Abstract
A recent study of CMIP5 models (Kok et al., 2018) shows the direct climate feedback by dust aerosols is about -0.04 to +0.02 Wm-2K-1, which accounts for a significant fraction of the direct climate feedback by all aerosols (-0.09 to -0.02 Wm-2K-1). In addition, dust is a major source of ice nuclei for aerosol indirect effects and supplies the key nutrients (iron/phosphorus) to the open ocean for ocean primary productivity. Therefore, quantification of dust life cycle and radiative effects in E3SMv1 has important implications on the water cycle and biogeochemical simulations in response to climate change. In the present work, we focused on evaluation of the simulated dust distributions and estimation of the dust direct radiative effects under the present-day conditions. The impact of increasing model resolution on dust simulations is also examined.
The global and annual mean dust aerosol optical depth (AOD) is 0.026 and 0.03 calculated with the low- (~1 degree) and high- (~0.25 degree) resolution E3SMv1, respectively. The model estimates of global dust AOD are similar independent of resolution, because dust emissions are tuned to match the observational constraint of global dust AOD at 0.03±0.01. Without tuning, increasing horizontal resolution by a factor of 4 from 1 to 0.25 degree would lead to about 42% increase of dust AOD. Although the global dust AOD in E3SMv1 is constrained, there are still substantial differences in the predicted dust AOD distributions depending on the model resolution, i.e., about 5 times different near the Taklimakan desert. The changes in dust AOD are more than ±10% near the source regions or in the downwind resulting from local changes of dust emissions, dry and wet removal efficiencies. They could further affect the simulated regional energy balance, especially in the dust-influenced high latitudes where the climate sensitivity to dust forcing is large.
Compared to the ground-based AERONET observations of AOD between 2006 and 2015, the low-resolution E3SMv1 underpredicts the total AOD by about 31% averaged over the 247 AERONET sites. However, the underestimation of AOD is mainly due to aerosol predictions in Asia associated with the anthropogenic emissions (year 2000). Over the selected 14 ‘dusty’ AERONET sites, the modeled mean AOD (0.299) agrees with the AERONET data (0.311), for a correlation coefficient of 0.91. This suggests that the EAMv1 model is able to simulate the AOD climatology of the ‘climate-state’ dependent dust aerosols near the source regions. The high-resolution model further improves the AOD comparisons over the ‘dusty’ sites. The vertical distribution of dust is compared with the CALIPSO satellite-retrieved profiles, indicating an underestimation of dust vertical transport. This is related to the short lifetime of dust simulated by the E3SMv1 for less than 2 days, due to excessive dry and wet dust removals compared with other global climate models.
The estimated annual and global mean of dust direct radiative effect at the TOA is -0.077 Wm-2 in the default E3SMv1 low-resolution model. Sensitivity studies with improved dust optics in the shortwave (AERONET-based) and size distribution (Kok, 2011) suggest a stronger TOA dust cooling effect of -0.4 Wm-2 with a weaker atmospheric warming, which could affect the simulated atmospheric energy balance significantly with the v1 model. Our estimate lies in the lower end of global estimates of the present-day dust direct forcing ranging from -0.6 to +0.1 Wm-2. The high-resolution model estimates a less global cooling effect by about 10% with larger regional differences. This study shows that the magnitude and possibly the sign of dust direct radiative effects in v1 are subject to large uncertainties for water cycle simulations. It requires more than constraint of the global dust AOD. Future improvement are need for simulation of dust vertical profiles and longwave radiation.