Spatial Heterogeneity of Clouds Impacts Rain Initiation in Global Models
Clouds do not produce rain uniformly, as illustrated by the photo. Because rain initiation rates are determined by cloud properties that vary in space, accounting for this variability is essential to improving the Earth system model’s performance.
Image courtesy of Mikhail Ovchinnikov, Pacific Northwest National Laboratory
When representing clouds in large-scale weather and climate systems, models often compute the rate of formation of rain droplets through cloud droplet collisions. When there are more aerosols that lead to smaller droplets, it becomes less efficient for cloud droplets to grow to drizzle size through collisions. This, in turn, prolongs the cloud’s lifetime. These collisions are represented as non-linear functions of cloud droplet number concentration and liquid water content averaged over the model’s horizontal grid. The grid size ranges between a few kilometers and 100 kilometers. An enhancement factor is often included in models to account for cloud variability at subgrid scales. This study assesses how well this process is captured in the Department of Energy’s Energy Exascale Earth System Model (E3SM) using advanced high-resolution cloud simulations that explicitly resolve cloud droplet spectra and spatial variability on scales from a few tens of meters to 25 km. The results reveal that E3SM reasonably represents the overall rain droplet formation rate, but this stems from combining an underestimated local rate and an overestimated enhancement factor accounting for unresolved, or subgrid, cloud variability. The overestimation is traced to neglecting horizontal variability, specifically in the cloud droplet number concentration, offering valuable insights into improving global atmospheric models.
The rate of conversion of cloud droplets to precipitation, known as the autoconversion rate, remains a major source of uncertainty in characterizing aerosol’s cloud lifetime effects and precipitation in global and regional models. The overall performance of an autoconversion representation is determined by the combined effect of the accuracy of the local process formulation and the treatment of the small-scale cloud variability not resolved by large-scale models. This study distinguishes the contributions of these factors to the autoconversion biases and uncertainties in the E3SM by analyzing an ensemble of high-resolution cloud simulations that explicitly resolve spatial variability of cloud properties and local autoconversion. The analysis identifies compensating errors in representations of the local autoconversion are biased low and the enhancement factor from the horizontal cloud variability biased high and strongly suggests that, to improve the overall model performance, these biases must be addressed together.
This study evaluates autoconversion representation in version two (v2) of E3SM using detailed simulations of low‐level clouds conducted with a new high-resolution atmosphere model called Predicting INteractions of Aerosol and Clouds in Large Eddy Simulation (PINACLES). An ensemble of 21 large eddy simulations (LESs) was used to better represent a broad range of cloudy conditions in the global model: seven cases cover a diverse range of maritime and continental environments, and each case includes three sensitivity simulations with different background aerosol concentrations. Both local autoconversion rates and the enhancement due to subgrid variability are evaluated in this study. In LES, the reference local autoconversion is computed by first separating the model-predicted droplet size distribution into the cloud and rain parts. The collision coalescence equation is then solved to estimate the cloud part of the size spectrum, followed by the determination of the autoconversion process rate as the rate of water mass transfer to the rain part of the size distribution. The E3SM representation of autoconversion can then be evaluated against this explicitly computed autoconversion. The study reveals that horizontal variabilities in both cloud droplet number and liquid water content play important roles in determining the area averaged autoconversion and that the correlation between them must be accounted for to improve the accuracy of rain initiation and, hence, aerosol’s impacts on clouds, in E3SM.