Current urban-greening programmes are all too often based on inadequate data (see, for example, C. T. Driscoll et al. BioScience 62, 354–366; 2012), and models for estimating the value of urban vegetation are largely untested. To make substantive progress towards urban sustainability, city managers and researchers need to know where, when, how and which greening programmes are appropriate for urban areas.

Simplified urban-forest models have been widely used to estimate the benefits of scattered planting of trees in city parks and avenues, but these mostly fail to build in estimates of uncertainty or to consider trade-offs and costs. For example, urban forests would be unlikely to reduce atmospheric concentrations of polluting particulates and nitrogen dioxide (H. Setälä et al. Environ. Pollut. 183, 104–112; 2013), and their high pollen density could exacerbate respiratory conditions such as asthma.

We suggest, therefore, that urban-greening strategies should be tailored specifically to their localities. Programmes need to be validated by testing against comparative studies that capture spatial and temporal variability in and among cities. This means that local urban data collection and ecosystem modelling will have to meet the same high standards as those applied to non-urban areas.