Climate-driven microclimate dynamics and pollution-induced degradation forecasting in painted cave environmental systems

Introduction Painted cave temples are highly vulnerable to environmental fluctuations, material degradation, and increasing anthropogenic pressures. Existing conservation approaches often lack an integrated framework that combines scientific monitoring, predictive analytics, international conservation standards, and indigenous conservation knowledge. This research presents an integrated four-phased framework for preventive conservation of painted cave temples, conceptually uniting Chinese traditional conservation philosophy with international environmental standards, including ISO 11799 and EN 15757. Methods Representative sites were selected through Multi-Criteria Decision Analysis (MCDA) based on accessibility, degradation intensity, and cultural value. Ethnographic surveys and advanced material characterization techniques, including X-ray Diffraction (XRD), Fourier Transform Infrared Spectroscopy (FTIR), handheld microscopy, and UV-fluorescence imaging, were employed to establish a scientific baseline and document indigenous conservation practices. A solar-powered, LoRaWAN-based IoT sensor network was deployed across key cave zones to continuously monitor temperature, relative humidity, and PM2.5 concentrations. Sensor calibration was performed bi-monthly, while data processing incorporated aggregation and Kalman smoothing. A standards-based dynamic threshold algorithm adapted from EN 15757 was developed to assess microclimatic risks, with thresholds adjusted according to local environmental variability, material vulnerabilities, and historical climate records. Furthermore, a SelfConvLSTM-based degradation risk forecasting model incorporating multi-modal inputs was developed, with feature selection optimized using Quantum Particle Swarm Optimization (QPSO). Results The integrated framework enabled continuous environmental monitoring, dynamic risk classification, and predictive assessment of degradation processes. The adaptive thresholding approach improved the identification of microclimatic risks by accounting for site-specific environmental conditions and material sensitivities. The SelfConvLSTM forecasting model effectively integrated sensor observations, material characteristics, and historical site patterns to support degradation risk prediction and early-warning capabilities. Discussion The proposed framework demonstrates how preventive conservation can be enhanced through the integration of traditional conservation knowledge, environmental monitoring, standards-based risk assessment, and artificial intelligence-driven forecasting. By supporting real-time monitoring, early intervention, and culturally informed decision-making, the framework offers a scalable and replicable model for sustainable heritage management of environmentally vulnerable cave temple sites.