Microbial communities are highly dimensional, with many species and many variable environmental factors. Macroecology, which studies communities as statistical ensembles, is a promising way to connect these complex data to mechanistic models. In this talk, I will discuss a minimal set of macroecological patterns that characterize the statistical properties of species abundance fluctuations across communities and over time. A mathematical model based on environmental stochasticity quantitatively predicts these three macroecological laws, as well as non-stationary properties of community dynamics. Building on these results, it is possible to disentangle the (statistical) properties that determine ecosystems' stability over time and reproducibility across communities.