Introduction
There has been considerable speculation in recent years (Reference ZubovZubov, 1948; Reference WexlerWexler, 1958; Reference FletcherFletcher, 1965; Reference Donn and ShawDonn and Shaw, 1966) about the climatic consequences of a warming trend in the Arctic over the last fifty years or so which, it is believed, is associated with a slight decrease in the ice cover of the Arctic Basin, and under what circumstances this would lead to an ice-free Arctic Ocean. A realistic study of the problem requires an examination of the various terms in the heat-balance equation for this area on a seasonal basis (e.g. Reference Vowinckel and OrvigVowinckel and Orvig, 1966, and other papers by these authors). At the moment, there is considerable uncertainty about the fraction of the incoming solar radiation that is absorbed at the surface in the Arctic summer. During this period, insolation is received at the surface continuously; an influx of 800 cal cm−2 d−1 (388 W m−2) is not uncommon in May or June (Reference LanglebenLangleben, 1966) and it is not surprising to find that the fluxes of latent and sensible heat are negligible compared with the radiative flux. The nature of the surface of the ice cover changes rapidly with the disappearance of the snow cover and with the formation of melt pools, and it is therefore of interest to make detailed measurements of temporal changes of albedo of the ice cover.
Time sequences of such observational data are scarce. Values of albedo of an ice cover during the summer have been reported by Reference MarshunovaMarshunova (1961), Reference ChernigovskyChernigovsky (1963) and Reference LanglebenLangleben (1966). There is an appreciable spread of values in these measurements, with little or no indication of the state of the surface, so that there is insufficient information to resolve the differences. Obviously the choice of site is the determining factor if the measurements are made from a height of about 2 m, which is typical in heat-budget studies.
There is considerable advantage to making albedo measurements from greater heights. Such measurements can be made from aircraft (Reference McFadden and RagotzkieMcFadden and Ragotzkie, 1967, over central Canada), flying the same track day after day if changes in albedo are to be observed. An alternate approach has been used by the author (Reference LanglebenLangleben, 1968) who suspended two radiometers at a height of 50 ft (15 m) between two towers which were 100 ft (31 m) apart. From that height, the response of the downward-viewing radiometer was almost entirely from a circular patch of radius 150 ft (46 m), an area containing tens of pools and hummocks and hence large enough to be representative of the surface. The work reported here is an extension of the previous investigation. The measurements were carried out at Tanquary Fiord, Ellesmere Island (lat. 81° 25′ N., long. 76° 50′ W.) from 12 May to 17 June 1968. The same arrangement was used to make the radiation measurements as was described by Reference LanglebenLangleben (1968). In addition frequent photographs were taken of the ice cover and it has been possible to relate the albedo values to the extent of pool formation.
Observations and Results
(a) Solar radiation and albedo. The albedo of the surface may be defined as the ratio of reflected to incident solar radiation. Measurements of incident and reflected solar radiation were made with two Kipp hemispherical radiometers. These were gimbal-mounted and supported at a height of 50 ft (15 m) above the surface by cables attached to two towers which were 100 ft (31 m) apart. Both pyranometers were recalibrated after use by the Meteorological Branch of the Department of Transport of Canada and there was no significant change in the calibration constants. Observations were recorded continuously on a time-sharing galvanometric recorder of recycling time 6 min; a sequence starting with a galvanometer zero or shorted input reading, followed in 2 min by a reading of incident radiation and again in 2 min by a reading of reflected radiation. Further details of the tower construction and calibration are provided in Reference LanglebenLangleben (1968). It is to be noted that the observations of incident and reflected radiation are not simultaneous and, since considerable variation in incident radiation may occur during cloudy or overcast conditions in an interval of 2 min, that this will tend to introduce error in the albedo values.
Mean hourly values of the incident radiation are shown as the top curve of Figure 1 starting on 17 May. (The first five days of the observations have been omitted as uninteresting because the state of the ice cover and the albedo both remained more or less unchanged until 20 May.) It is seen in Figure 1 that the diurnal variation in incident flux of radiation is such that the peak value is on the average about 0·75 cal cm−2 min −2 (525 W m−2) and the minimum value about 0·2 cal cm−2 min−1 (140 W m−2). Even though hourly averages have been used, the jaggedness in the radiation curve, caused by varying degree of cloudiness, is evident.
Fig. 1. Hourly means of incident short-wave radiation at Tanquary Fiord, Ellesmere Iceland in 1968. Hourly means of albedo of the ice cover. The two curves are of albedo ± root-mean-square deviation. (next to bottom). Percentage of surface covered by melt pools as a function of time. (bottom). Air temperature at screen height against time.
Hourly values of the albedo were calculated, each as the mean of ten ratios of reflected to incident radiation and the root-mean-square deviation was determined. Two albedo curves have been plotted in Figure 1, one of albedo plus root-mean-square deviation and the other of albedo minus root-mean-square deviation, to indicate the spread caused by non-simultaneity of the radiation readings. The typical root-mean-square deviation is about 0·03 and is comparable to the 3% error of the radiometers (Reference BenerBener, 1951). The temporal variation of albedo will be discussed below in relation to changes in the surface of the ice cover.
(b) Temperature. Air temperature at a height of 5 ft (1·5 m) was recorded on a thermograph inside a Stevenson screen located adjacent to one of the towers. Two thermometers inside the screen were read periodically to check the calibration of the thermograph. Temperatures have been plotted as the bottom curve of Figure 1. It is seen that the air temperature remained below 0°C until 1 June with the exception of a brief excursion on 29 May to abovezero temperatures. The transition to above-zero temperatures after 1 June initiated the melting phase.
(c) State of the ice cover. The ice cover in Tanquary Fiord in 1968 was mainly first-year ice of salinity about 5 parts per thousand by weight and of thickness 2·2 m. Initially the surface was partially covered with hard-packed patches of granular snow of thickness up to 15 cm and impregnated with dust wind-driven from the shore. Each wind storm between 12 May and 20 May redistributed the snow drifts without affecting the proportion of bare ice which remained at about 30%. As was to be expected, the albedo of the surface was relatively constant during this period and so the first five days of observations have not been included in Figure 1. A light snowfall on 21 May temporarily covered the ice cover and was accompanied by an increase of albedo to a value of 0·93. There were further snowfalls, in trace amount, on 23, 24, 27 and 28 May which again produced peaks in the albedo. Between 29 May and 1 June the extent of the snow cover rapidly decreased with the air temperature close to 0°C and by 2 June, when the temperature rose to 3°C, surface melt water was first observed. For the remainder of the period of observation, the air temperature was consistently at or above 0°C and melt pools, hummocks and drainage channels quickly developed.
During the melt, a programme of time-lapse photography of the surface was initiated. A remotely triggered camera was mounted on one of the towers at a height of 20 ft (6 m) with its axis inclined below the horizontal to point towards the base of the other tower. Although the automatically adjusted lens aperture did not function properly, some 43 usable photographs were obtained. The film was projected, frame by frame, and the percentage area of surface covered by melt pools was measured with a planimeter. These 43 values are shown as dots in Figure 1 and are seen to be reasonably spaced over the melt period. It is evident by comparing the albedo curve with these values that a large amount of puddling (6–10 June) corresponds to a low albedo value, whereas a lesser degree of puddling (11–15 June) is associated with a higher value of albedo. It is also apparent that during periods when the area occupied by melt pools is increasing, such as 3–4 June and 16–17 June, the albedo is at the same time decreasing.
(d) Functional relationship between albedo and fractional area of melt pools. It would appear from the observations made in the last paragraph that albedo is related to the fraction of the surface covered by melt pools. Values of albedo coinciding closely in time to the 43 photographs, have therefore been plotted against fractional melt-pool area in Figure 2. In spite of the scatter which is attributed largely to error in measuring off areas on the projected photographs, it is clear that a linear relationship exists between the two variables being considered. The straight line shown is a least-squares fit of slope −0.30±0.018 root-mean-square error. Its intercept on the albedo axis of 0.49±0.03 root-mean-square error gives the value of albedo for the ice cover just prior to the development of melt pools.
Fig. 2. Albedo of the puddled ice cover A plotted against the fraction of the surface covered by melt pools Sw/S. The line shown is a least-squares fit of the data.
It is simple enough to demonstrate the theoretical soundness of this linear relationship. Consider some reasonably large representative area S of the surface whose albedo is A containing a random distribution of melt pools. Let S w and A w be the area and albedo of the melt pools in the representative area and S i and A i the corresponding quantities associated with the ice surrounding the melt pools.
Now if I is the intensity of radiation incident on the representative area, then the incident flux is IS and the reflected flux is IA i S i+IA w S w.
The albedo of the representative area is
But since S = S i+S w, we have that
Therefore
which linearly relates albedo to fractional area of melt pools. The slope and intercepts in Figure 2 have been labelled in terms of this equation. The value of albedo A w = 0.19 associated with the melt pools is a reasonable one, although twice as large as the albedo of open water. The melt pools are not very deep and some fraction of the incident radiation would be reflected by the ice at the bottom of the pools.
Discussion
The albedo of an ice cover during the melt season has been found to be a sensitive function of the extent of water puddles present at the time of measurement. Analysis of the observational data indicated that the albedo decreased linearly with increasing area of puddle formation, and suggested limits in albedo of 0.19 for the melt pools and of 0.49 for the hummocks of ice surrounding the pools. Although the observations were made in an area subject during the winter to surface contamination by dust carried by off-shore winds, it is more than likely that the findings reported here have more general validity. It is true that the initial presence of dust on the surface would tend to hasten the disappearance of the snow cover and the initiation of the melting phase. However the dust soon sinks below the surface and hence would not be expected to influence the albedo of a melting ice cover.
It is much simpler to make broad aerial photographic surveys than measurements of albedo from aircraft where constant attitude is difficult to maintain. It should be possible to combine the results of this paper with aerial photographs to obtain detailed information of space and time variation of albedo in the Arctic Basin.
Acknowledgements
This work has been supported by the Defence Research Board under D.D.P. Contract GR.813007 and by the National Research Council under Grant No. A-4232. I should like to express my thanks to the Meteorological Branch, Department of Transport for calibrating the radiometers used in this study and to Dr Geoffrey Hattersley-Smith and Mr Harold Serson of D.R.B. for planning the logistics to get men and equipment into the field.
