River crossing span geometry mapping for aerial fibre clearance
Aerial cables crossing rivers need verified clearance above flood-stage water surfaces, not just bankfull averages. Stereo radar, laser altimetry and multi-temporal SAR combine to give span length, bank geometry and seasonal high-water elevation from orbit.
Sensors
- TanDEM-X (bistatic X-band SAR): Global DEM at 12 m posting (0.4 arcsec) with absolute vertical accuracy around 10 m at 90th percentile globally, better than 2 m over flat terrain with good coherence. Bistatic single-pass acquisition eliminates temporal decorrelation, making it the most consistent elevation source for river-bank geometry. Does not penetrate the water surface; returns the water-surface elevation directly when coherence holds.
- ICESat-2 (ATL03/ATL13 photon-counting lidar): Measures water-surface elevation along 14-beam ground tracks at roughly 0.7 m along-track spacing. Absolute vertical accuracy over inland water is typically 0.1 to 0.3 m. Revisit at a given latitude is 91 days for the exact repeat, but the dense track network means most river crossings in mid-latitudes are within a few kilometres of a pass. ATL13 is the published inland-water product.
- Planet SuperDove (PlanetScope): 3 m resolution multispectral imagery with near-daily revisit globally. Not an elevation sensor, but the NDWI-derived water-surface extent across a time series of 100-plus scenes per year gives the seasonal envelope of inundation width. Flood-stage width from SuperDove combined with bank-elevation from TanDEM-X or ICESat-2 constrains the worst-case span geometry.
- Sentinel-1 C-band SAR: 6-day repeat (12-day for single satellite) at 10 m resolution in Interferometric Wide Swath mode. Open archive back to 2014. Water surfaces return very low backscatter in C-band, making flood-extent mapping at each acquisition straightforward. The temporal density of the archive is what matters here: it captures multiple flood seasons and gives a statistical picture of high-water extent that a single survey cannot.
Why low-water surveys understate the engineering problem
A cable strung across a river at the end of dry season may clear the water surface by several metres. The same cable at peak monsoon or snowmelt may have a fraction of that clearance, or none. Sag increases with span length, and span length is measured bank-to-bank at the design flood stage, not at the survey date. If the bank geometry is wrong, the sag calculation is wrong from the first line.
Ground survey of river crossings is slow, access-dependent and often dangerous at high water, which is precisely when the measurement matters most. Satellite data does not replace a detailed structural survey, but it can establish the envelope of water-surface elevation and inundation width across several years of flood seasons before a single engineer visits the site. That changes the engineering brief rather than just confirming it.
What TanDEM-X gives you, and where it stops
The TanDEM-X global DEM, produced from bistatic X-band acquisitions between 2010 and 2015, is the most consistent publicly documented elevation dataset at river-crossing scale. At 12 m posting over flat floodplain terrain, relative vertical accuracy is typically better than 2 m, and in many areas better than 1 m. That is sufficient to characterise bank-top elevation and the slope of the approach terrain on both sides of a crossing.
The limit is important to state plainly. TanDEM-X measures the surface at the time of acquisition. Over a wide, slow river, the radar return comes from the water surface, not the riverbed, so the DEM captures the water-surface elevation at that moment. Over narrow channels, layover and shadow from bank vegetation can corrupt the elevation. Rivers narrower than roughly 30 to 50 m are poorly resolved. For those, ICESat-2 track intersections or airborne lidar remain the more reliable primary source, with TanDEM-X providing the surrounding floodplain context.
ICESat-2 as a water-surface elevation reference
ICESat-2's ATL13 inland-water surface product extracts water-surface elevation along each ground track using photon-counting returns. Published validation studies report absolute accuracy of 0.1 to 0.3 m over calm inland water, which is substantially better than any current spaceborne radar DEM. The catch is geometry: the satellite follows fixed ground tracks, and a given river crossing may not sit on one. The 91-day exact-repeat cycle means that even if a track does cross the river, the acquisition epoch may not coincide with peak flood.
The practical approach is to treat ICESat-2 as a calibration anchor. Where a track intersects the river near the crossing site, the measured water-surface elevation at that date calibrates the TanDEM-X datum offset. Multiple passes across different seasons build a partial record of stage variation at that location. Combined with a longer Sentinel-1 time series for flood-extent mapping, the result is a water-surface elevation envelope rather than a single snapshot.
Building the flood-stage envelope from SAR and optical time series
Sentinel-1 C-band backscatter over open water is typically below minus 15 dB in calm conditions, well separated from vegetated or bare soil returns. Automated thresholding or change-detection algorithms applied to the full Sentinel-1 archive since 2014 can map water-surface extent at each acquisition. Over a river crossing site, this produces a time series of inundation width at roughly 10 m resolution, 6-day cadence, across up to a decade of flood seasons.
Planet SuperDove adds spatial detail at 3 m. NDWI (Normalised Difference Water Index, using green and near-infrared bands) separates water from non-water reliably in clear-sky conditions. Cloud is the honest constraint: in humid tropical catchments, cloud cover during peak flood season can limit usable optical acquisitions to a small fraction of available dates. SAR is cloud-transparent, which is why Sentinel-1 carries the seasonal envelope analysis and SuperDove fills in the bank-geometry detail when skies allow.
The combined output is a flood-frequency map: for each pixel in the crossing corridor, the fraction of acquisitions over the archive period on which it was inundated. The 95th-percentile inundation width, read from this map, is the conservative span-length input for sag calculations.
Translating remote-sensing outputs into span-geometry inputs
Span length for sag calculation is the horizontal distance between the two attachment points. At a river crossing, the attachment points are typically on bank-top structures, and the critical constraint is the minimum clearance above the design flood-stage water surface at mid-span. The remote-sensing workflow delivers three numbers: bank-top elevation on each side (from TanDEM-X or ICESat-2), water-surface elevation at design flood stage (from the SAR-derived stage envelope, anchored by ICESat-2 where available), and inundation width at that stage (from the time-series flood-frequency map).
These three numbers do not replace a structural engineer's sag calculation, which also requires cable weight, tension, temperature range and attachment-height assumptions. What they do is constrain the geometry inputs to that calculation from orbit, before mobilisation, and flag crossings where the flood-stage width is substantially wider than the low-water channel. That is where span-length assumptions made from a single-date survey are most likely to be wrong.
Satellize has applied similar multi-sensor water-surface workflows in agricultural contexts, including the Tonga crop-estimation programme, where seasonal water-body extent feeds into irrigation and soil-moisture inference. The same sensor stack and temporal-compositing logic transfers directly to river-crossing geometry extraction.
Honest limits and what satellite data cannot settle
No current spaceborne sensor resolves individual cable sag directly. The workflow described here characterises the geometric envelope within which a cable must perform; it does not measure an existing cable's actual sag or condition. Rivers narrower than 30 m are poorly served by TanDEM-X and Sentinel-1 alike; for those, the bank-elevation inputs carry higher uncertainty. Dense riparian vegetation causes both radar layover and optical obstruction, degrading bank-top elevation estimates by a metre or more in the worst cases.
ICESat-2 track coverage is not uniform. Some crossing sites will have no usable ATL13 passes within a useful distance. In those cases the absolute water-surface elevation reference must come from a gauging station record or a field visit, with the satellite data providing relative geometry and seasonal width only. These are not arguments against the approach; they are the conditions under which a competent analyst scopes the deliverable honestly.
Typical figures
| TanDEM-X DEM spatial posting | 12 m (0.4 arcsec global product); 6 m available over selected areas |
| TanDEM-X vertical accuracy (flat terrain) | Relative: <1 m at 90th percentile; absolute: ~2 m typical over floodplain |
| ICESat-2 water-surface vertical accuracy | 0.1 to 0.3 m over calm inland water (ATL13 published validation) |
| ICESat-2 exact-repeat cycle | 91 days; dense track network gives partial coverage at shorter intervals |
| Sentinel-1 IW mode resolution | 10 m ground range; 6-day repeat (12-day single satellite); archive from 2014 |
| Planet SuperDove resolution | 3 m multispectral; near-daily global revisit; 8 spectral bands |
| Minimum detectable inundation width (Sentinel-1) | ~30 m reliably; narrower channels require SAR processing at higher resolution or optical confirmation |
| Archive depth for flood-season analysis | Sentinel-1: 2014 to present; Landsat: 1972 to present (30 m); Planet: 2016 to present |
| Delivery formats | GeoTIFF elevation layers, flood-frequency rasters, vector bank-line polygons, CSV stage-envelope tables |
Analytics Satellize can run
| Bank-top elevation profile | TanDEM-X DEM extraction with vegetation-mask filtering along a user-defined crossing corridor | GeoTIFF elevation layer and cross-section CSV, both banks, 12 m resolution |
| Water-surface elevation at design flood stage | Sentinel-1 SAR flood-extent time series combined with ICESat-2 ATL13 stage calibration where track coverage exists | Stage-elevation table with percentile breakdown (50th, 90th, 95th) and uncertainty range |
| Seasonal inundation-width envelope | Flood-frequency mapping from full Sentinel-1 archive; NDWI composites from Planet SuperDove for cloud-free epochs | Flood-frequency raster and per-crossing width statistics at 50th, 90th and 95th percentile flood stage |
| Span-length estimate for sag calculation input | Bank-to-bank distance measured at design-flood inundation boundary, derived from flood-frequency map | Per-crossing span-length estimate with confidence interval, delivered as GIS point layer and summary report |
| Minimum clearance risk flag | Differencing of cable-attachment elevation (from TanDEM-X bank-top) and 95th-percentile water-surface elevation at mid-span | Traffic-light risk classification per crossing, tabulated and mapped; crossings below a user-specified clearance threshold flagged for field verification |
| Multi-year flood-season change detection | Year-on-year comparison of peak inundation extent from Sentinel-1 archive to identify channel migration or floodplain change | Annual flood-extent polygons and channel-migration vector overlay, GIS layer |
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.