action response plan.

On the mining side, the safety case for instrumentation is measurable. NIOSH’s mining programme found that more than 40% of the most serious injuries — fatalities and permanent disabilities over 2000–2007 — involved machinery contact and powered haulage; surface collisions and driving over unseen edges caused 3 to 4 deaths a year in that period; pinning and striking incidents with continuous mining machines have caused 37 fatalities since 1984; and shuttle car and scoop incidents caused 16 deaths between 2000 and 2010, per NIOSH’s proximity detection topic page. Fatality totals themselves are refiled continuously by MSHA.

For context on the population those numbers describe: MSHA’s Mine Safety and Health at a Glance series (page last updated 27 February 2026) reports 12,568 mines in FY2025 — 886 coal and 11,682 metal/nonmetal — with 264,564 metal/nonmetal miners, and inspection intensity of 20–22 hours per metal/nonmetal mine against 204–242 hours per coal mine. Both series are refiled each fiscal year; the per-mine record, including employment, sits in the Mine Data Retrieval System, so look your own mine ID up there rather than quoting an article.

Mining and cloud: what moves up, what must stay down

The useful question is not “should mining use cloud” but “which decision runs where”. A defensible split:

• Stays on the machine or gateway (edge): anything that trips a physical outcome — proximity braking, gas alarms, dewatering pump interlocks, TARP trigger levels for an embankment. These must work with the WAN cut, and their logic must be readable without a login.
• Goes to the cloud: long-horizon trending, cross-site comparison, satellite deformation processing, model retraining, and the audit archive that inspectors and Engineers of Record read.
• Never leaves the site unencrypted or unlogged: the geotechnical time series. It is the evidence base if a facility ever moves.

The barrier is rarely bandwidth — it is format. The Global Mining Guidelines Group published its interoperability alignment report on 12 August 2019, built from interviews and workshops with 17 mining companies across six continents and drawing on wider workshops involving over 120 companies, precisely because point-to-point vendor integrations do not compose. Two practical consequences for a buyer: insist on a documented, exportable schema for every sensor stream before signing, and confirm that raw readings — not just vendor-computed indices — land in storage you control.

Cloud cover in oktas — and why radar exists

If your monitoring plan includes optical satellite imagery, the single variable that decides whether you get data is cloud amount, and meteorologists report it in oktas — eighths of the sky dome. The UK Met Office observation guide defines the scale as 0 for complete absence of cloud, 1 for one eighth or less but not zero, 7 for seven eighths or more but not full cover, 8 for full cover with no breaks, and 9 for sky obscured by fog or other phenomena. It is a coded scale, not a percentage, which matters when you write acceptance criteria: “reject scenes above 6 oktas” is testable; “reject cloudy scenes” is not.

Because an 8-okta day yields nothing optical, embankment deformation monitoring from space runs on C-band synthetic aperture radar, which sees through cloud. Per Copernicus SentiWiki, Sentinel-1 operates at 5.405 GHz (about 5.55 cm wavelength) with a 12-day repeat cycle of 175 orbits for one satellite and a 6-day revisit for a two-satellite constellation phased 180° apart, with Sentinel-1A and Sentinel-1C operating together, under an open and free data policy. Mode choice is the decision most people get wrong — it fixes both your footprint and your ground resolution.

For a single tailings facility a few hundred metres across, Interferometric Wide gives you the 5×20 m pixel and the revisit; Extra Wide’s 410 km swath is for ice and ocean, not crest movement. Budget for a stack — interferometry needs many passes, not one image — and expect to co-locate results with your GNSS monuments before you believe a millimetre-scale rate.

Dust, gas and the environmental sensor MRO problem

Environmental sensors are the layer where maintenance, repair and operations spending is chronically under-budgeted, because the obligation is not “have a sensor” — it is “produce valid data against a fixed averaging period, on a fixed reporting date”. Miss the calibration window and the reading is not merely imprecise, it is inadmissible.

The thresholds the data is measured against are published and stable. Per the US EPA NAAQS table (page last updated 4 November 2025), the PM_2.5 primary annual standard is 9.0 µg/m³ as an annual mean averaged over three years, the secondary annual standard is 15.0 µg/m³, the 24-hour PM_2.5 standard is 35 µg/m³ at the 98th percentile averaged over three years, and PM₁₀ is 150 µg/m³ over 24 hours, not to be exceeded more than once per year on average over three years.

The reporting calendar is equally fixed, and it is what drives service-visit scheduling. Queensland coal mine operators must submit respirable dust monitoring data within one month of each quarter’s end — Q1 due 1 May, Q2 due 1 August, Q3 due 1 November, Q4 due 1 February — with confirmation required even when no monitoring occurred, and revised occupational exposure limits applying to average exposures from 1 January 2021, per Business Queensland’s coal mine safety and health reporting page. Sampling numbers are set by Recognised Standard 14 and an official online calculator, so derive your sample count from the regulator’s tool rather than a rule of thumb.

On the market question — the size of the environmental sensor equipment MRO services market — we could not find a figure from a government or industry body that is free to verify, and the private research reports circulating on it disagree with each other by wide margins. Rather than repeat a number we cannot check, here is how to size the only figure that affects your decision: your own annual MRO cost per monitoring point. Four drivers set it — scheduled service visits, fully loaded cost per visit, hardware attrition, and the proportion of readings that end up invalid and therefore paid for twice.

Calculator: annual MRO cost per monitoring point

Enter your own network and service assumptions to see what a year of upkeep costs per node and per valid monitoring-day.

Where satellite mineral intelligence fits

Ground IoT tells you what is happening at a point you already chose. Satellite analysis tells you which points are worth choosing. Farmonaut works the second half: multispectral and hyperspectral analysis to map mineral prospectivity zones and produce structural interpretation, 3D subsurface models and drilling guidance without ground disturbance — projects spanning 18+ countries and 13+ mineral types, delivered as Premium and Premium+ report tiers. The practical sequence on a greenfield property is: satellite screening to rank targets, then geotechnical and environmental instrumentation only on the ground you are actually going to disturb.

A durable commissioning checklist

Ten questions that stay valid regardless of which sensors or platform you buy. Answer every one before the first invoice.

1. Which specific regulatory obligation does each sensor serve, cited by clause?
2. What is the averaging period the data will be judged against, and does the logger's sample rate support it?
3. What is the documented calibration or verification interval, and who is accountable for the record?
4. Which readings trip a physical action, and does that logic run without the wide-area network?
5. Where do raw values land, in what schema, and can you export the full history unaided?
6. Which two sections are the highest consequence, and are they instrumented redundantly?
7. What is your measured invalid-reading rate, and is it trending?
8. How is a satellite deformation signal reconciled against ground GNSS or prisms before anyone acts?
9. What okta threshold invalidates an optical scene in your acceptance criteria?
10. Who reviews alarms on a night shift, and what is the escalation clock in minutes?

FAQ

The short version

Instrument the physics, not the dashboard. Pore pressure trend, displacement rate and seepage turbidity are what tell you an embankment is moving; everything else is context. Keep trip logic at the edge, keep the audit archive and wide-area deformation processing in the cloud, and keep raw data egress in your contract. Then budget the MRO tail honestly, because a sensor whose calibration lapsed is worse than no sensor — it produces confident, inadmissible numbers.