Satellite Imagery and Aerial Analysis for Roof Damage Detection After Hailstorms

The Role of Satellite and Aerial Imagery in Hail Damage Assessment

While NEXRAD radar tells us where hail fell and how large it was, it cannot tell us what actually happened to roofs on the ground. Radar operates at atmospheric scales -- its resolution is measured in hundreds of meters, and it observes precipitation aloft, not damage at the surface. To bridge the gap between "hail fell here" and "this roof is damaged," we need eyes on the actual structures.

Historically, that meant sending an inspector to every potentially affected property -- a slow, expensive, and inconsistent process. Today, satellite and aerial imagery analysis offers a powerful complement to ground inspection, enabling large-scale damage screening that prioritizes the most affected areas and properties for follow-up.

At Hail Strike, satellite and aerial imagery analysis is a critical component of our data pipeline, providing the visual confirmation layer that transforms radar-based hail estimates into verified damage assessments. In this article, we explore the technology behind remote sensing for roof damage, its capabilities and limitations, and how it integrates with other data sources to produce reliable results.

Types of Imagery Used in Roof Damage Assessment

Commercial Satellite Imagery

Several commercial satellite constellations now provide imagery at resolutions sufficient for roof-level analysis:

Maxar (WorldView/Legion): Maxar operates the highest-resolution commercial satellites available, with the WorldView-3 satellite delivering panchromatic imagery at 31 cm resolution and 8-band multispectral imagery at 1.24 m resolution. Their newer Legion constellation provides sub-30 cm panchromatic imagery with faster revisit times.

Planet (SkySat): Planet's SkySat constellation provides imagery at approximately 50 cm resolution with daily revisit capability. While slightly lower resolution than Maxar, the daily coverage is valuable for ensuring pre-storm baseline imagery exists.

Airbus (Pleiades Neo): The Pleiades Neo constellation delivers 30 cm panchromatic and 1.2 m multispectral imagery, offering another high-resolution option with European and global coverage.

Aerial Imagery

In many cases, aerial imagery captured from manned or unmanned aircraft provides even higher resolution than satellite imagery:

Fixed-wing aircraft surveys: Companies like Nearmap and EagleView fly systematic aerial surveys of urban areas, capturing imagery at 5-8 cm resolution -- roughly 10 times finer than the best satellite imagery. These surveys are repeated multiple times per year, creating a rich archive for change detection.

Drone (UAS) surveys: Drones can capture imagery at 1-3 cm resolution for individual properties or small areas. While not scalable to large regions, drone surveys are invaluable for detailed inspection of high-value or high-priority properties. Many roofing contractors now use drone imagery as part of their inspection workflow.

Multispectral and Hyperspectral Imagery

Beyond standard visible-light (RGB) imagery, multispectral sensors capture data in additional wavelength bands that reveal information invisible to the human eye:

• Near-infrared (NIR): Healthy vegetation and certain roofing materials reflect strongly in NIR. Damaged areas show reduced NIR reflectance, especially where granule loss exposes darker underlayment.
• Short-wave infrared (SWIR): SWIR bands are sensitive to moisture content. Roofs with compromised waterproofing may show elevated moisture signatures in SWIR imagery.
• Thermal infrared (TIR): Damaged roof areas may exhibit different thermal signatures than intact areas, particularly during morning or evening periods when differential heating and cooling are most pronounced.

Multispectral analysis significantly improves damage detection accuracy compared to visible-light imagery alone, especially for subtle damage that is difficult to identify visually.

Computer Vision Techniques for Roof Damage Detection

Pre/Post-Storm Change Detection

The most powerful remote sensing approach for hail damage assessment is change detection -- comparing imagery of the same area captured before and after a storm event. The fundamental principle is simple: if a roof looks different after a storm than it did before, the difference is likely storm damage.

In practice, change detection involves several technical steps:

1. Image co-registration: Pre- and post-storm images must be precisely aligned to sub-pixel accuracy. Even small misalignment can produce false change signals, especially around building edges and roof ridges.

2. Radiometric normalization: Images captured at different times have different solar illumination angles, atmospheric conditions, and sensor calibration states. Normalization adjusts for these factors so that genuine changes in surface reflectance are not confused with imaging artifacts.

3. Change calculation: For each pixel or image segment, the algorithm calculates the magnitude and direction of change in relevant spectral bands. Common change metrics include spectral difference, spectral angle, and texture difference.

4. Change classification: Not all change represents damage. Trees may have leafed out or dropped leaves. Shadows shift with solar angle. Cars appear or disappear from driveways. Machine learning classifiers distinguish roof damage from other types of change based on spatial patterns, spectral characteristics, and context.

Object Detection and Semantic Segmentation

Modern deep learning approaches treat damage detection as an object detection or semantic segmentation problem:

• Object detection models identify and localize specific damage features such as missing shingles, tarp patches, or debris on roofs. These models output bounding boxes and confidence scores for each detected damage instance.

• Semantic segmentation models classify every pixel in the image into categories such as "intact roof," "damaged roof," "vegetation," "shadow," etc. This produces a complete damage map at the pixel level.

Hail Strike's AI models use a combination of both approaches, first segmenting roofs from their surroundings, then classifying roof condition at the pixel level within each detected roof polygon.

Texture Analysis

Hail damage to asphalt shingle roofs often manifests as granule loss -- the protective mineral granules are knocked off by hail impacts, exposing the darker asphalt underlayment. This granule loss creates a distinctive change in roof texture that can be detected in high-resolution imagery:

• Intact shingle surfaces have a relatively uniform, granular texture.
• Damaged surfaces show irregular dark spots (individual impacts) or broad areas of granule loss.
• Severely damaged roofs may show exposed underlayment, missing shingles, or visible holes.

Texture analysis algorithms quantify these differences using statistical measures of image texture (entropy, contrast, homogeneity) within each roof segment, flagging areas with significant texture changes as potential hail damage.

See how satellite imagery analysis works for your area. Sign up for Hail Strike to access post-storm damage maps, property-level assessments, and before/after imagery comparisons.

Integrating Imagery Analysis With Radar Data

The real power of satellite and aerial imagery for hail damage assessment emerges when it is combined with radar-derived hail data. Each data source addresses the weaknesses of the other:

Radar Provides Context for Imagery Analysis

Radar data tells the imagery analysis system where to look and what to expect:

• Geographic focus: Instead of analyzing imagery for an entire state, radar hail swath maps narrow the search to the specific areas where damaging hail occurred.
• Severity calibration: Radar-estimated hail size provides a prior expectation of damage severity, helping the imagery classifier set appropriate sensitivity thresholds. Areas with 2-inch hail should show more extensive damage than areas with 1-inch hail.
• Impact angle estimation: Wind data from radar volume scans indicates the likely angle of hail impact, which affects which roof faces are most damaged and helps the classifier focus on the correct roof slopes.

Imagery Validates and Refines Radar Estimates

Conversely, imagery analysis provides ground truth for radar-based predictions:

• Damage confirmation: If radar indicates 1.5-inch hail fell on a neighborhood and imagery shows widespread granule loss, the radar estimate is confirmed. If imagery shows no visible damage, the radar estimate may be biased.
• Spatial refinement: Radar hail swaths have resolution of 250 meters or more. Imagery analysis can refine the damage boundary to the individual property level, identifying which homes on a street were hit and which were spared.
• Damage quantification: While radar estimates hail size (a cause), imagery directly observes damage (an effect). Combining both yields a more complete picture than either alone.

At Hail Strike, this radar-imagery fusion is a core component of our StormClaim Score, which synthesizes all available data into a single, actionable metric for each property.

Limitations and Challenges

Temporal Coverage

The biggest challenge in satellite-based damage detection is ensuring imagery is available both before and after the storm event. While aerial survey companies like Nearmap capture urban areas multiple times per year, rural areas may have infrequent coverage. Cloud cover at the time of satellite overpass can also prevent useful imagery from being captured.

Hail Strike addresses this by maintaining subscriptions to multiple imagery providers and using the most recent available pre-storm image for each location, even if it was captured weeks or months before the event.

Resolution Limits

Even at 30 cm resolution, individual hail dents on a roof are too small to detect directly. Satellite imagery is best suited for detecting aggregate damage patterns -- widespread granule loss, missing shingles, or visible debris -- rather than individual impact marks. For properties where satellite analysis is inconclusive, ground-level or drone inspection remains necessary.

Roof Material Variability

Different roofing materials respond differently to hail and present different damage signatures in imagery:

• Asphalt shingles: Granule loss is relatively easy to detect as a darkening of the roof surface.
• Metal roofs: Denting does not change the color or texture of the roof in imagery, making metal roof damage nearly invisible to remote sensing.
• Tile roofs: Cracked or broken tiles can be detected if the fragments are displaced, but hairline cracks are not visible.
• Flat/membrane roofs: Punctures and tears may be detectable if they cause visible pooling or surface changes.

Hail Strike's models are trained to account for roof material type, adjusting detection sensitivity and confidence based on what type of damage is detectable for each material.

Lighting and Atmospheric Effects

Sun angle, cloud shadows, haze, and atmospheric scattering all affect imagery quality and can introduce artifacts that mimic or mask damage. Our processing pipeline includes atmospheric correction, shadow detection, and illumination normalization to minimize these effects, but some degradation in accuracy is unavoidable under poor imaging conditions.

The Future: Higher Resolution, Faster Revisit, Better AI

The capabilities of satellite and aerial imagery for roof damage detection are improving rapidly:

• Satellite constellations are growing: New constellations launching in 2026 and beyond will provide daily or even sub-daily revisit at 30 cm resolution, making it far more likely that clear pre- and post-storm imagery will be available.
• AI models are improving: As training datasets grow and model architectures advance, computer vision accuracy for damage detection continues to increase. Hail Strike's models are retrained with each major storm season's worth of new ground-truth data.
• LiDAR integration: Airborne LiDAR (Light Detection and Ranging) can detect 3D surface changes on roofs, potentially identifying dents and deformations that are invisible in 2D imagery. Hail Strike is evaluating LiDAR integration for future pipeline versions.

Practical Implications for Roofing Professionals

For roofing contractors, satellite and aerial imagery analysis offers several practical benefits:

• Pre-qualification: Before investing in a site visit, review Hail Strike's satellite-based damage assessment to determine which properties are most likely to need roof repair or replacement.
• Documentation: Before-and-after imagery provides compelling visual evidence for insurance claim supplementation.
• Efficiency: Focus your canvassing efforts on streets and neighborhoods where imagery confirms visible damage, rather than relying solely on radar hail swaths.
• Competitive advantage: Leverage Hail Strike's imagery analysis to reach high-probability damage properties before competitors who rely on less sophisticated methods.

For insurance adjusters, satellite imagery provides an objective, timestamped record of roof condition that can corroborate or challenge claim narratives.

Conclusion

Satellite and aerial imagery analysis represents a powerful complement to radar-based hail detection, bridging the gap between atmospheric observations and on-the-ground damage reality. By combining high-resolution imagery with computer vision, change detection, and multispectral analysis, platforms like Hail Strike can screen thousands of properties for hail damage in hours rather than weeks.

When fused with NEXRAD radar data and machine learning models, imagery analysis becomes even more powerful, enabling property-level damage assessments that are faster, more consistent, and more defensible than traditional manual inspection alone.

Ready to leverage satellite imagery analysis for your roofing business? Sign up for Hail Strike and gain access to post-storm damage maps, property-level imagery comparisons, and AI-powered damage scoring for your service area.