Weather influence on passive microwave brightness temperatures STlG M . LBVKS. IRENE RUBINSTEIN and CHRISTIAN ULSTAD LevBs. S. M., Rubinstein. I . & Ulstad, C. 1994: Weather influence on passive microwave brightness temperatures. Polar Research 13. 67-81, Sca ice charts produced using spacebornc passive microwave observations are used on routine basis at several ice forecasting centres and during sea ice research campaigns. The capability of passive microwave sensors to monitor thc earth. regardless of cloud cover or daylight, and the 1400 km swath width (SSM/I) makc thesc sensors well suited not only for ice forecasting but also for providing information needed for planning northern oceanic routes. The retrieval of sea ice parameters is carried out by utilizing 37 and 19 GHz brightness tempcraturcs measured by Special Sensor Microwave Imager (SSM/I) on the Defense Meteorological Satellite Program (DMSP) platform. The ice type identification uses a-priory established signature brightness temperatures for each ice type. The algorithm generating the sea ice information contains some climatological information on the weather dependence of the observed brightness tempera- ture. A comparison between passivc microwave (SSM/I) retrieved ice parameters and ice maps from the Nonvcgian Meteorological Institute (DNMI) indicates that the weather correcting procedure within the algorithm may need rcgionalized input. The correlation of variability of the observed brightness tem- peratures with the weather changes has to be derived prior to introducing any corrections to the existing algorithms. Thc observed brightness temperatures contain information not only about the ice surface but also the atmosphcrie contribution. One o f the crucial tasks is to establish a technique for flagging changes o n thc ice surface in order to scgrcgate them from the atmospheric influence on the passive microwave signal. To obtain more knowledge about the effects of regional weather on the rctricval of sea ice parameters, local climatological information from DNMI has been compared with the SSM/I ice charts. This paper focuses o n thc outcome of this comparison and how regional atmospheric information can be used in thc retrieval of sea ice information from passive microwave data. S . M . L @ v d s , S I N T E F N H L . N-7034 Trondheim. N o r w a y ; 1. Rubinstein. Institute f o r Space and Terrestrial Science ( I S T S ) , Y o r k University. 4850 Keele Street. North Y o r k . Ontario, C a n a d a M3J 3 K I : C. ULFtad, Norwegian Meteorological Institute ( D N M I ) , P . O . Box 43 Blindern, N-0313 Oslo. N o r w a y . Introduction The remote sensing of sea ice with active and passive microwave sensors has become an indis- pensable tool to the polar research community. The traditional sources of sea ice information are now augmented with the data obtained from interpretating Special Scanning Microwave Imager (SSM/I) multi spectral passive microwave observations. The study of physical properties of the earth’s surface and atmosphere with microwave radio- metry became possible with the launch of COSMOS-243 (Basharinov et al. 1971). Single frequency Nimbus-5 Electrically Scanning Micro- wave Radiometer (ESMR) observations from 1972 to 1976 provided for the first time a set of global observations that allowed sea ice studies on a synoptic scale (Staelin et al. 1973; Zwally 1984). Since June 1978, multispectral dually pol- arized observations (Scanning Multichannel Microwave Radiometer on SEASAT and later on NIMBUS-7, Special Scanning Microwave Imager on DMSP (Hollinger et al. 1987)) of the earth microwave radiance have been used in monitoring the extent and physical properties of the sea ice cover and other geophysical parameters. Algorithms for translating the microwave measurements into sea ice cover parameters evolved after 1978, when Scanning Multichannel Microwave Radiometer (on NIMBUS-7) data were used to establish the ice monitoring capa- bilities of passive microwave sensors. At the Insti- tute of Space and Terrestrial Science at York University, an algorithm known as the AES/York Algorithm has been developed. The algorithm utilizes the SSM/I channels 19 and 37 GHz (both polarizations) to produce sea ice information (Total ice concentration, First year ice fraction, Old ice fraction and Thin/New ice fraction) and wind speed and cloud cover information for ice- 68 S. M . L @ u h et al. free pixels. (Ramseier et a1 1989; Hollinger 1991). The AES/York Algorithm is calibrated against ground truth measurements from several ice- infested waters. Observations from several surveys (also in the Barents Sea) have shown generally good agreement with the total ice con- centration values provided by the algorithm. A study on the variations in ice type fraction in the Barents Sea based on SSM/I ice charts from September 1988-May 1989 (Lw& et al. 1991) showed that the amount of old ice and thin ice varied in reverse proportion. The existence of old ice is far more important to the design and operation of vessels and marine structures than new ice, hence a correct discrimination between old ice and new ice is important. Materials and methods Data sets The data used in this project were acquired from March 22 through April 10 1992 and are as follows: Table 1 . Remotely sensed data and derivatives. SSM/I data from the US Defence Meteorological AVHRR data from the NOAA-11 satellite. Atmospheric data from the atmospheric Limited Area Model, LAMSOS, of the Norwegian Meteorological Institute (DNMI). Passive Microwave Multispectral Measure- ments were obtained from Civilian Navy Data Distribution Services Computer (CNODDS) in Monterey, California and processed at the Earth Observations Laboratory of the Institute for Space and Terrestrial Science in Toronto, Canada. The dates of SSM/I imagery are listed in Table 1. Advanced Very High Resolution Radiometer data were obtained from and processed at the T r o m s ~ Satellite Station (low resolution) and DNMI (full resolution). The full resolution imagery has been processed and histogram enhanced at the Meteorological Institute SAT- ellite image processing system MISAT (MISAT is an 1% based image processing system with an IVAS station as display unit and VAX support. The system has been developed by Spacetec Satellite Program. Data sets SSM/I AVHRR Date Ice ( 1 9 2 ) Time Orbit Time (Low res.) Time (Full res.) analyses 23 March 0657 cc1636 06:29 0 24 March 06:44 ccl6SO 09:27 12:s2 25 March 06:32 cc1664 09: 13 0652 26 March 06: 19 cc1678 10:43 06:29 0 27 March 07:48 cc1693 10:32 28 March 09:17 cc1708 10:20 29 March 1008 30 March 07:lO dd1735 09:S6 05% 0 31 March 0 6 5 6 dd 1749 08:38 dd1637 09:39 1 April 08:25 dd 1764 11:12 04:42 3 April 06:47 cc1791 10:25 cc1792 09:w 4 April 06:3s cc 1805 10:37 5 April 0757 cc1820 10:25 6 April 09:24 dd1834 10: 13 7 April a7:37 cc1848 10:52 cc1849 1O:Ol 8 April 08:S9 dd1863 9 April m:10 dd1878 2 April 06:31 cc1777 11:w 08:ll 0 10 April 09:26 Weather influence on passive microwave brightness temperatures 69 A/S, Tromso, in cooperation with DNMI). The dates of AVHRR imagery are listed in Table 1. The atmospheric data were obtained from the LAMSOS atmospheric model of DNMI. Due to the “spin-up”-time when starting a large numeric model, realistic values on the humidity quantities are fully obtained after six hours of running. Therefore the data used here are twelve hour prognoses. The following quantities were used: Cloud cover in percent (4 levels) Relative humidity (7 levels) Convective precipitation (accumulated during six Frontal precipitation (accumulated during six Temperature at 2 m level (above the model top- A rectangle containing Hopen, Bj0rnoya and the southern part of Spitsbergen was chosen with 4 of a total of 48 grid points as corners. The time resolution is twelve hours, and the spatial resolution is 50 km. The sea-ice analyses were obtained directly from the daily update of the sea ice conditions in the MISAT system. They are produced by draw- ing curves on the displayed imagery, using the graphical system in MISAT. The dates of analysis are listed in Table 1. hours) hours) ography) Retrieval of sea-ice parameters from spaceborne passive microwave observations The microwave radiation from the earth is a com- plex function of the temperature, physical com- position and properties of the earth surface altered by the absorption, emission and scattering from the atmosphere. The quantitive deter- mination of the environmental parameters is obtained from a limited set of relatively noisy microwave radiation measurements. In general these measurements are not sufficient for unique determination of the environmental parameters without some a priori empirical knowledge and mathematical models on the relationship between these parameters and measured radiometric tem- peratures. The empirical information is used to impose limits within which these parameters can vary. The accuracy of the retrieved information is therefore affected by the noise in the empirical data and the uncertainties in assumptions used in model equations. The brightness temperature, TB, is defined as the temperature T i n K (Kelvin) to which a black body must be raised in order to radiate with the same intensity. The brightness temperatures measured by the sensor, for any of the sensor channels, contain contributions from the earth, atmosphere and space radiation. Assuming non- precipitating atmosphere, the observed bright- ness temperature can be expressed as: TB = TBsexp(-t) + TI + (1 - e)T2exp(-t) where TBs is the surface radiation, t is the atmospheric opacity, T I is the atmospheric upwelling radiation, e is the surface emissivity, T2 is the downwelling atmospheric radiation and Tc is the cosmic space contribution. The atmospheric components, under assumptions stated above, can be expressed in terms of an atmospheric mean temperature T,. In polar regions, for the fre- quencies used in the algorithms, the atmospheric opacity is small, but not negligible. Due t o the noncoherent nature of the natural microwave emission from the surface, the surface radiation can be written as a sum of contributions from different ice types and open water. Equation 1 can be rewritten for the following scenarios: 1) Single channel, dually polarized observations; 2) Dual frequency, same polarizations for both channels. (1): The surface radiation term, if C represents the fraction of the footprint covered with ice, can be written as: + (1 - e)T,exp(-2t) (1) where subscript P represents vertical or horizontal polarization, TBip is ice signature brightness tem- perature and TBwp is the open ocean signature brightness temperature. The surface emissivity can also be written as a sum of the emissivities from ice fraction and open ocean: ep = Ceip + (1 - C)ewp (3) Substitution of Equations 2 and 3 into Equation 1 yields Equations 4 and 5 for vertical and hori- zontal polarizations respectively: TBv = exp( - t)[ CTBiv + (1 - C)TBwv + 1 (Cewv - Ceiv)Tzl + TI + (Cewv - Ceiv)Tc x exp( -2t) (4) TBH = exp( - t)[CTBiH + (1 - C) TBwH + (cewH - ceiH>TZ1 + TI + (CewH - ceiH)Tc x exp(-2t) ( 5 ) I0 S. M. L@utfs et al. Taking the difference between Equations 4 and 5 and neglecting contribution from space yields following expression for C: c = [exp(r)DTBVH - DTBw]/[DTBi - DTBw + (Dew - Dei) T2] DTBVH = TBV - TBH DTBi = TBiv - TBiH DTBW = TBWV - T B ~ H Dew = eev - ewH (6) (7) (8) (9) (10) (11) C = A D T B w + B (12) where: De. = e - e. I IV IH General form of Equation 6 can be written as: Coefficients A and B are calculated using Equation 6. Equation 12 (when the assumption is made that the difference between vertical and horizontal sea ice brightness temperatures is inde- pendent of ice type) is known as Hughes algor- 1.6 _ _ _ _ _ . . - - - - - - - - - - . . _ . - - - - . 1.4 - 1.2 - 1 - B _ _ _ _ . - - - 2 g 0.8 - e 2 .- 0.6 - 0.4 - 0.2 - A 0 - _ - - - _ _ _ _ _ _ _ - - _ _ _ _ ________________________________________-------------- -0.2 sensitive to the sea ice presence within the pixel. In the absence of meteorological information, it can lead to false classification of an ice-free pixel for a wind-roughened ocean or an overcast sky. Errors in the calculated concentrations can also occur if an inappropriate value of DTBi is selected as a threshold value for the ice presence within a pixel. The dependence of the coefficients A and B on selection of typical optical opacity and DTB, is shown in Figs. 1 and 2. (2): For any two channels (frequency f l and f2) with different sensitivities to the atmospheric conditions and different sea ice and open ocean signatures, Equations 4 and 5 can be written as follows: TBI = exp(-ri)[mBil + (1 - C ) T B ~ I + (CewI - Ceil)T21I + + (Cewl - Ceil)Tc1 x exp(-2tl) (13) (Cew2 - CeidT221 + 7'12 + (Cew~-Cei~)Tc2 x exp( -2t2) (14) T B ~ = e x p ( - % ) [ C T ~ i ~ + (1 - C ) T B ~ ~ + ithm. From this equation one can analyse dependence of the total ice cover estimate on the variability in atmospheric conditions. The varia- bility of DTBi with the ice types present within the field of view can also influence the accuracy of the calculations. If the weather conditions are used as an input to calculations of A and B, this Assuming that only two ice types are present within the pixel ( C F ~ concentration of ice type one, CE! concentration of the second ice type) and Cw is the fraction of open water, the Equations 13 and 14 can be solved for the ice fractions under the following constraint: type of an algorithm can be tuned to be very CF, + Cm + cw = 1 (15) Weather influence on passive microwave brightness temperatures 71 Fig. 2. Dependency of eo- efficients A and B on D T B , . K Replacing (1 - C) by 1 - CF] - C,, one can show that solutions for CF] and cFz can be written as: cF1 = A I T B l + B l T B 2 + CI C F ~ = A ~ T B I -k B ~ T B ~ -k c2 (16) (17) c = CF] -k c, = ACTBl -k B c T B ~ -k cc (18) Coefficients A l , B , , C I , Az, B2, C2, Ac, Bc, CC are obtained from Equations 13 and 14. Good estimates of ice type fraction depends on validity of these coefficients. Hence, there is a need for ground truth from different areas and seasons. So far there has only been a few SSM/I ground thruth surveys in the Barents Sea. Such ground thruth measurements can easily be combined with other surveys into the ice-covered areas and would pro- vide calibration data to improving the retrieval of Barents Sea ice parameters. Equation 15 explains some of the reasons for the old/new ice fluctuations mentioned in the Introduction. If F2 (the second ice type) is classi- fied as old ice, then no new ice fraction can be calculated. If F2 does not meet the old ice criteria, then CFz is the new ice fraction. It can be shown that A,-, Bc, Cc are almost independent of optical opacity, whereas the dependence of A,, B,, CI, A2, B2, C2 on r leads to the uncertainty in old and first year ice type fraction estimates (Rubinstein 1986). Atmospheric influence on the retrieval of sea ice parameters The atmospheric influence on the observed brightness temperatures, and therefore on the retrieved information, can be evaluated by vary- ing the values for the transmittance of the atmos- phere and the atmospheric radiation. For non- precipitating clouds the optical opacity t can he calculated using the values of the attenuation coefficient, K ~ , by water clouds and fog as reported by Haroules & Brown (1969): r = K I . m J4.34 (19) The parameter m, is the water content of clouds in g/m3. The atmospheric contribution can be estimated using the following approximation (Swift et al. 1984): TI = (1 - exp( - r)) (1.12 . Tcloud - 50) We will describe the atmospheric influence on the passive microwave algorithms that contain linear combinations of the observed brightness temperatures. Due to the regional and seasonal variabilities of the ice signatures, the AES/ YORK algorithm contains several ice brightness temperature testing routines. Clustering of obser- vations containing data representative to different regions is used to define the range of pure sig- nature classes. 72 S. M. Lmods el al. SSM/I Brightness Temperatures 260. 2 5 0 . 240. C .- 2 230. N 0 .- L - 2 220. - 0 U . _ c 210. I N 3 200. m - 190. 180. 170. Pixels containing mixtures of pure classes can be identified as such if they are located within the limits of the following lines (Fig. 3): points along the O F line contain old and first year ice, while open ocean and first year ice mixtures are assumed to be along the WF lines. The WT line defines criteria for water plus new ice pixels. Fig. 3 shows several brightness temperatures outside the “ice” area. The cloud cover data points (0) can be separated into two groups. One group are pixels that are ice free, and changes in brightness tem- peratures are weather induced. For a calm ocean, the brightness temperature Tv3, can reach 220 for overcast sky (no rain), while TvlY can be 195. Hence, drawing a line from W (the open ocean point) with a slope of 0.75 will separate the two groups. Data points below this line is considered ice free, while data points between this line and the WF line probably contain (new/thinner) ice. Pixels can erroneously be classified as old ice (+) due to different footprint sizes at 19 GHz (50 Km) and 37 GHz (30 Km). This is more likely to happen for areas where there are changes in ice concentrations (i.e. diffuse ice to pack ice). Fig. 3. Scatter plot of T,,, vs. Tv3, for Old Ice (+). Thin Ice ( x ) and Cloud Cover (0) footprints in the Western Barents Sca in March/April 1992. The polygon (OWTF) shows thc valid area for ice footprints when no atmospheric con- tribution is present. The 19 GHz channel will be seeing some of the higher concentrations (higher brightness tem- peratures) as compared to 37GHz. If 19GHz brightness temperatures are higher than 37 G H Z , the algorithm may declare that old ice is present. Resampling the SSM/I data for all channels to represent the same size of footprints will remove some of the ice type classification errors. High ice surface roughness, giving an increased capability to trap snow, can also be the cause for some of the ice to be classified as old. Pixels not containing old ice can be identified by using the procedure described in the previous section. Since the calculated ice fractions should be real, positive numbers, special tests are per- formed on the incoming brightness temperatures if any of the calculated fractions are negative. One of the reasons for the fractions being negative is underestimation of the atmospheric opacity. New calculation of the ice fractions is then per- formed with optical opacity set to a different value. If that does not help, the pixel is classified as containing ice type of a positive fraction. The operational usefulness of this algorithm is Weather influence enhanced by the following additional procedures: The brightness temperatures for the pixels ident- ified as ice free are sent through an oceanic par- ameters algorithm. Ocean surface winds and cloud cover are calculated for these points. If any of the pixels contained ice and were not classified correctly, then the ocean algorithm will classify them as cloud covered (See Fig. 8) on passive microwave brightness temperatures 73 The modular structure of this algorithm allows the use of each of the modules as an individual algorithm. This retrieval technique, although it appears to be more complex than other methods used, can be used for global and regional moni- toring of the sea ice parameters. Equations 13 and 14 can be used for channels other than 19 and 37 GHz. The structure of the algorithm allows Table 2. Evaluation of the AES/YORK algorithm performance (From Ramseier et al. 1989). The geographic areas included in the evaluation were the Beaufort Sea (old and first year ice). the Labrador Sea (first year ice) and the Gulf of St. Lawrence (first year ice). The time differences (*) between the satellite orbits and the corresponding validation data were less than 1 hour (class A), bctween 1 and 3 hours (class B) or betwcen 3 and 6 hours (Class C). Ice concentration Ice edge displacement Mean diff. Accuracy Group category ("/.) 95% CI" (km) 95% CI" I Combined area Pooled I-A, class A I-B, class B I-C, class C LAB. class A + B 1-1, ice formation 1-2, winter 1-3, initial ice melt 1-4. ice melt -8.9 -4.2 -6.8 - 10.3 -6.3 -9.1 -5.5 Same as 11-3 Same as 11-4 0.9 3.1 1.6 1.1 1.4 1.3 1 .o Same as 11-3 Same as 11-4 I1 Arctic -0.2 I .9 - - -0.7 2.6 Same as 111-2 Same as 11-3 Same as 11-4 Same as 111-2 Same as 11-3 Same as 11-4 Pooled 11-A, class A 11-B, class B 11-C, class C 11-AB, class A t B 1 1 - 1 . ice formation 11-2, winter 11-3, initial ice melt 11-4, ice melt -8.6 0.4 -7.1 -10.0 -5.8 -9.1 -1.5 -5.1 -11.5 0.9 2.3 1.8 1.2 1.6 1.5 0.5 2.7 2.7 -3.1 - - - -5.1 -4.0 -4.3 - 2.1 - - - 2.9 9.5 3.6 - 111 Gulf of St. Lawrence Pooled 10.1 111-A, class A -12.1 111-B, class B -5.8 Ill-C, class C -12.1 Ill-AB, class A + B -7.8 111-1, ice formation -8.9 111-2, winter -11.6 1.6 4.5 2.5 2.2 2.2 2.3 2.9 7.4 - - - 6.2 21.1 4.2 - - - - 4.5 9.6 Ice fraction First year ice Old ice -8.0 6.5 2.7 2.6 'CI = Confidence interval 74 S. M . Lauds et al. AIR TEMPERATURE PRECIPITATION Western Rarents Sea - 1992 Western Harents Sea - 1092 4 4 2 n 1.5 -2 - 3 g :: E g -14 - : 2 hl - 8 5 2.5 5 -10 g -12 Y -18 .3 1.5 a -22 & I - c .- .- E - I 6 .& .20 2 .24 .26 0.5 -28 -30 91 83 85 87 99 91 93 95 97 9Y 101 81 83 85 87 89 91 93 95 97 99 101 Time (day number) Time (day number) CLOUD PERCENTAGE RELATIVE HUMIDITY Western Harents Sea - 1992 Western Barenrs Sea - 1992 110, I 105 I - too c g Ica : 90 - 0 4 80 4 :: 4 50 - 8 40 ; l5 5 10 m b O 0 - h) 85 p 70 tQ d A 80 3 7 0 2 65 30 .- I a 4 20 81 83 85 87 89 91 93 95 97 99 101 0 55 81 83 85 87 89 91 93 95 97 99 101 Time (day number) Time (day number) Fig. 4. Air temperature (top left). prccipitation accumulated over the last 6 hours (top right), cloud perccntagc (fog) at loo0 IiPa lcvcl (bottom left), and rclativc humidity at 1OOOHPa lcvel (bottom right) from LAMSOS model calculations. switching to different channels since most of the constants are calculated from the signature bright- ness temperatures. The performance of the algorithm described above was evaluated for different geographical regions using SSM/I data for 1987 and 1988 and sea ice information provided by the Canadian Atmospheric Environment Service (AES) Ice Centre. Results of the evaluation are shown in Table 2. The sea ice charts generated using this algorithm were used successfully for navigational support during sea ice research campaigns in the Greenland Sea, Barents Sea, Weddell Sea and North Atlantic Sea. Results Observations in the western Barents Sea, spring I992 For some parameters of the atmospheric data generated for this study, Fig. 4 shows the vari- ations as time series for all 48 grid points. The figures show a cold period from 25 March (Day 85) to 2 April. Prior to and early in this cold period there was some precipitation. During the period the cloud coverage varied. Generally, there was little fog, but there were some days with significant cloud percentages at the 300- 850 HPa levels. Weather influence on passive microwave brightness temperatures 75 Fig. 5. SSM/I total ice concentration on 26 March 1992. The frame shows the location of the ERS-1 SAR scene (4 hours later). The thick line shows position of the ice edge in the SAR image. 76 S. M. L @ v h et al. N70 N77 N76 N75 N74 N73 Fig. 6. Cloud cover percentages at 500 HPa level on 26 March. Weather influence on passive microwave brightness temperatures 77 v70 v77 V76 v75 v74 w 3 DNMI LAMSOS Cloud ercentage a1 500 HPa level Day: 0!2-L992 Time: 1 2 : O O : O O CUT Fig. 7. Cloud cover percentages at 500 HPa level on 1 April. 78 S. M. L ~ v h et al. N78 N77 "76 "75 N74 "73 I I- I- E l 6 e18 E20 E22 E24 E26 E28 E30 SSM/I [ ISTS Total ice concentration (1: Date: 01-04-1992 Orbit: Ib84 T h e : 08:47:37 GYT Fig. 8. SSM/I total ice concentrations (%. 0 ) and cloud coverage (/lo,+) on 1 April 1992. Weather influence on passive microwave brightness temperatures 79 In addition to the SSM/I and AVHRR data sets listed in Table 1, three ERS-1 SAR scenes have been acquired (23, 26 and 29 March 1992) to study the ice conditions. Ground observations were performed during a vessel survey and several helicopter flights in the period 29 March-3 April. During the helicopter flights the ice conditions were monitored simultaneously by S-VHS Video and a FLIR camera mounted on the underside of the helicopter. Fig. 5 shows the total ice concentrations on 26 March. The frame shows the location of the ERS-1 SAR scene and the location of the ice edge as observed in the SAR scene. The figure shows good agreement in ice edge position between the SSM/I and SAR data. This was also the case for the other two SAR scenes acquired. The SSM/I ice chart from 26 March shows more than 80 per cent total ice concentrations for most of the ice field. During the cold period, freezing occurred forming frazil, grease and pancake ice. The ice edge moved eastwards reaching an easterly maxi- mum 1 April. On this day there were 30-50 km frazil/grease ice between the open water and the compacted pancake ice edge. The air temperature increased rapidly from 30 March to 3 April due to the winds turning eastwards and further south- wards. Since the wind direction changed from off- ice to on-ice, the newly formed ice was com- pressed and the ice edge moved westwards. The easterly winds led t o an ice concentration close to 100 per cent from the ice edge into the ice field. During such conditions SSM/I seem to under- predict the ice concentrations slightly. Observations on 26 March and I April 1992 To study brightness temperatures during different weather conditions the data from 26 March and 1 April 1992 are selected. The daily SSM/I ice charts monitored total ice concentrations and the location of the compacted ice edge very well. Figs. 6 and 7 show the cloud percentages from LAMSOS at 500HPa level for 26 March and 1 April respectively. Despite the high cloud per- centages the AES/YORK Algorithm detected the compacted ice edge position very well on both days. The ERS-1 SAR image from 26 March showed only a few kilometres of frazil ice outside the compacted ice edge. Hence, the frazil ice will have less influence on the brightness temperatures on this day than on 1 April where observations from helicopter showed frazil/grease ice up to SSM/I Brightness T e m p e r a t u r e s 2 6 0 . 250. C 0 240. N h 230. .- .- a = 220. 0 .- e 2 2 1 0 . I 200. 0 190. m N I c 180. 170. 150. i a o . i C 0 . 2 0 0 . 2 i o . 2 2 0 . z 3 0 . 2 4 0 . 2 5 0 . z 6 o . 37 CHz - Vertical Polorizotion SSM/I B r i g h t n e s s T e m p e r a l u r e s 260. 250. .$ 240. 230. C 0 N .- - 0 a - 220. L 210. I 200. 0 U .- e 2 6 190. !?! N 180. 170. 37 CHz - Verticol Polorizotion Fig. 9. Scatter plots of T,,, vs. Tv3, for First Year Ice ( A ) , Old Ice (+), Thin Ice ( x ) and Cloud Cover (0) footprints in the area 74ON-77"N and 2OoE-35"E 26 March (top) and 1 April (bottom) 1992. The polygon ( O W ) shows the valid area for ice footprints when no atmospheric contribution is present. 40-50 km outside the compacted ice edge. This frazil/grease ice area was not reported as ice by SSM/I, but was interpreted as cloud cover (Fig. Figs. 9 and 10 show scatter plots of TVl9 vs. TV3, and Th37 vs. Tv3, respectively for First Year Ice, 8) 80 S. M. Lbvds et al. SSM/I B r i g h t n e s s T e m p e r a t u r e s 2 6 0 . 240. 0 .- c .! 2 2 0 . 0 0 a. 0 E - - 200. c .: 180. 0 L I 160. N S 0 r\ c) 140. 120. 37 - Verticol Polarization SSM/I B r i g h t n e s s T e m p e r a t u r e s 2 6 0 . E 240. .- L .- :: 2 2 0 . - 0 a ' 160. 0 2 b 1 4 0 . R 120. 37 CHI - Verticol Polorizotion Fig. 10. Scatter plots of ThS7 vs. TvJ7 for First Year Ice ( A ) , Old Ice (+), Thin Ice ( x ) and Cloud Cover (0) footprints in the area 74ON-77"N and 20"E - 35"E 26 March (top) and 1 April (bottom) 1992. The straight lines in the figures show the 0% and 100% total ice concentration lines and the WF line. Old Ice, Thin Ice and Cloud Cover (over ocean) footprints in the area 74ON-77"N and 2OoE-35"E on 26 March and 1 April 1992. The 37 GHz scatter plot from 26 March (Fig. 10 top) shows brightness temperatures along the WF line while the similar plot from 1 April (Fig 10. bottom) shows bright- ness temperatures along a line offset from the WF line. This offset is due to weather effects and results in some ice points being classified as ice free and underprediction of ice concentration. Discussion This paper has presented the basic principals and equations for retrieval of sea ice parameters from spaceborne passive microwave observations and has outlined some of the atmospheric influence on the retrieval of sea ice information. Figures showing data from 26 March and 1 April 1992 show some of these effects. The AES/YORK algorithm detects the compacted ice edge very well. Also the total ice concentrations estimates are reliable. Frazil/grease ice outside the com- pacted ice edge may be interpreted as cloud cover. As such ice is of minor importance to navigation this misinterpretation is also of minor importance for navigational use of the sea ice information derived from SSM/I data. The equations presented show that the bright- ness temperature from each satellite footprint can be split into three "ice" components; open water and two ice types. The most hazardous sea ice for navigation is old ice. Hence the AES/YORK algorithm first tests for old ice fraction. If the old ice criteria is met, the remaining ice fraction is the first year ice fraction. However, if no old ice is detected the algorithm calculates the fraction of first year and thin (new) ice. Changes in the weather (especially air temperature) may cause the algorithm to switch from identifying old ice one day to thin ice a day or two later (or vice versa). However, by following the changes in ice conditions from day to day such rapid changes in ice conditions will be revealed. If a satellite footprint is classified as ice free, cloud cover and wind speed information might be derived instead of ice fractions. The data sets derived for this study have not yet been completely analysed and we expect t o know more on how well regional climatological information can improve the interpretation of SSM/I data when this work is finished. References Basharinov. A . E . , Gurvich, A . S . . Yegorov, S. T . , Kurskaya. A. A . , Matveyev D. T. & Shutko. A . M. 1971: The results of microwave sounding of earth's surface according to exper- Weather influence on passive microwave brightness temperatures 81 imental data from Cosmos 243, Space Res., 11. Akademie- Verlag, Berlin. Haroules, G . G . & Brown. W. E. 1969: The Simultaneous Investigation of Attenuation and Emission by the Earth's Atmosphere at Wavelengths from 4 cm to 8 mm. J . Geophys. Res. 74, 4453-4471. Hollinger, J . , Lo. R. & Poe, G. 1987: Special Sensor Microwave Imager User's Guide. Naval Rcscarch Laboratory Report. Hollinger J . 1991: DMSP special sensor microwave/imager calibrdtion/validation. Final Report: Vol. 11. NRL, Wash- ington. D.C. L ~ v ~ s , S . M., Vefsnmo, S. & Ramseier, R. 0. 1991: Barents Sea Ice Conditions as observed by Passive Microwave and other techniques. 2nd WMO Operational Ice Remote Sensing Workshop, September 10-13. 1 9 9 1 , Ottawa, Ontario. Canada. Ramseier. R. O., Lapp, D., Rubinstein. I . G . & Asmus, K . 1989: Canadian Validation of the SSM/I and AES/York Algorithms for Sea Ice Parameters. ISTS/Microwave Group Report. Rubinstein, I . G. 1986: Selection of temperature independent dual frequency sea ice algorithm. Ph. D assoc. report for AES. Toronto, Ontario. 155 pp. Staclin, D. H.. Barctt. A. E.. Waters. J . W.. Barath. F. T . . Johnston, E. J . , Rosenkrantz. P. W., Grant, N. E., & Lenoir, W. R. 1973: Microwave spectrometer on the Nimbus-5 satel- lite. meteorological and geophysical data. Science 182, 133% 1341. Swift. C. T . , Fedor. S . & Ramscier. R . 0. 1984: An algorithm to measure sea-ice concentrations with microwave radio- meters. J. Geophys. Res. 90. 1087-1099. Zwally. H. J . 1984: Observing Polar Ice Variability. Annals of Glociol. 5 , 191-198.