Dataset for Multispectral PPFD and DLI Monitoring under Dynamic Plant-Factory Lighting

This dataset contains time-resolved multispectral measurements collected for the development and evaluation of a compact PPFD and DLI monitoring system based on the AS7265x spectral sensor under controlled plant-factory lighting conditions. The dataset comprises 45,675 paired observations collected across 315 measurement sessions, 35 date blocks, three AS7265x sensor units, and three dynamic lighting profiles: constant lighting, sinusoidal lighting, and DLI-feedback lighting. Each session consists of 145 observations acquired at 300-second intervals over a nominal 12-hour photoperiod, including both photoperiod endpoints. The dataset includes measurements from the 18-channel AS7265x multispectral sensor, embedded AS7265x-based PPFD estimates, paired PPFD measurements from an LI-180 spectrometer used as the operational reference, commanded PPFD, lamp-power measurements, timestamps, sensor/device identifiers, lighting-profile identifiers, and session/date identifiers. The dataset was used to evaluate physics-based PPFD estimation, linear recalibration, and multichannel ridge-regression models, including date-grouped cross-validation, leave-one-device-out validation, leave-one-profile-out validation, and chronological future holdout. The data also support analyses of instantaneous measurement agreement, temporal tracking, DLI reconstruction, sensor transferability, spectral composition, longitudinal behaviour, and sensitivity to sampling interval. All observations retained in the analysis-ready dataset passed the predefined quality-control criteria, including timestamp uniqueness, paired reference measurements, finite analytical values, chronological ordering, expected sampling cadence, and complete session coverage. This dataset is intended to support research on multispectral sensing, PPFD measurement, DLI monitoring, sensor calibration, measurement system validation, sensor transferability, controlled-environment agriculture, and data-driven spectral measurement models.