Constraining the volatile budget of the lunar interior has important ramifications for models of Moon formation. While many early and previous measurements of samples acquired from the Luna and Apollo missions suggested the lunar interior is depleted in highly volatile elements like H, a number of high-precision analytical studies over the past decade have argued that it may be more enriched in water than previously thought. Here, we integrate recent remotely sensed near-infrared reflectance measurements of several Dark-Mantle-Deposits (DMDs), interpreted to represent pyroclastic deposits, and physics-based eruption models to better constrain the pre-eruptive water content of lunar volcanic glasses. We model the trajectory and water loss of pyroclasts from eruption to deposition, coupling eruption dynamics with a volatile diffusion model for each pyroclast. Modeled pyroclast sizes and final water contents are then used to predict spectral reflectance properties for comparison with the observed orbital near-infrared data. We develop an inversion scheme based on the Markov-Chain Monte-Carlo (MCMC) method to retrieve constraints between governing parameters such as the initial volatile content of the melt and the pyroclast size distribution (which influences the remotely measured water absorption strengths). The MCMC inversion allows us to estimate the primordial (pre-eruption) water content for different DMDs and test whether their source is volatile-rich. Our results suggest that the parts of the lunar interior sampled by the source material of the DMDs investigated in this study range in water content from 400 to 800 ppm.