Evaluation of Low-Level Clouds, Temperature, and Surface Radiation

  Southern Ocean clouds are one of climate science's greatest challenges       

The Southern Ocean plays a pivotal role in regulating Earth's climate, yet the clouds that blanket this remote region remain among the least understood features in atmospheric science. By controlling both incoming sunlight and outgoing heat, these clouds strongly influence Earth's energy balance. Low-level clouds over the Southern Ocean (SO) exert a strong influence on surface radiation, yet their representation in reanalyses and climate models remains uncertain. Study evaluated cloud properties and surface radiative fluxes from ERA5, MERRA-2, and the CAM-ATRAS model using observations obtained over the SO by Japan's R/V Shirase. ERA5 and MERRA-2 overestimate the frequency of low-level clouds with a base below 1 km. Although CAM-ATRAS also overestimates very low clouds, it shows the best overall agreement with observations in cloud occurrence and phase. Despite the high frequency of low-level clouds, all data sets underestimate downward longwave (DLW) radiation. Cold bias and cloud phase bias likely reduce cloud-base emissivity and contribute to underestimated DLW radiation. Aerosol sensitivity experiments using CAM-ATRAS indicate that enhanced cloud condensation nuclei concentrations increase low-level cloud frequency but have a limited impact on surface radiative fluxes.

Even small errors in representing them can introduce significant uncertainties into weather forecasts, climate models and projections of future global warming, making them a long-standing challenge for climate scientists. To better understand why these clouds remain so difficult to simulate, researchers from the National Institute of Polar Research (Japan) and Nagoya University analyzed cloud observations collected during the 64th Japanese Antarctic Research Expedition (JARE64) aboard the research icebreaker R/V Shirase. Professor Jun Inoue explains, "Numerical models have been reported to exhibit limited skill in reproducing clouds. In particular, over the Southern Ocean and Antarctica, where cloud representation remains especially challenging, cloud-related biases have been shown to increase errors in the surface energy budget through biases in the radiative budget." Clouds over the Southern Ocean (SO) exert disproportionate influence on Earth's radiation budget by strongly modulating both shortwave and longwave radiative fluxes. Cloud macrophysical properties, such as horizontal and vertical cloud fractions, play critical roles in controlling downward shortwave (DSW) and downward longwave (DLW) radiation at the surface. Therefore, numerous studies have investigated the cloud fraction over the SO. Specifically, ice and liquid clouds produce distinct radiative effects because ice clouds have lower emissivity than liquid clouds, generally reflect less shortwave radiation, and emit weaker longwave radiation.

Many observational studies have been conducted focusing on the cloud phase over the SO. Although pure supercooled liquid water (SLW) freezes homogeneously below −38°C, ice nucleating particles (INPs), such as bioaerosols, mineral dust, and organic aerosols originating from both local and remote sources, can trigger freezing at much higher temperatures. Satellite and ship-based observations suggest that bioaerosols and other INPs contribute to mixed-phase and ice cloud formation at relatively high temperatures above −15°C over the SO and Antarctic coastal regions during spring and summer, which are seasons with high biological activity. Despite these findings, SLW clouds remain prevalent over the SO in summer, with occurrence frequencies of approximately 40% reported in active satellite products. Shipborne observations further indicate that SLW clouds are prevalent in the middle troposphere at temperatures higher than −25°C. Additionally, sulfate and organic aerosols, which comprise the greatest proportion of cloud condensation nuclei (CCN), play a central role in regulating low-level liquid cloud microphysics over the SO. Accordingly, many previous studies focused on aerosols and their role in low-level cloud formation over the SO. Numerical weather prediction and climate models are used widely to investigate cloud processes on the global scale. However, many models underestimate low-level cloud cover and the liquid water path, leading to excessive absorption of shortwave radiation at the surface over the SO. Such biases are commonly attributed to deficiencies in the representation of cloud phase and aerosol–cloud interactions, as demonstrated by comparisons with satellite products. The radiative effects of SO clouds remain a major source of uncertainty in climate projections. Indeed, substantial intermodel spread exists in the representation of cloud phase, and improved cloud phase representation with realistic INP concentrations can lead to reduced biases in cloud top radiative effects. Consequently, the SO remains one of the most challenging regions for cloud modeling, where persistent surface radiation biases have been identified in numerical models.

From December 2022 to March 2023, ship-based instruments continuously measured cloud properties, atmospheric temperature and humidity, surface radiation and aerosol concentrations, providing a comprehensive benchmark for evaluating model performance. The team evaluated two widely used atmospheric reanalysis data sets, ERA5 and MERRA-2, alongside the CAM-ATRAS climate model using observations throughout the expedition. Although all three data sets broadly captured cloud patterns over the Southern Ocean, important differences emerged. ERA5 and MERRA-2 consistently overestimated the occurrence of low-level clouds, whereas CAM-ATRAS most closely matched the observations, particularly in reproducing cloud occurrence and cloud phase. Surprisingly, despite simulating abundant low-level clouds, all three data sets underestimated the amount of downward longwave radiation reaching the surface. Comparison with observations showed that the reanalysis data sets contain higher aerosol concentrations than observed. Therefore, the researchers also conducted sensitivity experiments with CAM-ATRAS by increasing aerosol emissions over the Southern Hemisphere to examine how aerosols influence cloud formation and surface radiation. However, the aerosol sensitivity experiments further showed that increasing aerosol concentrations produced more low-level clouds but had only a limited effect on surface radiation.

The researchers traced this discrepancy to the physical properties of the simulated clouds rather than to cloud amount alone. In the models, clouds contained excessive ice, reducing the heat emitted toward the surface. However, these results demonstrate that biases in cloud representation alone cannot explain the underestimated DLW. Instead, the numerical models exhibit an inherent cold temperature bias, which also plays a role in the underestimation of DLW. These findings indicate that accurately representing both cloud phase and temperature is more important than simply reproducing cloud frequency when simulating the Southern Ocean's surface energy budget. By identifying the processes responsible for persistent cloud biases, the study provides valuable guidance for improving weather and climate models. Better representation of cloud microphysics, aerosol–cloud interactions and the background environment will help reduce uncertainties in simulations of Earth's energy balance, leading to more reliable predictions of future warming, sea ice change and climate variability. CAM-ATRAS explicitly simulates aerosol processes, including new particle formation, condensation, coagulation, activation to cloud droplets, aqueous-phase chemistry, dry and wet deposition, aerosol–radiation interactions, and aerosol–cloud interactions. Aerosols are represented using 12 size bins spanning diameters from 1 nm to 10 μm. Major aerosol species, including sulfate, black carbon, organic matter, SS, dust, marine organic aerosols, and bioaerosols are explicitly simulated, allowing detailed investigation of aerosol number concentrations and mixing states. INP number concentrations in clouds are calculated based on the simulated concentrations of dust, marine organic aerosols, and bioaerosols, the temperature dependence of ice-nucleation active site density per unit mass for each species, ambient temperature, and cloud fraction. These INP concentrations are then used to calculate ice nucleation within the cloud microphysical scheme.

In this study, the simulation period corresponding to the JARE64 cruise was used for analysis. Monthly sea surface temperature and sea ice distributions were prescribed as boundary conditions. The model was nudged toward MERRA-2 reanalysis fields for temperature and horizontal wind components in the free troposphere (pressure levels <800 hPa). The horizontal resolution was 0.9° × 1.25°, with 30 vertical layers extending from the surface to 40 km. In addition to the base simulation, a sensitivity simulation in which aerosol emissions over the SH were enhanced by two orders of magnitude (hereafter, referred to as the CAM-ATRAS SH × 100 experiment). The researchers emphasize that continued progress will require not only expanded observations of clouds across the Southern Ocean and Antarctica but also increased observations of fundamental atmospheric variables, particularly temperature, to reduce the cold bias in numerical models. As Assistant Professor Kazutoshi Sato notes, "Because observations over Antarctica remain sparse, numerical models still contain substantial uncertainties in their representation of the Antarctic atmosphere. Therefore, incorporating existing but currently underutilized observations into numerical models may provide an effective solution. For example, assimilating observations from the PANSY radar at Japan's Syowa Station, which are not yet routinely used in numerical weather prediction systems, could help reduce model biases and improve forecast accuracy." This study represents one of the most comprehensive observational evaluations of cloud and radiation simulations over the Southern Ocean using data collected during the JARE64 expedition.

During JARE64, clouds with a wide range of cloud base heights were observed over the SO and the Antarctic coastal regions. We calculated the frequency distribution of cloud base height over the entire observation period using 500-m vertical bins. Observations showed that clouds with a base height of below 1 km exhibit both primary and secondary peaks in occurrence frequency. Clouds with a base height in the middle troposphere were also frequently observed over the SO and Antarctic coastal regions. To compare cloud base height between observations and models, the modeled cloud base height was defined as the lowest model level at which the cloud liquid or ice mixing ratio exceeds 0.001 g kg−1. ERA5 overestimates the frequency of clouds with a base height of below 1 km. For CAM-ATRAS, although the frequency of clouds with a base height below 0.5 km is still overestimated, the occurrence frequency of clouds with a base height of below 1 km is closer to that of the observations than that of the two reanalysis data sets (ERA5 and MERRA-2). By revealing why current models struggle to reproduce these clouds and identifying the processes responsible for long-standing biases, the findings provide an important step toward more accurate weather forecasts, improved climate models and more confident projections of Earth's changing climate.

The National Institute of Polar Research (NIPR) was founded in 1973, is an inter-university research institute dedicated to advancing scientific research and observations in the Arctic and Antarctic regions. As one of the four institutes under the [Research Organization of Information and Systems (ROIS)], NIPR conducts comprehensive polar research through observation stations and international collaborations. The institute also promotes polar science by supporting collaborative research projects and providing access to scientific data, samples, and materials. NIPR remains Japan’s only institution devoted to comprehensive research activities in both polar regions. ROIS is a parent organization of four national institutes (National Institute of Polar Research, National Institute of Informatics, the Institute of Statistical Mathematics and National Institute of Genetics) and the Joint Support-Center for Data Science Research. It is ROIS's mission to promote integrated, cutting-edge research which goes beyond the barriers of these institutions, in addition to facilitating their research activities, as members of inter-university research institutes.