For human health studies, epidemiology has been established as important tool to examine factors that affect the frequency and distribution of disease, injury, and other health-related events in a defined population, serving the purpose of establishing prevention and control programs. On the other hand, gradient studies have a long tradition in the research of air pollution effects on plants. While there is no principal difference between gradient and epidemiological studies, the former address more one-dimensional transects while the latter focus more on populations and include more experience in making quantitative predictions, in dealing with confounding factors and in taking into account the complex interplay of different factors acting at different levels. Epidemiological analyses may disentangle and quantify the contributions of different predictor variables to an overall effect, e.g. plant growth, and may generate hypotheses deserving further study in experiments. Therefore, their use in ecosystem research is encouraged. This article provides a number of recommendations on: (1) spatial and temporal aspects in preparing predictor maps of nitrogen deposition, ozone exposure and meteorological covariates; (2) extent of a dataset required for an analysis; (3) choice of the appropriate regression model and conditions to be satisfied by the data; (4) selection of the relevant explanatory variables; (5) treatment of interactions and confounding factors; and (6) assessment of model validity.