We construct an innovative static–dynamic integrated workflow capable of bridging the gap between input geological data, inherent to a lacustrine carbonate outcrop containing karst geobodies, and the description of the flow patterns and quantification of the multi-well productivity index (MWPI) for a particular well configuration in the outcrop. The workflow incorporates additional features stemming from the use of Machine Learning-based methods to mitigate lack of data in the locations away from the sections of input signals, along with the construction of new upscaling methods to assess the MWPI matrix. The ML-enhanced geostatic model hinges upon shallow surface geophysical data collected using Ground-Penetrating Radar (GPR) techniques. Furthermore, by discretizing the flow equations and adopting a flow-based upscaling method, we construct correlations between well flow rates and pressure drawdown in a typical five-spot well configuration. In this setting, we analyze the sensitivity of each well productivity with respect to heterogeneity distribution and correlations in the karst system within the outcrop. Computational simulations illustrate the ability of the integrated workflow proposed herein to improve prediction of hydraulic-connectivity between well pairs, which appear manifested in the entries of the MWPI matrix, whose magnitude aims at quantifying the effects of the karst geobodies upon geofluid production.