Evaluation of Radiatively Active Frozen Hydrometeors Mass in CMIP6 Global Climate Models Using CloudSat-CALIPSO Observations

Abstract While most of models participated in CMIP6 do not consider precipitating ice (snow) radiative effects (NOS), there are several models implement prognostic snow in addition to cloud ice separately (CESM2-CAM6 and its family models: SON2) or combined as a parcel interacting with radiation (SON1). This study uses derived 2C-ICE frozen hydrometeors water path (IWP) and vertical profile of ice water content (IWC) estimates from CloudSat-CALIPSO satellite data to evaluate IWP/IWC simulations by CMIP6 models in terms of stratiform floating cloud (CIWP/CIWC), stratiform precipitating (Snow: SWP/SWC) and total IWP/IWC (TIWP/TIWC), which is the sum of CIWP/CIWC and SWP/SWC. The spatial map distributions of CIWP/SWP/TIWP, vertical zonally-average and regionally-averaged profiles of the subsets for NOS, SON1, and SON2 in CMIP6 are evaluated. The NOS models show better than SON2 agreements in comparison with CIWP, except for overestimated magnitudes over trade-wind regions over southern ocean. The SON2 models overestimate tropical convective zone of SWP/TIWP but underestimate over storm tracks. SON2 models simulate better global TIWP than NOS due to the inclusion of snow. The TIWC from all the subsets are underestimated over storm tracks but with a better simulation in SON2 relative to NOS and SON1. The TIWC in SON2 shows the best simulation above 400 hPa but slightly overestimated below 450 hPa that might be due to overestimated SWC below the 350 hPa over the tropics or due to attenuation of radar/lidar signals where precipitating ice and/or convective hydrometeors contribute greatly. The three key points: Key point #1: 2C-ICE CloudSat-CALIPSO cloud frozen hydrometeors and total frozen hydrometeor mass, including precipitating ice (snow) are used to evaluate CMIP6 models. Key point #2: Snow is simulated reasonably well from separate cloud ice and snow (SON2) against 2C-ICE estimates, but underestimated in the mid-troposphere over mid- and high-latitudes and overestimated in tropical convective zones. Key point #3: Models in SON2 subset have better total frozen hydrometeors mass against 2C-ICE estimates than no snow (NOS) and combined cloud ice and snow (SON1) but still underestimate over mid- and high-latitudes. 1. Introduction Radiation-active frozen hydrometeor mass in the atmosphere, including floating cloud or small particle cloud ice, precipitating (or falling or large particle) hydrometeors (snow), are important elements for atmospheric longwave (LW) and shortwave (SW) radiation computation in Coupled Model Intercomparison Project (CMIP) models (Li et al., 2012; 2013; 2023; Gettelman et al., 2010; Gettelman and Morrison, 2015; Michibata et al., 2019). In the global climate models (GCM) that misrepresent or do not consider falling part of frozen hydrometeor mass (snow) and its associated atmospheric radiative heating profiles are shown to cause biases in global radiation budget and cloud-radiation-circulation coupling due to underestimated total ice water path (TIWP) and content (TIWC) (Li et al., 2013; 2016b; 2021; 2022a; 2023, Gettelman and Morrison, 2015; Michibata et al., 2019). Li et al. (2012) focused on characterizing and diagnosing systematic biases associated with the frozen hydrometeors in the 5th phase of Coupled Model Intercomparison Project (CMIP5) that only few of models (CESM1-CAM5) implemented radiative effects of falling ice diagnostically in addition to cloud ice (Gettelman et al., 2010). Waliser et al. (2011) conducted radiative transfer calculations using CloudSat-CALIPSO data to examine the impact of excluding precipitating ice on atmospheric radiative fluxes and heating rates. They found that with exclusion of precipitating ice result in 5–10 Wm−2 underestimates of the top of the atmosphere (TOA) reflective shortwave flux and TOA longwave flux and overestimates of the downwelling surface shortwave in the Intertropical Convergence Zone (ITCZ), the South Pacific Convergence Zone (SPCZ) and warm pool. The differences between with and without precipitating ice in the vertical profiles of shortwave and longwave heating are non-trivial for about 10–25%. Similar biases to Waliser et al. [2013] in the radiation fields [Li et al., 2013; 2014a,b; 2016] also found the biases (can reach 20—30 Wm−2) of underestimate TOA reflected shortwave, excessive downward shortwave radiation at the surface and overestimated outgoing longwave radiation over precipitation active regions. These biases are partially due to the fact that models in CMIP5 do not consider the radiative effects of precipitating ice. The above-mentioned biases then produce biases in circulations, precipitation and water vapor against observations over the Tropical Pacific Ocean [Li et al. 2014a,b; 2016]. They found that for models without considering falling hydrometeors and their radiative effects (FIREs), the models would produce too cold surface temperatures over wintertime land and polar regions (Li et al., 2016b; 2022) but too warm sea surface temperatures over trade-wind regions (Li et al., 2014a, 2016a, b, 2021) and most importantly, excessive sea ice coverage and thickness (Li et al., 2022b). Frozen hydrometeors derived from CloudSat (Austin and Stephens, 2001; Austin et al., 2009) and CALIPSO provide retrievals of the total frozen hydrometeors of mass, including all sizes of particles such as floating cloud ice and precipitating ice (snow, graupel, and hail) in stratiform clouds (including detrained convective mass) and/or convective cores of mass (Li et al., 2012). In order to have sensible comparisons with the mass of floating and non-convective core area hydrometeors produced by most GCMs, Li et al. (2012) separated the floating cloud ice from precipitation and convective cores, using filtering methods to provide (floating) cloud ice water content (CIWC). These concepts and datasets have been widely employed by the scientific community to evaluate their floating ice, that is the variables of cloud ice water path “clivi” and content “cli” in CMIP5 data port output and several models with “diagnose” falling ice mass (snow), which is not an output in CMIP5 (Gettelman et al. 2010; Li et al., 2014b; Song et al. 2012; Wu et al., 2015; Ma et al., 2012). An alternative approach is to use satellite simulator software to evaluate the modelled properties (Bodas-Salcedo et al., 2011). The satellite simulator includes all the signals such as floating cloud, convective core area, falling ice (snow, graupels and hails), while in CMIP6 models, some models only have cloud-only hydrometeors is represented some have snow included. In this study, extended from Li et al. (2012) for CMIP5 model-data comparisons for floating cloud frozen hydrometeors only, here we provide observational estimates of hydrometeors of various types, including those of cloud ice (CIWP/CIWC), precipitating frozen hydrometeors that is snow for evaluating hydrometeors mass in the output of the CMIP6 models. We use 2C-ICE (Deng et al., 2010; 2013) estimates derived from CloudSat radar and CALIPSO lidar measurements for global retrievals of ice water path (IWP) shown in Figure 1 and vertically-resolved ice water content (IWC) shown in Figure 2, including small particles (floating cloud ice from stratiform and convective core area) to larger particles (snow+graupl+hails from stratiform and convective core area) as upper limit bounds so that a meaningful and robust observational mass estimate can be made for frozen hydrometeors evaluation. We will discuss these in detail in section 2. In the CMIP6, the majority of GCMs only consider the mass of floating stratiform including entrained convective clouds in to stratiform areas (few of the models consider diagnostic convective core of cloud ice) in radiative transfer calculation, but ignore radiatively important precipitating frozen stratiform hydrometeors, except the CESM2-CAM6 and its family of models using the same stratiform microphysics scheme with prognostic snow (Gettelman and Morrison, 2015), referred to as MG2 (Li et al., 2022a), with radiative effects. These models include CESM2-CAM6 (and its family), E3SM, GISS-E3 (Li et al., 2023), NorESM-LM and NorESM2-MM. This study applies the abovementioned 2C-ICE estimates to evaluate the counterpart IWP/IWC simulations by CMIP6 models in terms of stratiform floating cloud (CIWP/CIWC) and precipitating ice (SWP/SWC) and total IWP/IWC (TIWP/TIWC), which is the sum of CIWP/CIWC and SWP/SWC. We describe the data sources used for estimating upper limit of hydrometeor mass based upon retrievals of 2C-ICE CloudSat-CALIPSO measurements and the separation of different types of frozen hydrometeor in model simulations in Section 2 and 3. We discuss the results in Section 4. Section 5 summarizes the results and draws the conclusions. 2. Observational Path and Content of Frozen Hydrometeor Mass Estimates Eight types of frozen hydrometeors path (Figure 1) and vertical content profile (Figure 2) from the 2C-ICE (Deng et al., 2010; 2013) ice cloud retrievals from the combination of CloudSat radar and CALIPSO lidar measurements are adopted. They are total IWP (TIWP), non-precipitating and non-convective floating cloud ice (CIWP/CIWC), precipitating ice for stratiform (SWP/SWC) and convective core based on the “FLAG” method developed in Li et al. (2012) covering the period from January 2007 to December 2010. The “FLAG-method” distinguishes frozen hydrometeor mass associated with floating clouds with ice mass from precipitating and convective core frozen hydrometeor mass by separating any profile flagged as precipitating at the surface and any profile whose cloud type is classified as “deep convection” or “cumulus” (from CloudSat 2B-CLDCLASS dataset; Sassen and Wang, 2008). Excluding all the profiles from the above-mentioned conditions, the remaining profiles are associated with clouds with stratiform floating ice mass (CIWP/CIWC) commonly represented in most GCMs. The mass of the excluded profiles associated with stratiform precipitation is called stratiform precipitating ice (snow). The mass of the excluded profiles associated with convection core area is called convective snow. The total ice hydrometeor mass (TIWP/TIWC) is the sum of CIWP/CIWC and SWP/SWC. Note that, most models consider convective cloud entrained into stratiform area which is not the same as “convective core area ice”. This methodology was used to evaluate CMIP5 CIWP/CIWC (e.g., Li et al., 2012) and improve model cloud parameterizations in CAM5 (Gettelman et al., 2010; Song et al., 2012), as well as other applications mentioned in the introduction. Note that, the 2C-ICE estimates represent the total frozen hydrometeors mass, including “floating” cloud ice and the precipitating ice with variable sizes and terminal velocities (snow, graupel, and hail) as the combined measurements are sensitive to a wide range of particle sizes. In GCMs, however, it is generally assumed that convective core areas are small relative to model horizontal grid, and mass is expected to be small relative to a grid box in a typical GCM horizonal grid box size larger than a few hundred km. Such that the convective core clouds are generally not included as radiatively-active variables in most GCMs (e.g., Li et al., 2012; Waliser et al., 2009). However, as the resolution in the most current GCMs (such as in CMIP6) becomes higher, e.g., when the grid box reaches tens of km, the relative radiative contribution of convective mass increases and might need to taken into account either prognostically or diagnostically for models. The caveat of the aforementioned hydrometeors separation method that we need to keep in mind is that it is impossible to completely include all the particle sizes of frozen hydrometeors from the floating cloud, snow, hail, and graupel in a GCM. Furthermore, it is difficult to separate precipitating and floating cloud forms from measurement in a profile which only provide a snap shot, as they often coexist at some height intervals. In order to determine falling particles, we need at least a multiple-frequency radar and/or a Doppler radar capability, but yet to be available so far and is beyond the scope of this study. Shown in Figure 1 are the frozen hydrometeor path determined by 2C-ICE from CloudSat and CALIPSO data with the classification of precipitation and convection based on surface precipitation and convective core cloud flags, respectively. While Figure 2 is the same as Figure 1 but for the vertically-resolved frozen hydrometeor content. These frozen hydrometeors in Figures 1 and 2 include total ice IWP and IWC (TIWP and TIWC; Figure 1 & 2, panels a), which is the sum of stratiform falling ice (Figure 1, panel d), convective falling ice (Figure 1, panel g), stratiform floating cloud ice (SWP Figure 1, panel d), convective floating cloud ice (Figure 1, panel h). Figure 1 panels b and c and f show the sum of the stratiform and convective core of falling ice and floating cloud ice, respectively. Overall, the falling frozen hydrometers path dominates the TIWP over tropical convective zones and storm tracks (Figure 1b), and the mid-latitudes of both hemispheres (Figure 1d). The convective core falling ice water path contributes more in the tropical convective zones and tropical land mass over South America and Central Africa (Figures 1f, 1g, 1h). Stratiform cloud-only ice (Figure 1e), which is much smaller than the falling ice over the mid-latitudes and convective core floating cloud ice (Figure 1h) is also much smaller than falling ice in mass. For the vertical zonally-average frozen hydrometeors shown in Figure 2, the falling frozen hydrometers dominate the total ice IWC below 300 hPa in the tropics (Figure 2b), and stratiform falling ice below 500 hPa over the mid-latitudes of both hemispheres (Figure 2d). The convective cpre falling ice contributes more between 350—550 hPa from the tropical convective zones (Figure 2g). Stratiform cloud-only ice (Figure 2e), which is much smaller than the falling ice over the mid-latitudes and convective core cloud ice, is also much smaller in mass but with a maximum above 150 hPa over tropics because thin ice clouds can be detected by CALIPSO lidar.