Beyond SDI – Domen Mongus (Univ Maribor)

SDIs, undoubtedly, hold immense potential to support analytics of complex socio-environmental processes and assist in addressing some of the most important questions of our times. In particular, the paradigm of environmental intelligence is becoming the next big thing in future applications. It builds on:

• advanced Earth Observation infrastructures for nearly real-time monitoring (e.g. Copernicus),
• increasingly sophisticated data integration mechanisms and standards (e.g. OGC, INSPIRE), and
• advanced analytics capacities based on Artificial Intelligence (AI) and High-Performance Computing (HPC),

Environmental intelligence could indeed provide valuable decision support, in particular, with the inevitable shift of AI from deep learning concepts to explainable methods. Next generation systems will, thus, not only serve prediction and simulation support but will also allow for mining the reasons behind extremely complex interactions between humans and environments.

While contemporary SDIs obviously provide a backbone of such systems, their full utilisation is currently limited by the fact that they are mainly focused on professional end-users. Thus, existing SDIs are bulky when AI comes into play. When considering true environmental intelligence with an additional AI layer, machines will become their primary end-users, making a significant proportion of existing components obsolete. This should be compensated with substantial increase in the efficiency of data fusion, focusing on dynamics of the extraction of intermediate information layers, particularly when considering the time domain. As the Internet-of-Things (IoT) has triggered an unavoidable convergence of technologies, SDIs need to follow the trend of efficient spatiotemporal data processing, while avoiding human interaction.

While it is obvious that SDIs need to become AI friendly, there is no clear way on how to achieve that. Systematic mapping of AI requirements for SDIs is, thus, the first step towards achieving that.