Mature trees scattered throughout agricultural landscapes are critical habitat for some biota and provide a range of ecosystem services. These trees are declining in intensively managed agricultural landscapes globally. We developed a simulation model to predict the rates at which these trees are declining, identified the key variables that can be manipulated to mitigate this decline, and compared alternative management proposals. We used the initial numbers of trees in the stand, the predicted ages of these trees, their rate of growth, the number of recruits established, the frequency of recruitment, and the rate of tree mortality to simulate the dynamics of scattered trees in agricultural landscapes. We applied this simulation model to case studies from Spain, United States, Australia, and Costa Rica. We predicted that mature trees would be lost from these landscapes in 90-180 years under current management. Existing management recommendations for these landscapes - which focus on increasing recruitment - would not reverse this trend. The loss of scattered mature trees was most sensitive to tree mortality, stand age, number of recruits, and frequency of recruitment. We predicted that perpetuating mature trees in agricultural landscapes at or above existing densities requires a strategy that keeps mortality among established trees below around 0.5% per year, recruits new trees at a rate that is higher than the number of existing trees, and recruits new trees at a frequency in years equivalent to around 15% of the maximum life expectancy of trees. Numbers of mature trees in landscapes represented by the case studies will decline before they increase, even if strategies of this type are implemented immediately. This decline will be greater if a management response is delayed.