The Gulf of Mexico Loop Current (LC) System has been the subject of increased attention recently. Hazardous conditions caused by high velocities resulting from LC eddies are disruptive to industries in the region. In this work, we utilize our MIT Multidisciplinary Simulation, Estimation, and Assimilation Systems (MSEAS) and Error Subspace Statistical Estimation (ESSE) probabilistic modeling to estimate such velocity uncertainty, compute predictability limits, and optimize data collection. We consider several LC regimes and use our information-theoretic methodology to study the predictability of different LC features and events including detachments/reattachments and LC eddy dynamics. We show that hindcasts can have predictive capabilities for 1 to 3 months, with a slower loss of predictability in the quieter LC states. We find that uncertainty grows near the LC and its eddies, transferring to depth from the shelf and slope. We then showcase the assessment of industry risks using probabilistic risk theory. We present the likelihood and higher statistical moments of hazardous velocities over several time horizons, finding that the western Gulf is subject to the greatest risk of positive outliers, with this risk increasing over time. Using correlation and mutual information fields, we optimize future sampling by predicting the impacts and information content of observations. The most informative data are those that either directly sample dynamically relevant areas or sample coastal modes that are correlated with these areas. Subsurface data are shown to have more impact on forecasts of 30+ days. Lastly, we predict aircraft flight paths for the Remote Ocean Current Imaging System (ROCIS) that maximize data impact on 30+ day forecasts. We discuss flight constraints, candidate path construction with pruning, and MI evaluation schemes, and showcase our results for several dynamical regimes, using a variety of allowed bearing angles, path segments, and airports.