Increase the knowledge of Natural Systems through the evaluation of the uncertainty of environmental data: operational theory and application

Assessing the heterogeneity of environmental variables in a geometrically complex geological system, or more generally in a Natural System, represents a significant challenge in real-world applications where a quantitative approach is needed. The complexity arises from the paucity of samples in space, which collectively cover only a small percentage of the study area, ensuring a limited knowledge of the reality. Numerical modeling is a powerful support to expand knowledge available for the target domain, by producing distribution maps of samples. In this context, deterministic interpolation algorithms yield only one potential scenario without assessing the embedded uncertainty. Probabilistic approaches, on the other hand, incorporate spatial uncertainty inherent to any data-based modeling. At the operational level, evaluating uncertainty can boost the level of knowledge available to operators, thus enhancing the reliability of the final model and a more conscious approach to decisions. The Ph.D. Thesis aims to increase knowledge of environmental matrices by assessing uncertainty in the environmental data. Spatial uncertainty quantification derives from a stochastic approach based on geometric supports (structured/unstructured meshes), including the (either continuous or categorical) variables’ spatial distribution. The methodology supports both concave and convex domains spanning two or three dimensions, and it can be generalized for Natural Systems by treating environmental data as regionalized variables. The methodology is embedded into MUSE, a stochastic-based tool for Modeling Uncertainty as a Support for the Environment. The code, developed in C++ represents the main result of this Thesis. It is designed and implemented as a modular tool, incorporating geometry processing techniques to handle the morphology of target domains, geostatistics for estimating spatial dependence laws (automatically and supervised), and stochastic approaches to evaluate multiple, equiprobable scenarios and uncertainty. To guarantee transparency and replicability of entire processes, computational workflows are formalized and traced through an output metadata system, thereby ensuring results’ reproducibility, persistence over time, and tracking of calculation histories (Environmental Workflow Persistence, EWOPE). The Ph.D. Research Project is built on the PON "Ricerca e Innovazione" 2014-2020, Asse IV "Istruzione e ricerca per il recupero", Azione IV.5 "Dottorati su tematiche green" DM 1061/2021 and developed in collaboration with DIGIMAT company. MUSE, the main outcome of the thesis, meets the requirements of the PON-GREEN Research Line by achieving TRL 9. To reach such a TRL, MUSE has been proven in operational environments, highlighting its main features, such as generality, multi-dimensionality, flexibility, and computational efficiency. In particular, MUSE is applied in applications varying from geological modeling to assessment of geochemistry heterogeneity, geothermal resources evaluation, real-time monitoring, and preservation of cultural heritage. Thus conceived, MUSE makes available a clear and explanatory integrated representation of spatial variability of environmental phenomena to the widest possible audience, spanning environmental theorists, Environmental Agencies, Research Institutions, and policy- or decision-makers.