Using causal inference to investigate anthropogenic aerosol impacts on the diurnal temperature range

Near surface air temperature is a primary variable to track global climate change. While mean temperature is often used to quantify global warming, the diurnal temperature range (DTR) - defined as the difference of daily minimum and maximum temperature - can provide additional information on changes to the diurnal cycle and temperature extrema, which is important for impacts of climate change. Different to increasing global mean temperature, observations have shown a decrease of global mean DTR over recent decades. This trend has been attributed to human emissions of greenhouse gases (GHG) which increase daily minimum temperature (Tmin), usually measured at night, more than daily maximum temperature (Tmax), observed during the day, by trapping outgoing longwave radiation. Aerosol radiative forcing has been associated with absorbing and scattering incoming (solar) shortwave radiation; thus, aerosols are assumed to reduce Tmax more than Tmin, decreasing the DTR. However, historical single and ALL forcing simulations from models that are part of phase 6 of the Coupled Model Intercomparison Project (CMIP6) model a decrease of the DTR for GHG and ALL forcings but show an increase in the DTR for anthropogenic aerosols due to a larger reduction of Tmin than Tmax. We investigate this discrepancy in aerosol contributions to changes in the DTR by applying causal inference methods to quantify the impact aerosols have on the DTR in Europe. To address the various effects aerosols have on the climate system, by interacting with both radiation and clouds, we include Tmin, Tmax, aerosol optical depth (AOD), cloud cover, cloud height, incoming shortwave radiation (SW) and outgoing longwave radiation using observational satellite and gridded station data for the past decade in our analysis. First results agree with the cooling effect of aerosols on Tmax by reducing SW radiation and show a positive effect of AOD on low cloud cover. They further suggest that a decrease of Tmax causes a decrease in Tmin, possibly explaining the CMIP6 model results.