This thesis quantifies the operational impact of two interventions at the Flandersbach limestone mine: a tuned heuristic dynamic dispatch strategy and autonomy-related fleet changes. A discrete event simulation digital twin was developed and calibrated using telemetry-derived empirical inputs and site-confirmed operating rules, and scenarios were evaluated over two representative operating weeks. The model is used to compare throughput, waiting and congestion behavior, and target adherence under consistent assumptions.

The results show that the implemented dynamic dispatch policy is not an effective mechanism to increase production volume in its current form. While the compliance metrics indicate that dynamic dispatch can enforce relatively tight adherence to shift-level bench targets under the thesis definition, it does so at a throughput cost. Compared to the fixed-assignment baseline, dynamic dispatch reduces total moved tonnage by about one percent and increases congestion, most notably through substantially higher loader queue time and headway-related waiting.

Single-agent autonomy produces asymmetric benefits across subsystems. Introducing a single autonomous hauler or loader yields limited gains in a mixed fleet because productive time remains bounded by shared schedules and downstream constraints. In contrast, the autonomous load-and-carry scenario produces the largest throughput increase in the core scenario set, nearly eight percent, consistent with the fact that this intervention expands the effective operating window by operating through periods that are otherwise constrained by breaks and shift transitions. This result should be interpreted in light of scope, since the load-and-carry unit is modeled as a dedicated crusher-feeding unit, whereas in practice it also performs auxiliary duties.

Overall, the findings indicate that improving a single component often yields diminishing system-level returns when the haulage cycle remains constrained by shared resources, schedules, and interaction effects. The discrete event simulation digital twin provides a validated sandbox for relative comparison of dispatch and autonomy concepts, and the results motivate future concepts that change system-level constraints more directly.

The global transition to green energy has intensified demand for critical raw materials, increasing interest in deep-sea mineral resources such as cobalt-rich ferromanganese (Fe-Mn) crusts on seamounts like Tropic Seamount in the NE Atlantic. This thesis investigates the relationship between water depth and Fe-Mn crust composition and evaluates whether incorporating this relationship through Co-kriging (CK) improves resource estimation compared to Ordinary Kriging (OK). Geochemical and bathymetric data were analyzed using exploratory data analysis. The estimation workflow included block model generation, isometric log-ratio (ILR) transformation, Landmark-ISOMAP embedding for locally varying anisotropy, variogram modeling, and geostatistical estimation with Ordinary Kriging, Simple Co-Kriging (SCK) and Intrinsic Collocated Co-Kriging (ICCK) as well as Inverse Distance Weighted (IDW) Estimation. Model performance was assessed using Quantitative Kriging Neighborhood Analysis (QKNA) metrics and leave-one-out cross-validation (LOOCV), with a critical evaluation of LOOCV’s limitations.

Results demonstrate a significant relationship between water depth and Fe-Mn crust composition, with a significant improvement in estimation accuracy and confidence when water depth is used as the secondary variable in Co-Kriging, with SCK and ICCK providing more accurate and confident resource estimates than OK. ICCK also performed better than IDW. For the first time, tonnage calculations for metals in Fe-Mn crusts based on a 3D block model and real geochemical data are presented, highlighting Tropic Seamount’s potential as a substantial mineral resource for Europe. The workflow developed in this study, including ILR transformation and L-ISOMAP embedding, proved effective for handling compositional data and spatial anisotropy. The study also identifies methodological limitations, such as the need for improved variography, better sampling distribution, consideration of non-metallic elements, and more advanced cross-validation techniques. The assumption that the top 1 cm of crust represents the entire deposit is noted as a simplification, and future research should address vertical stratigraphy and sampling distribution.

The findings indicate that integrating ILR transformation and Landmark-ISOMAP embedding with Co-Kriging leads to better resource estimation for Fe-Mn crusts, enabling more confident and accurate assessments. This methodology offers significant potential for mineral resource exploration and future research in both marine and terrestrial environments.