Dragline and shovel cycle counting from shadow geometry
Sub-metre optical imagery captured at sub-daily revisit can resolve the shadows cast by large draglines and rope shovels, allowing swing cycles to be counted and correlated with published equipment specs to estimate tonnes moved per shift.
Sensors
- Maxar WorldView Legion: 30 cm panchromatic GSD, up to 15 revisits per day over a target latitude depending on constellation deployment; the sharpest commercially available shadow geometry for boom-tip localisation
- Planet Pelican: 50 cm GSD, designed for sub-daily tasking at high cadence; slightly coarser than Legion but sufficient to resolve dragline boom shadows on machines with boom lengths above roughly 90 m
- Airbus Pléiades Neo: 30 cm GSD, stereo and tri-stereo capable; useful for confirming machine position in three dimensions and cross-checking shadow-derived swing-arc estimates against stereo point clouds
- Planet SkySat: 50 cm GSD with video mode up to 90 seconds; video clips can resolve individual swing cycles directly without relying on shadow-azimuth differencing between discrete frames
What a shadow actually encodes
A dragline boom is a fixed-length steel lattice, typically 90 m to 115 m on large machines such as the Marion 8750 or Bucyrus 2570. At any given solar elevation angle, a boom of known length casts a shadow of predictable length on flat ground. The azimuth of that shadow is the azimuth of the boom. Across two images taken minutes apart, the change in shadow azimuth is the swing angle completed in that interval.
Solar elevation and azimuth at the image acquisition time are calculable to sub-degree accuracy from published ephemeris for any location and timestamp. That means the geometry is fully determined: boom length from public equipment specifications, sun angles from ephemeris, shadow length and azimuth from image measurement. The only unknowns are ground slope (which stereo DEM can correct) and whether the machine is actually carrying a load on the return swing. The method does not require any ground sensors or operator reports.
Counting cycles and converting them to tonnes
A full dragline cycle consists of a drag pass, a swing to the spoil pile, a dump, and a return swing. Published cycle times for large electric walking draglines run from roughly 55 seconds to 90 seconds depending on swing angle and operator practice. If imagery is collected at intervals shorter than one cycle, individual swings can be counted directly. At longer intervals, fractional shadow-azimuth change between frames can be used to estimate the number of completed cycles statistically.
Payload per cycle is published in manufacturer specifications. The Marion 8750, for example, has a rated bucket capacity of 107 cubic metres. Applying a swell factor for the material type (typically 1.25 to 1.35 for overburden) and a fill factor (commonly 0.85 to 0.95 in practice) gives a per-cycle tonne estimate. Multiplying by counted cycles over a shift produces a material-moved figure. The honest uncertainty on this is probably plus or minus 15 to 25 percent, driven mainly by fill-factor variability and the discrete sampling of imagery rather than any flaw in the geometry itself.
Rope shovels are smaller and faster, with cycle times around 25 to 35 seconds and bucket capacities from 15 to 55 cubic metres on large electric models. Their shorter booms (typically 15 m to 20 m) produce shadows that are harder to resolve at 50 cm GSD, making them a better fit for 30 cm imagery or Planet SkySat video.
Limits you should price into any decision
Cloud cover is the primary operational constraint. Open-cut mines in tropical or monsoonal climates can lose 60 to 80 percent of potential collection days to cloud, making shift-level inference unreliable without a long averaging window. Arid mining regions such as the Pilbara or Athabasca are far more amenable.
Shadow geometry degrades near solar noon, when shadows shorten and azimuth changes per unit of swing become harder to measure. Early morning and late afternoon collections, when solar elevation is below about 45 degrees, give the longest shadows and the cleanest azimuth signal. This is worth specifying when tasking commercial satellites.
The method also assumes the machine is on reasonably flat ground. A bench slope of five degrees or more introduces a shadow-length error that, if uncorrected, biases the inferred swing angle. Stereo-derived terrain models can correct this, but that adds a processing step and requires a stereo pair, not just a single frame. Finally, the approach counts machine motion, not productive motion. A machine swinging without a load, or cycling during a maintenance check, looks identical to a productive shift from orbit.
Turning shadow counts into an operational intelligence product
The practical output is a per-shift or per-day cycle count with an associated material-moved estimate, delivered as a structured data feed or tabular report. Stacked over weeks and months, this becomes a production rhythm baseline. Deviations from that baseline, a sudden drop in cycle rate mid-shift, an unexpectedly short operating window, are detectable without any access to the mine's internal reporting.
For investors and commodity analysts, the value is in the independence of the signal. Mine operators self-report production figures with varying frequency and detail. A satellite-derived cycle count provides a cross-check that does not depend on the operator's disclosure schedule. For mining companies themselves, the same method applied to their own assets can validate contractor performance or benchmark equipment utilisation across multiple pits.
Satellize can run this analytic on tasked imagery from WorldView Legion or Pléiades Neo under client licence, delivering cycle-count reports alongside the broader activity-inference work it conducts for resource and government clients, in the same analytical framework used for its Tonga crop-estimation programme.
Archive depth and historical reconstruction
WorldView-1 and WorldView-2, predecessors to Legion, have been collecting sub-metre imagery since 2007 and 2009 respectively. Maxar's archive holds imagery over many active mining districts that predates current operational interest. Where historical tasking happened to capture a dragline site at a useful sun angle, retrospective cycle counting is possible, giving a production history that no other remote-sensing method can match at this resolution.
Planet's SkySat archive extends to around 2014. Coverage of specific mine sites in the early years is sparse and opportunistic, but grows denser as the constellation expanded. Combining archive depth with current high-cadence tasking allows analysts to build multi-year production trend lines, which is particularly useful for due-diligence work ahead of asset transactions.
Typical figures
| Spatial resolution (panchromatic) | 30 cm (WorldView Legion, Pléiades Neo), 50 cm (Planet Pelican, SkySat) |
| Revisit rate | Up to 15 times per day (WorldView Legion at mid-latitudes); 1 to 2 times per day typical for Planet Pelican tasked collection |
| Minimum resolvable boom shadow | Boom length approximately 90 m or greater at 50 cm GSD; approximately 50 m or greater at 30 cm GSD (solar elevation below 45°) |
| Cycle-count uncertainty | ±15 to 25% on material-moved estimate, driven by fill-factor variability and image cadence |
| Spectral bands used | Panchromatic (shadow geometry); multispectral optional for ground-condition context |
| Cloud sensitivity | Method fails under cloud cover; optical only, no SAR equivalent for shadow geometry |
| Archive depth | WorldView series from 2007; SkySat from approximately 2014; Pléiades Neo from 2021 |
| Delivery formats | Structured CSV or JSON cycle-count feed; GeoTIFF annotated frames; PDF shift report |
| Latency (tasked collection to delivery) | Typically 2 to 6 hours post-acquisition for priority tasking pipelines |
Analytics Satellize can run
| Per-shift cycle count | Shadow azimuth differencing using solar ephemeris and published boom geometry | Tabular report: timestamp, swing count, estimated material moved, confidence interval |
| Production rhythm baseline | Time-series aggregation of daily cycle counts over rolling 30/90-day windows | Trend chart and anomaly-flagged data feed in CSV or JSON |
| Equipment utilisation estimate | Shadow presence/absence detection across scheduled shift windows to infer operating hours vs idle time | Daily utilisation percentage per machine, GIS layer with machine position log |
| Historical production reconstruction | Retrospective shadow analysis on archive imagery from WorldView or SkySat | Multi-year production trend table for due-diligence or benchmarking reports |
| Swing-arc geometry map | Boom azimuth range extraction from multi-frame shadow sequence to characterise working face orientation | Annotated GeoTIFF showing arc envelope and pit-face alignment |
| Terrain-corrected shadow model | Shadow-length correction using stereo-derived DEM (Pléiades Neo tri-stereo or WorldView stereo pair) | Corrected cycle-count report with slope-adjustment metadata |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.