Large-scale plant mortality has far-reaching consequences for the water and carbon cycles. The role of belowground root-zone water storage (RWS) on the conditions that lead to mortality remains uncertain. It has been proposed that the RWS capacity, Smax, can determine ecosystem vulnerability to drought (Hahm et al., 2019a, Klos et al., 2018). However, incorporating information about RWS into prediction of vegetation dynamics has been limited due to the challenge of quantifying RWS at large scales (Dawson et al., 2020, Trugman et al., 2021). Here, we present a mass-balance framework for assessing forest resilience to year-to-year variability in precipitation, including megadroughts, by quantifying RWS. We use the relationship between RWS and annual precipitation to evaluate the sensitivity of woody ecosystems to precipitation variability by classifying them as either capacity limited, where RWS is nearly constant annually and set by Smax, or precipitation limited, where RWS varies annually based on precipitation amount. We applied this framework to seasonally dry forests and savannas in California and found that approximately 16-23% of the state's total biomass is found in precipitation-limited locations where plants commonly rely on carryover of moisture from one year to the next. These precipitation-limited areas experienced disproportionately high rates of mortality in recent drought. In contrast, approximately 51-58% of the state's biomass is found in capacity-limited locations and thus experiences annually reliable moisture supply. Using precipitation projections for the next century, the model framework reveals a tipping point by which 5,163 km2 (27 Tg aboveground carbon) of forest and savanna could transition from stable to unstable moisture supply. An additional 11,950 km2 (55 Tg aboveground carbon) where moisture supply is already annually unstable is projected to experience increased water stress, due to additional years where precipitation is not sufficient to refill moisture deficits generated in dry years. This framework provides a novel approach for assessing vulnerability of RWS, and thus woody ecosystems, to climate change.