Hydrologic models are often used to predict flooding risk driven by land surface features and meteorology. These models can be useful in estimating the consequences of the intersection of two ongoing events in the Catskill region: increased precipitation extremes and the rapid dieback of Eastern hemlock, a foundation tree species. However, simulation of transpiration in these models tends to be erroneous, with storage of water in the plants emerging as a cumbersome process to simulate. In order to improve the fidelity of modeled plant hydraulics, it is important to avoid errors originating from the simplification of the storage of water within plants. Research has found that simulating tree water storage improves model calibration. We investigate water storage in four common conifers as captured by StorAge Selection (SAS) functions generated via a machine learning-based model. We generate model inputs through stable water isotope-tracer based experiments conducted in both growth chamber and field site settings, examining how key environmental variables drive changes in SAS functions. We integrate the SAS framework, enhanced by our experimental data, into a hydrologic model, and assess whether model performance is improved. Finally, we utilize this model to simulate hydrological impact of hemlock loss under different climate scenarios.