Literature review• Role of IRI• Profiler accuracy• Prior validation of crowdsourced IRI

Study data• NIRA connected-vehicle IRI (Average, State Estimate)• ARAN inertial profiler IRI• Two networks: A6 & Waaksaam/Noord Holland

Preprocessing & cleaning• Geometry validation• Null handling• Weighted averaging (forward/backward)• Profiler Mean Roughness Index

Spatial alignment &amp; aggregation• 25 m base units (buffer rectangles)• Custom Python tool for snapping/centroid match• Aggregation to 100 m

Grouping &amp; binning• Geometry: Regular vs Not Regular• Regions: A6, Waaksaam, Combined• Severity bins from profiler distribution

Statistical framework• Outlier detection (IQR residuals)• Distribution checks (skewness, kurtosis, normality)• Non-parametric tests

Value-based analyses• Correlation & regression• Residual diagnostics

Agreement analyses• Bland–Altman (bias, LoA, RMSE)• Confusion matrices (RWS classes)• Agreement %, Cohen’s Kappa

Sensitivity analyses• Segment length: 25 m vs 100 m• Geometry: Regular vs Not Regular• Region: A6 vs Waaksaam

Results & discussion• Comparison with literature• Geometry-driven patterns• Operational interpretation

Conclusions & recommendations• Practical recommendations• Challenges & future directions

title