Snow accumulation mapping for roof structural risk in property portfolios
Passive microwave and optical satellite products can estimate snow water equivalent and flag rooftops approaching structural load limits, giving insurers and property managers a triage layer after major snowfall events.
Sensors
- AMSR2 (GCOM-W1): Passive microwave radiometer operating at 6.9 to 89 GHz. Retrieves snow water equivalent (SWE) over land at roughly 10 km spatial resolution, daily global coverage. Sensitivity to SWE degrades above approximately 150 mm and in wet-snow conditions where liquid water masks the microwave signal.
- MODIS MOD10 / MYD10 snow-cover products: Normalised Difference Snow Index derived from Terra and Aqua MODIS at 500 m resolution, daily. Distinguishes snow-covered from snow-free pixels but gives no depth or mass information. Cloud cover is the primary operational constraint; persistent overcast can blank coverage for days at a time.
- VIIRS snow cover (VNP10 / VJ110): Successor to MODIS snow products aboard Suomi-NPP and NOAA-20, 375 m resolution at nadir, daily. Improves on MODIS in edge-of-swath geometry and provides a cross-calibrated continuity record. Same cloud limitation applies.
- Sentinel-1 SAR (C-band, 5.405 GHz): Synthetic aperture radar at 10 m resolution, 6-day repeat at mid-latitudes with both Sentinel-1A and 1B. Dry snow is nearly transparent to C-band, but the transition to wet snow produces a sharp backscatter drop that can be detected reliably. Useful for identifying melt onset and confirming snow presence under cloud, not for direct SWE retrieval.
Why a roof doesn't know the forecast
Structural engineers design roofs to a ground snow load standard, adjusted for roof geometry and local climate. In many building codes, the design load for a flat commercial roof sits between 1.0 and 2.5 kN/m², depending on jurisdiction. A 200 mm accumulation of dense, wet snow can weigh close to 2 kN/m². The roof has no margin for error if drainage is blocked, if snow drifts pile against parapets, or if the building is older than its original load certification.
The problem for a portfolio manager or insurer is scale. After a significant snowfall event, an urban portfolio may contain hundreds of assets across a region. Ground inspection of each is impractical in the hours that matter. Satellite-derived snow products cannot replace a structural engineer, but they can sort a portfolio into a credible risk order before anyone puts on boots.
What passive microwave actually measures, and where it stops
AMSR2 retrieves SWE by comparing brightness temperatures at different microwave frequencies. Dry snow scatters higher-frequency radiation more strongly than lower-frequency radiation; the brightness temperature difference between the 18.7 GHz and 36.5 GHz channels is roughly proportional to SWE. The JAXA standard AMSR2 SWE product reports daily values globally at about 10 km grid spacing, with a published uncertainty of roughly 30 to 40 mm SWE under ideal dry-snow conditions.
The method has well-documented failure modes. Wet snow, which occurs during melt or rain-on-snow events, absorbs rather than scatters microwave energy and causes severe underestimation. Dense forest canopy attenuates the signal. Urban surfaces introduce thermal noise. And 10 km resolution means the product represents a landscape average, not a specific building. For rooftop risk, the satellite SWE estimate is best treated as a regional loading indicator rather than a per-asset measurement.
Combining products to close the gaps
No single sensor solves the problem. The practical approach fuses three data streams. AMSR2 SWE gives a mass estimate for the region surrounding each portfolio asset, updated daily. MODIS or VIIRS snow-cover fraction at 375 to 500 m confirms whether snow is present and provides a spatial pattern that the coarse microwave grid cannot resolve. Sentinel-1 SAR, compared against a pre-event baseline, detects the backscatter signature of wet-snow onset, which is operationally important because wet snow is both heavier and more likely to be underestimated by passive microwave.
Terrain and urban morphology matter too. Digital elevation data from the Shuttle Radar Topography Mission or Copernicus DEM identifies sites at higher elevation within the portfolio, where accumulation is typically greater. Wind-exposure proxies from terrain analysis can flag assets prone to drift loading. None of this replaces a site survey, but it structures the inspection queue.
The honest resolution problem for individual rooftops
It is worth being direct: no current free-access satellite product resolves individual rooftops for snow mass. AMSR2 at 10 km and MODIS at 500 m are landscape-scale instruments. Very high resolution optical imagery from commercial providers can show snow presence and rough depth by shadow geometry, but converting that to SWE requires assumptions about snow density that introduce substantial error.
What satellite analytics can reliably do is triage. A portfolio of 400 assets across a snowbelt region can be stratified by estimated regional SWE, elevation, roof type (flat versus pitched, from prior building-footprint data), and proximity to known drift-risk geometry. The top decile of that stratification, perhaps 40 assets, is where inspection resources go first. That is a meaningful operational output, even if the physics prevents a per-roof kilogram count.
Insurance triage after a snowfall event
The use case for insurers is time-sensitive. After a major snowfall, the window between peak accumulation and partial melt is typically 24 to 72 hours in temperate climates. AMSR2 data is distributed by JAXA with a latency of roughly 24 hours from observation. MODIS and VIIRS snow-cover products are available from NASA Earthdata within a similar window. Sentinel-1 passes repeat on a 6-day cycle, so a useful SAR acquisition may or may not fall within the critical window.
A practical workflow issues a portfolio-wide alert when regional SWE crosses a predefined threshold, ranks assets by combined risk score, and produces a prioritised inspection list as a GIS layer or structured report within hours of data availability. Satellize applies this kind of threshold-and-rank logic to open constellation data; the Tonga crop-estimation programme uses a similar fusion approach across optical and microwave products on a different agricultural problem. The underlying method is transferable.
The threshold itself requires calibration to local building stock. A 100 mm SWE alert is not equally meaningful in a portfolio of modern steel-frame warehouses and a portfolio of 1960s flat-roofed retail units. Integrating building-age and construction-type attributes from client data is what converts a generic snow map into a risk-ranked asset list.
What to ask before commissioning this analysis
Three questions determine whether satellite snow analytics will add value for a given portfolio. First, is the portfolio geographically concentrated in a high-snowfall region, or scattered thinly across a continent? Triage value is highest when many assets face the same event simultaneously. Second, does the client hold building-attribute data (construction year, roof type, design load rating) that can be joined to the spatial output? Without it, the risk ranking is coarser. Third, what is the acceptable latency? If the answer is under six hours, current open-data pipelines cannot reliably deliver; commercial SAR tasking and near-real-time VIIRS processing can compress that window but add cost and complexity.
If those conditions are met, the satellite approach offers something ground inspection cannot: simultaneous coverage of an entire portfolio at the moment of peak risk, with a documented, repeatable methodology that can be audited after a loss event.
Typical figures
| AMSR2 SWE spatial resolution | ~10 km grid (JAXA standard product) |
| MODIS / VIIRS snow-cover resolution | 500 m (MODIS MOD10), 375 m (VIIRS VNP10 at nadir) |
| Sentinel-1 SAR resolution | 10 m (IW mode ground range detected) |
| Revisit frequency | AMSR2: daily global; MODIS/VIIRS: daily (cloud-permitting); Sentinel-1: 6-day at mid-latitudes |
| Data latency (open products) | Approximately 12 to 24 hours from observation for AMSR2 and MODIS/VIIRS standard products |
| AMSR2 SWE uncertainty (dry snow) | Approximately 30 to 40 mm SWE under ideal conditions; degrades significantly for wet snow |
| Microwave frequency channels used | 18.7 GHz and 36.5 GHz brightness temperature difference (primary SWE retrieval); 89 GHz for grain-size sensitivity |
| Cloud penetration | Passive microwave and SAR: cloud-transparent. Optical (MODIS/VIIRS): blocked by cloud cover |
| Archive depth | MODIS snow products from 2000; AMSR2 from 2012; Sentinel-1 from 2014 |
| Minimum detectable SWE (AMSR2) | Approximately 10 to 20 mm SWE; shallow or patchy snowpacks near this threshold are unreliable |
Analytics Satellize can run
| Regional SWE surface for portfolio footprint | AMSR2 brightness temperature difference algorithm (18.7 and 36.5 GHz), mosaicked and clipped to portfolio bounding box | GeoTIFF raster updated daily during snow season, with per-asset zonal statistics in CSV |
| Snow-cover presence and fraction per asset | NDSI thresholding from MODIS MOD10 or VIIRS VNP10, aggregated to building footprint polygons | GIS polygon layer with snow-cover fraction attribute, refreshed daily |
| Wet-snow onset detection | Sentinel-1 C-band backscatter change detection against pre-event baseline; threshold on sigma-naught drop | Binary wet-snow alert layer per SAR acquisition cycle, flagging assets where melt-phase loading risk is elevated |
| Portfolio risk ranking after snowfall event | Composite score combining regional SWE estimate, elevation from Copernicus DEM, roof-type attribute (flat vs. pitched), and snow-cover fraction; ranked by percentile | Prioritised inspection list as PDF report and GIS layer, issued within 24 hours of triggering SWE threshold |
| Historical snowfall exceedance frequency per asset | Return-period analysis on AMSR2 SWE archive (2012 to present) and MODIS snow-cover record (2000 to present) at each portfolio location | Asset-level table of 10-year, 20-year and 50-year SWE exceedance estimates for underwriting reference |
| Drift-risk exposure flag | Terrain wind-exposure proxy from DEM aspect and curvature combined with building footprint geometry to identify parapet-bounded flat roofs at high-drift sites | Attribute flag in portfolio GIS layer, updated once per season or on portfolio change |
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.