Estimating the causal effect of temperature on ozone air pollution

Ground level ozone is an air pollutant which contributes to hundreds of thousands of premature deaths annually. Ground level ozone concentrations are controlled by physical and chemical processes, which can be sensitive to meteorological variables such as the local temperature. Understanding how ground level ozone concentrations are likely to change in future climates is important to understand the future impacts of ozone air pollution on human health. One factor that will influence how ground level ozone changes in the future is the sensitivity of ozone to changes in temperature, since temperature affects many of the processes controlling ozone. One approach to determining the causal effect of changing temperatures on ozone is to perturb the temperature in simulations that model ozone concentrations. However, this approach relies on the simulations accurately representing observed physical processes. Here, we study another approach, estimating the effect of changing temperatures on ozone using causal inference methods and observed data. We show that double machine learning allows us to make estimates of the causal effect of temperature on ozone concentrations using large observational datasets. We estimate both average treatment effects and conditional average treatment effects, and compare our estimates with those of sensitivity studies using chemical box models.