https://repository.cyi.ac.cy/handle/123456789/857 Thu, 30 Apr 2026 23:07:53 GMT 2026-04-30T23:07:53Z https://repository.cyi.ac.cy/handle/CyI/2631 Title: Hypothetical Reconstruction for the Conservation, Preservation and Valorisation of Cultural Heritage: the Kampanopetra Basilica in Salamis, Cyprus Authors: Faka, Marina; Orabi, Rahaf; Tsagka, Anastasia; Papageorgiou, Andreani; Vassallo, Valentina; Hermon, Sorin; Bakirtzis, Nikolas Editors: Campana, S; Ferdani, D; Graf, H; Guidi, G; Hegarty, Z; Pescarin, S; Remondino, F Abstract: This article describes a digital documentation and visualisation project pursued by the Andreas Pittas Art Characterization (APAC) Laboratories of the Science & Technology in Archaeology and Culture Research Center (STARC) in the framework of the work of the Technical Committee for Cultural Heritage (TCCH), funded by the EU and implemented by United Nations Development Programme (UNDP) in Cyprus. The project’s aim was to create a hypothetical 3D (virtual reconstruction and maquette) of the Kampanopetra basilica in ancient Salamis, one of the largest Early Christian churches in Cyprus. The basilica complex is an archaeological site excavated more than 50 years ago and is in need of continuous conservation and special protection. The 3D outcome is useful to map the present state of preservation, for its future conservation and cultural valorisation. The workflow included 3D on-site documentation with image and range-based techniques combined with topographic measurements. The 3D hypothetical reconstruction model included 3 main parts: the documentation process, the authoring process and the integration of the model within the collaborative platform. The 3D reconstruction benefitted from the plans and drawings included in the archaeological report, combined with the utilisation of the 3D documentation of the site along with comparative material - namely examples of contemporary basilica structures in Cyprus and the broader Eastern Mediterranean basin. The produced Reconstruction Models are hosted in two different Web Viewers, the 3D HOP and ATON. The research team pursues key questions, research problems and innovative approaches in archaeology and cultural heritage through the application of advanced science and technology and integrated expertise in humanities, digital heritage and visualisation. The hypothetical reconstruction provides a general visualisation which can be used to inform the general public but also to provide the basis for its systematic and archaeologically detailed representation in the future. Wed, 01 Jan 2025 00:00:00 GMT https://repository.cyi.ac.cy/handle/CyI/2631 2025-01-01T00:00:00Z https://repository.cyi.ac.cy/handle/CyI/2604 Title: Application of deep learning algorithms in aerosol science Authors: Kanawade, Vijay; Jokinen, Tuija; Pikridas, Michael; Sciare, Jean Abstract: Atmospheric aerosols influence the Earth’s radiative balance through scattering/absorption and by modifying cloud microphysics, air quality and human health. However, aerosol forcing remains the largest uncertainty in our ability to accurately estimate the effective radiative forcing. This uncertainty primarily stems from the spatial variability of precursor emissions and the complexity of the physicochemical mechanisms driving particle formation and growth. New particle formation (NPF) is the largest source of aerosol numbers. NPF involves nucleation and condensation of low-volatility vapors, such as sulfuric acid or highly oxidized organic compounds. Traditionally, identifying NPF events relies on the manual visualization of aerosol number size distributions, which is both time-consuming and subjective. To overcome this challenge, we applied a deep learning-based object detection algorithm, You Only Look Once (YOLO) version 8, to automatically identify and classify NPF events from contour plots of aerosol number size distributions. The algorithm was trained and tested using data from 20 globally distributed sites across Asia, Europe, Africa, and North America. The YOLO model effectively detected the characteristic banana-shaped patterns of NPF events and estimated event start times and particle growth rates in different size ranges. Across all sites, its accuracy ranged between 0.74 and 0.97 at lower confidence levels, demonstrating strong robustness and adaptability to diverse atmospheric conditions. This study shows that YOLO can serve as a fast, accurate, and scalable tool for automated NPF detection, enabling more efficient global monitoring of aerosol formation. The approach strengthens our ability to quantify aerosols, enhance climate modeling, and deepen our understanding of the role of fine particles in Earth’s atmospheric system. Fri, 19 Dec 2025 00:00:00 GMT https://repository.cyi.ac.cy/handle/CyI/2604 2025-12-19T00:00:00Z https://repository.cyi.ac.cy/handle/CyI/2536 Title: Identification Of New Particle Formation Events Using Deep Learning Algorithm Authors: Kanawade, Vijay; Pikridas, Michael; Sciare, Jean; Jokinen, Tuija Abstract: Attached.; Atmospheric new particle formation (NPF), involving the formation of nanometer-sized molecular clusters and their subsequent growth to larger particle diameters, significantly impacts air quality, weather, climate and human health. NPF events are primarily identified through visualization of particle number size distributions, which is a subjective and time-consuming process. Here, we utilized a fast object detection deep learning algorithm, You Only Look Once (YOLO) version 8, to identify the features of NPF events. The YOLO algorithm was trained to automatically identify and annotate the visual appearance of particle size distributions and classify a given day into NPF event. A few previous studies have attempted to detect NPF events using deep learning methods (Fogwill et al., 2024; Joutsensaari et al., 2018; Su et al., 2022). Joutsensaari et al. (2018) were the first to apply transfer learning with a convolutional neural network (CNN) to classify NPF events from particle size distributions using a 15-year dataset from San Pietro Capofiume, Italy. Later, Su et al. (2022) applied a Mask R-CNN to localize the spatial patterns of NPF events and Zaiden et al. (2018 ) applied Bayesian classification to long-term data from the Station for Measuring EcosystemAtmosphere Relations (SMEAR) II station in Hyytiälä, Finland. Recently, Fogwill et al. (2024) used a statistical algorithm, Hidden Markov Models (HMM), to detect NPF events objectively and efficiently with reduced complexity. The primary goal of this work is to demonstrate the applicability and robustness of YOLO in detecting NPF event features in global data. This approach can be extended to new datasets from additional locations, further improving the algorithm's performance over time. Fri, 29 Aug 2025 00:00:00 GMT https://repository.cyi.ac.cy/handle/CyI/2536 2025-08-29T00:00:00Z https://repository.cyi.ac.cy/handle/CyI/2525 Title: Setting Up an Infinite Slope Stability Analysis on a High-Resolution DEM (0.21 × 0.21 m2) of a Mechanically Terraced Slope in Cyprus Authors: Djuma, Hakan; Zoumides, Christos; Keleshis, Christos; Constantinides, Christos; Leonidou, Andreas; Faka, Marina; Bruggeman, Adriana; Papageorgiou, Andreani Abstract: For centuries, the landscape of Mediterranean mountains has been shaped by the construction of terraces often supported by dry-stone walls, to allow farming activities. Nowadays, traditional dry-stone wall terraces are often replaced or combined with mechanically constructed terraces. Compared to dry-stone terraces, these behave differently regarding water retention, soil erosion and slope stability. The aims of this study are (i) to characterize a mechanically terraced slope (0.1 km2) partially protected by dry-stone walls and (ii) to set up a high-resolution stability analysis to verify its capabilities in predicting instabilities. The study site is located in the Troodos Mountains (Cyprus). Field observations were made between October 2021 and April 2022. Measurements of depth, hydraulic conductivity and bulk density, as well as the collection of samples for texture analyses, were performed for the physical characterization of soils. The study site was visited after several storms to map areas affected by instabilities. Rainfall was measured by a local meteorological station. A photogrammetric survey was made with a GoPro9 camera mounted on an unmanned aerial vehicle developed by the Unmanned System Research Laboratory of The Cyprus Institute both in firmware and software. The collected data were used as input into FSLAM, an open-source model that couples a simplified hydrologic model with an infinite slope stability analysis. The soil was a loamy sand, with a depth ranging between 0.30 and 0.80 m, an average bulk density of 1.70 g/cm3 and a hydraulic conductivity of 2.05 × 10–3 m/s. From the UAV flights, a DEM with a horizontal resolution of 0.035 m was produced. However, for modelling purposes the DEM was resampled at 0.21 m, to reduce calculation times and avoid high depth-to-length ratios at the single cell (infinite slope assumption). Preliminary results from multiple runs showed that the model responds reasonably well to flow accumulation and variations in soil depth. The model predicts higher instabilities for wetter deep soils. However, at this resolution it was not able to identify specific locations of failure. To improve the model output, FSLAM could be integrated with a routine able to process variable soil depths (now constant within soil units) and to route surface runoff. Mon, 01 Jan 2024 00:00:00 GMT https://repository.cyi.ac.cy/handle/CyI/2525 2024-01-01T00:00:00Z