Comparisons of Radiation-Circulation-Precipitation Coupling over the Tropical Pacific Ocean Between AMIP6 and CMIP6

AbstractMost global weather forecast and climate models represent small “cloud ice” particles and their interactions with radiation prognostically but neglect the radiative contribution from large-size falling ice (snow) particles or “falling ice radiative effects” (FIREs). In this study, we compare the impact of FIREs on radiation, low-level winds at 10m and precipitation fields between the uncoupled AMIP6 and the fully-coupled CMIP6 historical runs. The ensemble means from models without FIREs produce low biases of total ice water path (TIWP) up to 80-85%. This generates underestimated biases in upward longwave radiation (RLUT) by 4-12 W m2 over the convectively active regions over the Pacific and Atlantic, leading to more unstable convective columns. This leads to low-level divergence of anomalous flows over convective zones, weaker trade winds compared to QuikSCAT surface winds and excessive precipitation against GPCP. Compared to the models with FIREs, the models without FIREs have higher values of biases with underestimated RLUT by 4—20 W m2, overestimated reflected shortwave at the top of the atmosphere and underestimated downward shortwave at the surface with values of 12–20 W m2 over the trade wind regions. The impact of FIREs is generally similar between AMIP6 and CMIP6 models although the former has smaller biases than the latter. The above-mentioned findings from AMIP6 and CMIP6 models are generally consistent with our previous sensitivity studies by turning on and off FIREs using CESM1-CAM5 global climate model and a weather forecast model, suggesting that the FIREs are also important in different models.The three main points of the article:Main point #1: Observations are used to examine the impacts of falling ice radiative effects (FIREs) over the tropical Ocean from AMIP6 and CMIP6 modelsMain point #2: Lack of FIREs in AMIP6 and CMIP6 have overestimated outgoing longwave, leading to larger biases in circulation and precipitationMain point #3: Inclusion of FIREs in AMIP6 and CMIP6 reduces the impact of ocean-atmosphere coupling1. IntroductionClouds, convection, precipitation and their effects on radiation play important roles in global weather and climate. It is a great challenge to represent them properly using available observations and modeling approaches [e.g., Stephens, 2005; IPCC 2007; International Panel on Climate Change (IPCC) Fifth Assessment Report (AR5), Working Group 1, http://www.ipcc.ch/report/ar5/wg1]. One of the uncertainties is the representation of cloud and precipitation masses, and their interactions with radiation in the general circulation models (GCMs), such as those in the previous Phase 5 of the Coupled Model Intercomparison Project [CMIP5: Taylor et al., 2012] as well as in the current phase 6 [CMIP6: Eyring et al., 2016]. Traditionally, these GCMs represent cloud and precipitation with a number of discrete hydrometeor categories (e.g. cloud liquid, cloud ice, falling ice and rain) in both stratiform and convective cloud parameterizations prognostically or diagnostically and only the cloud liquid and cloud ice masses in the stratiform regions are considered in radiative flux calculations in the models, i.e., the radiative impacts of precipitating ice (snow or falling ice) and liquid (rain) and all hydrometeors in the convective regions are neglected [Li et al., 2012; 2013; 2014a,b; Waliser et al., 2009; 2011]. Up to date, a growing number of the GCMs participating in the ongoing CMIP6 have considered the FIREs. This progress motivates the present study. Waliser et al. [2011] conducted an observationally-based radiative transfer study using derived ice water content from CloudSat observations partitioned into cloud ice and precipitating/convective snow categories to assess the contribution of the larger particle “snow” category on the radiative fluxes and heating profiles for the northern hemisphere summer season. Their study found that ignoring the falling ice radiative effects (FIREs) results in biases of 5-10 W m-2 at the top-of-atmosphere (TOA) emitted longwave flux (RLUT) and differences in the vertical profiles of radiative heating by up to 25% over the Inter-tropical Convergence Zone/South Pacific Convergence Zone (ITCZ/SPCZ). Neglecting the FIREs also causes overestimated TOA reflective shortwave radiation (RSUT) and underestimated downwelling surface shortwave (RSDS) over the trade wind regions.As highlighted in the Supplementary Information (SI), Li et al. [2013; 2014a, b; 2015; 2016] noted that persistently systematic biases (Figure 1) in CMIP3 (Phase 3 of CMIP) and CMIP5 models (compared to observations) occur in conjunction with an underestimation of the total ice water path (TIWP) defined as the sum of convective core ice path (CIP), precipitating ice water path (PIWC) and cloud ice water path (CIWP) (shown in Figure 13 of Li et al. [2013]). In particular, sensitivity tests using the Department of Energy (DOE) and National Center for Atmospheric Research (NCAR) coupled Community Earth System Model (CESM1) with Community Atmosphere Model Version 5 (CAM5) showed that ignoring the FIREs could contribute to systematic errors over the Pacific Ocean. TIWP underestimated by the CMIP5 models could be up to 85% against CloudSat-CALIPSO estimate (Figure S1b), producing excessive downward shortwave (SW) radiation at the surface (RSDS: Figure S1c), weaker reflected SW (S1a) and excessive TOA outgoing longwave (LW) radiation (Figure S1d) in the heavily precipitating regions such as ITCZ/SPCZ. The FIREs also impact the vertical profile of radiative heating rate and the overturning circulation based upon the sensitivity tests with CESM1-CAM5. The LW radiatively unstable gradient (Figure S3) produces the excessive upper-level ascending motions aloft and slightly descending motions below required by the cumulus parameterization. The increased updraft is compensated by enhanced downdraft, leading to grid-box anomalous outflows shown in Figure S4. The compensating moist condensational heating/cooling at 300/850 hPa (black profile) is associated with the ascending motions above 650 hPa whereas the condensational cooling (rain re-evaporative cooling associated with convective downdraft) is associated with the descending motions below 650 hPa. This low-level descending motion produces anomalous low-level divergence (Figures S4 and S5) and, as a result, leading to anomalous advection of low-level moist and warm air originated from the warm pool and the ITCZ/SPCZ to the north/south central Pacific regions. The FIREs are particularly apparent over these strongly precipitating and/or convectively active regions (e.g., mid-latitudes storm tracks, warm pool and ITCZ/SPCZ). For more details, see Supplementary Information (SI).In this study, we focus on the tropical Pacific and Atlantic basins where the observed SST pattern displays large spatial variations and is strongly coupled to surface radiative flux, surface latent and surface sensible heat fluxes, and surface wind stress. Among them, the ocean surface wind stress is a major forcing of ocean circulation, surface evaporation (latent heat fluxes), and the distribution of SST. Despite the efforts of extensive previous works, most coupled GCMs still have difficulties in simulating realistic SST [e.g., Meehl et al., 2005; Dai, 2006; Lin, 2007; Randall et al., 2007; de Szoeke and Xie, 2008], surface heat fluxes [e.g., Li and Xie, 2014] and surface wind stress [e.g.,Li et al., 2015]. The aforementioned biases are expected to be associated with biases in clouds, radiation and precipitation in the atmosphere. Previous studies indicate that these biases can be caused by the surface wind stress bias and/or that of surface heat flux lost via evaporation from a coupled atmosphere-ocean feedback, or these biases could be resulted from internal atmospheric processes only [Li and Xie, 2012]. They showed that tropical mean SST biases in coupled GCMs originate from cloud biases in atmospheric models by comparing the pairs of models from CMIP and uncoupled AMIP. They found that the double ITCZ precipitation errors and excessive easterly wind biases are absent in AMIP simulations and resulted from the interaction with the thermocline. In the past few years, the importance of FIREs has been gradually noticed by modeling groups worldwide, which results in more GCMs participating in CMIP6 than in CMIP5 that have considered either diagnostically or prognostically the effects of falling ice (snow) and falling liquid (rain) in the radiation calculations. This progress provides an opportunity for the first time to compare the changes in the simulation fields between two groups of models with FIREs (SON) or without FIREs (NOS) that have performed both uncoupled AMIP6 and fully-coupled CMIP6 simulations. The influences of the FIREs from the ocean coupling could be isolated from examining the biases in AMIP6 against CMIP6, although other changes in the parameterizations and dynamics cannot be isolated. The results of this comparison can help to identify the degree of the gross impacts of the FIREs in the ocean-atmosphere coupling by comparing AMIP with CMIP and reconfirm the importance of FIREs found earlier with a weather forecast model and sensitivity tests with a single coupled GCM. It should be, however, pointed out that the differences between AMIP6 and CMIP6 as well as in their NOS and SON model groups also include the indirect changes related to the FIREs, for example, the local circulation changes, horizontal resolution and other changes related to model improvements. Therefore, the focus of this study is to examine the radiation-circulation-precipitation coupling that is resulted from the inclusion of FIREs in AMIP6 and CMIP6 over tropical oceans and to see the differences with and without ocean coupling, including Pacific and Atlantic oceans between the latitude belts of 60 oS – 60 oN.Observations will be used to examine systematic biases among the modeling groups, i.e., AMIP6 without FIREs (AM6NOS) and with FIREs (AM6SON), and CMIP6 without FIREs (CM6NOS) and with FIREs (CM6SON). We will use the following data products as references: the shortwave and longwave radiative fluxes at the surface and TOA from CERES-EBAF (2001—2018), near-surface winds (u10m, v10m) from QuikSCAT (2000—2010) and total precipitation from GPCP. In section 2, we describe the general information of AMIP6 and CMIP6 models. Observational data used in this study are described in section 3. In section 4, we illustrate and discuss the patterns of radiation fields, 10m winds, and precipitation against observations. Section 5 provides a summary and conclusions.