Further improvement and evaluation of nudging in the E3SM Atmosphere Model version 1 (EAMv1): simulations of the mean climate, weather events, and anthropogenic aerosol effects

[摘要] A previous study on the use of nudging in E3SM Atmosphere Model version 1 (EAMv1) had an unresolved issue; i.e., a simulation nudged to EAMv1's own meteorology showed non-negligible deviations from the free-running baseline simulationover some of the subtropical marine stratocumulusand trade cumulus regions. Here, we demonstrate thatthe deviations can be substantially reduced by(1) changing where the nudging tendency is calculated in the time integration loop ofa nudged EAM simulation so as to improve consistency with the free-running baseline and(2) increasing the frequency of the constraining data so as to better capture strong sub-diurnal variations. The fact that modification (2) improves the climate representativeness of the nudged simulations has motivated us to investigatewhether the use of newer reanalysis products with higher data frequencycan help improve nudged hindcast simulations by better capturing the observed weather events.To answer this question, we present simulations conductedat EAMv1's standard horizontal resolution (approximately 1 ∘ )with nudging towards 6-hourly ERA-Interim reanalysis or6-hourly, 3-hourly, or hourly ERA5 reanalysis.These simulations are evaluated againstthe climatology of free-running EAMv1 simulations as well asreanalyses, satellite retrievals,and in situ measurements from the Atmospheric Radiation Measurement user facility. For the 1 ∘ EAMv1 simulations, we recommend using the relocatednudging tendency calculation and the ERA5 reanalysis at 3-hourly or higher frequency. Simulations used for estimating the anthropogenic aerosol effects oftenuse nudging to help discern signal from noise.The sensitivity of such estimates to the configuration of nudging is investigatedin EAMv1, again using the standard 1 ∘ horizontal resolution.We find that, when estimating the global mean effects, the frequency of constraining data has relatively small impacts,while the choice of nudged variables can change the results substantially.The nudging of air temperature (in addition to horizontal winds) has two non-negligible effects. First, when the constraining data come from reanalysis, the nudging-induced mean bias correction can cause significant changes in the simulated cloudsand hence substantially different estimates of the aerosol effects.The impact of the mean bias correction on ice cloud formation has been noted in previous studies and is also seen in EAMv1.For applications like ours, where the preferred configurations of nudging are those capable of providing results consistent with the multi-year free-running simulations, the consequence of the mean bias correction is undesirable.The second important impact of temperature nudging is a significant suppression of adjustments to aerosol forcing, which also causes changes in the estimated aerosol effects. This effect can be seen in simulations nudged to either reanalysis or EAM's own meteorology.These results suggest that nudging horizontal winds but not temperature is a better choice for estimating the anthropogenic aerosoleffects.

[学科分类] 土木及结构工程学