Warm and moist atmospheric flow caused a record minimum July sea ice extent of the Arctic in 2020

The satellite observations unveiled that the July sea ice extent of the Arctic shrank to the lowest value in 2020 since 15 1979, with a major ice retreat in the Eurasian shelf seas including Kara, Laptev, and East Siberian Seas. Based on the ERA-5 reanalysis products, we explored the impacts of warm and moist air-mass transport on this extreme event. The results reveal that anomalously high energy and moisture converged into these regions in the spring months (April to June) of 2020, leading to a burst of high moisture content and warming within the atmospheric column. The convergence is accompanied by local enhanced downward longwave radiation and turbulent fluxes, which is favorable for initiating an early melt onset in the areas 20 with severe ice loss. Once the melt begins, solar radiation played a decisive role in leading to further sea ice depletion due to ice-albedo positive feedback. The typical trajectories of the synoptic cyclones that occurred on the Eurasian side in spring 2020 agree well with the path of atmospheric flow. Assessments suggest that variations in characteristics of the spring cyclones are conducive to the severe melt of sea ice. We argue that large-scale atmospheric circulation and synoptic cyclones act in concert to trigger the exceptional poleward transport of total energy and moisture from April to June to cause this new record minimum 25 of sea ice extent in the following July.

where is the number of pressure levels, 0 is the lowest pressure level.
corresponds to pressure at the nth pressure level, and represents the northward component of the wind speed and specific humidity at the nth pressure level, respectively.
We compared our estimated results of the vertical integrated northward water vapor flux against the existing dataset archived in ERA5. The results (not shown) show that the estimated results are highly consistent with the corresponding ERA5 filed both in the magnitude and change of all months across various latitudes (e.g., 70° N) during the period 1979 to 2020, which 120 lends credence to the direct use of the water vapor flux field obtained from ERA5. By the same token, we take advantage of the vertical integral of total energy from ERA5. The vertically integrated, atmospheric, northward energy transport consists of internal, potential, kinetic and latent energy.

Changes in sea ice thickness due to melt
Changes in surface energy budget related to energy and water vapor convergence affect sea-ice melting. The thickness of melt 125 caused by alteration of surface energy budget can be calculated via the sea-ice growth model (Maykut et al., 1992;Eisenman et al., 2007). Neglecting some smaller radiative fluxes, the changes in sea ice thickness can be written as a simplified function of the surface radiation and turbulent flux: where Δℎ represents sea-ice change, △ is the time step, represents the density of sea ice (917 kg/m 3 ), is the latent 130 heat of fusion for sea ice ( 333.4 kJ/kg ), ↓ and ↓ represent the surface net fluxes of longwave and shortwave radiation, respectively. ↓ corresponds to the sensible heat, and ↓ denotes the latent heat.

Cyclone identification and tracking
To examine the effects of cyclone activities on the anomalous energy and moisture transport in spring 2020, we use a revised automatic cyclone identification and tracking algorithm developed originally by Serreze et al. (1993) to diagnose the center 135 positions and trajectories of the cyclones from the 6-hourly SLP data (Serreze et al., 1993;Serreze, 1994;Serreze et al., 1997;Wang et al., 2006;Wang et al., 2013). The cyclone detection and tracking algorithm consists of two steps: (1) Inspecting the candidate center where the pressure is lower by 0.1 hPa than the surrounding grid points (Wang et al., 2013). If multiple cyclone center candidates are found within a radius of 1200 km, the one with the largest local Laplacian of SLP is determined as the exact cyclone center. (2) Tracking the centers between two consecutive time steps based on the "nearest neighbor" rule 140 to form trajectories, with further checks including the distance moved in specific directions and pressure tendency. Therefore, a cyclone track consists of a series of cyclone centers identified in sequential time steps at adjacent locations. In this study, https://doi.org/10.5194/tc-2021-159 Preprint. Discussion started: 26 July 2021 c Author(s) 2021. CC BY 4.0 License.
2, the average Arctic atmospheric condition from April to June 2020 was dominated by a persistent low-pressure anomaly centered over the north pole and extended southwards from the Barents-Kara Seas to the middle part of northern Eurasia. Two high-pressure anomaly centers were located in Eastern Siberia and around the Norwegian Sea, respectively. These SLP modes favor anomalous southerly winds, which transport moist and warm air mass from Eurasia into the Arctic through the entry in 160 the Kara Sea. After entering the Arctic Ocean, the air mass was deflected to move along the coast of Eurasia and influenced the shelf seas.
https://doi.org/10.5194/tc-2021-159 Preprint. Discussion started: 26 July 2021 c Author(s) 2021. CC BY 4.0 License. Figure 2. Spatial patterns of SLP anomalies (shading) during April to June 2020. The anomalies are computed as the difference between the averaged fields of the three months (April-June) and the corresponding climatology over the past four decades .

165
Using ERA5 reanalysis, we quantify the anomalies of the vertical integral of meridional total energy and water vapor flux.
As shown in Fig. 3, an anomalously large advection of energy and water vapor from lower latitudes, which is diverted by wind variations, prevailed in the region with conspicuous sea ice retreat ( Fig. 1) in spring 2020. Regions around the Laptev and Kara sea (45° E-120° E, 70° N) are the main entry channels for warm air-mass input from lower latitudes. It is estimated that the zonal mean of the meridional total energy flux (water vapor flux) through these main entry channels over the entire spring in 170 2020 reached up to 1.74×10 11 Wm -1 (1.51×10 3 kg m -1 s -1 ), producing a transport that was 2 (3) standard deviations larger than the 1979-2020 climatology. The pronounced poleward energy and moisture through the entry then converged into the Arctic.
As depicted in Fig. 3c and d, the major parts of the ice-retreated shelf seas in spring 2020 are characterized by positive convergence anomalies of the atmospheric moisture and energy transport. Particularly, the magnitude of the total energy and moisture flux convergence anomaly even exceeds 50 Wm -2 and 9×10 -6 kg m -1 s -2 , respectively, in the Kara Sea).  180 Figure 4 illustrates the meridional cross-sections of temperature and specific humidity anomalies spanning the regions with maximum convergence of the atmospheric fluxes (60° E-165° E, 60° N -90° N). Horizontally, elevated temperature and higher moisture content distributed widely from 60° N to 85° N near the surface. Vertically, the positive temperature and moisture anomalies extend conspicuously into the mid-troposphere (~750 hPa). The intrusion of moisture and energy leads to surface warming (damping) of up to 3-4 K (6-8×10 -4 kg kg -1 ) in the spring months. The vertical patterns of the anomalies 185 indicate that the great convergence ( Fig. 3c and d) of energy and moisture could contribute to the local increases in the atmospheric temperature and humidity, both at the surface and in the troposphere above the boundary layer, which is in agreement with the finding of Graversen et al. (2008). Noticeably, unusual conditions that higher moisture content and warming within the Arctic atmospheric column prevailed over the ice cover loss region. We also examine the role of local evaporation in the regional increase of moisture under a warmer Arctic climate. According to the ERA5 reanalysis, the spring 190 evaporation over the Arctic Basin exhibits a decreasing trend over the past four decades, except for the Barents and Norwegian Seas. In April-June 2020, below-normal evaporation dominated the Arctic with an averaged negative value of -1.5×10 -4 m in the regions with notable ice-retreat (not shown). The decline in evaporation indicates that the enhanced moisture contributing to the moister atmosphere over these regions is primarily provided by atmospheric transport from remote

Surface energy Budget
The surface energy budget that consists of thermal radiation, solar radiation, and turbulent fluxes is vital for sea ice melt and growth. An increase of humidity associated with the convergence of moisture flux may strengthen cloud formation (Johansson 205 et al., 2017), of which both contribute to the enhanced local greenhouse effect. In addition, the energy convergence in the atmosphere may partly be radiated directly to space in the form of longwave radiation, and partly radiated to the sea surface and turbulently mixed, contributing to the sea ice melt. Having shown the anomalously large convergence of water vapor and total energy transport in April-June 2020, in the following we will present the variations of different surface energy flux components. Note that the ECMWF convention for vertical fluxes is positive downwards. 210 In the Eurasian shelf seas with remarkable sea ice shrinkage, the surface gained more energy owing to both shortwave and longwave radiation, as well as turbulent fluxes, as the enhanced surface fluxes predominantly appeared in these regions ( convergence zones, with the largest amplitudes in the Kara Sea (~32 Wm -2 , Fig. 5a). The anomalies of the net longwave radiation (Fig. 5b) are roughly similar in spatial distribution to that of the downward component over the Arctic marginal seas 215 on the Eurasian side. The difference between Fig.5a and b indicates that part of thermal radiation was radiated upwards to increase the surface air temperature before the melt commenced. Downward and net shortwave radiation anomalies are portrayed in Fig. 5d and e. The downward component of the solar radiation was below-normal in most parts of the ice-retreated area (Fig. 5d), which is presumably attributed to increased cloudiness associated with the convergence of moisture. In contrast, remarkably positive anomalies of the net solar radiation were found in the Eurasian shelf seas where the extensive loss of sea 220 ice is observed. This is a result of the substantial formation of open water due to sea ice loss which reduces the albedo and thereby enables the upper ocean to absorb more heat (i.e., the ice-albedo feedback).
Additionally, sensible and latent turbulent surface flux anomalies both make a contribution to the energy surplus at the surface in the spring months of 2020 ( Fig. 5c and f). The positive (downward) anomalies of turbulent surface fluxes were detected in the regions which coincide well with the seas with contracted ice cover ( Fig. 5c and f). Intuitively, more turbulent 225 fluxes would be released to the atmosphere as more open water prevailed. That is, a negative (upward) value over the Arctic shelf seas is expected. However, reduced upward, or even downward, sensible and latent heat fluxes are detected in the study region during April-June in 2020. This can be attributed to the anomalously high moisture advection and convergence which as a result could reduce the gradient of the water vapor pressure at the surface. As implied in Fig.4, positive temperature and humidity anomalies extend from surface even to mid-troposphere, peaks at around 925 hPa. These changes would result in a 230 decreased vertical gradient in air temperature and humidity in the lower atmosphere, reducing the hypothesized upward turbulent fluxes from the ocean surface to the overlying atmosphere.  Figure 6a presents the time series of SIE, the anomalies of atmospheric energy transport convergence and the surface fluxes averaged over the study area (enclosed by the green polygon in Fig. 3c and d) during 2020. Indeed, the energy convergence event started at the end of March and lasted for three months, peaked in early June. This is followed by an augment in the downward thermal radiation plus turbulent fluxes (smaller) by ~30-40 Wm -2 (Fig. 6a, green line). The almost simultaneous 240 response of downward thermal radiation highlights that the convergence of the total energy and moisture flux has a significant imprint on the increased surface energy fluxes. With the enhanced downward infrared radiation, sea ice cover began to decrease gradually (Fig. 6a, grey line). The time development underlines that the positive anomalies of longwave radiation plus turbulent fluxes played a significant role in initiating an early than usual melt in 2020. As estimated from fields of melt dates archived in NASA, persistent melt conditions in the study area were observed in May 2020 (Fig. 6, red vertical line), 245 which occurred about 15 days earlier than the average value of the period 1979-2020. As the melt commenced, the formation of open water decreased the surface albedo, which in turn acted to increase the absorption of solar radiation (Fig. 6a, red line).
That is, the earlier melt onset could foster stronger ice-albedo feedback (Hall, 2004), leading to an accelerated decline in SIE in June-July when the anomaly of net solar radiation reached its maximum.

255
To quantify the thermodynamic impact of atmospheric energy of spring 2020 on the sea ice melt, we calculate the changes in sea-ice thickness due to the variations of surface energy fluxes via the sea-ice growth model (Maykut et al., 1992;Eisenman et al., 2007). According to equation (3), a 1 W/m -2 increase in surface energy budget during three spring months (April to June) would melt approximately 2.60 cm of sea ice. The spatial pattern of sea ice thickness change due to surface energy fluxes variations, calculated by equation (4), is portrayed in Fig. 7a. SIT anomalies due to radiative forcing are mostly 260 negative (i.e., melting) in the Kara, Laptev, and East Siberian Seas during the three spring months of 2020, with a particularly https://doi.org/10.5194/tc-2021-159 Preprint. Discussion started: 26 July 2021 c Author(s) 2021. CC BY 4.0 License. large value (-1.2m) in the Kara sea (Fig.7a). The region with significant SIT reduction agrees well with that with distinct SIC anomalies.
To reiterate the long-term changes, we examine the trend of SIT over the past four decades in the study area. As estimated from the thickness data provided by PIOMAS, the average thickness of spring (April-June) sea ice in the study 265 area has a remarkable decreasing trend of -0.27 m per decade (significant at the 99% confidence level) in the past four decades (Fig. 7b). SIT was persistently lower than 2.50 m since 2000 and dropped sharply to only 1.20 m in the spring of 2020. Thinner ice is more susceptible to changes thermodynamically forcing, thus prone to melt earlier, which in turn could foster a stronger summer ice-albedo feedback through the formation of open water areas. In other words, without the extensive coverage of thin, first-year ice in spring 2020 in the study area (Fig. 7b), the unusual atmospheric energy and moisture 270 transport would not have been nearly as effective in reducing ice extent as was observed (Fig. 1). ice decline experienced in the study area over recent decades, understanding the underlying effects of the cyclones on the poleward transport of energy/moisture is especially crucial. We identify and tracking cyclone systems that occurred in the spring months (April -June) during the period 1979-2020 using the automated algorithm (Serreze et al., 1993;Serreze, 1994;Serreze et al., 1997;Wang et al., 2006;Wang et al., 2013). To cross-check the cyclone systems diagnosed from the ERA5 SLP, we also analyzed data from ERA-interim. Despite some differences in certain regions, a high agreement exists in terms of 285 interannual variability and climatological geographical distribution of cyclone characteristics (not shown). To some extent, this consistency gives credence to the method and datasets utilized.  It is noteworthy that, in this study, we use a range of latitudes (50° N-70° N, with a step length of 1° ), other than a single 310 one, to define the poleward cyclones. For instance, poleward cyclones are defined as those that are generated south of a certain latitude within the range (50° N-70° N) and traveling northward through it. All of these cyclones may play a non-negligible role in carrying energy and water vapor to the Arctic in the form of a relay. As shown in Fig. 9, spring 2020 saw many low-pressure systems moving poleward from Eurasia and some of them entered the study area through the main entry channels in the Kara sea (Fig. 9, green thin lines). Besides, in the Eurasian shelf seas with great convergence of the total energy and 315 water vapor transport ( Fig. 3c and d), the majority of the cyclones are featured with trajectories in a zonal direction (Fig. 9, blue thin lines). Furthermore, we retrieve the typical trajectory paths of these cyclones following Gaffney (2004). The trajectory clustering was done using a polynomial regression mixture model where each cyclone trajectory is approximated as a second-order polynomial. The detected cyclones during the spring months (April-June) of 2020 are clustered in two categories, which are schematically explained in Fig. 9 with thick polylines. One track represents the cyclones that are 320 generated in the lower latitude of Eurasia with a poleward moving tendency, while the other denotes cyclones in the marginal seas that are characterized by an eastward movement toward or through the Kara, Laptev, and East Siberian Seas (Fig. 9). In general, the trajectories of these cyclones as observed during the spring months (April-June) of 2020 coincide well with the path of total energy and water vapor transport ( Fig. 3a and b). The good agreement implies that these extratropical cyclones in spring, as shown in Fig. 9, served as a vital carrier of the anomalously large amount of energy and moisture into the study 325 area. To sum up, the synoptic cyclones act in concert with the large-scale atmospheric circulation to cause anomalous energy and moisture fluxes into the study area and to change the characteristics of the Arctic climate system.  We further investigate the connection between the long-term changes in poleward cyclones and meridional transport. trends are observed in the northward transport of total energy and moisture together with more intense poleward cyclones ( Fig. 10a). As for the spring months of 2020, stronger and more frequent cyclones are detected in the Arctic (Fig. 10b and c).
The density of cyclone tracks is higher than normal in many parts of marginal seas and the central Arctic Ocean, with the largest positive values centered over the Taymir Peninsula. Most cyclones throughout the Arctic Basin have unusually high intensity than the climatology of the years 1979-2020, especially in the Barents, Kara, Laptev, and Beaufort Seas, indicating 345 lower-than-normal SLP in these regions (Fig. 10c). The spatial pattern of CAI anomalies is roughly in line with those of track density and intensity (Fig. 10d). In general, the Eurasian shelf seas had more frequent and stronger cyclones in the spring, especially in the Kara and Laptev 355 seas (Fig. 10). The cyclone variations could alter the spatiotemporal characteristics of the critical near-surface atmospheric parameters (wind stress, temperature, and humidity). As a result, the atmospheric conditions could have a significant impact on sea ice in the study area through control on ice motion, deformation, and melt associated with both dynamic and thermodynamic processes of a cyclone.
From the thermodynamical view, the enhancement of the total energy and water vapor transport in the Eurasia side (Fig. 3a  360 and b) is associated with the regional increases both in the number (Fig. 10b) and intensity (Fig. 10c) of synoptic cyclones that occurred in the main entry channels and some part of the study area. The warm and moist air mass carried by cyclones in spring ( Fig. 9) could alter the surface energy balance, thus initiating the earlier melt onset of sea ice as observed in the study area.
Moreover, the cyclones traversing the Arctic can trigger a spatially extensive sea ice melting with their associated frontal systems (Stramler et al., 2011). The cyclones and the associated frontal systems can also affect the formation of low-level and 365 midlevel clouds over the Arctic Ocean (Curry et al., 1996). All these thermodynamic factors may contribute to the significant SIT decline in Eurasia shelf seas as shown in Fig. 7. Dynamically, the extreme loss of SIE in July 2020 was accompanied by a strong pattern of anomalous cyclonic SIM in spring (Fig.11), with Ekman drift out of the central Arctic toward the marginal seas. On one hand, the cyclonic SIM anomaly in cold seasons serves to enhance the production of new ice within leads https://doi.org/10.5194/tc-2021-159 Preprint. Discussion started: 26 July 2021 c Author(s) 2021. CC BY 4.0 License. because of the increase in sea ice divergence. On the other hand, as the melt season commences, while ice divergence increases 370 extent, it can also accelerate melt by exposing more dark open water areas in the cracks, leads, and polynyas. More frequent and intense cyclones in the Arctic during spring 2020 (Fig. 10) may provide additional cyclonic wind anomalies which are superimposed on that of the large-scale atmospheric circulation as depicted in Fig. 2, promoting the above processes. Based on our results, the thermodynamical other than dynamical effects of cyclones seem to play a dominant role in regulating the changes in SIE during spring 2020, as the expansion of sea ice cover due to divergence was offset by the significant shrinkage 375 due to melt. Figure 11. Anomalies of the spring (April to June) SIM relative to the climatology of the years 1979-2020.

Discussion and Conclusions
An unprecedented reduction in SIE was observed in July 2020 since the satellite era , especially in the Eurasia 380 shelf seas covering the Kara, Laptev, and East Siberian Seas (60° E-165° E, 70° N-82° N). By utilizing global reanalysis datasets and satellite observations, we address the mechanisms of the extreme event. The variations of the total energy and moisture transport toward the study area are obtained and analyzed. We investigate the associated surface energy budget during spring (April to June) of 2020 to disentangle the driving effects of different energy components on sea ice in July.
Moreover, the influences of large-scale atmospheric circulation and synoptic cyclones on the poleward energy and moisture 385 transport are outlined.
Our results reveal that anomalously high advection of energy and water vapor prevailed during spring (April-June) in 2020 over the regions where conspicuous sea ice retreat occurred in the following July. The enhanced energy and moist transport converged into the study area through the main entry channels in the Laptev and Kara Seas from lower latitudes, which https://doi.org/10.5194/tc-2021-159 Preprint. Discussion started: 26 July 2021 c Author(s) 2021. CC BY 4.0 License.