This paper provides an overview of the United States (U.S.) Department of Energy’s (DOE’s) Energy Exascale Earth System Model version 2.1 with an Arctic regionally refined mesh (RRM), hereafter referred to as E3SMv2.1-Arctic, for the atmosphere (25 km), land (25 km), and ocean/ice (10 km) components. We evaluate the atmospheric component and its interactions with land, ocean, and cryosphere by comparing the RRM (E3SM2.1-Arctic) historical simulations (1950-2014) with the uniform low-resolution (LR) counterpart, reanalysis products, and observational datasets. The RRM generally reduces biases in the LR model, improving simulations of Arctic large-scale mean fields, such as precipitation, atmospheric circulation, clouds, atmospheric river frequency, and sea ice dynamics. However, the RRM introduces a seasonally dependent surface air temperature bias, reducing the LR cold bias in summer but enhancing the LR warm bias in winter. The RRM also underestimates winter sea ice area and volume, consistent with its strong winter warm bias. Radiative feedback analysis shows similar climate feedback strengths in both RRM and LR, with the RRM exhibiting a more positive surface albedo feedback and contributing to a stronger surface warming than LR. These findings underscore the importance of high-resolution modeling for advancing our understanding of Arctic climate changes and their broader global impacts, although some persistent biases appear to be independent of model resolution at 10-100 km scales.

 Assessing Radiative Feedbacks and their Contribution to the Arctic Amplification Measured by Various MetricsY. Huo1, H. Wang1, J. Lu1, Q. Fu2, A. K. Jonko3, Y. J. Lee4, W. Ma1, W. Maslowski4, Y. Qin11Atmospheric, Climate, and Earth Sciences Division, Pacific Northwest National Laboratory, Richland, WA, USA.2Department of Atmospheric Sciences, University of Washington, Seattle, Washington, USA3Earth and Environmental Science Division, Los Alamos National Laboratory, Los Alamos, NM, USA4Department of Oceanography, Naval Postgraduate School, Monterey, CA, USACorresponding author: Y. Huo (yiling.huo@pnnl.gov), and H. Wang (Hailong.Wang@pnnl.gov)Key Points:·      Arctic amplification is quantified by applying various metrics to observational data, reanalysis products, and model simulations.·      Radiative feedback analysis shows that albedo and lapse rate feedbacks are the main contributors to AA for all metrics.·      Water vapor and cloud feedbacks are most heavily influenced by internal variability. AbstractArctic amplification (AA), characterized by a more rapid surface air temperature (SAT) warming in the Arctic than the global average, is a major feature of global climate warming. Various metrics have been used to quantify AA based on SAT anomalies, trends, or variability, and they can yield quite different conclusions regarding the magnitude and temporal patterns of AA. This study examines and compares various AA metrics for their temporal consistency in the region north of 70°N from the early 20th to the early 21st century using observational data and reanalysis products. We also quantify contributions of different radiative feedback mechanisms to AA based on short-term climate variability in reanalysis and model data using the Kernel-Gregory approach. Albedo and lapse rate feedbacks are positive and comparable, with albedo feedback being the leading contributor for all AA metrics. The net cloud feedback, which has large uncertainties, depends strongly on the datasets and AA metrics used. By quantifying the influence of internal variability on AA and related feedbacks based on global climate model ensemble simulations, we find that water vapor and cloud feedbacks are most heavily affected by internal variability.Plain Language SummaryThe Arctic has been observed to warm at a faster pace than the rest of the world, especially over the recent few decades. This phenomenon is known as Arctic amplification (AA), which has connections with different components of the Earth’s climate system. Various metrics are used in the science community to measure AA in different datasets and to assess the performance of climate models in reproducing the observed history of AA. In this study, we apply a collection of AA metrics from the literature to the same historical datasets, based on observations and climate model experiments, to identify differences/similarities and pros/cons of these metrics. We also study how radiatively active processes in the climate system, like changes in snow/ice surface reflectivity (albedo), air temperature changes with height (lapse rate), humidity increase, and cloud changes, can in turn amplify (positive feedback) or dampen (negative feedback) this overall Arctic warming. Our study shows that changes in the Arctic surface albedo and lapse rate are the main drivers of AA. We also find that natural climate variations can greatly influence Arctic warming in climate models and their ability to reproduce the observed AA by affecting regional water vapor and cloud patterns.1 IntroductionArctic amplification (AA), characterized by a more rapid surface air temperature (SAT) warming in the Arctic than global average, is a major feature of both contemporary and historical global climate warming (Wood and Overland, 2009; Previdi et al., 2020; Thoman et al. 2020; Moon et al., 2021; Rantanen et al., 2022; Chylek et al., 2023a, b). Such changes in temperature have also been accompanied by substantial and rapid changes in various aspects of the Arctic climate system (Previdi et al., 2021; Taylor et al., 2022). The decline of sea ice in the Arctic Ocean stands out as a prominent example of this ongoing climate change primarily driven by atmospheric near-surface warming (Olonscheck et al., 2019) and can, in turn, have a positive feedback on SAT warming (Screen et al., 2012). Meanwhile, the amplified warming observed in the Arctic region has been attributed to various other feedback processes, including the Planck feedback, and feedbacks associated with lapse rate (LR), water vapor (WV), clouds, large-scale atmospheric circulation changes and ocean heat transport (Serreze and Francis, 2006; Serreze and Barry, 2011; Screen et al., 2012; Feldl et al., 2020; Wendisch et al., 2023; Hajjar and Salzmann, 2023; Singh et al., 2017; Zhang et al., 2018). These combined feedback processes lead to a higher sensitivity of the Arctic climate system to greenhouse gas increases, when compared to lower latitudes (Rantanen et al., 2022).Quantifying AA is crucial for understanding the pace and magnitude of climate change in the Arctic, which has significant implications for the entire planet. AA metrics serve as valuable tools for identifying periods characterized by the most significant differences in SAT responses to climate change between the Arctic and the whole globe. However, AA quantification has long been a challenging issue, and this difficulty was systematically discussed by Bekryaev et al. (2010). Their work laid the foundation for subsequent research, which has aimed to fill the knowledge gap in our understanding and to improve the metrics for AA quantification. Various metrics have been used to quantify AA based on SAT anomalies, trends, or variability, and these metrics can yield quite different conclusions regarding the magnitude and temporal patterns of AA (Table 1). One commonly used metric is the difference in spatially averaged SAT anomalies between the Arctic and a reference region (Francis and Vavrus, 2015), such as the Northern hemisphere or the globe. Another similar metric involves the ratio of these two anomalies (Ono et al., 2022). These two metrics, directly based on temperature anomalies, tend to have large temporal variability owing to the large internal climate variability of the Arctic SAT (Wood and Overland, 2009). Another type of AA metric, which has lower temporal variability, is defined as the ratio of SAT linear trends over the Arctic and the reference region (Johannessen et al., 2016). This metric is particularly useful for assessing Arctic climate change on multi-decadal and longer timescales. However, uncertainties in the linear trends can affect the magnitude of the metric and result in large spread, especially for periods without statistically significant trends. Moreover, the two metrics based on a ratio of two variables can have abnormally large values when the denominator is close to zero (e.g., near-zero global SAT anomaly or its trend), as also indicated by Hind et al. (2016). An alternative metric uses the ratio of SAT variability (e.g. standard deviation) in the Arctic and the reference region (Kobashi et al., 2013), providing a measure of the sensitivity of AA across different timescales. Moreover, the latter metric has an advantage over the former two based on ratios, as its denominator, the global SAT variability, does not approach zero. Furthermore, as Bekryaev et al. (2010) highlighted, the commonly used ratio of long-term Arctic and global SAT trends for AA quantification has significant limitations because strong multidecadal variability complicates the separation of trends from long-term fluctuations, making it challenging to quantify their individual contributions. Thus, Bekryaev et al. (2010) suggested using linear regression coefficient between Arctic and global SAT anomalies as a more stable metric compared to those linking polar and global temperature anomalies where denominator can approach zero. Note that metrics based on linear trends or inter-annual variability have no independent values for every year due to the need of a running window by design. Consequently, they are unsuitable for studying temporal behavior at periods shorter than the window length. On the other hand, the first two metrics can generate independent values for each year or season and thus is the best choice for year-to-year variability investigation. Finally, Davy et al. (2018) argued that it is crucial to consider the difference in variability of the Arctic and the reference region to provide an equitable comparison of climate change in these two regions. One approach to this problem is to introduce appropriate weighting by scaling the SAT anomalies to the magnitude of variability in each region (Przybylak and Wyszyński, 2020).Each of the metrics discussed above has its advantages and disadvantages for studying AA and thus one of the goals of this study is to compare these established AA metrics and assess their consistency during different time periods. We apply the climate feedback framework reported in previous studies (Soden et al., 2008; Zhang el al., 2018; Qin et al., 2022) to investigate the local and remote processes and mechanisms driving AA and to understand how the climate system responds to external radiative forcing. We decompose atmospheric radiative flux changes due to feedbacks from temperature (Planck and lapse rate feedbacks), surface albedo, WV, and clouds. Unlike previous studies, which mainly focused on the equilibrium temperature change in response to radiative forcing (Pithan and Mauritsen, 2014; Goosse et al., 2018; Stuecker et al., 2018; Hahn et al., 2021), we also quantify contributions of different radiative feedback mechanisms to AA based on short-term climate variability in reanalysis data and model outputs using the Kernel-Gregory approach (Gregory et al., 2004). We aim to examine radiative feedbacks that result in the amplified warming in the Arctic relative to the global or tropical average SAT increases by quantifying and comparing the radiative feedbacks on AA characterized by the various metrics.We also examine the role of natural climate variability, which strongly influences Arctic sea ice (Wu et al., 2021; Bonan et al., 2021). Hence it can also be partly responsible for Arctic SAT and AA changes (Rantanen et al., 2022; Chylek et al., 2022; Sweeney et al., 2023) by affecting heat/moisture transport and radiative feedbacks. In other words, AA is not solely attributable to externally driven global warming; rather, it encompasses patterns resulting from a combination of internal variability and various forcings operating during the examined period. Model simulations starting from different initial conditions, which represent distinct realizations of internal variability in each ensemble member associated with small perturbations to initial conditions, can be used to estimate the response of the Arctic climate to both external forcings and internal variability (Kay et al., 2015). This study utilizes Community Earth System Model 2 (CESM2) Large Ensemble (LE) simulations (Rodgers et al., 2021), generated by small perturbations to initial conditions of each member, to estimate the response of Arctic climate to both external forcings and internal variability (Kay et al., 2015) and to quantify the influence of internal variability on AA and feedback analysis.2 Datasets and Methodology2.1 DatasetsObservational and reanalysis datasets and Earth system model outputs are used in this study. We use the Met Office Hadley Centre/Climatic Research Unit global surface temperature dataset version 5 (HadCRUT5) global historical observational surface temperature monthly dataset (version 5.0.2.0) from January 1850 onwards with a resolution of 5° (Morice et al., 2021). The reanalysis data is the fifth generation European Centre for Medium-Range Weather Forecasts (ECMWF) reanalysis (ERA5; Hersbach et al., 2023a; Hersbach et al., 2023b), covering January 1940 to present. It provides hourly data on a 0.25° grid, and is produced by assimilating both satellite and in-situ observations. CESM2 LE simulations (Rodgers et al., 2021) are also used to quantify the influence of internal variability on AA and feedback analysis. The CESM2 LE consists of 100 members at 1° resolution and is available from 1850 to 2100 at monthly, daily, 6-hourly and hourly frequency. The ensemble spread was generated by employing various initial states for both the oceanic and atmospheric components. Only monthly data over the Coupled Model Intercomparison Project Phase 6 (CMIP6) historical (through 2014) period, which are subject to historical external forcing induced by changes in greenhouse gases, aerosol, and land cover, are used. Note that our studies only use the first set of 50 ensemble members of the CESM2 LE, which follow the original protocol for CMIP6 historical forcings.2.2 AA metricsSeven AA metrics (Table 1) have been applied to the three datasets discussed in Section 2.1, henceforth referred to as A1 - A7. SAT monthly anomalies are first computed by subtracting the climatological monthly means of the reference period 1950–1980 for the area-weighted average SATs. These monthly anomalies are then averaged to derive annual and seasonal (June-July-August or JJA for summer and December-January-February or DJF for winter) anomalies. Area-weighted averaging is applied to determine the time series of the Arctic (north of 70°N) and global SAT anomalies, which are the basis for AA quantification using the various metrics.Table 1. Summary of seven AA metrics used in this study. For a given period considered, SAT anomaly in A1, A5and A6 is the average of monthly anomalies in each calendar year (season) for the yearly (seasonal) anomaly. The unit for A1 is K, whereas remaining metrics are dimensionless.MetricDefinitionProsConsScientific purposeReferenceA1Difference between Arctic SAT anomaly and global SAT anomalyQuantified at any temporal resolutionLarge temporal variability Most commonly used for year-to-year variability investigationFrancis and Vavrus (2015)A2Ratio of the absolute value of Arctic SAT linear trend to the absolute value of global SAT linear trendLow temporal variabilityNo values for every individual year; concern of the zero denominator Assessing Arctic climate change on multi-decadal and longer timescalesJohannessen et al. (2016)A3Ratio of the Arctic SAT interannual variability, measured by standard deviation of yearly/seasonal anomalies, to the global SAT interannual variability No concern of the zero denominator No values for every individual yearAssessing inter-annual and multi-decadal temporal variability of AA togetherKobashi et al. (2013)A4Coefficient of linear regression between Arctic and global SAT yearly/seasonal anomaliesNo concern of the zero denominatorNo values for every individual yearAssessing inter-annual and multi-decadal temporal variability of AA togetherBekryaev et al. (2010)A5Ratio of the Arctic-mean to the global-mean SAT anomaliesQuantified at any temporal resolutionLarge temporal variability; concern of the zero denominatorYear-to-year variability investigationOno et al. (2022)A6Ratio of the Arctic standardized SAT anomaly to the global standardized SAT anomalyEquitable comparison of the Arctic and global climate change with the respective variability consideredLarge temporal variability; concern of the zero denominatorConsidering the variability in comparable areas; better distinguish between the forced change from the internal variabilityPrzybylak and Wyszyński (2020)A7Ratio of the Arctic trends of standardized SAT yearly/seasonal anomaly to the global trends of standardized SAT yearly/seasonal anomalyEquitable comparison of the Arctic and global climate change with the respective variability consideredNo values for every individual year; concern of the zero denominator Considering the variability in comparable areas; better distinguish between the forced change from the internal variabilityPrzybylak and Wyszyński (2020)For A2, linear least squares regressions are first fitted to the time series of SAT anomalies, and the significance of these regressions is tested using a two-tailed Wald Test with t-distribution. Autocorrelation is not considered here, and the significance might be overestimated. A3 based on the standard deviation of yearly SAT anomalies is dominated by interannual variability but also affected by the multi-decadal variability and trends. Linear least squares regression based on yearly/seasonal anomalies is applied to determine the coefficient of linear regression between the Arctic and global SAT anomaly time-series in A4, and the statistical significance of each regression result is assessed using a two-tailed Student-t test. The standard deviation from the time series is employed for the standardization of SAT anomalies in A6and A7.2.3 Radiative feedback analysisWe utilize the combined Kernel-Gregory method (Gregory et al., 2004; Block and Mauritsen, 2013; Zhang et al., 2018; Kramer et al., 2019b) for calculating radiative feedbacks. Monthly anomalies (denoted by 

In high-latitude environments such as the Arctic Ocean, phytoplankton growth is strongly constrained by light availability. Because light penetration into the upper ocean is attenuated by snow and ice cover, it was generally believed until recently that phytoplankton growth was limited to areas of open water, with negligible growth under the ice. However, under-ice phytoplankton blooms have been reported multiple times over the past several decades [e.g. Fukuchi et al. (1989); Legendre, Ingram, and Poulin (1989)]. In July 2011, Arrigo et al. (2012) observed a massive phytoplankton bloom beneath sea ice in the Chukchi Sea. Observational evidence suggests that this bloom was not an isolated case, and that under-ice blooms maybe widespread on Arctic continental shelves (Arrigo et al., 2014; Lowry, van Dijken, & Arrigo, 2014). Arrigo and van Dijken (2011) estimate the total primary production north of the Arctic Circle to be 438 +/- 21.5 Tg C yr -1. However, due to observational limitations, this estimate did not include under sea ice production. Therefore, an open question remains: How important are under-ice phytoplankton blooms to the total Arctic primary production? RASM is a high-resolution, fully-coupled, regional model with a domain encompassing the entire marine cryosphere of the Northern Hemisphere, including the major inflow and outflow pathways, with extensions into North Pacific and Atlantic oceans. The components of RASM include: atmosphere, sea ice, ocean, biogeochemical, and land hydrology (Maslowski et al. 2012, Roberts et al. 2015, DuVivier et al. 2016, Hamman et al. 2016, Hamman et al. 2017, Cassano et al. 2017). The ocean BGC component in RASM is a medium-complexity Nutrients-Phytoplankton-Zoo-plankton-Detritus (NPZD) model (Jin et al. 2018). The model has three phytoplankton categories: diatoms, small phytoplankton and diazotrophs. RASM results show that under-ice pelagic chl-a and primary production values can at times be very high, particularly during the spring and early summer. Our numerical model results produce a mean of 495 Tg C yr -1 north of the Arctic Circle during 1980-1998 (and 507 Tg C yr -1 during 1980-2018). We also see an increase in primary production over the last several decades. This increase is attributed to the reduced sea ice cover, which increases light availability to the upper ocean. We conclude that under-sea-ice pelagic primary production makes up a large fraction of the total production and cannot be considered negligible.

A set of decadal simulations has been completed and evaluated for gains using the Regional Arctic System Model (RASM) to dynamically downscale data from a global Earth system model (ESM) and two atmospheric reanalyses. RASM is a fully coupled atmosphere - land - ocean - sea ice regional Earth system model. Nudging to the forcing data is applied to approximately the top half of the atmosphere. RASM simulations were also completed with a modification to the atmospheric physics for evaluating changes to the modeling system. The results show that for the top half of the atmosphere, the RASM simulations follow closely to that of the forcing data, regardless of the forcing data. The results for the lower half of the atmosphere, as well as the surface, show a clustering of atmospheric state and surface fluxes based on the modeling system. At all levels of the atmosphere the imprint of the weather from the forcing data is present as indicated in the pattern of the monthly and annual means. Biases, in comparison to reanalyses, are evident in the ESM forced simulations for the top half of the atmosphere but are not present in the lower atmosphere. This suggests that bias correction is not needed for fully-coupled dynamical downscaling simulations. While the RASM simulations tended to go to the same mean state for the lower atmosphere, there is a different evolution of the weather across the ensemble of simulations. These differences in the weather result in variances in the sea ice and oceanic states.

As the sea-ice modeling community is shifting to advanced numerical frameworks, developing new sea-ice rheologies, and increasing model spatial resolution, ubiquitous deformation features in the Arctic sea ice are now being resolved by sea-ice models. Initiated at the Forum for Arctic Modelling and Observational Synthesis (FAMOS), the Sea Ice Rheology Experiment (SIREx) aims at evaluating current state-of-the-art sea-ice models using existing and new metrics to understand how the simulated deformation fields are affected by different representations of sea-ice physics (rheology) and by model configuration. Part I of the SIREx analysis is concerned with evaluation of the statistical distribution and scaling properties of sea-ice deformation fields from 35 different simulations against those from the RADARSAT Geophysical Processor System (RGPS). For the first time, the Viscous-Plastic (and the Elastic-Viscous-Plastic variant), Elastic-Anisotropic-Plastic, and Maxwell-Elasto-Brittle rheologies are compared in a single study. We find that both plastic and brittle sea-ice rheologies have the potential to reproduce the observed RGPS deformation statistics, including multi-fractality. Model configuration (e.g. numerical convergence, atmospheric forcing, spatial resolution) and physical parameterizations (e.g. ice strength parameters and ice thickness distribution) both have effects as important as the choice of sea-ice rheology on the deformation statistics. It is therefore not straightforward to attribute model performance to a specific rheological framework using current deformation metrics. In light of these results, we further evaluate the statistical properties of simulated Linear Kinematic Features (LKFs) in a SIREx Part II companion paper.

Simulating sea-ice drift and deformation in the Arctic Ocean is still a challenge because of the multi-scale interaction of sea-ice floes that compose the Arctic sea ice cover. The Sea Ice Rheology Experiment (SIREx) is a model intercomparison project formed within the Forum of Arctic Modeling and Observational Synthesis (FAMOS) to collect and design skill metrics to evaluate different recently suggested approaches for modeling linear kinematic features (LKFs) and provide guidance for modeling small-scale deformation. In this contribution, spatial and temporal properties of LKFs are assessed in 36 simulations of state-of-the-art sea ice models and compared to deformation features derived from RADARSAT Geophysical Processor System (RGPS). All simulations produce LKFs, but only very few models realistically simulate at least some statistics of LKF properties such as densities, lengths, or growth rates. All SIREx models overestimate the angle of fracture between conjugate pairs of LKFs and LKF lifetimes pointing to inaccurate model physics. The temporal and spatial resolution of a simulation and the spatial resolution of atmospheric forcing affect simulated LKFs as much as the model’s sea ice rheology and numerics. Only in very high resolution simulations (≤2\,km) the concentration and thickness anomalies along LKFs are large enough to affect air-ice-ocean interaction processes.