SEARCH

SEARCH BY CITATION

Keywords:

  • Arctic;
  • melt ponds;
  • melting

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Observations
  5. 3. Ice and Ocean Conditions
  6. 4. Relationship Between Ice Concentration and Water Temperature
  7. 5. Role of Heat Transmission Through Ponds and Ice on the Ice-Upper Ocean System
  8. 6. Summary and Discussion
  9. Acknowledgments
  10. References

[1] To observe sea ice and ocean conditions in the Arctic in summer, a trans-Arctic research cruise of the U.S. Coast Guard Cutter Healy was conducted from 5 August to 30 September 2005. The relationship between the ice concentration observed by the on-board ice-watch and the temperature above the freezing point (ΔT) measured by expendable conductivity-temperature-depth (XCTD) sensors had a negative correlation (CT-relationship) before the onset of freezing. This means that as ice concentration decreases, ΔT increases due to the larger absorption of solar radiation. However, ΔT in high ice-covered regions (>90%) remains more than 0.1 K during the melting season, suggesting that sea-ice and melt-pond areas work as heat source areas as well as leads. By separating the effects of heat input from open water, melt ponds, and ice on the heating of mixed layers, we found that the contribution of the transmitted heat through ponds and ice on the ΔT-gain is large in highly ice-covered regions. To examine the effect of such heating on ice melting, a simplified ice-ocean-coupled model was applied. By changing the heat input to obtain the analyzed ΔT-gain for each surface category, the transmittances of ponds and ice were indirectly estimated as 55% and 9%, respectively. After including the effects of transmitted heat through ponds and ice, the modeled results agreed with the observed CT-relationship. Comparisons between the results of turning on and off the effect of transmitted heat through ponds and ice showed that it amplified the open water-albedo feedback mechanism in the highly ice-covered region.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Observations
  5. 3. Ice and Ocean Conditions
  6. 4. Relationship Between Ice Concentration and Water Temperature
  7. 5. Role of Heat Transmission Through Ponds and Ice on the Ice-Upper Ocean System
  8. 6. Summary and Discussion
  9. Acknowledgments
  10. References

[2] The interaction between sea ice and heat input from the atmosphere is one of the most important processes contributing to declining ice cover during summertime in the Arctic [e.g., Curry et al., 1995]. Because shortwave (SW) radiation is a main source of heat input in the Arctic, it is important to determine how SW radiation is distributed within the atmosphere-ice-ocean system and how this distribution affects heat and mass exchanges within the system. Because of the importance of albedo to the surface heat budget of sea ice, a considerable number of studies for a wide range of ice types and conditions have demonstrated that the variability of the albedos is linked to a strong dependence on surface conditions and the structure of the ice [e.g., Grenfell and Maykut, 1977; Grenfell and Perovich, 1984, 2004; Hanesiak et al., 2001].

[3] The SW radiation that penetrates into open water can become the dominant heat source for ocean surface warming and consequent lateral and bottom melting of sea ice [Maykut and Perovich, 1987; Maykut and McPhee, 1995]. Using a drifting buoy, McPhee et al. [2003] showed that heat flux below the Arctic sea ice is dominated by the storage and release of SW radiation energy in the ocean boundary layer during summer. The effect of wind speed on lateral and bottom melting has also been investigated [e.g., Richter-Menge et al., 2001; Inoue and Kikuchi, 2006]. While much of the solar energy is transmitted to the ocean through leads, substantial portions are also transmitted through bare ice and ponded ice. Perovich [2005] showed that the energy transmitted to the ocean reached a maximum near the end of the melt season in mid-August when the albedo and ice thickness were at their minima and lead and pond fractions were at their maxima. He also showed that observed pond transmittance jumped from near 0 to 0.2 after the onset of melting and monotonically increased up to 0.4–0.5 during the summer as the ice beneath the pond thinned. In his observations, a sharp decrease in pond transmittance was also observed after the onset of freezing and snowfall. Bare ice transmittances, on the other hand, were smaller than pond transmittances (∼0.2).

[4] Although satellite passive microwave sensors are very useful for revealing Arctic sea ice changes, their resolutions are too coarse to distinguish open water from ponded ice. Hence total ice concentrations determined by these devices are not as accurate during the summer melt season as in winter [Cavalieri et al., 1984; Fetterer and Untersteiner, 1998]. While satellite visible-band sensors have higher resolution, the persistent cloud cover over the ice pack during the summer [e.g., Intrieri et al., 2002; Inoue et al., 2005a] severely limits the utility of these sensors to observe melt ponds. To understand the relationship between recent ice decay and the ice-upper ocean coupled system, observations of ice and ocean conditions in various areas of the Arctic must be obtained.

[5] In the Antarctic, ice concentration is negatively correlated with temperature and positively correlated with salinity for spatially averaged data, suggesting that the local balances of heat and salt nearly hold in a bulk area [Ohshima et al., 1998]. This relationship has been investigated and explained by a simplified coupled model of the ice and upper ocean [Ohshima and Nihashi, 2005; Nihashi et al., 2005]. In contrast, very few observational studies have examined the ice-upper ocean-coupled system in the Arctic, although a simple ice-upper ocean-coupled model has been developed [e.g., Steele, 1992]. Using one-dimensional sea ice model, Ebert et al. [1995] found that melt ponds increase the penetration of SW radiation into the ice interior by 50%, depending on the fractional coverage and depth of ponds, and increase the transmission of incoming SW radiation into the upper ocean relative to unponded ice. As for climate modeling, sea-ice processes are influenced by differences in the albedo parameterizations [Liu et al., 2006]. Therefore an explicit treatment of melt ponds in the models is needed for a correct simulation of the sea-ice albedo feedback [Curry et al., 1995, 2001]. However, there has not been an adequate ocean data set to evaluate the effects of surface conditions on sea-ice melting.

[6] To understand the interaction of ice with ponds in the upper ocean system, observations from an ice-breaker are essential to simultaneously monitor the characteristics of the ice surface and upper ocean. From the beginning of August 2005 until the end of September 2005, we observed sea-ice cover and upper ocean conditions across the Arctic Ocean from the USCGC Healy (Figure 1). Here we describe sea-ice melting processes based on various sea-ice and upper ocean data collected in the summertime Arctic and compare the data to the results of a simple ice-upper ocean-coupled model.

image

Figure 1. Cruise track of USCGC Healy and the XCTD stations. Light and dark shadings denote mean ice concentrations during August 2005 greater than 30% and 90%, respectively. The area enclosed by the dashed line is shown again in Figure 3.

Download figure to PowerPoint

2. Observations

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Observations
  5. 3. Ice and Ocean Conditions
  6. 4. Relationship Between Ice Concentration and Water Temperature
  7. 5. Role of Heat Transmission Through Ponds and Ice on the Ice-Upper Ocean System
  8. 6. Summary and Discussion
  9. Acknowledgments
  10. References

[7] The USCGC Healy trans-Arctic cruise in 2005 presented a unique opportunity to obtain sectional data across the Arctic Ocean [Darby et al., 2005]. The Healy departed Dutch Harbor, Alaska, on 5 August and tracked across the western and northwestern perimeter of the Canadian Basin. The ship passed across the North Pole on 11–12 September and arrived at Tromso, Norway, on 30 September. This track enabled us to observe oceanic and sea-ice conditions across the Arctic Ocean, and to record basic meteorological conditions (e.g., air temperature, humidity, and wind).

[8] Expendable conductivity-temperature-depth (XCTD) sensors and processing equipment (Tsurumi-Seiki Co., Ltd., Yokohama, Japan) measured temperature and conductivity (i.e., salinity) from the sea surface to a depth of 1100 m. An XCTD probe launched from the ship into the water sank at a constant rate while measuring temperature and conductivity. We launched XCTD sensors at 71 observational sites across the Arctic Ocean (see Figure 1). The first 38 observations were conducted on the western and northwestern perimeter of the Canadian Basin. These sites were located over or across the major ridges around the Canadian Basin. The final 33 stations were located across the central and eastern Arctic Ocean.

[9] Bi-hourly visual observations of sea-ice cover were made according to the method of Worby and Allison [1999] from the bridge while the ship was moving through the pack ice (Figure 2). For each visual observation, the sea-ice cover was divided into three thickness categories for which we estimated the ice concentration, ice type, and mean snow depth on unridged ice. In this study, we used data for the fractions of open water (Ao), ponds (Ap), ice (Ai), and ice thickness, as well as daily ice concentration data derived from the Special Sensor Microwave Imager (SSM/I) on the Defense Meteorological Satellite Program (DMSP) F13 at a grid resolution of 25 km.

image

Figure 2. Panoramic photograph of the ice cover taken from the flying bridge of the Healy on 14 August 2005 at 78.193°N, 153.548°W.

Download figure to PowerPoint

3. Ice and Ocean Conditions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Observations
  5. 3. Ice and Ocean Conditions
  6. 4. Relationship Between Ice Concentration and Water Temperature
  7. 5. Role of Heat Transmission Through Ponds and Ice on the Ice-Upper Ocean System
  8. 6. Summary and Discussion
  9. Acknowledgments
  10. References

[10] Figure 3 shows the time-latitude cross-section of the ice concentration averaged between 150°W and 180°W, corresponding to the cruise to the North Pole. Near the Alaskan coast (72°N), the ice concentration started to decrease in the middle of July, and the highly ice-covered area (i.e., 90%) retreated rapidly northward until mid-August. The cruise of USCGC Healy started when sea-ice melting was predominant. We moved northward up to Northwind Ridge (78.5°N) under moderate ice concentrations (70%). After traveling westward along the same latitude (Figure 1), the ship turned northward again and finally entered the highly ice-covered area (over 90%) after 23 August. During this period, the incoming SW radiation, which can be approximated as one-third the value of SW radiation at the top of the atmosphere [Inoue et al., 2005b], decreased from 130 to 90 W m−2. However, the latitudinal gradient was very weak, suggesting that the heat input into the upper ocean was almost the same during the summer.

image

Figure 3. Time-latitude cross-section of the ice concentration (%) averaged between 150°W and 180°W, derived from the DMSP SSM/I. The thick solid and dashed lines denote the cruise track of the Healy and one-third the value of the daily mean shortwave radiation (W m−2) at the top of the atmosphere, respectively.

Download figure to PowerPoint

[11] Figure 4 shows some of the vertical XCTD profiles. The water structure was not uniform across the Arctic Ocean. The Canadian Basin (e.g., St. 5 at 77.2°N, 153.9°W) had a complex temperature profile as noted by Shimada et al. [2001]. In almost all areas, however, temperature and salinity were approximately uniform from near the surface to a depth of at least 10 m. That is, the summer surface mixed layer in the Arctic Ocean had a thickness of more than 10 m. From these vertical profiles, we could assume that the temperature and salinity between 5 and 10 m approximately represented the values averaged over the surface mixed layer as a whole because the layer is commonly affected by sea-ice melting for all profiles.

image

Figure 4. Vertical profiles of the water temperature (thin line) and salinity (thick line) obtained by the XCTD sensors. Corresponding axes of temperature and salinity are indicated at the top and bottom, respectively, of each panel. Locations of the stations are indicated in Figure 1.

Download figure to PowerPoint

[12] Figure 5 shows the time series of the observed atmosphere, ice, and ocean characteristics along the cruise tracks. The sea-ice condition was complicated because of the melt ponds. The ponded area covered approximately 25% of the sea-ice area during August (Figure 5c) but then decreased due to freezing as the air temperature dropped below zero (Figure 5b). Although two major algorithms provided the basic temporal/spatial evolution of ice concentration near the ship, the high ponded fraction made satellite-derived ice cover estimations unreliable. The ice concentration derived from the Bootstrap Algorithm [Comiso, 1990] appears to include the ponded area (open squares), which is the same feature reported by Inoue et al. [2008], while that from the NASA Team Algorithm [Cavalieri et al., 1990] seems to be partly affected by melt ponds (closed circles). In general, ice thickness increased from 1 to 2 m as the ship moved northward (Figure 5d). The proportion of FYI and MYI observed by ice watch during August was 59% and 41%, respectively. MYI dominated after 18 August (>78°N).

image

Figure 5. Time series of (a) the latitude cruised by the Healy, (b) air temperature obtained on the ship, (c) ice concentration derived from SSM/I data at the closest grid to the ship (open square, Bootstrap Algorithm; closed circle, NASA Team Algorithm), the ice-watch data (dark gray, open water; light gray, melt pond; white, sea ice), (d) ice thickness obtained by the ice-watch, and (e) temperature above the freezing point between 5 and 10 m derived from the XCTD data. The air temperature and ice-watch data were smoothed using a 1-day running mean.

Download figure to PowerPoint

[13] To determine how much the upper ocean was heated by the atmosphere (i.e., incoming solar radiation), the temperature above the freezing point (ΔT) is a useful parameter, particularly in areas where salinity is not horizontally uniform (see Figure 4). ΔT, averaged from 5 to 10 m, was relatively high in the southern region below 80°N (Figure 5e) where the ice concentration was relatively low (e.g., 50–60%). This suggests that incoming solar radiation was absorbed in the open water and some of this radiation was used in the bottom and lateral melting of sea-ice, while the remainder increased the temperature.

4. Relationship Between Ice Concentration and Water Temperature

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Observations
  5. 3. Ice and Ocean Conditions
  6. 4. Relationship Between Ice Concentration and Water Temperature
  7. 5. Role of Heat Transmission Through Ponds and Ice on the Ice-Upper Ocean System
  8. 6. Summary and Discussion
  9. Acknowledgments
  10. References

4.1. CT-Plot

[14] We examined the relationship between the ice concentration (ponds and ice fractions: Ap + Ai) observed by the on-board ice-watch and the upper ocean XCTD data, which is useful for understanding the heat balance of atmospheric heat inputs, sea-ice melting, and heat storage in the ocean mixed layer (hereafter termed the local heat balance). Figure 6 shows the ice concentration plotted against ΔT (hereafter called the CT-relationship) for the XCTD casts before the onset of freezing, i.e., during August (Sts. 1 and 38). The visually observed ice concentrations were averaged to daily values to reduce any observer bias. ΔT increased as the ice concentration decreased, which is consistent with results obtained from an Antarctic cruise off Syowa Station [Ohshima et al., 1998] and in the Ross Sea [Nihashi et al., 2005]. Their results have shown that the relationship between ice concentration and the mixed-layer temperature converges with time and that the local heat balance in the ice-upper ocean system approximately holds over that temporal/spatial scale [Ohshima and Nihashi, 2005]. In such a case, a process in which ice concentration decreases as ΔT in the upper ocean increases is continued if there is a large absorption of solar radiation (a kind of ice-albedo feedback).

image

Figure 6. Scatterplots of the ice concentration (Ap + Ai) versus temperature above the freezing point (ΔT). Daily mean ice concentration was derived from the on-board ice-watch, and ΔT data were calculated from XCTD observations. The dashed line is a regression based on the observations. R and RMSE indicate the correlation coefficient and root mean square error, respectively. The solid line is the same line subtracted from the Δ T-bias (=0.09 K) from the regression line weighted by ice concentration. The shaded area denotes the Δ Tp + Δ Ti.

Download figure to PowerPoint

[15] Interestingly, the CT-relationship in our case shows a positive bias (∼0.1 K) where the ice concentration is almost 100% denoted by a regression line in Figure 6 (dashed line). This bias is significantly larger than errors resulting from the accuracy in salinity measurements (±0.03 psu causes ±0.02 K in ΔT). These results suggest that the upper ocean gains heat from other sources in addition to the open water area. Solar radiation transmitted through melt pond and sea-ice areas may also affect the ice-upper ocean system.

[16] Because the ice concentration is defined as the sum of pond and ice fractions (Ap + Ai), the increase of ΔT (hereafter termed ΔT-gain) by the open water area [Ao = 1 − (Ap + Ai)] can be explicitly separated from the total ΔT-gain which is the sum of contributions of open water (ΔTo), melt ponds (ΔTp) and ice (ΔTi). By subtracting the bias resulting from ΔTp + ΔTi(= 0.09 K when Ap + Ai = 100%) from the dashed line in Figure 6 and weighting by Ap + Ai, an adjusted regression line for ΔTo [= ΔT-gain −0.09 × (Ap + Ai)] is obtained (the solid line in Figure 6). The area enclosed between these two lines indicates the ΔTp + ΔTi induced by Ap + Ai. This contribution increases as Ap + Ai becomes large.

4.2. Partitioning of Heat From Open Water, Ponds, and Ice

[17] The benefit of ice-watch data in this research is that Ap is also independently observed. Therefore we also obtained the contribution of ΔTi to the ΔT-gain explicitly by removing the effect of ΔTo + ΔTp in the same manner as in Figure 6. Figure 7 shows the relationship between Ai and the ΔT-gain. There is a negative correlation (dashed line), and a bias from ΔT still exists as 0.06 K at the completely ice-covered area without ponds (Ai = 100%). The ΔTi induced by Ai is also obtained as the shaded area the same way as in Figure 6.

image

Figure 7. As in Figure 6, but showing Ai versus ΔT. The shaded area denotes ΔTi. The dotted line is the same as the dashed line in Figure 6, but the range of the lower abscissa is expanded.

Download figure to PowerPoint

[18] Here we consider the heat partitioning of the ΔT-gain among Ao, Ap, and Ai using the CT-plots. As a specific case, for example, to gain 0.15 K of ΔT when 80% of the area is covered by ice with ponds (Ap + Ai: Figure 6), the ΔT-gain is partitioned into ΔTo = 0.08 K and ΔTp + ΔTi = 0.07 K, respectively. The ΔT-gain can be also partitioned as the following combination when Ai equals 53% (Figure 7): ΔTo + ΔTp = 0.12 K and ΔTi = 0.03 K. Eventually, the ΔT-gain (0.15 K) consists of ΔTo = 0.08 K from Ao(=20%), ΔTp = 0.04 K from Ap(=27%), and ΔTi = 0.03 K from Ai(=53%). In the same way, the heat partitioning under different combinations of surface fractions among Ao, Ap, and Ai is obtained as in Figure 8. Although the contribution of Ao is the main source of heat over most of the area, the roles of Ap and Ai are also relatively large beneath the highly ice-covered area. For example, half of the ΔT-gain comes from ponded and ice-covered areas when Ap + Ai equals 83%. Basically, ΔTp seems to be nearly constant (∼0.05 K) at any Ap + Ai due to less change in Ap as shown in Figure 5c. Considering that Ap is significantly smaller than Ai, the heat transmission through ponded areas may be larger than in ice-covered areas, and smaller than in open water areas because of the difference in albedos between melt ponds and bare ice [e.g., Grenfell and Maykut, 1977; Grenfell and Perovich, 2004].

image

Figure 8. ΔT-gain from open water, melt ponds, and ice as a function of Ai + Ap. Closed circles are model results by RUNO using Fn = 110 W m−2 after 15 days of the time integration.

Download figure to PowerPoint

5. Role of Heat Transmission Through Ponds and Ice on the Ice-Upper Ocean System

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Observations
  5. 3. Ice and Ocean Conditions
  6. 4. Relationship Between Ice Concentration and Water Temperature
  7. 5. Role of Heat Transmission Through Ponds and Ice on the Ice-Upper Ocean System
  8. 6. Summary and Discussion
  9. Acknowledgments
  10. References

[19] In this section, we developed a simple model to understand the air-ice-sea-coupled system in the summertime Arctic by focusing on the role of transmitted heat through ponds and ice on the melting processes.

5.1. Model Description

[20] Because heat input into the ice-upper ocean system mainly occurs in open water when ice concentration is relatively low as we showed in the previous section, we began by examining the observed CT-plot using a simple ice-upper ocean-coupled model in which sea-ice bottom and lateral melting is caused only by heat input through open water. Here, we briefly describe the model that was proposed by Ohshima and Nihashi [2005]. The upper ocean is simply represented by a layer of thickness H with a uniform temperature T and salinity S. Heat and water exchanges with the ocean below this layer are assumed to be zero due to the strong stratification during the Arctic summer. We implicitly assumed that the open water is well mixed with the water just beneath the ice. The net heat budget on the ice is assumed to be zero; thus, the surface net heat flux would only be supplied in the open water area, 1-Ai. If sea-ice melting is caused by this heat input, the heat balance of the upper ocean can be given by

  • equation image

where Fn is the net heat input at the surface, cw(=3990 J Kg−1 K−1) is the heat capacity of seawater, ρw(=1026 kg m−3) and ρi(=900 kg m−3) are the densities of seawater and sea ice, respectively; t is time, and Lf is the latent heat of fusion for sea ice. The salinity of sea ice in the Arctic is assumed to be 2 psu based on observations in the Arctic multiyear ice [Eicken et al., 1995]. We use a fixed value of Lf = 0.316 MJ kg−1. The ice thickness h is defined as the average thickness; αo(=0.06) is the surface albedo of open water.

[21] Sea ice melts at the bottom and lateral faces at a rate proportional to the difference between the water temperature and the freezing point (i.e., ΔT). Bottom melting is parameterized as

  • equation image

where ch(= 0.0057) and u*(= 0.0056 m s−1) are the heat transfer coefficient and interface friction velocity obtained from the drifting buoy program of the North Pole Environmental Observatory (NPEO) [McPhee et al., 2003], respectively. Lateral melting is parameterized in a way similar to that of Hibler [1979], as

  • equation image

In this equation, h* is defined as an effective ice thickness averaged over each of the grid cells, including the open water fraction. Thus h* can be regarded as the total ice volume per unit area. Note that the actual averaged thickness is h = h*/Ai in this model.

5.2. Outline of the Baseline Simulation

[22] Using the model, we discuss the relationships between the ice fraction (Ai), ice thickness (h), and temperature above the freezing point (ΔT). Here, all parameters are set to values suitable for the observational region. We set H at 15 m on the basis of the temperature and salinity profiles from Figure 4. The net heat input at the surface Fn is set to a constant value.

[23] Initial ΔT and h0 equaled 0.0 K (i.e., the freezing point) and 1.5 m, respectively. However, several calculations were done using different initial ice concentration (Ai0) and heat input (Fn) by changing from 30% to 99%, and from 10 W m−2 to 150 W m−2, respectively. Hereafter, we call this experiment set RUNO, in which we only included the heat input from open water. To compare the modeled results with our observation, the output from RUNO using Fn = 110 W m−2 (baseline case) is suitable because the averaged incoming SW radiation during August is nearly 110 W m−2 around the region [Persson et al., 2002; Inoue et al., 2005b; Figure 3].

5.3. Results Without the Effects of Heat Transmission Through Ponds and Ice

[24] CT-relationships converge asymptotically to a single curve with a timescale of ∼10 days regardless of initial conditions of Ai and ΔT [Ohshima et al., 1998; Ohshima and Nihashi, 2005]. Closed circles in Figure 8 show the CT-plot after 15 days of the integration of RUNO. In this calculation, ΔT is zero under the completely ice-covered situation due to the lack of heat input through open water. In other words, the modeled ΔT-gain is comparable to the analyzed ΔTo. Both lines are closely matched, suggesting that the model can appropriately describe the heat transfer through open water.

[25] To examine the sensitivity of Fn, the results with 110 ± 10 W m−2 are plotted in Figure 9 (note the lower abscissa in reverse). The differences of ΔT from the baseline case is ±0.01 K under a 30% fraction of open water (observed Ao ranges from 0 to 40%). Therefore the sensitivity of Fn seems to be small in the highly ice-covered region.

image

Figure 9. ΔT-gain induced by open water, melt ponds, and ice. These relationships are obtained by the regression lines in Figure 8. The lower abscissa shows each fraction for Ao, Ap, and Ai, respectively. Dashed lines denote model results (RUNO) induced by different heat inputs Fn.

Download figure to PowerPoint

[26] Since we already know the relationship between Ap and ΔTp, and between Ai and ΔTi (Figure 8), we can indirectly estimate how much heat input is required to obtain the ΔT from ponded and ice-covered areas by using the results of RUNO (Figure 9). The heat input corresponding to the relationship between Ap (from 20 to 25% in the observation) and ΔT-gain ranges between 40 and 80 W m−2, while that for Ai ranges around 10 W m−2. These values for Ap and Ai are considered to be the transmitted heat through ponds and ice that should be basically much smaller than Fn(=110 W m−2), because most of the incoming SW radiation is reflected at the pond and ice surfaces.

[27] Here, we calculated the transmittance of ponds (τp) and ice (τi) using these values. To gain the same amount of ΔT only through ponds and ice, τpFn(=60 ± 20 W m−2) and τiFn(=10 W m−2) are the respective heat inputs into the ocean (Figure 9). Thus the heat transmittances of ponds (τp= 60 ± 20/110 = 0.55 ± 0.18) and ice (τi = 10/110 = 0.09) are obtained. These two values suggest that sea-ice transmits little SW radiation, while melt ponds have an intermediate characteristic between open water and sea-ice. Therefore even if the Ap is smaller than Ai as we observed, the pond contribution to the ΔT-gain is almost the same (Figure 8).

5.4. Effects of Heat Transmission Through Ponds and Ice on Melting

[28] Using τp and τi, the effect of melt ponds and ice on the melting processes can be examined by including theadditional terms related to these parameters into equation (1) as,

  • equation image

where the transmittance of open water (τo = 1 − αo = 0.94), ponds (τp = 0.55), and ice (τi = 0.09) is a fixed value, respectively. In this model framework, Ap is also given as a function of Ai obtained from Figure 7 (Ap ranges from 20% to 25% at any Ai). In addition, the depth of melt ponds is not treated in this model. Other initial conditions are the same as in RUNO. Hereafter, this experiment set is referred to as RUNALL and includes the heat input from all surface categories.

[29] Figure 10 shows the convergence curves of RUNALL (solid line) and RUNO (dashed line) after 15 days of the integration, respectively. Clearly, RUNALL has a bias of about 0.06 K under highly ponded and ice-covered situations (e.g., Ap + Ai = 99%), which almost agrees with the observed CT-relationship (dots in Figure 10). Without the heat transmission from ponds and ice (RUNO), ΔT would not be so large (dashed line).

image

Figure 10. As in Figure 6, but showing model results with heat transmittance by melt ponds and ice (solid line, RUNALL), and without their effects (dashed line, RUNO). Dots show the observational results; doted lines show the results calculated using Fn = 110 ± 10 W m−2.

Download figure to PowerPoint

[30] Because this higher ΔT in RUNALL is a consequence of heat balance in the ice-upper ocean system, the ice concentration and ice thickness should also vary differently between RUNO and RUNALL. Figure 11 shows the changes over time of ice thickness and ice concentration (Ap + Ai) when the initial open water fraction (Ao0) is set to 3% for RUNALL and RUNO. The difference in the decrease in ice thickness between the two runs is less than 0.1 m after 15 days of the integration. Interestingly, the decrease in ice concentration is linked to the decrease in ice thickness, particularly for RUNALL. A possible explanation for this may be that because the total extent of the lateral faces of relatively thin ice is small, less heat is used in lateral melting, thus enhancing the increase in ΔT. As ΔT becomes greater, Ap + Ai and h decrease further. The lower Ap + Ai and thinner h further enhance the increase in ΔT, promoting the open water-albedo feedback [Ackley et al., 2001].

image

Figure 11. Evolution of ice thickness (top) over time and ice concentration (Ap + Ai: bottom). Solid and dashed lines denote the cases where the effects of heat transmittance of melt ponds and ice are turned on (RUNALL) and off (RUNO), respectively. The same runs as RUNALL but H is varied 15 ± 10 m (dotted lines) and a realistic Fn is given (thin solid line).

Download figure to PowerPoint

[31] In the case of RUNO, on the other hand, ice thickness and concentration hardly change due to the lack of heat input from ponds and ice as well as from open water. Therefore the effect of transmitted heat through ponds and ice significantly affects the open water-albedo feedback process, particularly in highly ice-covered regions.

5.5. Effects of Transmitted Heat Through Ponds and Ice on the Open Water-Albedo Feedback

[32] How is the open water-albedo feedback amplified by the transmitted heat through ponds and ice? To assess the strength of this feedback, we additionally compared perturbed simulations that included heat input through ponds and ice areas with a corresponding perturbed simulation that did not include these effects. The strength of the feedback may be quantified by defining the feedback gain ratio Rf [Curry et al., 2001]:

  • equation image

where h is the ice thickness as the climate parameter, Δ denotes the change in h associated with a specified forcing, and the subscripts on and off indicate that the heat input through ponds and ice areas are turned on and off, respectively. A positive value of Rf(h) greater than 1.0 indicates a positive feedback mechanism; Rf(h) varies with the sign and the magnitude of the forcing. Four different model simulations are required to obtain Rf(h):

[33] (1) a baseline simulation using equation (1), representing the current unperturbed conditions without heat transmission of ponds and ice;

[34] (2) a simulation using equation (1) in which the climate is subject to an external perturbation and heat transmission of ponds and ice are turned off;

[35] (3) a simulation using equation (4), representing the current unperturbed conditions with heat transmission of ponds and ice;

[36] (4) a simulation using equation (4) in which the climate is subject to an external perturbation and heat transmission of ponds and ice is operative.

[37] A warming perturbation is given as 4 W m−2, which is comparable to the doubled CO2 forcing [Ramanathan et al., 1989]. The feedback gain ratio is then evaluated from

  • equation image

where the numerical subscripts refer to the above enumerated model simulations.

[38] Figure 12 shows Rf(h) as a function of initial ice concentration (Ap + Ai) after 15 days of the integration. In the case of relatively high initial ice concentration (e.g., ≥95%), the large Rf(h) suggests that bottom melting was significantly amplified (i.e., Δ hon ≫ Δ hoff), because the heat input through ponds and ice overwhelms that from open water. In contrast, in the case of lower ice concentration, the feedback by this effect is small because heat input from open water is the main source of energy, although positive feedback still operates (Rf(h) ≥1). The result of Rf(ice) is also the same as Rf(h) (data not shown), suggesting that the heat transmitted through ponds and ice areas significantly affects the open water-albedo feedback in the highly ice-covered region. Of course, this feedback would be largely affected by the ice thickness distribution. Ebert et al. [1995] showed that the net transmission of solar radiation into the ocean is roughly 30% greater for a distribution of ice thicknesses categorized differently than for a single ice slab (e.g., this study) under the same mean ice thickness. Thus the feedback might be stronger in fact than modeled one.

image

Figure 12. Feedback gain ratio for the effect of transmitted heat through ponds and ice on the bottom melting as a function of the initial ice concentration (Ap + Ai). Results are based on 15 days of the integration with perturbed heat flux.

Download figure to PowerPoint

6. Summary and Discussion

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Observations
  5. 3. Ice and Ocean Conditions
  6. 4. Relationship Between Ice Concentration and Water Temperature
  7. 5. Role of Heat Transmission Through Ponds and Ice on the Ice-Upper Ocean System
  8. 6. Summary and Discussion
  9. Acknowledgments
  10. References

[39] The XCTD observations and ice monitoring during the trans-Arctic cruise in the summer of 2005 revealed that the temperature above freezing (ΔT) in the surface mixed layer is a function of the ice concentration (CT-relation: Figure 6). This is a consequence of heat balance in the ice-upper ocean system, which consists of heat input, ice melting, and heat storage in the mixed layer. In the Antarctic Ocean, this relationship is clear when 20–30-km running mean data were used [Ohshima et al., 1998]. This suggests that the local heat balance in the ice-upper ocean system approximately holds over that spatial scale, while at smaller spatial scales, ice advective effects predominate. In our case, the spatial running mean is not effective because the XCTD stations are sparsely located; instead, daily running mean sea-ice data were used. If we assume the free drift of ice as 2% of wind speed (typically ∼5 m s−1), the ice travels ∼10 km per day. Considering the ship's speed (around 100 km per day), the spatial scale to interact with the upper ocean by ice drifting corresponds to 10 days or more.

[40] In the analyzed CT-relationship, there is a positive bias of ΔT in the highly ice-covered area, suggesting that melt ponds and sea ice allow significant heat input to the ocean by transmitting SW radiation. By partitioning the gain of ΔT into the components of open water, ponds, and ice, we found that the role of ponds and ice areas is significantly large in the highly ice-covered area (Figure 8), which enhances bottom melting without heat input from open water. From these results, the respective transmittance of ponds (0.55) and ice (0.09) was evaluated. These transmittances should vary as surface conditions change depending on the melting stages (e.g., before melt onset, melting period, and freezeup).

[41] Considering that our analysis was based on observations made during August, the transmittances for ponds (0.55) and bare ice (0.09) have reached a maximum, which is consistent with results of Perovich [2005]. In fact, at the beginning of the cruise, we saw very mature ponds with a low albedo, and thin ice at the bottom of the pond (Figure 2). Although the transmittances might vary from place to place depending on the ice thickness and pond depth, our results represent the case of relatively thin ice with a minimum ice thickness of ∼1.5 m during late summer (Figure 5d).

[42] Using a simple ocean-ice-coupled model, we examined the effects of transmitted heat through ponds and ice on the open water-albedo feedback mechanism. A comparison between the modeled results by turning the effects on and off showed that the decrease in ice thickness is enhanced by these effects and amplifies the open water-albedo feedback. Because our model is very simplified, time evolutions of the ponded fraction (Ap), thickness of the mixed layer (H), seasonal changes in heat input (Fn), and horizontal oceanic heat advection may be sources of errors. In general, the ponds begin forming in early June shortly after snowmelt, and a discrete jump in the pond fraction from 5% to 20% occurs during mid-June. Then there is a steady and gradual increase in the pond fraction exceeding 20% [Perovich, 2005]. Therefore our modeled results using the fixed value of Ap is probably applicable under a short time integration (e.g., less than 1 month), particularly by focusing on the ice-upper ocean system during late summer, although the sensitivity of FYI ablation depends on pond fraction and the associated spatial variability in surface albedo [Hanesiak et al., 2001].

[43] As for the thickness of the mixed layer H, we have checked its sensitivity by perturbing H = 15 ± 10 m (dotted lines in Figure 11). Ice melts relatively quickly when H is relatively shallow because ΔT readily increases, and vice versa in the case where H is deeper. However, the differences in ice thickness and ice concentration are very small between the runs. With regard to the seasonal change of Fn, we conducted an additional sensitivity run using one-third of the time-dependent SW radiation at the top of atmosphere, which has been verified by Inoue et al. [2005b]. Although Fn decreases from 140 to 80 W m−2 during August as in Figure 3, changes in ice thickness and ice concentration are very small compared to RUNALL (thin solid line in Figure 11). We also did not consider the contribution of net longwave radiation and turbulent heat flux to Fn because they are near zero during August [Persson et al., 2002]. Therefore the fixed value of Fn given as SW radiation is presumably appropriate unless the time integration is too long (∼2 weeks). Although the role of horizontal oceanic heat advection is not treated in our model, its effect is the secondary one because the difference between the observed ΔT(=0.09 K) and modeled ΔT(=0.06 K) in the completely ice-covered area is smaller than the effect of heat transmission.

[44] In conclusion, in modeling Arctic sea-ice melting, the inclusion of heat input through melt ponds and ice areas is of major importance in highly ice-covered regions, and even a simplified model that neglects other effects can describe the fundamental features of ice melt during a short timescale. CT-relationship in the Arctic Ocean introduced in this study could be a good indicator to verify the heat balance in the upper ocean system simulated by climate models used in the Intergovernmental Panel on Climate Change (IPCC), for example.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Observations
  5. 3. Ice and Ocean Conditions
  6. 4. Relationship Between Ice Concentration and Water Temperature
  7. 5. Role of Heat Transmission Through Ponds and Ice on the Ice-Upper Ocean System
  8. 6. Summary and Discussion
  9. Acknowledgments
  10. References

[45] We sincerely thank all participants on-board USCGC Healy. We are deeply indebted to Kazutaka Tateyama, Hirokatsu Uno, Bruce Elder, Tom Grenfell, and Jeremy Harbeck for their help with the ice-watch. Suggestions and comments from anonymous reviewers and associate editor were very helpful. Support for this study was provided by the Japan Agency for Marine-Earth Science and Technology (JAMSTEC) and the National Science Foundation (NSF) under Grant No. ARC-0454900.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Observations
  5. 3. Ice and Ocean Conditions
  6. 4. Relationship Between Ice Concentration and Water Temperature
  7. 5. Role of Heat Transmission Through Ponds and Ice on the Ice-Upper Ocean System
  8. 6. Summary and Discussion
  9. Acknowledgments
  10. References