Radiatively Active Hydrometeors Frequencies Derived from CloudSat-CALIPSO Data for Evaluating Cloud Fraction in Global Climate Models

Abstract This study derives radiatively-active hydrometeors frequencies (HFs) from CloudSat-CALIPSO satellite data to evaluate cloud fraction in present-day simulations by CMIP5 and CMIP6 models. Most CMIP5 models do not consider precipitating and/or convective hydrometeors but one of CMIP5 models, CESM1-CAM5, has diagnostic snow and one of CMIP6 models, CESM2-CAM6, has prognostic precipitating ice (snow) included. However, the models do not have snow fraction available for evaluation. Since the satellite-retrieved hydrometeors include the mixtures of floating, precipitating and convective ice and liquid particles, a filtering method is applied to produce observational estimates of cloud-only HF (or NPCHF) from the total radiatively-active HF (THF), which is the sum of NPCHF, precipitating ice HF and convective HF. The reference HF data for model evaluation include estimates of liquid-phase NPCHF from CloudSat radar-only data (2B-CWC) and ice-phase THF from CloudSat-CALIPSO 2C-ICE combined radar/lidar data. The model evaluation results show that cloud fraction from CMIP5 multi-model mean (MMM) is significantly underestimated (up to 30 %) against the total HF estimates, mainly below the mid-troposphere over the extratropics, in the upper-troposphere over the midlatitude lands and a few tropical convective regions. The CMIP5 cloud fraction biases are reduced dramatically when compared to the cloud-only HF estimates, but CMIP5 models overestimate the cloud-only HFs from the tropical convective regions to mid-latitudes in the lower and upper troposphere. The implication of these results on model representations of cloud fraction is that models should include radiatively active precipitating ice and convective hydrometeor types besides the cloud-only type. The three key points: Key point #1: Deriving non-precipitating and non-convective (cloud only) and total radiatively-active hydrometeor frequency (HF) from CloudSat-CALIPSO data. Key point #2: Cloud fractions from CMIP5 multi-model-mean compare well to cloud-only HF estimates implying severely underestimated against total HF estimates. Key point #3: Hydrometeors frequency estimates from CloudSat-CALIPSO provides a reference for GCM’s cloud fraction in stratiform and convective forms. 1. Introduction Both the frequency and mass of radiatively active hydrometeors, including floating cloud ice and liquid, precipitating hydrometeors (snow), and convective ice and liquid, are important for atmospheric shortwave (SW) and longwave (LW) radiation computation (Li et al., 2013, 2018; Waliser et al., 2011; Gettelman et al., 2010; Gettelman and Morrison, 2015; Michibata et al., 2019). However, most general circulation models (GCMs), such as those participated in the 5th phase of Coupled Model Intercomparison Project (CMIP5) (Taylor et al., 2001; Gleckler et al., 2011), and the 6th phase (CMIP6) (except CESM2-CAM6 family that considers snow-radiative effects) only consider the mass and frequency of floating cloud ice and liquid, ignoring radiatively important precipitating hydrometeor and convective core hydrometeor. Thus, the modeled atmospheric heating profiles and possibly the global radiation balance may be impacted by the missing hydrometeors because atmospheric radiation is sensitive to the broader range of hydrometeors (Li et al., 2012; Waliser et al., 2009). The miscounted or misrepresented mass of precipitating ice and convective core hydrometeors result in underestimated total ice water content and path (Li et al., 2012), which are expected to contribute to model biases of radiation budget (Li et al., 2013). Our previous studies have been focusing on characterizing and diagnosing systematic biases in CMIP3/CMIP5/CMIP6 models that are associated with the precipitating ice radiative effects and in weather forecast models such as the European Centre for Medium-range Weather Forecast (ECMWF) (Li et al., 2014b). For example, these biases produce underestimated land surface temperature (Li et al., 2016b), overestimated sea ice concentration (Li et al., 2022) and have impacts on the modeled sea surface temperatures (Li et al., 2014a, 2016a, b, 2021). While the aforementioned systematic biases contributed by ignoring the precipitating hydrometeors mass exist in many GCMs, it is essential to evaluate their performance in terms of the frequency (fraction) of radiatively active hydrometeors because it also contributes to atmospheric radiation in GCMs. However, satellite observations (e.g., CloudSat and CALIPSO) only provide retrievals of the total water mass for liquid and ice, which is the sum of floating water/ice and precipitating water/ice in stratiform clouds and convective cores (Li et al., 2012). Therefore, they are not suitable for direct comparisons with the mass and frequency of non-precipitating and non-convective hydrometeors produced by most GCMs. To separate the floating cloud ice from precipitation and convective cores, Chen et al. (2011) and Li et al. (2012) developed filtering methods to provide (floating) cloud ice water content (CIWC). These concepts and datasets have been widely employed by the scientific community. For example, Gettelman et al. (2010) used CIWC to evaluate new ice cloud microphysical approaches for Community Atmosphere Model version 5 in the Community Earth System Model version 1 (CESM1-CAM5) and to develop a new convection scheme with convective cloud ice mass included in CAM5 (Song et al., 2012). Zhang et al. (2014) investigated ice nucleation in cirrus clouds. The dataset has also been used to evaluate the IWC representation in the UCLA GCM (Ma et al., 2012), the Weather Research and Forecasting (WRF) model (Wu et al., 2015), and the Goddard Multiscale Modeling System (Tao et al., 2009). Another approach is to use satellite simulator software for model assessment (Bodas-Salcedo et al., 2011), such as using the GCM-Oriented CALIPSO Cloud Product (CALIPSO-GOCCP) (Cesana et al., 2016), and to evaluate model’s cloud phase transition and low cloud feedback (Cesana et al., 2019). But this approach does not separate the different types of hydrometeors frequency and might miss the frequency of large particles, which are detected by CloudSat radar but not by CALIPSO lidar (Cesana et al., 2019). It is noted that the aforementioned studies have focused on the mass and radiative effects of cloud and precipitating hydrometeors. In this study, we turn our perspective to the occurrence frequency of the radiatively active hydrometeors, or hydrometeors frequency (HF), which is generally considered equivalent to the cloud fraction except for sampling cloud fields at a fixed location in time (Clothiaux et al., 2009; Xu et al., 2012) or on a narrow satellite swath in space such as CloudSat and CALIPSO. The objective of this study is to provide an observational estimate of different types of HF for evaluating cloud fraction from model output, including cloud ice, precipitating ice, and cloud liquid. Three retrieval algorithms, either using CloudSat radar or CALIPSO lidar or both, provide global retrievals of ice water content (IWC), including small particles (floating cloud ice) to larger particles (snow), and liquid water content (LWC), as well as the effective radius (Re) and the extinction coefficient from the thinnest cirrus (seen only by the lidar) to the thickest ice cloud (Austin et al., 2001; Hogan et al., 2006; Delanoë and Hogan, 2008, 2010; Macc et al., 2009; Young and Vaughan, 2009; Sassen et al., 2009; Deng et al., 2010; Stein et al., 2011). In this study, we use cloud liquid HF from CloudSat-only 2B-CWC-RO5 product (Austin et al., 2009; Li et al., 2018), combined with CloudSat-CALIPSO ice water products from 2C-ICE (Deng et al., 2010, 2013) and DARDAR (raDAR/liDAR) (Hogan, 2006; Delanoë and Hogan, 2008, 2010) for obtaining the total HF (THF), non-precipitating and non-convective HF (NPCHF), precipitating ice HF (PIHF), and convective HF (CHF), so that a robust and meaningful observational HF estimate can be made for model evaluation. In Section 2, we describe the observational resources for the estimated hydrometeor frequency from CloudSat-CALIPSO datasets, the separation of different types of hydrometeor frequencies and the cloud fractions in model simulations. In Section 3, we discuss the results with a summary and conclusions drawn in Section 4.