Registration Number
0226U000739,
(0125U001223)
, R & D reports
Title
DT4LC: Developing Scalable Digital Twin Models for Land Cover Change Detection Using Machine Learning
popup.stage_title
Розгортання та налаштування Open Data Cube (ODC) для України, збір навчальних і супутникових даних, розробка семантичної онтології для аналізу часових рядів і виявлення змін земного покриву.
Head
Shelestov Andrii Yu., д.т.н.
Registration Date
15-01-2026
Organization
National Technical University of Ukraine «Igor Sikorsky Kyiv Polytechnic Institute»
popup.description1
The main goal of the project is to develop scalable digital twins that will combine satellite data and artificial intelligence to detect complex land cover changes, and to provide decision-makers in Ukraine and Switzerland with advanced and practical products to monitor the effects of war and climate change, using modular machine learning architectures that are optimized for different territories.
popup.description2
During the first year of the DT4LC project, the conceptual, architectural, methodological and infrastructural foundations of a digital twin for the analysis of land use and land cover changes were laid. A systematic analysis of current digital twin initiatives in the field of Earth sciences, in particular DestinE, NASA Earth System Digital Twin and BioDT, was carried out, focusing on their suitability for the analysis of land use changes. A two-temporal concept of a digital twin was developed, which distinguishes between rapid changes (floods, military actions, extreme weather events) and gradual long-term transformations of land use. A hierarchical architecture of a digital twin, consisting of individual Digital Twin Instances (DTI) integrated through a Digital Twin Aggregator (DTA), was proposed. Fundamental models for the analysis of Earth observation data were integrated, including visual encoders, temporal transformation models and multimodal data fusion models. They were adapted to multispectral satellite images and regional conditions of Ukraine and Switzerland. A hybrid approach was developed that combines machine learning methods with PINN to ensure physical consistency of forecasts. The digital twin was applied to analyze land use changes caused by military actions and crisis events. The example of Ukraine demonstrated land degradation, disruption of agricultural activities and changes in vegetation cover. The concept of a cognitive interface for the digital twin was developed, which uses generative fundamental models to interact with users in natural language. The interface provides adaptive visualization of results, generation of explanations and transformation of analytical outputs into understandable insights.
Product Description
popup.authors
Alla M. Lavreniuk
Hanna O. Yailymova
Oleksandr A. Yavorskyi
popup.nrat_date
2026-01-15
