Moisture Monitoring in Historic Buildings: A Coordinated Image-Based Protocol for Time-Lapse Thermal and Visual Data

Abstract

Moisture is one of the main deterioration agents affecting historic buildings, yet current diagnostic approaches based on visual and thermal imagery mainly rely on one-off surveys. To address this gap, this study proposes a coordinated image-based protocol for periodic moisture monitoring, combining a supervised Random Forest (RF) classifier for mapping visible surface alterations on photogrammetric orthoimages, and an unsupervised Otsu thresholding method for segmenting moisture-related thermal anomalies on thermograms. The protocol was demonstrated on an abandoned convent in southern Italy affected by rainwater infiltration problems, through repeated photogrammetric and thermographic surveys conducted between December 2024 and December 2025 under comparable conditions, and integrated with environmental monitoring data. RF-based mapping highlighted marked spatial differentiation and seasonal variability in biological colonization, with winter maxima, summer minima, and autumn reactivation. Otsu-based segmentation similarly showed a clear seasonal trend in damp-area extent, closely related to rainfall distribution, with maximum values in winter and minimum values in summer. The combined interpretation of visual, thermal, and environmental data confirmed the persistence and seasonal recurrence of the phenomenon, supporting identification of its active dynamics over time. The results show the potential of relatively simple and accessible image-processing routines for multi-temporal moisture assessment in heritage contexts.