Data and code for "Calibration of simulated spectral indoor daylight models"

This repository contains the data and code for the PhD chapter <strong>"Calibration of simulated spectral indoor daylight models."</strong><strong>Summary of the project:</strong> This chapter introduces and tests a practical calibration workflow for indoor spectral daylight simulation. The workflow aims to account for external sources of uncertainty—namely (1) sun–sky conditions, (2) exterior/context obstructions, and (3) window spectral transmittance—and restricts calibration to the latter. Using short in situ measurements in a University of Toronto classroom, simulated 18-channel irradiances from <em>Radiance</em> v6.0 are compared to field data, and the transmittance spectrum is scaled. Across options, simulated–measured spectral shapes agree well (Pearson <em>r</em> ≈ 0.89–0.92); the best window yields MAE₍room,w₎ = 0.0093 W m⁻². Post-calibration, systematic bias on the control sensor is largely removed for photopic and melanopic illuminance (MBE₍rel₎ ∈ [−5.79%, 2.71%]), with overall error magnitude and dispersion reduced to RMSE₍rel₎ ≈ 18% and σ₍rel₎ ≈ 18%. The melanopic/photopic (M/P) ratio remains stable (∼ 9–10% errors). Errors increase with distance from the windows. Limitations include a single room, short overcast monitoring, and ambiguity in two reported GHI series. The results show that scaling transmittance alone is an effective approach to reduce bias in <em>Eᵥ</em> and <em>EML</em> while preserving spectral balance.