Evaluate Autoconversion and Accretion Enhancement Factors in GCM Warm-rain Parameterizations using Metrics from Ground-based Observations

Recent studies show that GCMs simulate more warm rain events but with less precipitation amount compared to observations. Warm rain process, in GCMs, is parameterized as functions of grid-mean cloud water mixing ratio (q_c), number concentration (N_c), and rain water mixing ratio (q_r). Autoconversion (E_auto) and accretion (E_accr) enhancement factors are introduced to account for the sub-grid variance of cloud properties. However, only a few studies have assessed the dependence of enhancement factors on sub-grid scale microphysical variabilities and none of them used ground-based observations in the evaluations.

In this study, we use ground-based observations and retrievals over the Azores to calculate E_auto and E_accr with different temporal variations (0.5 h to 5 h) that correspond to different model spatial scales. Compared to the constant value in GCMs, the calculated E_auto in this study increase from 1.79 (0.5 h) to 3.15 (3.5 h) and remain relatively stable afterwards and close to the assumed value of 3.2 in GCMs. On the other hand, the calculated E_accr increases from 1.25 (0.5 h) to 1.6 (5 h), both are greater than the assumed value (1.07) in GCMs. The large E_auto used in GCMs can be used to partially explain why most of the GCMs can produce too frequent precipitation events, while the small E_accr could be used to explains why most of the GCMs can produce more light precipitation amounts.

Furthermore, we classified the MBL clouds into different boundary layer conditions using lower tropospheric stability (LTS): stable (LTS > 18K), mid-stable (13.5K ≤ LTS ≤ 18K), and unstable (LTS < 13.5K). Both E_auto and E_accr increase with decreased boundary layer stability. E_auto and E_accr in nonprecipitating regimes (LWP ≤ 75 g m^-2) are smaller than those in precipitating regimes (LWP > 75 g m^-2). Using constant values of E_auto and E_accr in GCMs cannot reveal the variability of cloud properties in different grid sizes and under different boundary layer conditions. Therefore, it is necessary to implement the results found in this study into GCMs simulations in the future.