Does PV-FSI Vary Systematically with UAV Inspection Altitude? Implications for Fault Spatial Morphology

UAV infrared thermography campaigns routinely image photovoltaic (PV) modules at multiple inspection altitudes to satisfy IEC TS 62446-3 quality requirements, yet whether the Photovoltaic Fault Severity Index (PV-FSI) varies systematically with altitude, and what that variation encodes, has not been examined. This study answers both questions. Using the DTU Drone Infrared Thermography Dataset (640 acquisitions, 43 modules, 7 defect types, 4 altitudes), this study shows that PV-FSI does vary systematically with inspection altitude and that the direction of variation encodes the spatial morphology of the underlying thermal fault. We define the Altitude Response Index (ARI) as the linear slope of module-mean PV-FSI on inspection altitude. Crack-family defects (cell cracks, interconnects, PID, and glass cracks) exhibit positive ARI (mean = +0.171 PV-FSI per meter) because spatial averaging at lower resolution amplifies distributed thermal disorder; high-intensity family defects (bypass diode and soiling) exhibit negative ARI (mean = -0.299) because hotspot pixels are diluted as spatial resolution decreases with altitude. The family-level ARI difference is statistically significant (Mann-Whitney U = 379.0, p < 0.001; Kruskal-Wallis H = 26.26, p = 0.0002 across seven defect types). Leave-one-out cross-validation classification of the 43 modules into defect families using ARI-based features achieves an accuracy of 0.977 and an AUC of 0.998. Combined with the snapshot PV-FSI score, ARI defines a two-dimensional severity-morphology diagnostic space that enables defect-family identification and differentiated maintenance action from a single standard multi-altitude UAV inspection campaign, without additional sensing, supervised training, or changes to the flight protocol.