”Rock moisture” (exchangeable water stored in weathered bedrock beneath the soil) is a key and yet overlooked component in hydrologic cycles. It can be partitioned to free water and capillary-bound water. Determining dynamic partitioning of rock moisture is crucial for conceptualizing critical zone functions and climate and hydrologic modeling. However, the quantification of rock moisture partitioning is challenging, especially in rocks with complex pore structures and weathering patterns. Laboratory nuclear magnetic resonance (NMR) measurements are performed on heterogeneous bedrock samples from a merokarst vadose zone to quantify the dynamics of rock moisture partitioning during the drying process. By fitting a multi-Gaussian function, NMR [Formula: see text] distributions are autodecomposed into multiple [Formula: see text] peaks representing different pore sizes and environments. This spectral analysis enables us to track the change of position, width, and area of peaks at any given saturation stage, shedding light on water depletion rates and patterns, water residence time, and partitioning and redistribution of the water in drying rocks. The changes in [Formula: see text] peaks associated with drying among our samples show strong correspondence with mineralogy, and [Formula: see text]-[Formula: see text] measurements indicate that rock moisture depletion and redistribution are closely related to the pore structures. Limestone with well-connected macropores shows a sequential water loss from large to small pores, whereas limestone with poorly connected macropores simultaneously loses water from all pore sizes. The [Formula: see text] peak decomposition analysis can be extended to the field scale to track rock moisture partitioning in the pore network. This capability has implications for documenting critical zone processes, including quantifying water storage dynamics, estimating plant available water, and monitoring weathering processes.