Evaluation of Upper Tropospheric Geopotential Height Anomalies over the Tropical and Subtropical Oceans in CMIP6 Models Using GNSS Radio Occultation Observations

Abstract We investigate the influence of hydrometeor radiative effects on winter season (DJF) geopotential height anomaly (ZA) over the subtropical and tropical Pacific Oceans in the Coupled Model Inter-Comparison Project phase 5 (CMIP5) and phase 6 (CMIP6) models utilizing satellite observations from GPS radio occultation (RO). We assess the average ZA biases in historical climate simulations by examining the methods employed in CMIP6 models for calculating the radiative properties of floating and falling frozen hydrometeors. CMIP6 models are categorized based on their treatments: cloud ice only (NOS), combined treatments (SON1), and separate treatments (SON2) for cloud ice and falling ice (snow) radiative properties. NOS models exhibit an overestimation of averaged absolute biases (ZA) in the upper troposphere. SON2 models, on the other hand, mitigate these biases by 30—80 m. However, this improvement is not observed in SON1 models compared to NOS. Spatially-averaged absolute biases in ZA for SON1 models are similar to those of NOS, indicating that the combined treatment of frozen hydrometeors radiative effects might not yield the anticipated impacts. The above-mentioned biases in CMIP6 also align with sensitivity tests conducted with CESM1-CAM5 and CESM2-CAM6, where turning off falling ice radiative effects results in significantly larger absolute biases in ZA compared to simulations with these effects included. Overall, progress from CMIP5 to CMIP6 is somewhat constrained, due to improvement from SON2 models but not from SON1 models. These findings imply that a separate treatment of frozen-hydrometeor radiative properties could be crucial in minimizing the variability among CMIP6 models.  The three key points: Key point #1: Satellite radio occultation observations are used to evaluate geopotential height in present-day coupled ocean-atmosphere model simulations. Key point #2: Geopotential height anomaly (ZA) is improved with falling ice radiative effects included (SON) compared to without (NOS) in CESM1 and CESM2. Key point #3: CMIP6 models with separate frozen hydrometeors produce superior performance in ZA than with total hydrometeors for radiative effects and NOS.   1. Introduction Global Positioning System (GPS) radio occultation (RO) measurements can be used to estimate geopotential height (GPH, zg or Z) for climate model evaluation and monitor climate changes (Ao et al., 2015; Kursinski et al., 1997; Goody et al., 1998; Wielicki et al., 2013). Furthermore, the structural uncertainty of RO is demonstrated to be low in the Upper Troposphere and Lower Stratosphere (UTLS) over tropical and subtropical oceans, as evidenced by meticulous comparisons of retrievals from various RO processing centers (Ho et al., 2009, 2012; Steiner et al., 2013). Another way is to use the retrieved pressure profile, geopotential height can be derived on constant pressure surfaces (Leroy, 1997). This information, in turn, enables the estimation of winds through the geostrophic approximation, offering novel insights into atmospheric dynamics such as the locations of jet streams and storm tracks (Guo et al., 2009; Davis and Birner, 2013; Scherllin-Pirscher et al., 2014; Verkhoglyadova et al., 2014). Additionally, the difference in geopotential height between two pressure levels is directly proportional to the (log-pressure weighted) layer-averaged temperature. Consequently, changes in the bulk tropospheric temperature can be inferred from geopotential height near the tropopause. This study will use the RO-observed geopotential heights between 100 and 300 hPa to evaluate global climate models (GCMs). While all Coupled Model Inter-Comparison Project phase 5 (CMIP5) models (Taylor et al., 2012) and CMIP6 (Eyring et al., 2016) models incorporate cloud ice radiative effects, most of them either misrepresent or ignore the radiative effects associated with falling ice mass (Kodama et al., 2015; Li et al., 2012, 2013, 2014a, b). The significance of the impacts resulting from frozen hydrometeors-radiation interactions has been established in numerous previous studies (Kodama et al., 2015; Gettelman & Morrison, 2010; Li et al., 2013, 2016, 2020a, 2020b, 2021b, 2022, 2023a, 2023b; Michibata et al., 2019; Stephens, 2005; Stephens et al., 2008). Research exploring the changes of tropical climate, particularly by considering or excluding falling ice radiative effects (FIREs) in a GCM like CESM1-CAM5 (Li et al., 2014a) and across subsets of CMIP6 models, has consistently demonstrated the substantial impact of FIREs. These effects influence various climatic aspects, including sea surface temperature (SST), radiation, precipitation, and circulation. Overall, the incorporation of FIREs contributes to an improved simulation, particularly over the Pacific Ocean (Li et al., 2014a,b, 2020a, 2020b, 2021a,b,c, 2022, 2023a,b; Chen et al., 2016, 2018; Gettelman & Morrison, 2010; Kodama et al., 2015; Michibata et al., 2019). These investigations have highlighted the substantial influence of FIREs and the indirect effect of hydrometeor-radiation-circulation interactions (Li et al., 2014a, 2016). We briefly outline the mechanism as follows and more details can be found in Supplementary information (SI). Research by Li et al. (2014a, 2016) using CESM1-CAM5 demonstrated that within deep convective zones, the exclusion of FIREs increases upward thermal emission (outgoing longwave or RLUT) and reflected shortwave (RSUT) at the top of the atmosphere, and an excessive downward shortwave at the surface (RSDS) over regions such as the Inter-tropical Convergence Zone (ITCZ), South Pacific Convergence Zone (SPCZ), and tropical western Pacific (TWP) leading to unstable convection. Consequently, there is a low-level divergence anomaly away from convective zones, relative to the inclusion of FIREs. These anomalous winds, when away from the ITCZ/SPCZ and TWP, typically reduce the prevailing trade winds and surface wind stress, giving rise to ocean temperature bias patterns reminiscent of El Niño–Southern Oscillation (ENSO) (Chen et al., 2018; Li et al., 2018) on the flanks of the ITCZ. Similar sensitivity tests are carried out using the newer NCAR-DOE CESM2-CAM6. Figure 1 illustrates the disparities in SST and surface wind stress observed in a pair of sensitivity tests conducted using CESM2-CAM6 with FIREs on (SON_p) and off (NOS_p). In Figure 1c, a weaker surface wind stress is evident in NOS compared to SON, leading to warmer SST over the trade-wind regions, as depicted in Figure 1a. The incorporation of FIREs results in increased surface wind stress in SON, contributing to improved SSTs (Ts) compared to NOS, as illustrated in Figure 1b. These results are consistent with sensitivity tests using CESM1-CAM5 (Li et al., 2014a, 2016). Additionally, these anomalous winds induce excessive precipitation and generate unrealistic middle- and high-level clouds over the trade-wind regions where shallower clouds are frequently observed in the real-world atmosphere (Li et al., 2021b, 2023a,b). In CMIP6, a greater number of models, in comparison to CMIP5, depict the radiative effects associated with falling ice mass. These models include both the falling ice mass radiative effects (FIREs) and cloud ice radiative effects (Gettelman et al., 2010; Li et al., 2020a; Michibata et al., 2019). Li et al. (2020a) classified CMIP6 models into two groups (SI Table S1): the Snow-On (SON) group with FIREs, and the No-Snow (NOS) group without FIREs. Models in the NOS group only include cloud ice radiative effects. Li et al. (2020b) further examined the impacts of two different treatments of FIREs in CMIP6 models on precipitation simulation, one with combined (SON1) and the other with separate treatments (SON2) of cloud ice and falling ice radiative properties, and compared them to the NOS group. They found that FIREs improved the precipitation simulation significantly in SON2 compared to SON1 and NOS. These prior findings motivate us to examine simulated geopotential height (Z) fields in CMIP6 in the context of hydrometeor-radiation interactions treated in GCMs. We will evaluate 7 SON1 models, 6 SON2 models and 13 NOS models (Table S1) against RO observations in terms of the boreal winter (Dec-Jan-Feb: DJF) mean as the summer variations of geopotential height anomaly (ZA) are small (not shown) and the seasonal cycle of ZA biases in particular over the southeast Pacific trade-wind regions which is indirectly influenced by FIREs. We will use the RO product developed by Ao et al. (2015). Ao et al. (2015) conducted a comprehensive analysis by comparing geopotential height fields from the CMIP5 models with RO data to examine the annual mean and seasonal cycle of the 200 hPa geopotential height over the period from 2002 to 2008 utilizing a set of atmosphere-only (AGCM) model runs (AMIP) from the CMIP5 archive. This means that the feedback from ocean was ignored. They found that most models agree well with the observations in the tropics in both the annual means and interannual variabilities. In this study, we will examine CMIP5 and CMIP6 models that are fully coupled to explore the impacts from fully coupling of SSTs with wind stress and the atmosphere, in particular over the trade-wind regions by examining both the DJF mean Z anomaly (ZA), in which the global mean is removed, and their seasonal cycle across individual models and different subgroups at pressure levels 300, 200 and 100 hPa. In addition, we will also use the ZA derived from RO to contrast with the model results obtained through two sets of sensitivity tests using the NCAR-DOE-CESM1 and NCAR-DOE-CESM2. These sensitivity tests will be compared to CMIP5 and CMIP6 models. The objective of this study is to explore the impact of falling ice radiative effects in geopotential height in both CMIP5 and CMIP6 models, focusing on seasonal mean climatologies and seasonal cycle. It's important to note some caveats of this study, acknowledging that the identified differences between subgroups of CMIP6 models may be influenced by other processes and model configurations beyond the treatment of FIREs. These additional factors could include model resolution, aerosol loading, air-sea interactions, and representations of warm cloud physics (Li et al., 2021a; 2023b). In Section 2, we outline the model sensitivity tests concerning the treatments of radiative properties of frozen hydrometeors (cloud ice and falling ice) using CESM1, CESM2, and CMIP5/CMIP6 models categorized into subgroups. The RO products are also briefly described. In Section 3, we show results of seasonal-mean spatial patterns of ZA fields and their seasonal cycles compared to observations. Section 4 provides a summary and discussions.