Surface characteristics and atmospheric footprint. of springtime Arctic leads at SHEBA

Transcription

1 Surface characteristics and atmospheric footprint of springtime Arctic leads at SHEBA James O. Pinto, A. Alam, J. A. Maslanik, and J. A. Curry Program in Atmospheric and Oceanic Sciences Department of Aerospace Engineering Sciences University of Colorado, Boulder, CO Robert S. Stone NOAA/Environmental Technology Laboratory 325 Broadway, Boulder, CO Journal of Geophysical Research Special Issue on SHEBA revised 23 March 2001

2 Abstract Observations of several freezing leads that occurred in spring near the SHEBA (Surface Heat Budget of the Arctic) ice station were made. The leads that formed during this study were between 3 and 400 m wide. Ice production in the leads less than 20 m wide was predominantly through congelation growth while both frazil ice production and congelation ice growth was observed in the wider leads. The production frazil ice and its advection downwind allowed open water to persist in the m wide leads for between 5 and 24 hours depending on the crossing angle of the wind. The surface energy budget of a wide lead is estimated from observations and with a model that resolves the coupling between surface turbulent fluxes and ice growth across the lead. The modeled lead-average net heat flux deficit tends to be significantly larger than that estimated from observations because of the magnitude of the modeled turbulent fluxes. Both estimates tend to be negative throughout the day despite strong input from solar radiation. Sensitivity studies indicate that the surface roughness length used for thin congelation ice in the model may be much smaller than that used in previous studies. The model results are also sensitive to the treatment of solar absorption by open water and ice. The atmospheric influence of a wide freezing lead was also observed. Under lead-perpendicular winds, the atmospheric influence of a 400-m wide lead extended more than 2.5 km downwind. Sensible heat fluxes observed 70 m downwind of the lead were a strong function of across lead fetch, upwind stability, and open water fraction. The sensible heat fluxes measured at this site were elevated above background values for nearly two days despite 11.5 cm of ice growth in the lead. 1

3 1. Introduction Leads, which are cracks in the sea ice, play an important role in determining the surface energy budget and salinity of the Arctic Ocean. The components of the surface energy budget (SEB) differ markedly for leads and the surrounding ice pack. Here, leads are defined as those consisting of open water and/or snow free ice less than 20 cm thick. Although new leads occupy a small fraction of the arctic ice pack, they have a strong influence on the area-integrated heat budget. During most of the year, sensible and latent heat fluxes from leads are two orders of magnitude greater than that found over the surrounding ice pack [e.g., Maykut, 1978]. The relatively warm surface of a new lead produces an upwelling longwave flux that is much larger than that typically found over the surrounding ice pack. At the same time, the surface of a new lead reflects much less shortwave radiation than the surrounding snow-covered sea ice. However, much of the incident solar radiation is transmitted into the ocean mixed layer and does not affect the SEB of the lead. The amount of heat and moisture emanating from recently open leads determines the horizontal and vertical extent of their atmospheric influence. Turbulent fluxes of heat and moisture into the air above a lead may result in the production of clouds which can influence downwelling radiation over a large area [Schnell et al., 1990]. In addition, the production of brine during ice growth in leads has a significant impact on the salinity of the ocean mixed-layer [Gow et al., 1990]. The area-average SEB of the Arctic is a function of the fractional coverage of different ice types and the local SEB of each ice type. During the Surface Heat Budget of the Arctic (SHEBA) field project, the SEB of several surface types including snow-covered multiyear ice, bare thick first-year ice, multiyear ice with melt ponds, and summertime leads was monitored. Few measure- 2

4 ments of the SEB over thin ice were obtained despite the importance of this ice type in the areaaverage mass balance and SEB of the Arctic. Observations of various components of the SEB over thin ice or freezing leads have been made [e.g., Andreas et al., 1979; Ruffieux et al., 1995]. Andreas et al. calculated sensible and latent heat fluxes over leads from data collected during the AIDJEX (Arctic Ice Dynamics Joint Experiment) Lead Experiment (ALEX). Using this data, Andreas and Murphy [1986] developed an empirical relationship between the heat and moisture fluxes from leads and the fetch across the lead. The SEB of an ice-covered springtime lead was estimated by Ruffieux et al. [1995] using data collected during LEADEX. In their calculation, the surface skin temperature was specified using a simple model and the surface albedo was assumed to increase linearly with time. In addition, the transmission of shortwave radiation through the thin ice was neglected and surface properties were assumed to be constant across the lead. Their estimated SEB was positive for over 10 hours due to a large net shortwave flux. Despite this gain in heat, the observed ice thickness did not decrease during this time. Ruffieux et al. [1995] explained this discrepancy by noting the uncertainties associated with the surface albedo and skin temperature. The calculation of surface turbulent fluxes above a lead usually neglects the effect of ice growth [e.g., Andreas and Murphy, 1986; Alam and Curry, 1995, 1997]. Models that consider freezing assume that the lead is uniformly covered with congelation ice [e.g., Maykut, 1986]. This assumption is only valid for narrow leads and/or low wind speeds when frazil ice production can be neglected. At larger fetches, leads freeze over starting from the downwind side as frazil ice is advected downwind and accumulates against the lead edge [Martin, 1981]. Using a detailed model of frazil ice, Bauer and Martin [1983] determined the depth of frazil ice at the downwind edge of a lead as a function of fetch and wind speed. Alam and Curry [1998; hereafter AC98] have 3

5 developed a surface turbulent flux/ice growth model that considers the influence that both congelation and frazil ice growth have on the SEB across a lead. Recent observations indicate that leads may have a greater impact on the heat and moisture budgets of the Arctic than previously thought. The length of time a lead remains open, the fraction of open water, and the thickness of new ice determines the horizontal extent of its atmospheric influence. Using airborne lidar data, Schnell et al. [1990] observed clouds emanating from wide, open leads in the Arctic. These clouds spanned horizontal distances of several hundred kilometers and attained heights of up to 4 km. The cumulative moistening from a network of open leads may be sufficient to produce large-scale regions of low-level cloud cover as described by Fett et al. [1994]. Models ranging from simple buoyancy-argument schemes to large eddy simulation (LES) models have been employed to determine the atmospheric influence of leads. Serreze et al. [1992] estimated the vertical extent of a lead-induced plume by determining the height of neutral buoyancy attained by air passing over a lead. The heights obtained by this method may be considered as maximum vertical displacements, since the air parcels were assumed to remain in the surface layer while passing across the lead and to be non-entraining as they rise to their level on neutral buoyancy. Pinto et al. [1995] simulated the internal boundary layer that develops over an infinitely wide lead using a 1-D model that treated turbulence, cloud microphysics, and radiative transfer. The height of their simulated lead-induced plume tended to be much lower than that predicted by the simple model of Serreze et al. [1992] because it included non-adiabatic effects (e.g., entrainment) which limited plume growth. Results from a dry 2D model developed by Alam and Curry [1995] indicated that under low wind conditions, thermally direct circulations may develop over a lead which can enhance the vertical transport of heat and moisture into the atmosphere. 4

6 Glendening and Burk [1992; hereafter GB92] used an LES model to determine the 3D turbulent response of the atmosphere to a wintertime ice-free lead. They found that, for moderate across lead winds, the maximum turbulence and updrafts occurred downwind of the lead and that significant warming of the boundary layer occurred five lead widths downwind. Simulations performed by Burk et al. [1997] with a 2D model, which included a second order closure for turbulence and a bulk microphysical parameterization, indicate that the horizontal and vertical extent of leadinduced clouds are very sensitive to the upwind atmospheric stability. Using a cloud-resolving model, Zulauf and Krueger [1999, 2000] found that including treatments of cloud microphysics and radiative transfer increased the plume depth by 25% over that found for dry simulations. The effect that rapid freezing of a lead has on the surface fluxes and hence the evolving internal boundary layer downwind is investigated. In situ measurements and results from the surface turbulent flux/ice growth model of AC98 are used to determine the lead-average SEB for a 400-m wide lead observed during SHEBA that featured both frazil ice production and congelation growth. The effect that resolving horizontal variations in surface properties across the lead has on calculations of the lead-average net heat flux is discussed. Finally, the atmospheric footprint of this lead is characterized using data collected at three platforms that were located downwind. 2. Measurement platforms and instrumentation Observations were obtained during the SHEBA field experiment which took place from October 1997 through October The Canadian ice-breaker, Des Groseilliers, served as the central headquarters during SHEBA. It was frozen into a multiyear ice floe and allowed to drift with the pack ice in an anticyclonic direction. Freezing leads were monitored with surface-based instrumentation between April 18 and May 10. During this time, the mean location of the ship 5

7 was in the Chukchi Sea at 76 N, 165 W. Lead activity was low from April and high from April 27-May 10. Characteristics of the leads that formed near the Des Groseilliers during the active period are given in Table 1. The leads are identified by the nearest landmark ( Atlanta and Seattle denote two NCAR flux-pam (Portable Automated Mesonet) stations). Much of the observational analysis presented in this paper was collected at the Atlanta lead site. Four flux-pam stations were located at remote sites within 5 km of the Des Groseilliers. The Atlanta flux-pam station was approximately 70 m southwest of a lead that opened on April 28. Relevant instrumentation at the flux-pam stations includes: sonic anemometers, hygrothermometers (TRH), and upward- and downward-looking pyranometers and pyrgeometers. Sensible heat fluxes and the mean wind were measured with the Applied Technology (ATI) sonic anemometers. This brand of sonics performed well under riming conditions. The pyranometers were standard Kipp and Zonen CM21 models while the pyrgeometers were NCAR-modified Eppleys. The TRH instruments were nominally between 1-2 m above the snow surface while the sonic anemometers were 2-3 m above the surface. The flux-pam stations were located on snow-covered ice of varying thickness and roughness characteristics. The flux-pam data used in this study are 5- minute averages. Wind speed and sensible heat fluxes collected at the Atmospheric Surface Flux Group (ASFG)* 20-m tower are also used in this study. The ASFG 20-m tower was roughly 2.4 km southwest of the Atlanta lead (Figure 1). The wind speed, wind direction, and sensible heat flux were measured at 5 levels on the tower using ATI sonic anemometers. Ten-minute mean winds and hourly sensible heat fluxes from the ASFG tower are used in this study. The ASFG tower *The Atmospheric Flux Group (ASFG) was a collaboration of scientists from NOAA/Environmental Technology Laboratory (ETL), Army Cold Regions Research and Engineering Laboratory (CRREL) and the Naval Postgraduate School. 6

8 measurements are described in great detail by Persson et al. [2001]. Surface-based field measurements were also made using a sled-mounted platform instrumented with a KT19 radiometer (8-14 µm), a LICOR pyranometer ( µm) with a limited field of view, and a shielded thermistor. The LICOR data have been scaled to broadband values using side-by-side comparisons with a calibrated Kipp and Zonen pyranometer at the Atlanta flux-pam station. The sled-mounted platform could be transported to remote sites via snow machine, hence the name Mobile Radiometric Platform (MRP). The MRP s portability was essential due to rapidly changing ice conditions in the vicinity of the ship. The MRP had an extendable boom on which the radiometers and shielded thermistor were mounted. These probes were typically located about 2 m above the new ice surface in the lead and about 1.5 m from the edge of the lead (see Maslanik et al. [1999] for a complete description of the MRP). Five minute averages have been obtain from 10-second data collected in the field. Ice thickness and snow depth measurements were made periodically for each lead. Ice thickness measurements were made using a steel plumb on a graduated string. These measurements were made by dropping the plumb through a hole in ice and pulling it tight against the bottom of the ice. This method was accurate to within 0.25 cm. Snow depths were measured with a cm-ruler. Several ice thickness and snow depth measurements were made in the vicinity of the MRP to characterize the local variability in these quantities. A hand-held LICOR pyranometer, with a hemispheric field of view, was used to characterize local variations in the surface albedo. The albedo was calculated from measurements of the upwelling and downwelling shortwave radiation made 1 m above the surface with the hand-held LICOR pyranometer. Care was taken to ensure that the instrument was level during each measurement using a bi-directional bubble level. The irradiances obtained with the hand-held LICOR can 7

9 be used to obtain the albedo between 0.4 and 1.2 µm. Data from the Canadian Convair-580 are used to characterize horizontal variations in surface temperature and albedo in the lead. The Canadian Convair overflew the "Atlanta" lead between 0030 and 0050 on 29 April 1998, performing several overpasses about 6 hours after the lead had opened. Solar radiation was measured with a pair of Eppley pyranometers (0.295 to 2.85 µm) and the surface temperature was measured with a Barnes PRT-5 IR sensor (9.5 to 11.5 µm). A complete description of the instrumentation on the Canadian Convair-580 is given by Gultepe et al. [2001]. 3. Lead characteristics 3.1. Overview of springtime lead characteristics Observations of freezing leads were made during a period of enhanced lead activity in late April and early May. Tschudi et al. [this issue] obtained a lead fraction for early May of between and 0.86 for the area immediately surrounding the ship. This lead fraction includes areas of open water and new ice; however, very little open water was observed (see Tschudi et al. [this issue] for details). The characteristics of each lead observed during this period are given in Table 1. Lead widths ranged from 3mtoover400m. TheSeattle lead opened, froze over, and dynamically closed twice between 27 April and 4 May while the Atlanta lead opened on 28 April, froze over, and then opened further on May 6. A schematic representing the geometry of the leads relative to the ship and the locations of the observation sites is given in Figure 1. The Atlanta and Runway leads were located north and south of the ASFG tower, respectively. These leads are prominent 8

10 and persistent features in the 90 GHz AIMR images obtained from the NCAR C-130 (see Figure 1ofTschudi et al. [this issue] for an example). These two wide leads were connected by a smaller lead (Seattle) for short periods of time between 30 April and 6 May Characteristics of ice growth Frazil production and congelation ice growth were observed in the Runway lead and both Atlanta leads while only congelation growth was observed in the Seattle leads. The production of frazil ice depends on the open water fetch and the wind speed [Bauer and Martin, 1983]. The 2- hour average wind speed at 8.9 m and the maximum width attained by each lead are given in Table 1. Frazil ice production was only observed in the leads greater than 20 m across. In these leads, the advection of frazil ice from the center of the lead allowed open water to persist for several hours (Table 1) particularly when the wind direction was nearly parallel to the lead (e.g., Runway and Atlanta2 cases). Frazil production was observed at mean wind speeds as low as 3.9 m s -1. When the winds had a significant cross-lead component, frazil ice quickly covered the lead (e.g., Atlanta 1 case). For this case, measurements of the frazil ice pile up depth were made Albedo characteristics The new ice albedo in late April and early May was found to be a function of the ice thickness, the occurrence and spacing of frost flowers, and the accumulation rate of snow. The cloud fraction also affects the observed surface albedo by changing the character (i.e., relative amounts of direct to diffuse radiation, spectral composition) of the incoming solar radiation. The snow 9

11 depth, occurrence of frost flowers, and ice thickness of each lead listed in Table 1 were monitored. The albedo in each lead was determined using both the hand-held LICOR and the continuous measurements from the LICOR mounted on the MRP. The impact of drifting snow, frost flowers, and ice thickness on the observed surface albedo is evident in Figure 2. The albedo of bare ice clearly increases with ice thickness. The observed bare ice albedos tend to be less than those obtained with the albedo parameterization of Ebert and Curry [1993; hereafter EC93]. Frost flowers eventually formed on most of the leads observed during this study. They were typically initiated within a few hours of the lead opening. The points labeled F in Figure 2 correspond with those cases that were snow-free but had a significant coverage of frost flowers. Not surprisingly, the albedo of bare ice with frost flowers tends to be greater than the observed bare ice albedo and that obtained from the EC93 albedo parameterization. The prevalence of frost flowers observed on new ice during SHEBA is similar to that found by Perovich and Richter- Menge [1994] for springtime leads. Because of the ubiquity of frost flowers on new ice in springtime leads and their high reflectivity, it is felt that their impact should be included in SEB calculations. The accumulation of snow on newly frozen leads is a function of the precipitation rate and the wind speed. The wind speed determines the rate of accumulation and horizontal extent of drifting snow onto a new lead. During late April and early May, the accumulation of snow on newly frozen leads was dominated by drifting. The impact of accumulating snow on the albedo of a lead is evident in Figure 2. The surface albedo approaches that of snow-covered multiyear ice at snow depths greater than 3 cm. 3.2.Description of Atlanta1 lead 10

12 The Atlanta1 lead opened at 1815 UTC on April 28, just 70 m from the "Atlanta" flux- PAM station. The lead was located approximately 2.4 km northeast of the ASFG 20-m tower, bisecting a typically-sized multiyear ice floe. The lead grew to an estimated width of about 100 m in 1.5 hours and continued to grow for several hours reaching a width of roughly 400 m after 5.5 hours as measured with a laser range finder. Surveys done by the Canadian Convair-580 between 0030 and 0100 UTC on April 29 indicated that the lead was about 10 km long. Photographs indicate that the lead was approximately 30% covered with new ice after 1.5 hours and 98% covered after 6 hours. Cold surface air temperatures, strong surface winds, and a long fetch resulted in the rapid production of frazil ice. The frazil ice organized in linear bands parallel to the wind via langmuir circulations in the ocean surface waters, consolidated into frazil mats, advected toward the downwind edge of the lead, and piled up. As the irregularly-shaped frazil mats came together at the downwind edge of the lead, gaps left between the consolidated mats formed 5-10 m wide pools of open water (shown in Figure 3) which persisted for several hours. These observations of frazil ice are similar to those described by Smith et al. [1990]. After approximately 5 hours, the frazil ice had accumulated to a depth of 2.2 cm at the downwind edge of the lead. Frost flowers began to form on the new ice surface after just 2.5 hours. They played an important role in determining the albedo of the new ice in the lead (Figure 2). Measurements of the surface albedo obtained with both the MRP coned LICOR and the hand-held, hemispheric, LI- COR (circled data points) indicate that frost flowers significantly elevated the surface albedo. The relationship between albedo and ice thickness for the Atlanta1 lead was typical of that observed for springtime leads at SHEBA, with bare ice generally having albedos smaller than those given by EC93 and frost flowers elevating the albedo above that given by EC93. 11

13 Several aspects of the Atlanta1 lead made it an ideal springtime lead case to study. Measurements of ice thickness, surface temperature, and upwelling shortwave flux were made at the downwind lead edge. Measurements began less than 2 hours after the lead formed and continued for 24 hours. The coverage of ice across the lead and the lead width were well documented. The ice near lead edge remained snow-free for an extended period of time as winds were off-lead for nearly two days. In addition, several instrumented platforms were fortuitously located downwind of this lead for an extended period of time. Data collected from these platforms are used to assess the atmospheric impact of a wide freezing lead Surface energy budget estimated from observations The SEB can be estimated from observations using a few assumptions. The net heat flux is given by: F net = (1 - τ)(1 - α)s 0 + F d - F u - H s - H L + F c (1) where S 0 is the downwelling shortwave flux, F d and F u are the downwelling and upwelling longwave fluxes, F c is the conductive flux, H s and H L are the sensible and latent heat fluxes, α is the surface albedo, and τ is the solar transmission through the upper 5 cm of the surface. Solar transmission through the upper 5 cm of sea water is assumed to be equal to 0.4 following EC93. Solar transmission through the upper 5 cm of ice is assumed to range between 0.4 and 0.7, the range of possible values obtained using the extinction coefficients given by Ebert et al. [1995]. The downwelling fluxes of shortwave and longwave are obtained from a nearby flux-pam station. The upwelling longwave flux is determined from F u = εσt 4 s, where ε is the surface emissivity (0.98 is used for water and ice), σ is the Stephan-Botlzman constant, and T s is the surface skin temperature. The skin temperature of ice is measured at lead edge by the MRP while that for open water is 12

14 set to -1.8 C. The surface albedo of open water is set at after Pegau and Paulson [2000] while that of ice is determined from measurements of the MRP and a nearby flux-pam station. The sensible and latent heat fluxes over open water and new ice in the lead are estimated using bulk aerodynamic formulae. The bulk expression for sensible heat flux is given by H s = ρc p- C H U(T a -T s ), where ρ is the air density, c p is the specific heat of air, C H is the turbulent exchange coefficient, T s is surface skin temperature, T a is the upwind air temperature at 2 m, and U is the wind speed at 2 m. A similar bulk expression can be written for the latent heat flux, where the surface is assumed to be saturated with respect to salt water. The turbulent exchange coefficient for moisture is assumed to be equal to that for heat. The upwind air temperature, specific humidity, and wind speed are obtained from a flux-pam station that was not affected by the lead. Separate exchange coefficients are calculated for open water and ice surfaces. The parameterization of Andreas and Murphy [1986] is used to calculated the exchange coefficients for ice and open water as a function of fetch and upwind stability. Andreas and Cash [1999] have shown that their relationship between exchange coefficient and fetch holds for leads or polynyas covered with thin ice. To treat the effect of upwind stability on C H it is assumed that the surface roughness is the same for open water and ice. The fetch across the lead is assumed to increase from 2 to 400 m at a rate of 3mmin -1. The lead is assumed to be ice free for the first 1.5 hours, and then 20% ice-covered until the total fetch reaches 400 m. Once fetch across the lead reaches 400 m, the open water fetch is assumed to decrease at a rate of 1.5 m min -1 so that the lead is covered with ice after 5 hours. The resulting open water exchange coefficients ranged from 2.0 to 3.4 x 10-3 with the largest values being calculated just after the lead opens and just before the lead is covered with ice. After 1.5 hours, the exchange coefficient for ice must also be calculated. Since ice collects on the downwind side of the lead, the exchange coefficient for thin ice is calculated using the entire 13

15 fetch across the lead (which includes open water and new ice). Because of the longer fetch and cooler surface, the exchange coefficients for thin ice ranged from 1.6 to 2.0 x Shortwave radiation plays an important role in the evolution of the SEB in the lead. The observed upwelling, downwelling, and net absorbed shortwave fluxes for open water and ice are shown in Figure 5a. The absorbed fluxes are estimated assuming absorptivities of 30% and 60% for ice and open water, respectively. During the first 6 hours after the lead opened, the surface albedo was small and cloud cover was limited, allowing copious amounts of solar radiation to be absorbed at the surface. The albedo of the new ice was similar to that for open water. Open water regions absorb more solar radiation than ice-covered regions due to the greater solar absorptivity. One day later, 50% less shortwave radiation is absorbed by the surface, as ice now covers the lead and has a much greater albedo than 24 hours earlier. In addition, cloud cover has increased, reducing the incoming solar flux at the surface. The absorption of solar radiation is partially offset by longwave radiative cooling (Figure 5b). The net longwave flux over open water and ice-covered regions of the lead is negative throughout the observation period with the largest longwave cooling occurring during the first 6 hours that the lead was open. During this time, the net longwave flux over ice averaged around - 110Wm -2 with variations of +/-20 W m -2 caused by fluctuations in cloud cover. The net longwave cooling by the open water surface was as much as 20 W m -2 greater than that estimated for the new ice surface due to differences in surface temperature. The smallest longwave cooling, of around -40 W m -2, is associated with the presence of thicker clouds and a warmer surface. Time-variations in the bulk turbulent heat fluxes over the lead are quite complex because they depend on the evolving fetch, upwind air temperature, upwind surface layer stability, and wind speed. Over the ice-covered portion of the lead, the bulk sensible and latent heat fluxes gen- 14

16 erally decrease with time (Figure 5c). This decrease is initially caused by a rapid increase in fetch. Warming and moistening of the upwind surface layer between and UTC also reduces the surface turbulent fluxes (Figure 4b). The sharp decrease around UTC is caused by an abrupt decrease in the wind speed (Figure 4a). The open water turbulent fluxes are up to 300% larger than the values obtained over new ice. The time-variations seen in the open water turbulent heat fluxes result from changes in fetch across the open water part of the lead. Despite warming surface air temperatures, the turbulent fluxes over open water increase with time between and UTC as the lead freezes over and the open water fetch decreases. The net heat flux over open water and ice is shown in Figure 5d. The net heat flux over ice includes the conductive heat flux which averages around 5 W m -2. The net heat flux was negative over both open water and new ice portions of the lead during the first 24 hours that the lead was open. The net deficit is much greater over open water than over ice. These large deficits allow for congelation ice growth in the ice-covered regions and frazil ice production in the open water regions of the lead. The net heat flux is sensitive to amount of solar radiation absorbed by the surface. When the lead was covered with ice, the net heat flux was negative for the range of solar absorptivities obtained from Ebert et al. [1995]. Frazil ice production was observed until the lead was covered with ice. The net heat flux deficit estimated for open water averaged around -400 W m -2 during the first 4 hours that the lead was open. The net heat flux deficit estimated for ice between hours 1.5 and 4 averaged between and -100 W m -2 depending on the assumed solar absorption. Ice growth rates may be calculated from the net heat flux deficit following F n =q i h i / t where q i is the volumetric heat of fusion (2.678 x 10 8 Jm -3 at 0 C) and h i is the change in ice thickness in time, t. Using the net heat flux over open water gives a change in ice thickness of 2.2 cm after 4 hours. This change in 15

17 ice thickness nearly matches the total depth of frazil observed at the ice edge after 5 hours. The ice thickness measurements can be used to infer the net heat flux deficit over the congelation ice observed at lead edge. The ice thickness was observed to increase by 2.5 cm between and UTC. This implies a time-average net heat flux deficit during this time of - 240Wm -2 which is comparable to the net heat flux estimated from the observations using a realistic absorptivity of 30%. This comparison indicates that the components of the net surface heat flux over the ice at lead edge have been determined reasonably well. However, the SEB for the entire lead is more difficult to estimate from the observations. Only two surface types were treated in the estimated surface energy budget: open water and new ice. In fact, during the first 6 hours that the lead was open, the ice varied across the lead from thin dark congelation ice at lead edge, to unconsolidated frazil, to grease ice and open water. These different surfaces types will have different albedos, surface temperatures, and roughness lengths which will cause significant variations in the upwelling radiative and turbulent fluxes across the lead. To account for the effect of these across-lead variations on the SEB, a sophisticated ice growth/surface flux model is employed. 4. Simulation of turbulent heat fluxes and ice growth in a dynamically-active freezing lead The sea ice growth/surface turbulent flux model developed by AC98 has been modified to study the evolution of ice cover and surface turbulent fluxes above a dynamically-active freezing lead. The model is evaluated with continuous measurements made with the MRP, periodic in situ measurements of ice thickness and ice coverage, and episodic remote measurements of surface temperature and albedo from overpasses of the Canadian Convair. 16

18 4.1. Model description The sea ice growth/surface flux model of AC98 determines the co-evolution of surface freezing characteristics and surface turbulent (sensible + latent) heat fluxes across a lead. The model includes congelation ice growth and frazil ice production and uses surface renewal theory to obtain the surface turbulent fluxes as a function of surface roughness length, wind speed, and upwind stability. The turbulent heat flux calculations account for modification of surface air temperature and moisture as air flows across the lead [Alam and Curry, 1997]. Under the frazil ice regime, ice that forms through heat loss by the open water regions of the lead is advected toward the downwind edge of the lead where it accumulates. The thickness of frazil ice that accumulates at the downwind edge is called the pile up depth. The pile up depth is parameterized as a function of lead width and the 10-m wind speed based on the detailed modeling results of Bauer and Martin [1983]. Frazil ice is produced when the open water fetch is greater than 20 m and 10-m wind speed exceeds 4.35 m s -1. It is assumed to be instantaneously advected to the downwind edge of new ice in the lead. As frazil ice production continues, the edge of new ice in the lead advances upwind, eventually covering the lead [Alam and Curry, 1998]. Heat loss by the frazil ice-covered portion of the lead results in consolidation of the frazil ice. Ice consolidation is handled by increasing the concentration of previously-generated frazil ice while keeping its thickness constant [Alam and Curry, 1998]. During ice consolidation, it is assumed that rejected brine does not affect the salinity of the frazil ice-salt water mixture. The salinity is fixed at the observed value of psu [Stanton, pers. comm.]. Once the frazil ice concentration reaches 40%, the surface of the ice/water mixture (called grease ice) begins to cool. The skin temperature of grease ice is calculated as a frazil ice concentration weighted mean of the sea water and frazil ice surface temperatures. The sea water surface temperature is held constant 17

19 at the freezing point of salt water, while the frazil ice temperature is a function of the SEB. Consolidation of the frazil ice continues until the frazil ice concentration reaches 100%. At this time, the thickness of this consolidated frazil layer (called pancake ice) begins to increase in response to further cooling at the surface. The observed downwelling solar radiation is used to determine the amount absorbed at the surface. It is assumed that only the solar energy absorbed within 5 cm of the water or ice surface affects the SEB. An albedo of is used for open water while solar absorption by open water is parameterized following EC93. The albedo of ice is parameterized following EC93 while solar absorption by ice is determined following Ebert et al. [1995]. Snow cover on the Atlanta lead was negligible for the first 24 hours; therefore, its radiative influence has been neglected in the model. The model has been modified to allow for an evolving lead width, upwind atmospheric conditions, and downwelling radiative fluxes. Based on observations, the fetch across the lead is assumed to increase from 20 to 400 m at a rate of 2 m min -1. The initial fetch is assumed to consist of open water. The upwind temperature, moisture and surface layer stability required for the sensible and latent heat flux calculations are estimated from a flux-pam station that was unaffected by the lead (Figure 4). The wind speed observed at 8.9 m by the ASFG tower is used to calculate the turbulent fluxes and the pile up depth. The downward shortwave and longwave radiative fluxes are given by the Atlanta flux-pam station observations shown in Figure 5. The roughness length varies with surface type as given by AC98 with one significant exception. The roughness length used for thin congelation ice or nilus is an order of magnitude less than that used in AC98 (0.045 mm vs mm). This new value was chosen based on comparisons of modeled and observed congelation growth rates for a variety of roughness lengths for runs made during a period of negligible solar radiation between and UTC. 18

20 The model is run for 24 hours using a 5 minute timestep and a horizontal resolution of 1 m. The number of grid points is allowed to increase while the lead dynamically opens Modeled sensible heat fluxes The modeled sensible heat flux generally decreases with time across the lead (Figure 6). Over the first 3 hours of the simulation, decreases in the sensible heat flux are caused by an increasing fetch. During this time, the sensible heat flux at lead edge is much smaller than the leadaverage value because the lead is mostly open water (Figure 7) and the surface roughness length of frazil ice ( mm) is much less than that for open water ( mm). As frazil ice covers the lead, the difference between the lead edge and lead average sensible heat flux decreases. The spike in the modeled sensible heat flux at lead edge is caused by an abrupt increase in the surface roughness length associated with a change from the frazil ice regime to the pancake ice regime [Alam and Curry, 1998]. The pancake ice regime is short-lived, because the lead soon becomes completely covered with frazil ice. At this point, congelation growth begins across the lead resulting in a much smoother surface, and thus much smaller sensible heat fluxes. During the congelation growth regime, the sensible heat flux at lead edge tends to be slightly greater than the lead-average value. As will be shown, the larger heat fluxes modeled at lead edge are due to the warmer surface temperatures found there. The modeled lead-average and lead edge sensible heat fluxes generally fall between the bulk estimates for open water and ice (Figure 6). The bulk estimates for open water and ice shown in Figure 6 were both calculated using the time-varying fetch across the lead described in section 3.3. The resulting bulk sensible heat flux over open water differs from that shown in Figure 5 and depicts an upper limit for the sensible heat flux. The modeled sensible heat fluxes are greater than 19

21 the bulk value for ice, H sb (ice), except during the first few hours of the simulation when the surface roughness for frazil in the model (i.e., mm) was much smaller than that assumed in fetch dependent parameterization of AM86 (i.e., 0.32 mm). After , the modeled sensible heat fluxes are larger than the H sb (ice), with the difference increasing with time. Since the surface roughness values used in the model for congelation ice (i.e., 0.45 mm) are similar to that assumed by AM86, this discrepancy may be related to differences in the treatment of stability in the two approaches Comparison with observations at downwind edge The modeled surface temperature at the lead edge tends to be too warm at the beginning and end of the simulation (Figure 8a). The warm bias between and UTC can be related to differences in the modeled and observed concentration of frazil ice. While the model takes 5 hours to reach a frazil ice concentration of 100% at the lead edge, observations indicate that the frazil ice concentration at lead edge reached 100% after about 1 hour. Thus, the modeled surface temperature, which is a weighted average of the frazil ice and sea water temperature, tends to be much warmer than observed. When the modeled concentration of frazil ice reaches 100%, the modeled surface temperature rapidly approaches the observed value. The warm bias at the end of the simulation can be related to problems with the treatment of solar radiation. During the second half of day 119, the modeled surface albedo tends to be 33% less than the observed value (Figure 8b) because frost flowers, which have been shown to significantly increase the surface albedo (Figure 3), are not treated by the model. The smaller albedo in the model contributed to the warm bias seen during the day. Additional errors in solar heating of the surface may arise from uncertainties in the treatment of solar absorption by ice. Sensitivity 20

22 studies revealed that this uncertainty may account for errors of +/- 0.5 C in the predicted surface temperature by the end of the simulation. The modeled ice thickness at the lead edge is a function of the initial frazil pileup depth, the rate of frazil ice consolidation, and the congelation growth rate. The frazil pileup depth of 1.88 cm, obtained for the initial lead width of 20 m, persists for 5 hours (Figure 8c) while the concentration of frazil ice remains below 100%. The modeled pileup depth is 1 cm greater than that observed 1.5 hours after the lead opened. Sensitivity studies (not shown) indicate that the modeled amount of frazil ice in the lead, and thus the open water fraction, is quite sensitive to the treatment of shortwave absorption by open water regions of the lead. Both the model and the observations indicate that congelation growth began at the downwind edge of the lead roughly 5 hours after the lead opened. The 2.2 cm of frazil ice observed at lead edge after 5 hours is comparable to that produced by the model in the first timestep indicating that the assumed instantaneous advection of frazil is reasonable at longer timescales. After 5 hours, the modeled and observed congelation ice growth rates are fairly similar Variations in surface properties across lead Horizontal variations in ice thickness, ice concentration, skin temperature, albedo, and turbulent heat fluxes are simulated. Comparison with observed horizontal variations are limited to skin temperature and albedo obtained by the Canadian Convair-580 during a leg flown over the center of the lead between 0030 and 0100 UTC on 29 April. Across-lead variations in the modeled ice thickness result from variations in the frazil ice pile up depth. The pile up depth and number of gridpoints covered with frazil ice after it is advected downwind are calculated at each timestep. The frazil ice pile up depth ranged from 0.9 to

23 cm (Figure 9a) with the largest depths corresponding with the greatest fetches. The smallest pile up depth, seen at 20 m, was caused by a brief lull in the winds which decreased to 5ms -1 (see Figure 4). At each gridpoint, the ice thickness remains constant for an extended period of time while the ice concentration increases. This may be unrealistic as one would expect loose frazil ice to continuously advect beneath the initial frazil ice layer via ocean currents and sea ice drift, potentially increasing the depth of frazil ice at the downwind gridpoints and smoothing out horizontal variations in ice thickness. In fact, the depth of the frazil ice at lead edge was observed to increase with time before congelation growth began. Unfortunately, detailed observations of frazil ice advection were not made during SHEBA, so that this remains an open problem and a source of uncertainty in the model. While open water is present, the concentration of frazil ice varies across the lead (Figure 9b). The concentration of frazil ice is largest toward the downwind edge of the lead where the smallest pile up depths were obtained (e.g., between 15 and 25 m). By 3 hours, the concentration of frazil ice has reached 100% between 15 and 25 m allowing the depth of frazil ice at these grid points to increase. Once the open water fetch decreases to less than 20 m (Figure 7), congelation ice growth begins across the lead. Horizontal variations in the modeled surface temperature (see Figure 9c) are caused by several factors. Initially, the coldest temperatures are simulated at the downwind lead edge of the lead where ice concentrations have increased above 40% allowing the surface to cool. After the lead is covered with ice and congelation growth has begun, the coldest temperatures are found in the center of the lead where the initial pileup depths were larger (Figure 9c). Between 5 and 6 hours (valid between 2300 on April 28 and 0000 UTC on April 29) the modeled ice surface temperatures varied between -7.5 and -3 C. This range of values is colder than that (-5 to -1 C) ob- 22

24 tained by the Canadian Convair during an along-lead overflight (Figure 10). Since the surface temperature is an indicator of the ice thickness, this comparison indicates that the modeled ice thicknesses in the center of the lead were too large. The modeled surface albedo varies across the lead depending on ice thickness and ice type. The modeled albedo at each grid point increases with time (Figure 9d). The modeled leadaverage albedo increase from after 3 hours to after 6 hours with horizontal variations of as much as 50%. The modeled lead-average albedo at 0000 UTC on April 29 is similar to that obtained from the along-lead transect flown over the center of the lead by the Canadian Convair which varied between 0.19 and The small range of albedos observed during the along-lead transect indicate that these measurements may be influenced by the surrounding ice pack. The actual albedos of the lead may have been considerably less than observed by the Canadian Convair. The modeled sensible heat flux varies across the lead due to variations in the surface conditions and the open water fetch. The fetch dependence over open water is evident at hours 1 through 3 with the largest turbulent fluxes occurring at the upwind lead edge (Figure 11). This fetch dependence is similar to that parameterized by Andreas and Murphy [1986]. Variations in wind speed and upwind stability cause the turbulent flux maximum at the upwind lead edge to vary in time between 550 and 680 W m -2. At hours 4 and 5, the maximum sensible heat fluxes are found near the downwind lead edge where pancake ice, which has a large surface roughness length, has formed in the model. By 12 hours, the modeled sensible heat flux is nearly constant across the lead at about 120 W m -2. The modeled lead-average net heat flux deficit obtained from the data shown in Figures 9 and 11 tends to be significantly greater than that estimated from observations using two surface types. The larger deficit arises from the much larger turbulent heat fluxes calculated with the 23

25 model. Since the model albedo tends to be larger than observed during day 118, the modeled net shortwave absorbed at the surface tends to be smaller than observed; furthering increasing the lead-average net heat flux deficit. 5. Atmospheric footprint The fortuitous location of the Atlanta1 lead allowed for continuous monitoring of its atmospheric influence at three downwind locations. The MRP measured the 2 m air temperature at the downwind edge of the lead. The Atlanta flux-pam station and the ASFG 20-m tower were located approximately 70 m and 2.4 km downwind of the lead, respectively. The 2 m air temperature, specific humidity and sensible heat flux were measured at the Atlanta flux-pam station. Air temperature, specific humidity and sensible heat flux were measured at five heights at the ASFG 20-m tower. Data collected from these platforms are used to determine the atmospheric impact of this freezing lead. The sensible heat flux observed at the Atlanta flux-pam station is much larger than background values (e.g., at Baltimore), but much smaller than that estimated over the lead (Figure 6). During the first 6 hours of the lead s existence, the modeled lead-average sensible heat flux was 2 to 3 times larger than that observed 70 m downwind. It was found that the sensible heat flux observed at the Atlanta flux-pam station was elevated above background values for nearly two days despite ice growth of over 10 cm. After two days, the sensible heat flux at Atlanta reverted to background values because the site was no longer downwind of the lead. Sensible heat flux from the lead warms the 2 m air temperature just downwind of the lead by as much as 2 C (Figure 12). Warming of only 1 C was observed just 70 m downwind from the 24

26 lead, indicating that air passing over the multiyear ice between the MRP and Atlanta flux-pam station was cooled by 1 C. This downwind temperature pattern is similar to that obtained by GB92 in their LES studies. The downwind specific humidity was also observed to increase at the Atlanta flux-pam station by as much as 0.1 g kg -1 (or 15%). The sensible heat flux was observed to decrease with distance downwind of the lead. Just 70 m downwind the sensible heat flux has decreased dramatically from that inferred over the lead (Figure 6). The 12-hour average sensible heat flux observed at the Atlanta flux-pam station was 80Wm -2 greater than the background value. At the ASFG tower, sensible heat fluxes were positive during the day and typically decreased with height away from the surface. Between and UTC, the sensible heat flux at the ASFG tower was observed to increase with height with all values being elevated over background values. The background values were estimated using a 30-day composite of the sensible heat flux data at each level of the tower. The sensible heat flux perturbation caused by the lead (found by subtracting background from instantaneous values) is shown in Figure 13. The vertical gradient in sensible heat fluxes observed at the tower reversed with the largest fluxes of around 30 W m -2 found at the highest level. The influence of the lead was felt at the tower for nearly 12 hours. The sensible heat flux observed at the Atlanta flux-pam station depends on several factors. It is a function of the surface layer stability found up wind of the lead, the fetch across the lead and across the snow-covered multiyear ice, the wind speed, and the open water fraction. In order to relate changes in the sensible heat flux observed at the Atlanta flux-pam station to changes in fetch, open water fraction, and upwind stability; a normalized value is computed. The local surface layer stability is affected by diurnal variations in the net shortwave flux absorbed by the surface. This effect is removed by subtracting the background diurnal cycle of sensible heat 25

27 flux (as observed at the Baltimore flux-pam station) from that observed at the Atlanta flux-pam station. The adjusted sensible heat flux is then normalized using the observed 2-m wind speed and the observed surface-air temperature difference over the lead. The adjusted, normalized sensible heat flux, H n, is given by: H n (f ow,ψ, Τ 0 ) = (H 1 - H 2 )/(U(T a - T s )) (2) where H 1 is the sensible heat flux at the Atlanta flux-pam station, H 2 is the background sensible heat flux, f ow is the open water fraction, Τ 0 is the upwind surface layer temperature jump (a proxy for stability), and ψ is the crossing angle, which is equal to 0 when winds are perpendicular to the lead. This normalized heat flux represents a fetch-dependent exchange coefficient that includes contributions from open water, new ice and snow-covered multiyear ice. The fetch across the lead is a function of the actual lead width and the wind direction relative to the lead orientation. The effect that varying the crossing angle and lead width has on the normalized heat flux is evident in Figure 14a. The lead width increased rapidly to greater than 100 m in less than 2 hours on day 118. During this time, H n increased dramatically from near 0 just after the lead opened to 0.6 by the end of day 118. A least-squares fit is calculated from the data collected after It is seen that the normalized heat flux increases by a factor of two when the crossing angle increased from 0 to 25 (using F = Xsec(ψ) gives a 10% change in the fetch). The correlation of 0.36 is significant. The least squares error estimates for this fit are less than 3% indicating fairly high confidence in the trend fit to the data. This relationship indicates that the sensible heat flux was quite sensitive to small changes in fetch despite the lead being covered with ice. The sensible heat flux from the lead is also affected by the upwind surface layer stability. The difference between the surface temperature and the 2-m air temperature at the Baltimore flux- 26

28 PAM station is used as a proxy for the upwind surface layer stability. A strong correlation (ρ = 0.48) between the surface layer stability and the normalized sensible heat flux is evident in Figure 14b. Using only the data collected after , a least squares fit is determined. The smallest normalized sensible heat fluxes are observed for negative surface-air temperature differences which denote stable surface layer conditions. Once again, the influence of increasing lead width on day 118 is evident, resulting in a more rapidly increasing H n as a function of surface layer stability. The open water fraction also affects the sensible heat flux measured by the Atlanta flux- PAM station. The adjusted heat flux (H 2 -H 1 ) is found to decrease by 40% between 2030 and 2315 UTC as the open water fraction decreases from over 50% to less than 5% (not shown). 6. Conclusions Leads observed during spring at SHEBA generally froze over in less than 24 hours. Under across lead winds and cold temperatures (i.e., ~ -20 C), a 400 m wide lead froze over in about 5 hours during peak solar heating. This duration would have been even shorter if the lead had opened during off-peak solar heating. Since this lead is larger than 98% of the leads typically observed in the Arctic Basin [McLaren, 1989] and because it was observed during peak solar heating, it can be assumed that most leads with some non-negligible component of across-lead wind freeze over in less than 6 hours in mid-spring. Wintertime leads freeze over much quicker as seen in AC98, owing to lack of solar heating and colder air temperatures. However, the length of time that a lead influences the atmosphere depends on the rate of ice growth across the lead. Comparisons with the wintertime simulations of AC98, indicate that solar heating increases the length of time that a lead will affect the atmosphere by limiting the growth of sea ice. 27

29 Unknown parameters required by the surface flux/ice growth model were adjusted so that the results compared more favorably with the observations. The production of frazil ice and thus, the period of time open water persisted, was very sensitive to the amount of solar radiation assumed to be absorbed by the upper 5 cm of sea water. A 10% increase in the solar absorption allowed open water to persist in the lead for an additional hour. Sensitivity studies also revealed that the roughness length for thin congelation ice had to be an order of magnitude smaller than that given in AC98 to accurately simulate the congelation ice growth. Simulations with the adjusted surface flux/ ice growth model revealed a warm bias in the modeled surface temperature. This warm bias has been related to an underestimated surface albedo caused by neglecting the effect of frost flowers. The simulated surface temperature was also sensitive to the absorption of solar radiation by the new ice in the lead. Uncertainties in the solar absorptivity of new ice resulted in errors of +/- 0.5 C in the modeled surface temperature by the end of the simulation. After the 400 m wide lead froze over, the heating and moistening of the atmosphere decreased with time, but was observed to persist for nearly two days just 70 m downwind and up to 12 hours 2.4 km downwind. Simulations by AC98 indicate that an elevated turbulent heat flux can persist over the lead for several days despite ice thicknesses exceeding 20 cm. Of coarse, major snowfalls or convergent ice motions can reduce the period of time that leads influence the atmosphere. The fact that the influence of this 400 m wide lead was detected at 14 m by the ASFG 20- m tower for 12 hours indicates that the observed sensible heat fluxes at this height are actually representative of an upwind area larger than 2.4 km. 28

30 Acknowledgments. This research was supported by National Science Foundation Grant no (SHEBA) and NASA (FIRE-ACE). Thanks is afforded to NOAA/CMDL for lending the mobile radiometric platform for use during SHEBA and especially J. Wendell who designed the data acquisition system of the platform. We also thank I. Gultepe for kindly providing data from the Canadian Convair-580. Data collected with the NCAR C-130 processing was preformed by K. Laursen of the Research Aviation Facility. We also thank P. Guest, Ola Persson, and C. Fairall for helpful discussions regarding the manuscript and analysis and interpretation of data collected by ASFG 20 m tower. Thanks are also offered to the NOAA and NCAR technicians who deployed and maintained the surface facilities at remote sites surrounding the ship. In addition, T. Horst and G. Maclean did much of the calibration and processing of the flux-pam data. Comments and suggestions provided by the anonymous reviewers greatly improved the manuscript. 29

31 REFERENCES Alam, A., and J. A. Curry, Lead-induced atmospheric circulations, J. Geophys. Res., 100, , Alam, A., and J. A. Curry, Determination of surface turbulent heat fluxes over leads in arctic sea ice, J. Geophys. Res., 102, , Alam, A., and J. A. Curry, Evolution of new ice and turbulent fluxes over freezing winter leads, J. Geophys. Res., 103, 15,783-15,802, Andreas, E. L., C. A. Paulson, R. M. Williams, R. W. Lindsay, and J. A. Businger, The turbulent heat flux from Arctic leads, Boundary Layer Meteorol., 17, 57-91, Andreas, E. L., and B. Murphy, Bulk transfer coefficients for heat and momentum over leads and polynyas, J. Phys. Oceanogr., 16, , Andreas, E. L., Estimation of heat and mass fluxes over Arctic leads, Mon. Wea. Rev., 108, , Bauer, J., and S. Martin, A model of grease ice growth in small leads, J. Geophys. Res., 88, , Burk, S. D., R. W. Fett and R. E. Englebretson, Numerical simulation of cloud plumes emanating from Arctic leads, J. Geophys. Res., 102, 16,529-16,544, Ebert, E. E., and J. A. Curry, An intermediate one-dimensional thermodynamic sea ice model for investigating ice-atmosphere interactions, J. Geophys. Res., , Ebert, E. E., J. L. Schramm, and J. A. Curry, Disposition of solar radiation in sea ice and the upper ocean, J. Geophys. Res., , Fett, R. W., S. D. Burk, W. T. Thompson, and T. L. Kozo, Environmental phenomena of the Beaufort Sea observed during the Leads Experiment, Bull. Am. Soc., 75, , Glendening, J. W., Horizontally integrated atmospheric heat flux from an Arctic lead, J. Geophys. Res., 100, , Glendening, J. W. and S. D. Burk, Turbulent transport from an arctic lead: A large-eddy simulation. Boundary Layer Meteorol., 59, , Gultepe, I., G. A. Isaac, S. G. Cober and J. W. Strapp, Ice crystal number concentration versus 30

32 temperature in Arctic clouds. J. Climate, Submitted, Gow, A. J., D. A. Meese, D. K. Perovich, and W. B. Tucker, The anatomy of a freezing lead, J. Geophys. Res., 95, 18,221-18,232, Martin, S., Frazil ice in rivers and oceans, Annu. Rev. Fluid Mech., 13, , Maslanik, J., R. Stone, J. Pinto, J. Wendell, and C. Fowler, Mobile platform observations of surface energy budget parameters at the SHEBA site, Proceedings, Fifth conference on Polar Meteorology and Oceanography, AMS, Dallas, TX, Jan 1999, , Maykut, G. A., The surface heat and mass balance, in The Geophysics of Sea Ice, edited by N. Untersteiner, pp , Plenum Press, New York, McLaren, A. S., The under ice thickness distribution of the arctic basin as recorded in 1958 and 1970, J. Geophys. Res., 94, , Pegau, W. S., and C. A. Paulson, The albedo of Arctic leads, Annals of Glaciology, submitted, Perovich, D. K., and J. A. Richter-Menge, Surface characteristics of lead ice, J. Geophys. Res., 99, 16,341-16,350, Persson, P. O. G., D. Ruffieux, and C. W. Fairall, Recalculations of pack ice and lead surface energy budgets during the Arctic Leads Experiment (LEADEX) 1992, J. Geophys. Res., 102, 25,085-25,089, Persson, P. O. G., C. W. Fairall, E. L. Andreas, P. S. Guest and D. K. Perovich, Measurements near the atmospheric surface flux group tower at SHEBA, 1. Site description, data processing, and accuracy estimates, J. Geophys. Res., submited. Pinto, J. O., J. A. Curry, and K. L. McInnes, Atmospheric convective plumes emanating from leads 1. Thermodynamic structure, J. Geophys. Res., 100, , Ruffieux, D., P. O. G. Persson, C. W. Fairall, and D. W. Wolfe, Ice pack and lead surfeace-energy budgets during LEADEX J. Geophys. Res., 100, , Schnell, R. C., R. G. Barry, M. W. Miles, E. L. Andreas, L. F. Radke, C. A. Brock, M. P. McCormick, and J. L. Moore, Lidar detection of leads in Arctic sea ice, Nature, 339, , Serreze, M. C., J. D. Kahl, E. L. Andreas, J. A. Maslanik, M. C. Rehder, and R. C. Schnell, Theo- 31

33 retical heights of buoyant convection above open leads in the winter arctic pack ice cover, J. Geophys. Res., 97, , Smith, S. D., R. D. Muench, and C. H. Pease, Polynyas and leads: An overview of physical processes and environment, J. Geophys. Res., 95, , Tschudi, M. A., J. A. Curry, and J. A. Maslanik, Characterization of springtime leads in the Arctic Ocean from airborne observations during FIRE/SHEBA, special J. Geophys. Res. issue SHEBA. Zulauf, M. A., and S. K. Kreuger, Two-dimensional numerical simulations of arctic leads, Preprints, Fifth Conference on Polar Meteorology and Oceanography, Dallas, TX, Amer. Meteor. Soc., , Zulauf, M. A., and S. K. Kreuger, Two-dimensional numerical simulations of Arctic leads: Plume penetration height, J. Geophys. Res., submitted,

34 Figure Captions Fig. 1. Lead geometry in vicinity of SHEBA on 29 April Dashed lines indicate flight tracks of the Canadian Convair-580 which flew several overpasses on 29 April Symbols indicate positions of measuring platforms including: Seattle flux-pam station (asterisk), Atlanta flux- PAM station (filled circle), Florida flux-pam station (triangle), ASFG 20-m tower (diamond), mobile radiometric platform at Runway lead, MRP (plus). The jagged thick line denotes approximate location of the Runway lead. The straight thick line denotes approximate location of the Seattle lead. The string of diamonds denotes skin temperature warmer than -5 C as measured by the Canadian Convair and indicates the actual length of the Atlanta lead. Fig. 2. Albedo versus ice thickness from data collected with both the hand-held LICOR instrument and the downward-looking, coned LICOR on MRP. Points labeled F had substantial frost flower growth, filled circles indicate minimal snow depth, diamonds indicate snow depths between 0 and 3 cm, asterisks represent snow depths greater than 3 cm. The circled points indicate those obtained for the Atlanta1 case. The dotted line indicates the albedo observed in summertime leads by Pegau and Paulson [2000] and the solid line is from the parameterization of Ebert and Curry [1993]. Fig. 3. Photograph looking across the lead taken at 0000 UTC 29 April Pools of open water scattered throughout the lead are evident. Frost flowers are seen on the new ice surface. Fig. 4. Time series of (a) wind speed at 8.9 m from the ASFG 20 m tower and (b) temperatures (surface and 2.5 m air) and specific humidity from the Baltimore flux-pam station. Fig. 5. Time series of (a) downward shortwave flux, upward shortwave flux over ice, absorbed shortwave fluxes over ice (τ = 70%) and open water; (b) upward and downward longwave fluxes, 33

35 and net longwave fluxes for open water and ice surfaces; (c) sensible and latent heat fluxes obtained using bulk formulae for open water and new ice as well as lead average values (lines); and (d) net heat fluxes over open water, and over ice with transmissivity of 40 and of 70%. The dashed line denotes the time that the lead began to open. Fig. 6. Modeled, estimated, and observed sensible heat fluxes over and downwind of Atlanta1 lead. The modeled lead-average sensible heat flux, <H s >, and the downwind lead edge sensible heat flux, H s0 are given as a bold solid line and a dotted line, respectively. The bulk sensible heat flux over open water, H sb (ow), and over ice, H sb (ice), are given as filled circles and diamonds, respectively. The thin solid line labeled Atlanta gives the flux observed at the Atlanta flux-pam site located 70 m downwind of the lead. The background sensible heat fluxes (bold solid line) were obtained at the Baltimore flux-pam station. Fig. 7. Timeseries of specified lead fetch and modeled open water fetch for the first 6 hours of the Atlanta1 lead simulation. Fig. 8. Modeled and observed (a) surface temperature, (b) ice thickness, and (c) albedo at the downwind lead edge of the Atlanta1 lead. The vertical dashed line in (a) denotes when the concentration of frazil at lead edge reached 100%. The ice thickness calculated from the observed net heat flux deficit assuming a solar absorptivity of 30% is also given in (b). Fig. 9. Simulated across lead variation in (a) frazil ice concentration, (b) surface temperature, (c) surface albedo, and (d) ice thickness at hours 1 through 6 and 12 (odd hours labeled). The downwind edge of the lead is at 0 m. Fig. 10. Distribution of surface skin temperatures obtained during an along lead overflight of the Atlanta1 lead by the Canadian Convair-580 at 0050 UTC on 29 April The leg was flown at 34

36 100 m and was approximately 11 km long. The dotted line gives the cumulative distribution. Fig. 11. Simulated across lead variation in surface sensible heat flux at hours 1 through 6 (odd hours labeled). The downwind edge is at the origin of the axes. Fig. 12. Comparison of the evolution of 2-m air temperatures obtained at the downwind lead edge (MRP), the Atlanta flux-pam station (solid) and a background flux-pam station (dotted). Fig. 13. Timeseries of the perturbation in the sensible heat flux caused by the lead at 5 heights of the 20 m tower (2.1, 3.2, 5.1, 8.9 and 14 m). These values were obtained by subtracting a 30-day composite of sensible heat flux diurnal cycle from the instantaneous values. The vertical dashed line indicates the time the lead opened. Fig. 14. Comparison of normalized heat flux and (a) upwind surface/air temperature difference and (b) crossing angle (where a crossing angle of 0 indicates the wind is perpendicular to the lead). The data have been separated into 2 time periods to denote when open water was present (pluses). The lines indicate least squares fits to the data. Pearson correlation coefficients are also given. 35

37 Table 1: Summary of SHEBA springtime leads. Lead Case Date Initial Time on Site (UTC) Latitude/ Longitude ( ) Orien tation ( ) Max Width (m) Duration open water present (hours) Ice Growth Regime* 24 h Avg. T 2m ( C) 2 h Avg U 10m (m s -1 ) Runway , ~30 F/C Atlanta , < 5 F/C Seattle , < 2 C Seattle , < 2 C Atlanta , ~12 F/C * Observed type of ice production, where F indicates frazil production and C indicates congelation ice growth. 36

38 UTC ( N) Latitude Seattle Atlanta ASFG-tower Runway + MRP Longitude ( E) Figure 1 37

39 1.0 Albedo * + sd > 3, manual 0 < sd < 3, manual sd = 0, manual sd = 0, MRP F F F F 0.2 CE92-param 0.0 F= frost flowers present Ice Thickness (cm) Figure 2 38

40 Figure 3 39