Authors' reply

We thank Boucher and Quaas for their interest in our paper, but strongly disagree with their conclusions. In our analysis1, aerosol optical depth was on average 2.5 times higher in polluted compared with clean conditions. Here, we use basic hygroscopic growth and radiative transfer calculations to argue that aerosol humidification cannot account for this difference.

To bound the aerosol humidification effect — and calculate the relative humidity differential required to explain a 2.5-fold difference in aerosol optical depth — we performed detailed radiative transfer calculations2 for a range of coarse and fine aerosol size distributions and compositions, for a constant aerosol concentration. As hygroscopic growth is highly nonlinear with relative humidity, the difference in aerosol optical depth due to humidification depends on the relative humidity range selected. For a highly hygroscopic aerosol that readily takes up water, a 2.5-fold reduction in aerosol optical depth can be achieved if the average relative humidity is reduced; for example, from 95% to 77% or from 90% to 55%. For a less hygroscopic aerosol, the same reduction can be achieved by a relative humidity drop from 95% to 65%, or from 90% to 3%. Note that the relative humidity differential required to explain a 2.5-fold difference in aerosol optical depth grows as the background relative humidity declines.

We argue that such differences in relative humidity are not realistic for cloud fields that form in the same location and season, such as those examined in our study1. To validate our main statement here, we have analysed over 30 years of atmospheric sounding data in cloud fields, revealing an average relative humidity of around 76% for the maritime cloudy lower troposphere (±10%) and 62% (±15%) for the continental lower troposphere. The variance is further reduced if the data is limited to similar meteorological conditions, as we do in our study by sorting the data into similar updraft or relative humidity regimes1,3. Such relative humidity differences result in AOD differences on the order of 10%, as opposed to 250%, as observed in our study1.

Furthermore, relative humidity declines exponentially with increasing distance from a cloud field2. As a result, the influence of aerosol humidification will be greatest within the first few tens of metres around each cloud. The MODIS algorithm filters out pixels within 1 km of detectable clouds, where the influence of hygroscopic growth would be maximal4. Furthermore, 25% of the brightest pixels are rejected within each 10 × 10 km aerosol retrieval box. Both of these measures would significantly reduce the effect of relative humidity on aerosol optical depth retrievals5.

We argue that if, as suggested by Boucher and Quaas based on their model output, one could produce all or most of the trend shown in our paper — that is, an increase in aerosol optical depth of 250% — then it raises the fundamental question of how well a climate model is able to represent the overall effect of aerosol hygroscopic growth without resolving small-scale variations in relative humidity, clouds and aerosols.