Research Institute for Global Change, Japan Agency for Marine-Earth Science and Technology, Yokosuka, Japan
Department of Earth and Environmental Science, Graduate School of Science and Technology, Hirosaki University, Hirosaki, Japan
Corresponding author: K. Sato, Department of Earth and Environmental Science, Graduate School of Science and Technology, Hirosaki University, 3 Bunkyo-cho, Hirosaki, Aomori 036-8561, Japan. (email@example.com)
 Cloud-base observations over the ice-free Chukchi and Beaufort Seas in autumn were conducted using a shipboard ceilometer and radiosondes during the 1999–2010 cruises of the Japanese R/V Mirai. In comparison with cloud-base heights in an ice-covered case (the Surface Heat Budget of the Arctic Ocean project in 1998), our ice-free results showed a 30% decrease (increase) in the frequency of low clouds with a ceiling below (above) 500 m. Temperature profiles revealed that the boundary layer was well developed over the ice-free ocean in the 2000s, whereas a stable layer dominated during the ice-covered period in 1998. The change in surface boundary conditions likely resulted in the difference in cloud-base height, although it had little impact on air temperatures in the mid- and upper troposphere. Data from the 2010 R/V Mirai cruise were investigated in detail in terms of air-sea temperature difference. Stratus clouds near the sea surface were predominant under a warm advection situation, whereas stratocumulus clouds with a cloud-free layer were significant under a cold advection situation. The threshold temperature difference between sea surface and air temperatures for distinguishing the dominant cloud types was 3 K. Anomalous upward turbulent heat fluxes associated with the sea-ice retreat have likely contributed to warming of the lower troposphere.
 Arctic clouds are key elements of the Arctic climate system. However, the retreat of Arctic sea ice has led to changes in the fraction and vertical structure of Arctic clouds. The recent trend of perennial sea-ice extent has remained strongly negative at −12.2% per decade [Comiso, 2012]. During autumn, the ice-free ocean provides a large amount of turbulent heat flux to the lower troposphere, resulting in the deepening of the boundary layer [Schweiger et al., 2008]. Changes in ice cover and heat fluxes may lead to large changes in the Arctic clouds.
 Satellite observations revealed a increase in cloud frequency below 2-km level [Palm et al., 2010]. Kay and Gettelman also analyzed Arctic clouds during 2006–2008 using combined satellite data and reported that the amount of low-level clouds increased over the ice-free ocean in early autumn. Additionally, surface observations showed a slight increase in total cloud cover during all seasons, particularly in spring and autumn [Eastman and Warren, 2010]. A 10-year record (1998–2008) of the cloud fraction derived from radar-lidar and ceilometer measurements revealed that the fraction increased from March to May and remained relatively high from May to October [Dong et al., 2010]. The vertical distribution of the cloud fraction was found to have a lower peak between 1 and 2 km and a higher peak between 8 and 11 km [Xi et al., 2010].
 Mixed-phase clouds consisting of both ice and supercooled liquid occur over the Arctic region throughout the year, affecting the surface heat budget [Morrison et al., 2011]. Year-round observations of these features were conducted in an ice-covered area during the Surface Heat Budget of the Arctic Ocean (SHEBA) project in 1997–1998 [Uttal et al., 2002]. In SHEBA cloud radar and lidar observations, the maximum averaged cloud occurrence was 97% in September, and the minimum was 63% in February [Intrieri et al., 2002]. The highest frequency of low cloud with base height below 0.5 km occurred during summer and early autumn. Radiative cooling at the cloud top and shear mixing near the surface under strong wind conditions induced by cyclones contributed to long-lived cloud systems over the ice surface [Inoue et al., 2005].
 However, no previous study has fully reported direct observations of cloud-base height over the ice-free Arctic Ocean. Cruises by the Japanese R/V Mirai have been conducted in the Chukchi and Beaufort Seas during September and October since 1999 (1999, 2000, 2002, 2004, 2006, 2008–2010). The Mirai is classified as an ice-strengthened ship and thus is limited to areas without ice cover, as shown inFigure 1. However, the Mirai has frequently entered the Arctic Ocean and has broken her northernmost records in each successive cruise (76.4°N in 2002, 76.6°N in 2004, 78.9°N in 2008, 79.0°N in 2009, and 79.1°N in 2010) due to the recent sea-ice retreat. Up-to-date observation systems on board the Mirai have allowed for investigations of unique meteorological events [Inoue et al., 2010, 2011; Inoue and Hori, 2011]. GPS radiosondes launched by the ship's automatic balloon launcher and the shipboard ceilometer have been used to detect Arctic cloud-base height over the ice-free areas. These observations are suitable for comparison with data obtained previously over an ice-covered area. This study compares these data, reveals the characteristics of cloud-base height over the ice-free ocean, and discusses the impact of sea-ice retreat on the recent change in cloud-base height.
2.1. Overview of Arctic Cruises by the R/V Mirai From 1999 to 2010
 The R/V Mirai is equipped with meteorological instruments such as the ceilometer and automatic balloon launcher for radiosondes, which have been used to conduct observations over ice-free areas since 1999 (http://www.godac.jamstec.go.jp/cruisedata/mirai/e/index.html). During cruises from 1999 to 2010, the ceilometer (Vaisala CT-2K) automatically observed the cloud-base height every minute with 10-m vertical resolution at 905 nm wavelength. Radiosonde observations (Vaisala RS92-SGPD) were regularly performed every 6 hours (00, 06, 12, and 18 UTC). During intensive observational periods (IOPs), 3-hourly observations were made. In the 2010 cruise, we also set three special IOPs along the 162°W line from 71°N to 75°N at 0.5° latitudinal intervals every 2 weeks, i.e., on 14 (IOP-1) and 28 (IOP-2) September and 11 October 2010 (IOP-3). In this study, we used the data obtained north of 70°N. The near-surface air temperature (Ts) at 20 m above sea level and sea-surface temperature (SST) were observed every minute.
Table 1lists the cruise periods in the Arctic and the number of radiosonde observations. For example, in 1999 and 2006, the cruise periods were about 1 week. Even during longer cruises (e.g., 2000 and 2004), the number of radiosonde stations was very limited, and the number of radiosonde data was insufficient for investigating year-to-year variability of the cloud base. Continuous cloud-base observations were conducted by the ceilometer during all cruises. To determine the mean cloud-base height over the ice-free area, we compared the cloud-base heights detected by the radiosondes and by the ceilometer. To detect the cloud base in radiosonde data, we set the threshold for the cloud layer as the layer reaching 100% relative humidity with respect to ice.Figure 2ashows time series of cloud-base heights derived from the radiosondes (red dots) and ceilometer (black line) observations for each year. The temporal change in cloud-base height by the ceilometer agreed with that by radiosondes, suggesting that the ceilometer is useful for monitoring Arctic clouds, even in cases when radiosonde data are unavailable.Figure 3ashows frequency distribution of the cloud-base heights detected by the ceilometer (black bars) and radiosondes (gray bars). The difference is less than 5% in the lowest cloud-base category. With regard to upper clouds, the difference becomes large partly due to the time lag resulted from two observational methods (i.e., the ship with the ceilometer moves when a radiosonde is ascending).
Table 1. Cruises of the R/V Mirai: Year of Cruise, Period, Days North of 70°N, and Number of Radiosonde Stations (Nsonde)
14 Sep.–23 Sep.
07 Sep.–29 Sep.
05 Sep.–06 Oct.
04 Sep.–07 Oct.
31 Aug.–05 Sep.
29 Aug.–05 Oct.
11 Sep.–11 Oct.
05 Sep.–12 Oct.
Figure 2bshows the time series of Ts and SST for each year. In most of periods, Ts was lower than the SST. Although the temperature difference (SST-Ts; hereafter, ΔT) was within 2 K, it sometimes exceeded 5 K. ΔT in September tends to be smaller than that in October. Considering the lower (higher) cloud-base height in September (October), as shownFigure 2a, ΔT likely affects the cloud characteristics.
2.2. Data Over Ice-Covered Area Obtained by SHEBA and “North Pole” Drifting Stations
 SHEBA observations were conducted from a drifting station in Arctic Ocean pack ice from 1997 to 1998 [Uttal et al., 2002], providing valuable year-long data over an ice-covered area. During SHEBA, a Canadian Coast Guard Ship was frozen in sea ice in the Chukchi and Beaufort Sea area (Figure 1), and measurements were made using various instruments. Cloud-base height data were obtained by a depolarization and backscatter unattended lidar (DABUL) at a green wave length (523 nm) [Intrieri et al., 2002]. Sounding data from a GPS/Loran Atmospheric Sounding System (GLASS) were also obtained.
 The Soviet Union began deploying its ‘North Pole’ (NP) drifting ice stations in the Arctic Ocean in 1937. Radiosonde observations were performed at 21 drifting stations during the period 1950–1990, and more than 20,000 soundings were made in total [Kahl et al., 1999]. The Molchanov-type RS-049 Soviet radiosonde was used until the mid-1960s and was then replaced with the A-22 model. For present study, to investigate recent changes in air temperature over previously ice-covered areas, we selected three stations where ice drifting NP measurements had been made in the 1980s (1983, 1985, and 1989). Their positions were close to the cruise areas of the R/V Mirai in the 2000s and to SHEBA in 1998. We manually excluded the profiles which were not reliable to use in the analysis due to the vertical discontinuity and/or abnormal values. In addition, we did not calculate the cloud-base height using the data because the relative humidity data often included unreasonable values.
 As ancillary data, monthly mean air temperature (1979–2011) in the ERA-Interim reanalysis [Dee et al., 2011] was used to validate the NP data and calculate the temperature trend.
3. Difference in Cloud-Base Height and Temperature Profiles Between Ice-Free and Ice-Covered Situations
 To understand the impact of ice cover on the cloud-base height, the frequency distributions of the cloud-base height as a function of altitude were calculated for the ice-free (8-year mean of R/V Mirai data) and ice-covered (September 1998 during SHEBA) cases (Figure 3).
 The frequency of the cloud-base height (below 0.5 km) was 60% in the ice-free cases (black bars inFigure 3a: R/V Mirai) and 90% in the ice-covered case (black bars inFigure 3b: SHEBA). This tendency was a common feature for the cloud-base height calculated by radiosondes (gray bars inFigure 3). The lower frequency over the ice-free ocean means that boundary-layer clouds were well developed as a result of heat and moisture supplies from the ocean. Well-mixed layers were clearly seen in radiosonde observations (Figure 3a, right). The frequency of cloud-base heights above the 0.5-km level was 20% higher in the ice-free than the ice-covered situation. This agrees with the subsequent decrease in static stability and deepening of the atmospheric boundary layer, which contribute to the rise in cloud-base height [Schweiger et al., 2008]. On the other hand, in the ice-covered case, a surface stable layer (Figure 3b, right) was likely formed by warm advection [e.g., Inoue et al., 2005].
 To illustrate the difference in temperature profiles up to the lower stratosphere, Figure 3cshows averaged profiles for the NP stations (the 1980s), SHEBA (1998), and R/V Mirai (the 2000s). We used the data obtained north of 75°N during each September. For reference, the area-averaged temperature profiles using the ERA-Interim (75–82.5°N and 150–170°W) were also shown by dashed lines. As noted above, clear warming of the lower troposphere has occurred in the 2000s due to the recent sea-ice retreat. Profiles above mid-troposphere are almost the same for 1998 and the 2000s. In contrast, the profiles for 1980s are 4 K colder than those in 1998 and the 2000s, even in the mid- and upper troposphere, suggesting that polar amplification had already begun in the 1990s prior to the rapid sea-ice retreat in the 2000s. The temperature trend over the period 1979–2011 calculated by the ERA-Interim supports our observational results (Figure 3d). Changes in atmospheric heat transport may play an important role in these features [e.g., Yang et al., 2010]. Increases in tropopause height and decreases in minimum temperature from the 1980s to the 2000s are also evident both in observations and temperature trend (Figures 3c and 3d). Therefore, the effect of sea-ice retreat on the warming trend during autumn is likely limited to the lower troposphere.
4. Discussion and Conclusion
 Comparison of our observations with SHEBA observations suggest that the change in cloud-base height resulted from the difference in sea-ice cover. The temperature difference between SST and Ts (= ΔT) is an important thermodynamic parameter for understanding stability in the lower troposphere. Here, we present results from the IOP sections along 162°W during the 2010 cruise under different meteorological situations. As shown inFigure 2, IOP-1 (14 September) was characterized by low cloud-base height (about 130 m) with small ΔT (nearly zero) and high cloud cover (completely overcast). This is a typical feature of a shallow stable boundary layer during warm advection over the cold ocean. During this period, warm southeasterly winds prevailed. On the other hand, during IOP-2 (28 September) and −3 (11 October), convective clouds with higher cloud bases above 500 m and a cloud lower fraction around 60% dominated (visual observations). Cold northwesterly winds from the sea-ice area kept ΔT large, exceeding 3 K. This situation was highly favorable for the development of the atmospheric boundary layer associated with large heat release from the ocean to the atmosphere.
 To further investigate the relationship between cloud-base heights and ΔT, we made a scatter plot of ΔT versus cloud base-height obtained by the ceilometer (crosses) and radiosondes (dots) during the 2010 cruise (Figure 4). There are two typical distributions: one is lower cloud base near the surface, with low ΔT from −1 to 3 K, and the other is higher cloud base above 300 m. The cloud-base heights during IOP-1 are classified in the former group, whereas those in the other IOPs are classified in the latter group. This feature can be seen not only for the IOPs (colored dots and crosses) but also for the whole period (black dots and crosses). This result suggests that change in the characteristic cloud-base height depends strongly on surface boundary conditions and resultant surface turbulent heat fluxes. Although our results are based only on data obtained over the ice-free ocean, the findings related to a thermodynamic response at the lower troposphere can be extended to the impact of the difference in sea-ice cover. In fact, the SHEBA ΔT value from September 1998 was about 0.6 K, suggesting a situation very similar to that of IOP-1. Therefore, the ice-covered condition is more favorable for generating low stratus clouds associated with a stable boundary layer [Inoue et al., 2005], whereas the ice-free condition tends to generate stratocumulus clouds with a well-mixed layer [Inoue and Hori, 2011]. A change in cloud types from stratus to stratocumulus causes a decrease in the cloud fraction [e.g., Cuzzone and Vavrus, 2011]. As a result, the enhancement of sea-surface cooling due to the decrease in downward longwave radiation may contribute to rapid sea-ice recovery during autumn.
 Our continuous ceilometer observations showed a 30% decrease in the frequency of low cloud-base events over the ice-free area as compared with the SHEBA period. In contrast, the frequency of cloud-base height exceeding the 500-m level has increased by more than 30%. Although the SHEBA data is very limited to compare with Mirai data due to the amount of data, same features have been found in the literature. For example,Palm et al. reported a decrease in low clouds below the 500-m level and an increase in higher clouds between 800 and 1800 m heights over the ice-free ocean.Kay and Gettelman also found that low-level cloudiness (∼2 km) increased over open water with low static stability near the surface and larger turbulent heat transport. Although other factors (e.g., wind speeds and direction) might also contribute to changes in cloud-base height, our results show the precise thermodynamic environments for different cloud-base heights based on in situ observations.
 We do not yet have sufficient data to discuss the year-to-year variability. Similar observations of cloud-base height using ceilometers are recommended. Compared with radiosondes, cloud radars, and other instruments, ceilometers offer a simpler way to observe cloud-base height and can be installed on research vessels and ice breakers. The obtained data will help clarify near-future changes in Arctic clouds.
 We are greatly indebted to S. Okumura, S. Sueyoshi, N. Nagahama, A. Doi, and W. Tokunaga for conducting the radiosonde observations. We also thank the crews of the R/V Mirai, and are grateful for the data collected during the SHEBA project and by the NP drifting stations. We also thank M. E. Hori and two anonymous reviewers for their useful comments. This work was partly supported by KAKENHI(A)24241009. PMEL contribution #3857.
 The Editor thanks the two anonymous reviewers for assisting with the evaluation of this paper.