Ice accretion frequency mapping for antenna and tower structural risk assessment

Combining ERA5 reanalysis meteorology with MODIS cloud liquid water path lets network engineers quantify glaze, rime and wet-snow accretion frequency at tower locations before a structural or radiation-pattern failure forces the question.

Sensors

MODIS Terra / Aqua (NASA): Provides cloud liquid water path (LWP) retrievals at 1 km pixel scale from the 1.6 µm and 2.1 µm shortwave-infrared bands. Twice-daily combined overpass frequency (Terra descending ~10:30 local, Aqua ascending ~13:30 local) gives a coarse but long-archive (2000-present) climatology of supercooled liquid water cloud occurrence, the primary driver of in-cloud rime accretion.
AMSR2 on JAXA GCOM-W: Passive microwave radiometer at 6.9–89 GHz. Retrieves integrated cloud liquid water and columnar ice water path at roughly 5–60 km footprint depending on channel. Penetrates cloud cover that blinds optical sensors, making it useful for corroborating MODIS LWP in persistently overcast icing climates. Daily global coverage.
ERA5 reanalysis (ECMWF): Hourly gridded meteorological fields at approximately 31 km horizontal resolution, 1940-present. Supplies wet-bulb temperature, wind speed and direction, specific humidity, and precipitation phase discrimination needed to separate freezing-rain, in-cloud rime and wet-snow regimes. The 31 km grid smooths orographic gradients substantially; ridge-top sites in complex terrain can differ from the nearest ERA5 cell by several degrees Celsius.
Sentinel-3 SLSTR (ESA): Sea and Land Surface Temperature Radiometer, dual-view, 500 m and 1 km resolution in thermal and shortwave-infrared bands. Useful for detecting persistent cold-surface anomalies on high terrain and for cloud-top temperature mapping that supports supercooled-layer identification. Two satellites (Sentinel-3A and 3B) give sub-daily repeat at mid-latitudes.

Three kinds of ice, three kinds of damage

Glaze ice forms when freezing rain or drizzle strikes a surface warmer than the droplet: the water spreads before freezing, producing a dense, clear accretion (density roughly 700–900 kg/m³). It is mechanically the most damaging because it coats every surface uniformly, including the windward and leeward faces of antenna radomes, and adds load faster than most tower designs anticipate.

Rime accretes from supercooled cloud droplets or fog that freeze on contact. Soft rime (density 100–300 kg/m³) is feathery and grows into the wind; hard rime (300–700 kg/m³) is denser and more adhesive. Both alter antenna radiation patterns by changing the effective dielectric constant of the radome surface and by building asymmetric protrusions that shift beam azimuth. In-cloud icing is the dominant regime above the cloud base on exposed ridges and masts.

Wet-snow accretion occurs in a narrow temperature band around 0 °C. Wet snow is sticky and can reach densities of 400–800 kg/m³. It is particularly hazardous to dish antennas and flat-panel arrays because it accumulates on horizontal surfaces rather than distributing around a cylindrical profile. A 1-metre dish antenna can accumulate several hundred kilograms of wet snow in a single event. Knowing which regime dominates at a given site determines whether the engineering response is a heating element, a hydrophobic coating, a structural upgrade, or some combination.

What the satellite and reanalysis data actually measure

MODIS cloud liquid water path is a column-integrated quantity retrieved from the contrast between solar reflectance at 0.86 µm (cloud-top scattering) and absorption at 2.1 µm (droplet size proxy). It identifies where supercooled liquid water clouds exist and, combined with cloud-top temperature from the 11 µm thermal band, flags the altitude bands where in-cloud icing is probable. It does not directly measure accretion rate on a structure; that requires combining LWP with wind speed and droplet size assumptions from the Makkonen accretion model or equivalent ISO 12494 methods.

ERA5 provides the meteorological context: wet-bulb temperature thresholds for freezing rain (roughly 0 °C at surface, warmer aloft), wind speed for rime accretion rate, and precipitation phase. The reanalysis assimilates radiosonde, aircraft and satellite observations, so it captures synoptic-scale icing events reliably. What it misses is the local acceleration of wind over a ridge crest, the temperature inversion that keeps a valley floor frost-free while the tower above it ices, and the persistent fog layer that sits below the ERA5 vertical resolution. These gaps are real and must be stated plainly to any client specifying anti-icing systems.

AMSR2 adds value in regions where persistent cloud cover makes MODIS optical retrievals sparse. Its passive microwave channels are sensitive to liquid water in the column even through thick overcast, though the coarse footprint (15–60 km at most channels) limits its use to corroborating ERA5 rather than resolving site-level differences.

Building a frequency climatology from 20-plus years of data

The practical output is an icing-hours climatology: for each tower location, the annual expected hours of each accretion regime, the return period of the design-load event (typically the 50-year return period specified in IEC 60826 for overhead line design, adapted for towers), and the seasonal distribution. This feeds directly into anti-icing system duty-cycle specification and structural load calculations under EN 50341 or national equivalents.

The method stacks ERA5 hourly fields over the full available archive (1940-present, though satellite-era data from 1979 onward is better constrained) and applies temperature-wind phase-space classifiers to assign each hour to a glaze, rime, wet-snow or no-icing category. MODIS LWP climatology (2000-present) is used to validate and adjust the in-cloud rime frequency where ERA5 cloud fraction fields underestimate fog persistence. The combined product gives a per-site probability distribution rather than a single design value, which is more useful for risk-tiering a network of hundreds of towers.

Where the method breaks down

The 31 km ERA5 grid is the binding constraint for complex terrain. A ridge-top tower at 1,400 m sitting in a grid cell whose centroid is at 900 m will have its icing frequency systematically underestimated. The wet-bulb temperature error alone can shift the glaze/rime boundary by tens of hours per year. Downscaling using a digital elevation model and a lapse rate correction reduces this error but does not eliminate it; in-situ ice load sensors or ice detection cameras remain the only way to validate the climatology at a specific mast.

MODIS has a known gap in polar night and in persistently overcast regimes where optical retrieval fails entirely. For sites above 60° latitude or in maritime climates with near-continuous low cloud, AMSR2 and ERA5 must carry more of the analytical weight, with correspondingly wider uncertainty bounds. Sentinel-3 SLSTR thermal data can partially fill the cloud-top temperature gap but is not a substitute for optical LWP retrieval.

Accretion models also require droplet size distributions that neither ERA5 nor MODIS provides directly. The standard approach uses climatological median volume diameter assumptions from the literature, which introduces perhaps ±30% uncertainty in accretion rate estimates. For a first-pass structural risk screen across a tower portfolio this is acceptable; for a single critical site, it is not.

From risk map to engineering specification

The icing climatology translates into two engineering inputs. First, an ice load map expressed in kg/m of conductor or antenna aperture, keyed to return period, which feeds structural calculations for tower legs, cross-arms and guy wires. Second, an icing-hours distribution that specifies the required duty cycle and energy budget for resistive or hot-air anti-icing systems. A site with 200 icing-hours per year concentrated in three or four events needs a different system design than one with 200 hours spread across the winter as persistent light rime.

Radiation pattern distortion is a separate but related concern. Rime and wet-snow accretion on a radome or directly on an exposed array shifts the phase centre, increases side-lobe levels and can detune resonant elements. The severity depends on accretion geometry, which is partly a function of which regime dominates. A glaze-ice event that coats a radome uniformly degrades gain less than an asymmetric rime plume that grows into the prevailing wind. Knowing the dominant wind direction during icing events, available from ERA5 directional statistics, allows antenna orientation to be chosen to minimise the worst-case distortion.

Satellize runs this analysis as a portfolio screening product, processing ERA5 and MODIS archives against a client-supplied tower coordinate list to produce per-site risk tiers and design-input tables. The same pipeline was adapted for agricultural microclimate work in the Kingdom of Tonga crop-estimation programme, where wet-bulb temperature and cloud liquid water fields drove frost-risk rather than ice-load outputs.

Typical figures

ERA5 spatial resolution	~31 km horizontal grid; 137 pressure levels vertically; hourly timestep
ERA5 archive depth	1940-present (satellite-constrained assimilation from 1979)
MODIS LWP spatial resolution	1 km at nadir; degrades to ~2 km at swath edges
MODIS revisit / archive	Twice daily (Terra + Aqua combined); archive from 2000-present
AMSR2 footprint	5–60 km depending on channel (6.9 GHz coarsest, 89 GHz finest); daily global coverage
Sentinel-3 SLSTR resolution	500 m (shortwave channels), 1 km (thermal channels); sub-daily repeat at mid-latitudes with 3A+3B
Icing regime classification accuracy	Synoptic events: high confidence; ridge-top orographic icing: ±30–50% frequency error without downscaling
Ice load uncertainty (accretion model)	Approximately ±30% in accretion rate from droplet-size assumptions alone
Deliverable formats	GeoTIFF climatology grids, CSV per-site risk tables, PDF engineering summary report
Design-standard compatibility	Outputs structured for IEC 60826 and EN 50341 load inputs; national standards on request

Analytics Satellize can run

Per-site icing-hours climatology	ERA5 wet-bulb temperature and wind speed phase-space classification against published ISO 12494 regime thresholds	CSV table of annual icing hours by regime (glaze, rime, wet-snow) per tower coordinate, with 10-, 25- and 50-year return period estimates
Ice load design values	Makkonen cylinder accretion model applied to ERA5 wind and temperature fields; calibrated against MODIS LWP for in-cloud rime component	Per-site ice load (kg/m) at specified return periods, formatted for structural engineer input
Supercooled liquid water cloud frequency map	MODIS Collection 6 cloud optical properties product (MOD06/MYD06); cloud-top temperature filter to isolate supercooled layers	GeoTIFF seasonal frequency grids at 1 km, covering the client's tower network bounding box
Dominant icing wind-direction statistics	ERA5 hourly wind direction conditioned on icing-classified hours; circular statistics per site	Wind rose plots for icing events per site, supporting antenna orientation decisions
Portfolio risk tier ranking	Composite score weighting icing frequency, dominant regime severity and tower elevation relative to ERA5 cell centroid	Ranked tower list with high/medium/low risk tiers and flagged sites requiring in-situ validation
Orographic downscaling adjustment	DEM-based lapse-rate correction to ERA5 temperature fields; ridge-exposure index from terrain analysis	Adjusted icing-hours estimates for high-elevation sites with documented uncertainty bounds

Sources

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.