Mature tree mortality rates are poorly documented in desert riparian woodlands. I monitored deaths and calculated annual survivorship probability (P_s) in 2 groups of large (27–114 cm DBH), old (≥40 years old) Fremont cottonwood (Populus fremontii Wats.) in a stand along the free-flowing Yampa River in semiarid northwestern Colorado. P_s = 0.993 year^-1 in a group (n = 126) monitored over 2003–2013, whereas P_s = 0.985 year^-1 in a group (n = 179) monitored over the same period plus 3 earlier years (2000–2003) that included drought and a defoliating insect outbreak. Assuming P_s was the same for both groups during the 10-year postdrought period, the data indicate that P_s = 0.958 year^-1 during the drought. I found no difference in canopy dieback level between male and female survivors. Mortality was equal among size classes, suggesting P_s is independent of age, but published longevity data imply that either P_s eventually declines with age or, as suggested in this study, periods with high P_s are interrupted by episodes of increased mortality. Stochastic population models featuring episodes of low P_s suggest a potential for an abrupt decline in mature tree numbers where recruitment is low. The modeling results have implications for woodland conservation, especially for relictual stands along regulated desert rivers.