Ali Deger Ozbakir

Managing residential energy efficiently without compromising comfort is challenging. The key difficulty lies in predicting occupancy in real time. Intelligent control strategies are widely used for residential energy management. As sensors become increasingly common in homes, occupancy prediction has gained significant attention as one such approach. Traditional occupancy prediction methods analyze temporal patterns within individual rooms but overlook spatial connectivity. This lack limits the accuracy of occupancy-forecasting methods. Incorporating spatial information could improve prediction accuracy and enhance the effectiveness of occupancy-based energy management. One promising approach to represent such spatial information is through the use of embeddings, which encode spatial connectivity in a data-driven vector space. Another complementary method is to represent the building layout as a graph, where nodes correspond to rooms and edges capture spatial connectivity. In this study, we evaluate multiple occupancy prediction architectures, including Convolutional Neural Networks (CNN) and Long Short-Term Memory networks (LSTM). These models are further enhanced by positional encoding (PE) and Graph Neural Networks (GNN) that explicitly encode spatial connectivity among rooms. Results indicate that spatial embeddings improve predictive performance, with graph-based methods providing the highest performance. Among all models, the hybrid GNN + PE + CNN yielded the best overall performance. However, considering practical heating control scenarios, balancing comfort, energy savings, and computational complexity, we found that LSTM + PE emerges as the most suitable choice.

Landslide mapping is essential for hazard assessment and disaster response, and methods based on Earth observation (EO) enable the mapping of large areas impacted by major disasters. These methods, however, often rely on cloud-free optical images, which are rarely available in high-rainfall areas prone to landslides, delaying timely detection. Furthermore, mosaicking multiple consecutive images to eliminate cloud cover discards valuable temporal information, such as the actual timing of landslide occurrences and the progression of their extents over time. To address these challenges, we introduce a novel method that processes successive partially cloudy images to detect visible landslide extents and automatically aggregates this information for rapid first detection and accurate spatiotemporal mapping of landslides. The model-agnostic method supports various EO-based landslide detection models from the literature. It uses binary model outputs (landslide / no landslide), associated uncertainty levels (if available), and cloud mask data together with cloud uncertainty to classify individual image cells into four states: landslide, background, unknown (e.g., cloud covered or other unusable data), and anomaly (e.g., identified as landslide despite cloud cover). A confidence level is also calculated for each cell state. The method continuously refines cell states by analyzing time series data from successive images, reducing unknowns and anomalies to improve landslide detection accuracy. Alternating labels are considered as an indication of uncertainty, whereas cells without a clear pattern are classified as unknown. The method generates robust, time-aware landslide maps by integrating spatial classification from model outputs and cloud masks with temporal consistency checks. We present an overview of the developed method and demonstrate its practical application through a case study conducted in Adıyaman, Türkiye. The study focuses on landslides triggered by the February 2023 Türkiye earthquake sequence and a subsequent rainfall event in March 2023. Using various landslide detection methods (e.g., an NDVI-based approach and a deep learning model) and optical EO data with different ground resolutions (e.g., Sentinel-2, Planet SuperDove), the case study showcases the method’s ability to enhance temporal insights into landslide occurrence and progression. These results underline its potential as a valuable tool for rapid hazard monitoring and disaster response.

The Sürgü-Cardak fault ruptured ~9 hours after the Mw 7.8 mainshock on the East Anatolian Fault zone (EAFZ) during the 2023 Kahramanmaraș earthquake sequence. With a moment magnitude Mw 7.5, involving up to 11 m slip, this event featured comparable slip magnitudes as the mainshock. Published strain rate fields based on geodetic observations do show strain accumulation around the EAFZ, but strain accumulation around the Sürgü-Cardak fault appears to be absent. We therefore reexamine the GNSS-based interseismic strain rate field to see whether, or not, the Sürgü-Cardak fault accumulated significant slip deficit prior to the earthquake.We use GNSS data from eastern Anatolia. To estimate strain rates and their uncertainties in regions that experience both fast and slow deformation rates, we employ a tailored stochastic interpolation technique. With this method we show that the strain rate peaks around the Sürgü-Cardak fault. To better interpret 2D strain rate fields around faults, we develop a novel decomposition of the strain rate tensor and its covariance, that allows us to estimate the strain rate in a fault-oriented frame. The decomposition method is analogous to descriptions of deformation in structural geology, and allows for direct comparison with slip types from focal mechanisms. Not only does the strain rate peak around the Sürgü-Cardak fault exceed the uncertainty, the direction of interseismic slip deficit accumulation is also compatible with the coseismic slip direction.We conclude that interseismic slip deficit accumulation on the Sürgü-Cardak fault was previously missed. Coseismic slip is consistent with the loading history. The Sürgü-Cardak earthquake thus has most likely been triggered by the mainshock.

Landslides cause severe damage to the built environment and communities, requiring effective hazard management. Landslide catalogs, which provide essential data on past landslide occurrences, are the primary data sources for this purpose. Furthermore, they enable training and validation of predictive landslide models. Although landslide catalogs are largely compiled through manual mapping based on expert judgement, various advanced techniques using optical Earth observation (EO) imagery have been developed to automate and enhance the creation of such inventories. These methods, however, are mostly tested in specific case studies and they are not put into operation to detect landslides on a regular basis. Moreover, they rely on cloud-free imagery that can be time-consuming to gather, resulting in delays in the timely detection of landslides. This is especially true in regions with frequent rainfall, such as mountainous areas, where landslides are more prevalent.The Landslide Hunter is a prototype online platform designed to reduce the gap by addressing cloud-cover-related omission in optical imagery, reducing delays in landslide detection, and providing an environment for testing and benchmarking of different EO-based landslide detection methods through a simple plug-and-play method. The platform monitors online resources for events that could trigger landslides, such as major earthquakes, and identifies regions where landslides are likely to have occurred in their aftermath. It then collects and analyzes consecutive optical EO images for these areas to identify visible landslide extents using various landslide detection models, ranging from simple index-based approaches (e.g., NDVI) to advanced machine learning techniques utilizing image segmentation. Proximity to cloud cover is used to assess whether a landslide extent is partially visible, with partial extents being marked for further tracking until complete landslide coverage is achieved through successive analyses. This enables the timely first detection and effective monitoring of landslides, even under cloudy conditions. The results are made available in an open-access landslide catalog through a user-friendly web portal, offering faster updates than traditional catalogs. Users are notified when new landslides are detected, facilitating rapid damage assessment efforts that can ultimately enhance the safety of communities and the built environment.We present a detailed overview of the design principles and operational framework of the Landslide Hunter platform, highlighting its core features, functionalities, and user interface. We also provide a thorough explanation of the data access methods developed to improve interoperability and ensure seamless integration with other systems. A live demonstration will illustrate how the platform automatically identifies and tracks landslides under cloudy conditions, enabling timely detection and monitoring of landslide progression.

The aim of this study is to develop a valid and reliable measurement tool for obtaining information about the ecological literacy of adults. For this purpose, a 72-item and three-part measurement tool was created in accordance with information in the literature and presented to experts for their opinions. After finalizing the scale in accordance with the feedback received, the 25-item 5-point Likert-type Ecological Literacy Awareness Scale, the 17-item true/false-type Ecological Literacy Knowledge Test (1), and the 17-item multiple-choice Ecological Literacy Knowledge Test (2) were created. The 59-item draft measurement tool was administered to two study groups. These study groups were selected by convenience sampling method, a non-probability sampling method. The data of the first group were used for exploratory factor analysis and the data of the second group were used for confirmatory factor analysis. It was determined that the finalized 37-item and three-part Ecological Literacy Measurement Tool for Adults (ELMT) was valid and reliable in determining the awareness and knowledge of teacher candidates about ecological literacy.

The Anatolia–Aegean domain represents a broad plate boundary zone, with the deformation accommodated by major faults bounding quasi-low deforming units. The main characteristics of the Anatolia-Aegean deformation were identified using a GNSS-derived velocity field. Recent advancements in GNSS measurements and networks have improved the spatial resolution of the Anatolia-Aegean deformation field, however, for a better understanding of the deformation, interstation distances that are smaller than fault-locking depth and consistent data processing using a single reference system are needed. Our goal is to address this gap and produce a uniform velocity solution.In this study, we processed the time series of 836 stations, of which 178 are published for the first time with sub-millimeter accuracy. With a period of up to 28 years, we present the most accurate velocity field with increased spatial and temporal resolution and homogeneity. We used the improved coverage of the velocity field to calculate strain accumulation on the North and East Anatolian Faults. Modeled slip rates vary between 20 and 26 mm/yr and 9.7 and 11 mm/yr for the North and East Anatolian faults, respectively. Further analysis of the data can help better understand the kinematics of continental deformation in general, and test outstanding hypotheses about the kinematics and dynamics of the Anatolia - Aegean domain in particular.

Continental deformation can be described in two end-member approaches: block (or microplate) and continuum models. The first considers a strong lithosphere with deformation localized in fault zones. For the latter, however, the lithosphere is weak and deforms as a thin viscous sheet. The Anatolia – Aegean domain represents both continuum and plate-like deformation. Furthermore, recent modeling studies suggest a dynamic support mechanism of the Anatolian plateaus, with dynamic topography estimates ranging from 1 to 3 km for various crustal models and geodynamic scenarios, although the gravity and crustal thickness data support predominant Airy isostasy. The solution to both intricacies relies on the thermal structure of the crust and the lithosphere. Available thermal considerations stem from either the uppermost mantle velocity structure or thermal modeling with assumptions on radiogenic heat production and boundary conditions. Yet, homogeneous and independent constraints on the lithospheric structure are scarce. We aim to contribute to this knowledge gap by providing Curie Point Depths (CPDs), which corresponds to the depth at which rock-forming minerals lose their magnetization at the Curie temperature, ~580 oC.Resolution of deep magnetic sources requires spectral methods with large windows, which reduce the CPD resolution. Moving & overlapping smaller windows have been used in order to increase the resolution, but these introduce spectral leakage and bias. In previous studies, subjective wavenumber ranges of the magnetic anomaly spectra were used, often combined with wrong scaling factors between map units and the equations. This resulted in generally erroneous CPD estimates. Furthermore, CPD uncertainties have often been unquantified for the study area. We use a wavelet transform method, which overcomes the artifacts due to segmentation of magnetic signal to finite windows, results in higher spatial resolution as well as enabling uncertainty estimation. We used as large an area as possible for constraining the edge effects away from the study area. The resultant CPD map spatially correlates well with low Pn velocity areas, locations of volcanoes, and thermal springs.