Several recent longitudinal studies have found that exposure to the natural environment is associated with lower non-accidental mortality. However, most of these studies used the normalized difference vegetation index (NDVI) as an exposure metric; and because NDVI might not be sensitive enough to adequately capture changes in urban vegetation, these studies might lack true longitudinal variation in exposure. Therefore, we used a natural experiment to assess the impact of 30 years of tree planting by the nonprofit Friends of Trees on non-accidental, cardiovascular, lower-respiratory, and accidental mortality in Portland, Oregon (mortality data were provided by the Oregon Health Authority). We estimated autoregressive mixed models of Census-tract level mortality rate (deaths per 100,000 population) associated with trees planted, including a tract-level random effect. All models used data from the American Community Survey to control for year, race, education, income, and age. Each tree planted in the preceding 15 years was associated with significant reductions in non-accidental (-0.21, 95 % CI: −0.30, −0.12) and cardiovascular mortality (-0.066, 95 % CI: −0.11, −0.027). Furthermore, the dose–response association between tree planting and non-accidental mortality increased in magnitude as trees aged and grew. Each tree planted in the preceding 1–5 years was associated with a reduction in mortality rate of −0.154 (95 % CI: −0.323, 0.0146), whereas each tree planted in the last 6–10 and 11–15 years was associated with a reduction in mortality rate of −0.262 (95 % CI: −0.413, −0.110) and −0.306 (95 % CI: −0.527, −0.0841) respectively. Using US EPA estimates of a value of a statistical life, we estimated that planting a tree in each of Portland’s 140 Census tracts would generate $14.2 million in annual benefits (95 % CI: $8.0 million to $20.4 million). In contrast, the annual cost of maintaining 140 trees would be $2,716–$13,720.