Changes in observed tropical mean precipitation and Northern Hemisphere surface temperature show a clear, well-understood response to increases in atmospheric aerosols during the second half of the 20th century. Based on simple theory and climate modelling, a robust negative response to aerosols is also expected in Northern Hemisphere mid-latitude (NHML) precipitation, but none is detectable in the surface-based record. In fact, a weak positive response seems to occur.

In this study, we explore possible reasons for this using corresponding surface runoff observations and a new ensemble of land surface models taken from the TRENDY project that simulate changes in runoff in response to known physical processes when driven by observed precipitation data and changes in land use. We find that observed and modelled changes in runoff anomalies are similar in the second half of the 20th century, but that all-century values differ markedly. Given that observed values are much more consistent with the expected response to aerosol forcing than modelled values, a possible explanation is that the precipitation observations driving the TRENDY models is in error. A range of other possible explanations for the discrepancy are discussed.

In common with previous studies, it is recommended that pre-1950 NHML precipitation observations are not used for climate model validation work. An exception is Europe, where we find excellent correspondence between precipitation and runoff observations and model simulations.