Clustering analysis is used to partition data into similarity groups based on mathematical relationships between the measured variables. These relationships (or prototypes) are derived from the specific correlation required between the measured variables (data) and an environmental property of interest. The data points are partitioned into the prototype-driven groups (i.e., clusters) based on error minimization. Once the data is grouped, quantitative predictions and sensitivity analysis of the property of interest can be derived based on the computed prototypes. Additionally, the process inherently minimizes prediction errors due to the rigorous error minimization during data clustering while avoiding overfitting via algorithm parameterization. The application used to demonstrate the power of the method is pressure gradient analysis.