Spatiotemporal algorithms for microclimatic parameters assessment

In the past few years, extreme microclimate patterns in various locations throughout the world have caused extreme damage, both environmentally and economically. These issues have also been addressed, due to their importance, within the UN sustainable development goals, primarily in climate action (Goal 13) and secondarily in sustainable cities and communities (Goal 11). Furthermore, the Paris Agreement addresses rapid urbanization as an increasing cause for anthropogenic climate change. As the EU climate policy in place remained insufficient for achieving the Paris Agreements’ temperature increase limit, the EU Commission presented the European Green Deal (EGD) that aims for the EU to be climate neutral by 2050. Global climate changes are important driver of local microclimatic conditions (e.g., air pollution and high temperature), which directly affect human well being. Therefore, it is imperative to better understand the microclimatic parameters to improve local decision-making processes for climate action. At the same time, data acquisition by large-scale Earth Observation (EO) using remote sensing and in-situ sensing, has increased more than tenfold in the past few years. This enables new opportunities for better decision making and monitoring capabilities of microclimate parameters (e.g., snow cover, temperature, and air pollution). Environmental simulations and machine learning using EO data are nowadays among the most promising solutions to assess more complex environmental phenomena (e.g., heat transfer and wind dynamics) more accurately, in spatial and temporal dimensions. These approaches provide a foundation for predictive and prescriptive analytics and, therefore, yield further improvements in decision support systems within the cities and reduce the decision-making and monitoring costs. In recent years, there has a lot of attention given to machine learning (especially deep learning) approaches, which require large amount of learning datasets and still cannot fully explain complex microclimatic physical processes. State-of-the-art environmental simulations algorithms provide better explicit understanding of these processes; however, they are mostly done in low resolution over large-scale areas due to high computational complexities. With the proposed basic interdisciplinary basic research project Spatiotemporal Algorithms for Microclimatic Parameters Assessment (SAMPA), the aforementioned challenges shall be tackled efficiently by structuring large-scale and high-resolution EO spatiotemporal data into a suitable 4D surface representation, which shall be used as multiresolution input data for the newly developed environmental analytical and numerical simulation methods that would be parallelized using High-Performance Computing (HPC). By data fusion of multiple environmental simulations with structured EO data, the assessment of environmental microclimate parameters with sufficient spatial accuracy (up to 1 m2) over a larger area (at least 10 km2) will be possible through both spatial and temporal dimensions (4D). This shall be validated through three pilots (land surface temperature assessment, snow cover changes, and air quality assessment) executed on the new HPC centre RIVR at University of Maribor, by utilizing General Purpose Computing on Graphics Processing Units (GPGPU). Each pilot’s results shall be disseminated on the state-of-the-art Geographic Information Systems (GIS) infrastructure. SAMPA interdisciplinary team shall consist of three research groups, namely the core research group from Laboratory for geospatial modeling, multimedia and artificial intelligence at University of Maribor (UM) for managing the project activities and developing new algorithms, the group from Laboratory for power engineering at UM responsible for in-situ environmental sensors processing and placement, and Geodetic Institute of Slovenia group responsible for EO data processing and pilots’ results validation.