Technical White Paper

RoofRecon Pro: Measurement Methodology and Data Integrity Framework

Version 1.0  ·  April 2025  ·  For distribution to insurance carriers, independent adjusting firms, and engineering reviewers

Overview

RoofRecon Pro is a satellite-derived roof measurement and claim documentation platform designed for use by roofing contractors and insurance professionals. This document describes the technical methodology underlying the platform's measurement outputs, storm event correlation, and claim documentation package — with specific attention to the data sources, processing steps, accuracy characteristics, and known limitations relevant to insurance claim evaluation.

The platform is not a replacement for physical inspection. It is a documentation tool that produces independently verifiable data to support the inspection process, reduce supplemental requests, and establish a structured evidentiary record for claim files.

Important: Satellite-derived measurements are subject to the limitations described in Section 5. Physical inspection by a qualified professional remains the authoritative source for damage assessment and scope of loss determination. RoofRecon measurements are intended to corroborate, not replace, field inspection findings.

Measurement Methodology

2.1 Imagery Source

All roof measurements are derived from high-resolution satellite imagery provided through the Google Maps Platform. Imagery resolution varies by location and availability, with urban and suburban areas typically receiving imagery at 15–30 cm ground sample distance (GSD). The imagery date is recorded in every report and reflects the most recent available capture for the property coordinates.

Imagery quality is rated on a four-tier scale (Excellent / Good / Fair / Poor) based on resolution, cloud cover, shadow coverage, and seasonal vegetation interference. Reports with Fair or Poor imagery quality receive a reduced confidence score and are flagged for manual review.

2.2 Segmentation Process

Roof plane segmentation is performed using a computer vision pipeline trained on a dataset of verified residential and commercial roof structures across climate zones in the continental United States. The pipeline identifies individual roof planes, classifies edge types (ridge, hip, valley, eave, rake), and assigns pitch estimates based on shadow analysis and structural pattern recognition.

Each identified segment is assigned a confidence score reflecting the model's certainty in the boundary definition and pitch classification. Segments with confidence below the user-configured threshold are flagged in the report with a specific notation.

2.3 Measurement Outputs

The following measurements are produced for each report:

MeasurementUnitMethod
Total roof areaSquare feetSum of all segment areas, slope-adjusted
Area with waste factorSquare feetTotal area × waste multiplier (10–15% based on complexity)
Ground footprintSquare feetProjected horizontal area
Pitch per segmentX/12Shadow analysis and structural pattern recognition
Ridge lengthLinear feetDetected ridge line segments, summed
Hip lengthLinear feetDetected hip line segments, summed
Valley lengthLinear feetDetected valley line segments, summed
Eave lengthLinear feetDetected eave line segments, summed
Rake lengthLinear feetDetected rake line segments, summed
Segment countIntegerNumber of distinct roof planes identified

2.4 Confidence Scoring

Each report receives an overall confidence score from 0 to 100, computed as a weighted composite of imagery quality, segmentation certainty, pitch estimation reliability, and geometry validation results. The score is displayed prominently on the report and in the contractor's dashboard.

Contractors may configure a minimum confidence threshold below which reports are withheld from clients pending manual review. The default threshold is 40/100. Reports with scores below 70 are considered lower-confidence and should be supplemented with physical inspection findings before use in claim documentation.

Storm Event Data

3.1 NOAA NEXRAD SWDI

Storm event data is sourced from the NOAA Severe Weather Data Inventory (SWDI), which aggregates NEXRAD (Next Generation Radar) detections of hail and high-wind events across the continental United States. The SWDI is maintained by the NOAA National Centers for Environmental Information (NCEI) and is publicly accessible at ncdc.noaa.gov/swdiws.

For each report, the platform queries the SWDI API for hail and wind events within a 35-mile radius of the property coordinates over the preceding 24 months. The 35-mile radius is the standard search radius used in insurance hail claim investigations and corresponds to the typical detection range of a single NEXRAD station for hail events.

3.2 Hail Data

Hail detections are reported as MESH (Maximum Expected Size of Hail) estimates derived from NEXRAD reflectivity data. MESH values represent the estimated maximum hail diameter in inches at the detection point. The platform records the event date, MESH value, and the NEXRAD station ID for each detection.

The insurance industry standard threshold for compensable roof damage from hail is 1.0 inch diameter. Events at or above this threshold are highlighted in the report with an insurance documentation note. Events below 1.0 inch are included in the event history but are not flagged as damage-causative.

Verification: All NOAA NEXRAD hail detections can be independently verified at ncdc.noaa.gov/swdiws using the event date, coordinates, and NEXRAD station ID included in the report.

3.3 Wind Data

Wind events are sourced from NEXRAD Tornado Vortex Signature (TVS) detections, which identify rotational wind signatures associated with severe thunderstorm wind events. Maximum velocity is reported in meters per second and converted to miles per hour in the report interface. The standard threshold for shingle uplift damage is 58 mph (EF0 tornado equivalent), which corresponds to approximately 25.9 m/s.

3.4 Historical Weather Context

In addition to NOAA NEXRAD data, the platform queries the Open-Meteo historical weather API for 5-year daily weather data at the property coordinates. This data is used to compute freeze-thaw cycle counts, annual precipitation totals, and long-term wind exposure metrics — providing context for age-related deterioration patterns that may be relevant to claim causation analysis.

Accuracy and Validation

4.1 Measurement Accuracy

Internal validation against manual takeoffs on a sample of 500 residential structures across five climate zones produced the following accuracy characteristics:

MetricMean Error95th Percentile
Total roof area±1.8%±4.2%
Individual segment area±3.1%±7.8%
Pitch classification±0.5/12±1.5/12
Ridge/hip/valley length±2.3%±5.9%
Eave/rake length±1.9%±4.7%

Validation sample: 500 residential structures, 2022–2024. Structures with confidence score below 50 were excluded from the accuracy analysis.

4.2 Confidence Score Correlation

The confidence score is a reliable predictor of measurement accuracy. Reports with confidence scores above 80 show mean total area error of ±1.2%. Reports with scores between 50 and 80 show mean error of ±2.8%. Reports below 50 show mean error of ±6.1% and should be treated as preliminary estimates requiring physical verification.

Carriers and adjusters are advised to apply additional scrutiny to reports with confidence scores below 70, and to require physical inspection confirmation for reports below 50.

Limitations

The following limitations are inherent to satellite-derived measurement and should be understood by all parties using RoofRecon reports in claim documentation.

Imagery Currency

Satellite imagery is not captured in real time. The imagery date in each report reflects the most recent available capture, which may be weeks to months prior to the inspection date. Structural changes, storm damage, or repairs made after the imagery date will not be reflected in the measurements.

Vegetation Interference

Tree canopy, overhanging branches, and dense vegetation can obscure roof planes and reduce segmentation accuracy. Affected areas are typically flagged in the geometry validation section of the report, but partial occlusion may not always be detected.

Complex Roof Geometries

Highly complex roof structures — particularly those with multiple dormers, turrets, curved surfaces, or non-standard intersections — may produce lower segmentation accuracy. The confidence score will typically reflect this complexity, but manual verification is advisable for structures with more than 12 distinct roof planes.

Pitch Estimation

Pitch is estimated from shadow analysis and structural pattern recognition, not from direct measurement. Low-slope roofs (below 3/12) are particularly susceptible to pitch misclassification and should be verified by physical inspection before use in material quantity calculations.

NOAA Storm Data Coverage

NEXRAD coverage is not uniform across the continental United States. Rural areas, particularly in the western states, may have gaps in radar coverage that result in underreporting of storm events. The absence of NOAA-recorded events does not preclude storm damage; physical inspection remains the authoritative method for damage determination.

NOAA Data Latency

NOAA SWDI data is typically available within 24–72 hours of a storm event. Reports generated immediately after a storm may not include the most recent events. The data fetch date is recorded in the report and should be considered when evaluating storm history completeness.

Scope of Loss Determination

RoofRecon measurements provide area and linear footage data. They do not assess damage severity, material condition, or scope of loss. Damage assessment and scope determination require physical inspection by a qualified roofing professional or licensed adjuster.

Data Sources

Google Maps Platform — Satellite Imagery

maps.google.com

High-resolution satellite imagery for roof segmentation and measurement. Imagery date is included in every report.

NOAA NEXRAD SWDI — Storm Events

ncdc.noaa.gov/swdiws

Radar-verified hail and wind event detections. Publicly accessible for independent verification.

Open-Meteo — Historical Weather

open-meteo.com

5-year daily weather history for freeze-thaw cycle analysis and long-term exposure assessment.

NOAA National Weather Service — Active Alerts

api.weather.gov

Current NWS alerts for the property location at time of report generation.

NOAA Solar Resource Data

nsrdb.nrel.gov

Solar irradiance data for solar potential assessment. Not used in claim documentation.

Use in Claim Documentation

7.1 Appropriate Uses

Establishing baseline measurements for scope of loss estimation
Corroborating contractor-submitted measurements
Providing independently verifiable storm event history for causation analysis
Documenting adjuster access and annotation history for the claim file
Generating structured Xactimate input data from verified measurements
Creating a timestamped photo documentation record for the claim package

7.2 Uses Requiring Supplemental Verification

Scope of loss determination (requires physical inspection)
Damage severity assessment (requires physical inspection)
Material condition evaluation (requires physical inspection)
Causation determination for claims in areas with known NEXRAD coverage gaps
Measurement-based settlements on structures with confidence scores below 70

7.3 Report Integrity

Each RoofRecon report is assigned a unique report number and generation timestamp. The report number is traceable to the specific imagery date, processing parameters, and data sources used in its generation. Reports cannot be retroactively modified after generation; any re-analysis of the same address produces a new report with a new report number.

The adjuster access log records the name, company, email, claim number, access timestamp, and IP address of every party who accessed the report through the shared link. This log is available to the contractor and can be provided to the carrier as part of the claim file.

RoofRecon Pro Measurement Methodology and Data Integrity Framework, Version 1.0

© 2026 RoofRecon Pro. This document may be reproduced for insurance claim documentation purposes with attribution.