NASA GISS-E3 calibrated physics ensemble (CPE) cloud properties and feedbacks outputs

Description This dataset contains cloud radiative feedbacks and cloud property diagnostics derived from SST-prescribed AMIP simulations performed with the NASA GISS-E3 climate model using a Calibrated Physics Ensemble (CPE). The dataset supports the analysis presented in the manuscript “Improved representation of cloud properties and feedbacks in the NASA GISS-E3 climate model” (Cesana et al., 2026). The Calibrated Physics Ensemble (CPE; Elsaesser et al., 2025) consists of 300 model configurations obtained by systematically perturbing uncertain cloud-related parameters and then calibrating them against a set of satellite observations using machine learning (ML) techniques. As such, the CPE members sample diverse parameter combinations while remaining close to top-of-atmosphere radiative balance and consistent with multiple hydrological, energy, and radiative metrics derived from satellite observations. The framework also explicitly accounts for observational biases during the inference process. Overall, this ML- and observation-driven tuning approach enables an improved representation of cloud properties and their associated radiative effects, while providing a framework to quantify parametric uncertainty in the model. Here we present a dataset made of a subset of the original CPE, 50 configurations, which we used to evaluate the impact of this ML- and observation-driven tuning approach on cloud properties, cloud vertical structure, and cloud radiative feedbacks in GISS-E3 in Cesana et al. (2026). Diagnostics are derived using satellite simulators to facilitate direct comparison with observations, including CALIPSO (lidar-based) and ISCCP cloud products, and the simulations are AMIP simulations from 2007 to 2010. Contents The dataset consists of multiple NetCDF files containing diagnostics for the 50 ensemble members. Variables are provided on a common horizontal grid (2.5° × 2°) and include the following dimensions: longitude (lon) latitude (lat) time (time) model (CPE ensemble member) level (for vertical profile variables) All profile/temperature variables have been averaged over the analysis period (2007-2010) to produce climatological fields. File descriptions 1. Cloud Radiative Feedbacks File: GISS-E3_CPE_52_cloud_feedbacks.nc.zip This file contains decomposed shortwave (SW) and longwave (LW) cloud radiative feedbacks. The decomposition separates contributions from cloud amount, altitude, and optical depth using the cloud radiative kernels of Zelinka et al. (2012) and based on 4 years (2007-2010) of AMIP and AMIP-p4K simulations. Variables include: dRlw_true, dRsw_true: total LW and SW cloud feedbacks dRlw_prop, dRsw_prop: cloud amount feedbacks dRlw_dctp, dRsw_dctp: cloud altitude feedbacks dRlw_dtau, dRsw_dtau: optical depth feedbacks Each component is further partitioned into: low-cloud contributions non-low-cloud contributions Units: W m⁻² K⁻¹ 2. Cloud Fraction (2D fields) File: GISS-E3_CPE_52_cloud_cover_simulator.nc.zip This file contains cloud fraction diagnostics derived from inline satellite simulators, enabling consistent comparison with observations. Variables include: cllcalipso: CALIPSO-simulator low-cloud fraction clmcalipso: CALIPSO-simulator mid-cloud fraction clhcalipso: CALIPSO-simulator high-cloud fraction cltcalipso: CALIPSO-simulator total cloud fraction cltisccp: ISCCP-simulator total cloud fraction Units: % (0-100) 3. Cloud Radiative Effect (2D fields) File: GISS-E3_CPE_52_TOA_CRE.nc.zip This file contains cloud radiative effect fluxes at TOA. Variables include: lwcrf_toa: LW cloud radiative effect at TOA swcrf_toa: SW cloud radiative effect at TOA Units: W m⁻² K⁻¹ 4. Omega 500 (2D fields) File: GISS-E3_CPE_52_Omega500.nc.zip This file contains the vertical velocity at 500 mb. Variables include: omegacp: vertical velocity at 500 mb Units: hPa day⁻¹ 5. Cloud Fraction Profiles (3D fields) File: GISS-E3_CPE_52_cloud_fraction_profiles_simulator.nc.zip This file provides vertical profiles of cloud fraction from the CALIPSO simulator. Variables include: clcalipso: total cloud fraction profile clcalipsoliq: liquid cloud fraction profile clcalipsoice: ice cloud fraction profile These variables are functions of: longitude latitude vertical level (CALIPSO simulator levels, 40 levels from 0 to 19.2 km every 480 m) Time-dependent data have been averaged over the 2007-2010 period to produce climatological profiles. Units: % (0-100) 6. Cloud Fraction Temperature Profiles (3D fields) File: GISS-E3_CPE_52_cloud_fraction_temperatures_simulator.nc.zip This file provides temperature profiles of cloud fraction from the CALIPSO simulator. Variables include: clcalipsotmp: total cloud fraction profile clcalipsotmpliq: liquid cloud fraction profile clcalipsotmpice: ice cloud fraction profile These variables are functions of: longitude latitude temperature bins (40 bins from -93˙C to 27˙C every 3˙C) Time-dependent data have been averaged over the 2007-2010 period to produce climatological profiles. Units: % (0-100) Notes All variables are provided on a common horizontal grid (2.5° × 2°). Vertical levels correspond to the CALIPSO simulator configuration within GISS-E3. Missing values are represented as NaNs. The “model” dimension indexes individual CPE ensemble members. The time period used is 2007-2010 for AMIP simulations (SST-prescribed). All information can be found in Cesana et al. (2026). References Cesana et al., 2026: Improved representation of cloud properties and feedbacks in the NASA GISS-E3 climate model, J. Adv. Model. Earth Syst., submitted. Elsaesser et al., 2025: Using machine learning to generate a GISS ModelE calibrated physics ensemble (CPE). J. Adv. Model. Earth Syst., 17, no. 4, e2024MS004713, doi:10.1029/2024MS004713.