1400JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGYVOLUME 11Differences between Radiosonde and Dropsonde Temperature Profiles over the Arctic Ocean SUZANN- M. SKONY AND JONATHAN D. W. KAHLDepartment of Geosciences, University of Wisconsin--Milwaukee, Milwaukee, Wisconsin NINA A. ZAITSEVACentral.derological Observatory, State Committee for Hydrometeorology, Dolgoprudny,'Moscow Region, Russia10 September 1993 and 26 January 1994 ABSTRACT The boundary layer structure measured by 402 pairs of approximately co/located radiosonde and dropsondetemperature profiles over the Arc'fie Ocean during the peried'1957-61 is examined. The radiosonde profileswere obtainexi at the Russian drifting ice camps "North Pole 7" and "North Pole $," and the dropsonde profil~were measured during the United States Air Force "Ptarmigan" series of weather reconnaissance flights. Theboundary layer structure is characterized by the features of the low-level tropospheric temperature inversion. The results indicate that the dropsonde soundings, although containing relatively few measurement levels,contain sufficient vertical resolution to characterize the temperature inversion. Systematic differences werenoted in wintertime inversion features and near-surface temperatures as measured by dropsondes and radiosonde~These differences are attributed to contrasting temperature lag errors accompanying ascending and descending1. Introduction Atmospheric measurements over sea ice have historically been extremely difficult to obtain due to logistical problems arising from inaccessibility and harshclimates. This is especially true for upper-air measurements, for which automated sampling is particularlychallenging and costly. Despite these difficulties, however, significant numbers of vertical soundings havebeen made over the Arctic Ocean during the past sevend decades. These historical upper-air datasets consistprimarily of radiosondes released at drifting ice campsin the Arctic Ocean during 1954-90, and dropsondesreleased from aircraft during 1950-61. The radiosondereleases were included in the meteorology componentof the "North Pole" expeditions, a series of Russiandrifting ice stations established in 1937 (Voskresenskyet al. 1983). The dropsonde measurements were partof the United States Air Force "Ptarmigan" weatherreconnaissance program (Kahl et al. 1992b). In all,approximately 27 000 meteorological soundings overthe Arctic Ocean during the period 1954-90 are available for analysis (Kahl et al. 1993). Corresponding author address: Jonathan D. W. Kahl, Dept. ofGeosciences, University of Wisconsin--Milwaukee, P.O. Box 413,Milwaukee, WI 53201.E-mail: Kahl~.onvex.csd.uwra.edu These profile measurements are of great valuecause they comprise the only long-term, in situ observational records over the Arctic Ocean, an otheva4sedam-sparse region with an area approximately equivalent to that of the continental United States. The Arcfie is thought to be a sensitive indicator of global climatechange, thus diagnostic analyses of these data may re-veal important insights on climate change processes.Indeed, a number of recent studies have utilized thesemeasurements to investigate various features of thenorthern high4atitude climate (Kahl et al. 1993; Set.reze et at. 1992a; Serreze et at. 199213; Skony 1992;Nagurnyi et aL 1991; Timerev and Egtrov 1991 ). Efforts are currently under way to include these measurements in global upper-air data archives (Eskfidgeand Sterin 1993) where they may be utilized for modelreanalysis (e.g., Kalnay and Jenne 1991 ) and verification studies. We have noted two potential problems ~hat may affect analyses using ~hese datasets, particularly withspect to the temperature structure in the lowest fewkilometers of the atmosphere. The first problem concerns the vertical resolution. The dropsonde data contain an average of seven measurement levels betweenthe surface and 700 hPa, about half the number usuallypresent in archived radiosonde data. To what extentdoes the reduced vertical resolution limit the ability ofthe dropsonde data to describe the boundar,y layertemperature structure? The second problem concernsc 1994 American Mcteorologieai SocietyOCTOB-R 1994 NOTES AND CORRESPONDENCE 1401a possible systematic bias caused by lag errors. Temperature profiles measured by radiosonde and dropsonde are subject to systematic lag errors caused bythe failure of the instrument to respond instantaneouslyto changes in temperature during ascent or descent.Lag errors corresponding to ascending and descendingtemperature sensors are expected to be of opposite sign(all other factors being equal). In this paper we investigate these problems by comparing approximatelycollocated temperature profiles over the Arctic Oceanmeasured by radiosonde and dropsonde instruments.2. Data and methodsa. Temperature profile data Two independent sets of temperature profile dataover the Arctic Ocean were utilized. The first consistsof 16 850 radiosonde ascents conducted at the "NorthPole" series of Russian drifting ice stations during theperiod 1954-90. The radiosondes were released one totwo times daily at locations along the drift path, mostlyin the central Arctic Ocean (Fig. la). The data wereobtained from the State Hydrometeorological Committee, Moscow, Russia. The second set utilized is the "~armigan dropsondearchive" (Karl et al. 1992b). This database containsover l0 000 lower-tropospheric temperature profilesover portions of the Beaufort Sea and western ArcticOcean (Fig. lb). The dropsondes were released byUnited States Air Force weather reconnaissance aircraftduring the period 1950-61. Temperature profiles fromboth sounding datasets were subjected to quality control procedures as described by Skony (1992).b. Determination of temperature inversion features A central feature of the Arctic boundary layer is thelow-level temperature inversion (Serreze et al. 1992a;Bradley et al. 1992; Kahl 1990; Kahl et al. 1992a). Theboundary layer structure may be characterized bythe features of the low-level temperature inversion-specifically, the height of the inversion base, the inversion depth, and the temperature difference betweenthe top and base of the inversion. Low-level temperature inversions, that is, those witha base below the 700-h?a level, were objectively identified using the algorithm developed by Kahl(1990).The procedure is to scan each sounding from the surface upward, defining the inversion base as the bottomof the first layer in which temperature increases withaltitude. The inversion top is defined as the bottom ofthe first subsequent layer in which the temperature decreases with altitude. Since Arctic temperature inversions often exhibit a complicated vertical structure (e.g.,Belmont 1957), lapse layers are frequently encounteredin the lower levels of the sounding. If these layers arethin (< 100 m), they are considered to be embeddedwithin the overall inversion layer. The inversion depthwas then computed as the difference in altitude between FiG. 1. (a) Locations of 16 850 temperature profiles measured by radiosonde at Russian drifting ice stations during 1954-90. The stationsdrift with the prevailing winds and surface currents in the Arctic Ocean, typically moving about 100 km each month. (b) Locations ofI0 326 temperature profiles measured by dropsonde from United States Ptarmigan weather reconnaissance aircraft during 1950-61. Theaircraft typically flew diamond-shaped patterns extending from central Alaska to the North Pole. Portions of the flight paths are visible as"streaks" of dropsonde positions. Drops were often made at predetermined locations, with as many as 500 soundings made at a few specificpoints. Over I00 soundings were made at the North Pole.1402 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY -OLUME 11 "3 ; ,.. ,',., ,,.. ~ ,,' ,.." ,.,, ... - st, ; ',.. .,.' - ~/.~..., ' ' ,._ .... , ....... ..,., FIG. 2. Positions of matched Russian ice station and Ptarmigandropsonde profiles. The asterisks denote the 402 ice-station positionsthat were determined to be approximately collocated with Ptarmigansoundings. Most of the collocated Ptarmigan soundings occurred atthree specific locations: 85.9-N, 156.2-W (247 points, symbol "1 ");85.6-N, 149.9-W ( 103 points, symbol "2"); and 80.6-N, 149.9-W(30 points, symbol "3"). The 84--88-N, 135-W-180- sector, usedfor the mean profiles in Fig. 3, is also shown.inversion top and inversion base. The temperature difference across the inversion was similarly computed as.the difference in temperature between inversion topand inversion base.c. Calculation of mean monthly temperature profiles using dropsonde data Using the entire dropsonde archive we computedmean monthly soundings for various subregions withinthe Ptarmigan area (Fig. lb). Each sounding was firstinterpolated to 50-m increments from 0 (surface).to6000 m. The interpolated soundings were then averaged to obtain a mean sounding for each month.d. Comparison of inversion features determined from radiosonde and dropsonde profiles We searched the two datasets for spatially and temporally collocated temperature profiles. As can be seenin Fig. 1, most of the ice-station soundings lie oulsideof the Ptarmigan region. Furthermore, the overlap inreporting periods is only 8 years: 1954-61. Determination of paired soundings was therefore hmited 'to icestations "North Pole 7" and "North Pole 8." The driftpaths of these stations traversed, the western ArcticOcean region from May 1957 to December 1961, aperiod coinciding with the Ptarmigan reporting period. An acceptable match between a Ptarmigan dropsonde profile and an ice-station radiosonde profile wasdefined if the two reports were separated by less. than300 km and taken within 12 h of each other. Thesecriteria were established through an iterative processin which we tried to maximize the number of matcheswhile keeping the distance (in space and time) betweenthe two profiles as small as possible. If these criteriadid not yield a unique match between the two damsets,the pair of soundings taken closest in time was retainedand all others discarded. A total of 466 matches wereidentified, but this number was reduced to 402 after~,,e ........... ,~,(,...~ ................................... Sit 5~ I~1~ g~ ~ X ltll -~1 -41 -~1 -21 -~t - 11 Te~eflti~e (~- ~INam= 2e~eFiG. 3. Mean monthly temperature profiles for the region extending from 84- to 88-N and 135-W to 180- (shown in Fig. 2), computedfrom 1638 dropsonde soundings during 1950-61. The dry-adiabatic lapse rate is shown in the lower left comer for reference.OCTOBER 1994 NOTES AND CORRESPONDENCE 1403 T^BLE I. Summary of temperature inversion features as determined from approximately collocated ice-station radiosonde ("Ice") and Ptarmigan dropsonde ("Ptarm") temperature profiles.Jan Feb Mar Apt May Jun Jul Aug Sep Oct Nov Dec TotalBoth surface based 13 21 21 27 4 5 6 1 6 6 9 30 149Both elevated 0 0 I 3 19 17 3 4 2 6 1 2 58Ice surface basedPtarm elevated 6 8 2 5 7 2 5 0 1 1 4 10 51Ice elevatedPtarm surfacebased 1 3 0 ~ 8 11 4 1 _Q0 4 3 6 41Inversion in both 20 32 24 35 38 35 18 6 9 17 17 48 299Ice surface basedPtarm noinversion 2 0 4 3 1 0 2 1 0 I 2 0 16Ice elevatedPtarm noinversion 0 0 0 0 1 1 I 0 0 I 0 0 4Ice no inversionPtarm surfacebased 2 1 0 1 3 5 4 1 0 1 0 1 19Ice no inversionPtarm elevated O ~0 ~0 O 15 15 21 _~5 2 O .__Q0 _~0 4~9Inversion in one 4 1 4 4 11 21 28 7 2 3 2 I 88Both no inversion O ~0 ..__Q0 O 2 4 5 1 3 O _~0 ~0 1.__~5Total 24 33 28 39 51 60 51 14 14 20 19 49 402Ice stationTotal inversions 22 32 28 38 40 36 21 7 9 19 19 48 319Surface based 21 29 27 35 12 7 13 2 7 8 15 40 216Elevated 1 3 I 3 28 29 8 5 2 11 4 8 103Inversionfrequency 92% 97% 100% 97% 78% 60% 41% 50% 64% 95% 100% 98% 79%Percent surfacebased 95% 91% 96% 92% 30% 19% 62% 29% 78% 42% 79% 83% 68%Percent elevated 5% 9% 4% 8% 70% 81% 38% 71% 22% 58% 21% 17% 32%PtarmiganTotal inversions 22 33 24 36 47 55 43 12 11 18 17 49 367Surface based 16 25 21 28 15 21 14 3 6 11 12 37 209Elevated 6 8 3 8 32 34 29 9 5 7 5 12 158Inversionfrequency 92% 100% 86% 92% 92% 92% 84% 86% 79% 90% 89% 100% 91%Percent surfacebased 73% 76% 88% 78% 32% 38% 33% 25% 55% 61% 71% 76% 57%Percent elevated 27% 24% 13% 22% 68% 62% 67% 75% 45% 39% 29% 24% 43%eliminating soundings with poor vertical resolution(four or fewer levels) or a missing surface report. Thelocations of the paired soundings are shown in Fig. 2.Inversion characteristics were determined for all pairedsoundings.3. Resultsa. Mean dropsonde temperature profiles Mean monthly soundings for the region extendingfrom 84- to 88-N and 135-W to 180- are shown inFig. 3. This region contained 1638 soundings includingover I00 of the ma~.ches shown in Fig. 2. The low-levelinversions can easily be seen during the winter months,with the coldest temperatures and deepest inversionsoccurring from December to March. Summertimeprofiles are nearly isothermal in the lowest kilometer,with a transition to wintertime conditions beginningin September. The annual progression of the temperature inversionas depicted in the mean monthly soundings is in goodagreement with earlier analyses of Arctic sounding data(Serreze et al. 1992a; Bradley et al. 1992; Kahl 1990;Kahl et al. 1992a; Nagurnyi et al. 1991; Timerev andEgorov 1991 ). The vertical resolution of the dropsondeprofiles, an average of seven levels between the surfaceand 700 hPa, thus appears to be sufficient to identifythe gross characteristics of the inversion layer. Similarresults (not shown) were obtained for mean monthlysoundings in other regions within the Ptarmigan measurement area.b. Frequency of inversion types A summary of the percentage of inversion typesfound in the paired soundings, by month, is given in1404JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGYVOLUME 1 1..~ 80- -" Lcr 40h_ 90 0 F M A M d J A S 0 N D Month FIG. 4. Monthly inve~ion frequencies ~ dete~ined from pairedice-station radiosonde and ~igan dropsonde ~un~n~. The~xes ~ve inve~ion fr~uencies for the dropsonde soundings: Mdeboxes repre~nt sufface-ba~d inversions, na~ow boxes repre~nt elevated invemions, and the sum of both box ty~s ~ves the tot~ inve~ion frequency. The lines ~ve invemion frequencies for the icestation m~o~nde ~un~n~: the ~h~ (u~r) ~ne repre~n~ to~inversion fr~uency, the solid (lower) line ~ves sufface-b~d inversion frequency, and the differen~ ~tween the two lines representsthe fr~uency of elevated inversions.Table 1. An inversion was present in both the ice-station radiosonde and Ptarmigan dropsonde profiles in299 cases (74% of the total number of sounding pairs).In 15 cases (4%) no inversion was present in eithersounding, and for the remaining 88 pairs (23%) aninversion was present in one of the two soundings. The annual progression of the frequencies of different inversion types is shown graphically in Fig. 4. During the winter months the frequency of surface-basedinversions in the Ptarmigan soundings is 10%-20% lessthan the corresponding frequencies over the ice stations. During summer the total inversion frequenciesfor the ice-station soundings are considerably less thanthe corresponding frequencies for the Ptarmigansoundings (Fig. 4). This difference reaches 43% in July.The minimum in the ice-station total inversion frequency, 41% in July, is inconsistent with previousanalyses of inversions over the Arctic Ocean. Sen'ezeet al. (1992a) reported a summer total inversion frequency of 89%, Belmont (1957) reported a frequencyof 90%, and Vowinckel and Orvig (1970) reported aJune minimum total inversion frequency of 61%(partly based on North Pole 7 soundings). The summertime inversion frequencies given by the Ptarmiigansoundings, however, are in good agreement with theprevious studies. In an attempt to determine the reasons for the surprisingly low frequency of summertime inversions depicted by the ice-station soundings, we looked raoreclosely at the summer cases. There were 68 cases yearround in which inversions were identified in the Ptarmigan soundings but not in the ice-station soundings;of these, 62 cases occurred in the months May-September (Table 1 ). There were 106 cases during thesesame months in which inversions were identified inboth soundings. Examination of the Ptarmigan inversions corresponding to these groups of cases revealedthe following. For the 62 summer cases with invemionsin the Ptarmigan but not in the ice-station soundings,the median inversion base height, depth, and temperature difference was 357 m, 433 m, and 2.9-C, respectively. For the 106 summer cases with inversionsidentified in both soundings, the corresponding median3030 20-~0-2(I-30-40-5O0 10 20 30 40 50 60 70 80 90 100 Percentile- Tsfc (ice station) + Tsfc (Ptarmigan) -~- dT.(ice station)2O-40-5O FIG. 5. Cumulative frequency distributions of surface temperature (Tsfc; fight scale) and temperature difference across the inversion (dT; left scale) for the 299 cases for which inversions arepresent in both ice-station radiosonde and Ptarmigan dropsonde profiles.-ta- dT (Ptarmigan)OCTOBER 1994 NOTES AND CORRESPONDENCE 1405~0 t 2 2 2 $ 1 3 4 2 2 2 4 11 2 22 3 2 I 1 23 3 2 2 3J F M A M J J A S O N O Month 0-20J F M A M J J A S O N D Month FIG. 6. Monthly distiSbutions of differences in (a) surface temperature and (b) temperature difference across the inversion for the 299cases for which inversions were present in both ice-station radiosonde and Ptarmigan dropsonde profiles. Differences are computed as icestation value minus Ptarmigan value. The plotted numbers denote the number of paired soundings, and the line connects the median valuefor each month.inversion base height, depth, and temperature difference was 11.6 m, 421 m and 3.6-C, respectively. Thiscomparison indicates that the ice-station radiosondesoundings tend not to contain inversions when thecorresponding Ptarmigan inversion is elevated andweak (i.e., with a small temperature difference acrossthe inversion). In addition, of the 62 summertime casesin which inversions were present in the Ptarmigan butnot in the ice-station soundings, 33 of the ice-stationsoundings depicted an elevated isothermal layer withdepth greater than 100 m. This raises the possibilitythat weak, elevated inversions are being "missed" bythe ice-station soundings and instead are incorrectlyreported as elevated isothermal layers. This possibilityis discussed further in section 4b.c. Inversion features and temperatures We prepared cumulative frequency distributions oftemperatures (surface, 850 and 700 hPa) and inversionfeatures (base height, depth, and temperature difference) for the. 299 cases in which inversions were presentin both the ice-station and dropsonde soundings. Distributions for most variables were similar for bothsounding types, but systematic differences were evidentfor surface temperature and temperature differenceacross the inversion (Fig. 5). As shown in Fig. 5, Ptarmigan soundings consistently depicted higher surfacetemperatures and weaker inversions (smaller temperature differences) than their ice-station counterparts. Monthly distributions of the differences in thesevariables (ice-station value minus Ptarmigan value) aregiven in Fig. 6. The largest differences and variancesin surface temperature occurred in winter (Fig. 6a),with Ptarmigan surface temperatures about 6--12-Cwarmer than the ice-station surface temperatures. Thedifferences in the temperature difference across the inversion (Fig. 6b).is anticorrelated with the differencesin surface temperatures (Fig. 6a), indicating thatweaker inversions tend to accompany the warmerPtarmigan surface temperatures. This relationship isalso evident in Fig. 7, which confirms that warmerPtarmigan surface temperatures tend to accompanyweaker Ptarmigan inversions (smaller temperature differences). We note that warmer Ptarmigan surfacetemperatures are also observed in 75% of the 149 casesin which both sounding types indicate surface-basedinversions. These pairs of surface-based inversions occur primarily in the winter months (Table 1 ).4. Discussiona. Temperature lag errors Temperature profiles measured by radiosonde anddropsonde are fundamentally different in that the radiosonde measurements are made while the instrumenttravels upward, and the dropsonde measurements aremade while the instrument travels downward. Bothmeasurement types are subject to systematic lag errorscaused by the failure of the temperature instrument torespond instantaneously to changes in temperatureduring ascent or descent. Temperature lag errors,though well understood in principle (e.g., Middleton1943), have often been ignored in inversion analyses(Houvilla and Touminen 1990).1406 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 11252015.E.~_-10.-15-20-25 -25 1I 1 1 1 11 111 1 I 1 2 1 1111 I 21 1 1 1 12112 1 11212 1121 l 1 3111 2211 11 2 1 1 11~234 2221 1 1111 11 112221111 11 1 I 2115 12342 1 21 11 1121 1 3 221 11 2 21 111111 1 121 1 132 2 1 21123 1 I 11 I 1 221 I I 1 1211 1 1 I 2 1 21 21 112 11 11 I 1 1 1 I 2 1221 1 I 1 11 1 31 . 11 1 1 1 11 1 21 11 1 111 11I 11 112 111 1 1 2 1 1 - 1 .1 1. 1 - i I I I I I I I -2o -is -lO -s 5 ~o 15 20 25 Difference in Temperature Difference Across the Inversion FiG. 7. Relationship between difference in surface temperature vs difference in temperaturedifference across the inversion for the 299 cases for which inversions were present in both icestation radiosonde and Ptarmigan dropsonde profiles. Differences are computed as ice-stationvalue minus Ptarmigan value. The plotted numbers denote the number of paired soundings. The expected effects of temperature lags on dropsonde and radiosonde instruments traveling througha surface-based inversion are illustrated in Fig. 8. Thedescending dropsonde instrument will "remember"the cooler temperatures above the inversion, thus artificially lowering the inversion top and decreasingthe inversion depth. As the instrument responds tothe warmer temperatures in the inversion layer it willartificially increase the surface temperature, thus decreasing the temperature difference across the inversion. The expected systematic lag error of an ascendingradiosonde instrument, on the other hand, has theopposite effect. The warmer temperatures sensedwithin the inversion layer tend to raise the inversiontop and increase the temperature difference, artificiallyproducing a deeper and stronger inversion. In thepresence of arf elevated inversion, the temperature lagcould result in either a positive or negative bias, depending or/the lapse rates within and above the inversion layer. The magnitude of the temperature lag error in anyactual dropsonde or radiosonde is dependent on anumber of factors including sensor type, ventilation,humidity, and radiation corrections. Operationalmethods for correcting early Ptarmigan soundings fortemperature lag errors have been documented (AWS1952); however, we do not know whether any suchcorrections were incorporated into the raw data. usedto assemble the Ptarmigan dropsonde archiw~. Ananalysis of early Ptarmigan data by Poage (1954) suggested that errors introduced by temperature lag wouldtypically be less that 1 -C. Our analysis presents conflicting evidence on thepresence of a systematic lag bias in boundary layertemperatures measured by ice-station radiosondes.While the wintertime Ptarmigan soundings depictwarmer surface temperatures and weaker inversionsthan their ice-station counterparts, they do not consistently depict smaller inversion depths and lower temperatures at 850 hPa, a level that roughly correspondsOCTOBER 1994 NOTES AND CORRESPONDENCE 1407 radiosonde %.....'"".~' ~ actua, .?....~/' dropsonde - _ , Temperature FIG. 8. Effect of temperature lag errors on a descending dropsonde(dashed line) and an ascending radiosonde (dotted line). The solidline represents the actual temperature profile.tO the top of the inversion layer. Based on the information available, we estimate that systematic temperature lag errors are responsible for a portion of thedifferences in surface temperatures and inversion features as measured by Ptarmigan dropsondes and icestation radiosondes. Lag or calibration errors in pressure sensors could also contribute to the bias, as themeasured pressure is used in the hydrostatic determination of geopotential height.b. Possible instrumental problems associated with North Pole 7 and North Pole 8 radiosondes The unexpectedly low frequency of summertime inversions depicted by the ice-station radiosonde sounding lacks an obvious physical explanation. Rather, itmay reflect a systematic failure of the radiosonde instrument to detect weak elevated inversions. The firstSoviet radiosondes, model RS-049, were used at icestations North Pole 7 and North Pole 8 (Zaitseva1990). The bimetal temperature sensor supplied withthis early radiosonde instrument was evidently not ableto respond quickly to temperature changes experiencedwhile ascending through elevated inversion layers. Asdiscussed earlier, this problem apparently caused weakinversion layers to be incorrectly recorded as isothermallayers. By I963 the RS-049 instrument had been superseded at virtually all Soviet radiosonde stations(Zaitseva 1990). This apparent problem should notaffect surface temperatures, as standard measurementprotocol requires acclimating the instrument with theambient air temperature prior to release. We were unable to ascertain the specific instrument types carriedby the Ptarmigan dropsondes during the comparisonperiod ( 1957-61 ).5. Summary a~ c~nc~us~ons We examined 402 pairs of approximately collocatedradiosonde and dropsonde temperature profiles overthe Arctic Ocean during the period 1957-1961. Theradiosonde profiles were obtained at the Russian drifting ice camps North Pole 7 and North Pole 8, whereasthe dropsonde profiles were measured during theUnited States Air Force "Ptarmigan" series of weatherreconnaissance flights. Our principle results are as follows. 1) The relatively small number of lower-tropospheric measurement levels reported in the Ptarmigandropsonde soundings provides sufficient resolution todetect the main features of the Arctic temperature inversion. The ice-station radiosonde data contain approximately twice as many levels in the lowest few kilometers; however, many of these levels appear to havebeen interpolated and do not necessarily provide bettervertical resolution. 2) Wintertime surface temperatures measured bydropsondes are typically 6--10-C larger than corresponding temperatures measured by ice-station radiosondes. 3) Dropsonde soundings consistently depict weakerinversions during winter. Median temperature differences across the inversion were 5-C less than thosedepicted by the corresponding ice-station soundings. 4) The frequency of summertime elevated temperature inversions depicted by ice-station radiosondes wasup to 50% lower than similar frequencies found in thedropsondes and in previous studies. Our results suggest that the systematic differences inwintertime inversion features and near-surface temperatures are partially attributable to contrasting temperature lag errors from ascending and descendingsounding instruments. Unfortunately, it is impossibleto quantify this bias because our paired soundings werenot truly collocated (the match criteria specified thatthe two soundings be within 300 km and 12 h of eachother). Other possible factors contributing to these differences include spatial and temporal inhomogeneitieson scales smaller than 300 km and 12 h and differencesin sounding data reduction algorithms (e.g., Elliott andGaffen 1991 ). Acknowledgments. This research was sponsored by the NOAA Climate and Global Change Program (Contract NA90AA-D-AC794), the National Science Foundation (Contract DPP-8822472), and the Electric Power Research Institute (Contract RP2333-07). It was conducted as part of the US-USSR Joint Committee on Cooperation in Environmental Protection, the In fluence of Environmental Change on Climate (Work1408 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 11ing Group VIII). Helpful comments were provided byJ. Overland (NOAA/PMEL), E. Andreas (CRREL),and two anonymous reviewers. REFERENCESAmerican Weather Service (AWS), 1952: AWSM 105-23.Belmont, A. D., 1957: Lower tropospheric inversions at ice-island T-3. Proc. the Polar Atmosphere Symp., Part I. J. Atmos. Terr. Phys., (Suppl.), 215-284.Bradley, R. S., F. T. Keimig, and H. F. Diaz, 1992: Climatology of surface-based inversions in the North American Arctic. J. Geo phys. Res., 97, 15 699-15 712.Elliott, W. P.,' and D. J. Gaffen, 1991: On the utility of radiosonde humidity archives for climate studies. Bull. Amer. Meteor. Soc., 72, 1507-1520.Eskridge, R. E., and A. Sterin, 1993: Detection of climate change using the CARDS data set. Proc. Eighth Symp. on Meteorological Observations and Instrumentation, Anaheim, CA, Amer. Me teor. Soc., J83-J84.Houvilla, S., and A. Touminen, 1990: On the influence of radiosonde temperature lag error on upper air climatological data in Finland, 1950-1988. Meteorological Publication No. 14, Finnish Me teorological Institute, Helsinki, 29 pp.Kahl, J. D., 1990: Characteristics of the lowqevel temperature in version along the Alaskan Arctic coast. Int. J. Climatol., 10, 537-548. , M. C. Serreze, and R. C. Schnell, 1992a: Low-level tropospheric temperature inversions in the Canadian Arctic. Atmos.-Ocean, 30, 511-529. --, S. Shiotani, S. M. Skony, and R. C. Schnell, 1992b: In situ meteorological sounding archives for Arctic studies. Bull. Amer. Meteor. Soc., 73, 1824-1830.--, D. J. Chafievoix, N. A. Zaitseva, R. C. Schnell, and M. C. Serreze, 1993: Absence of evidence for greenhouse warming over the Arctic Ocean in the past forty years. Nature, 361, 335-337.Kalnay, E., and R. Jenne, 1991: Summary of the NMC/NCAR re analysis workshop of April 1991. Bull. Amer. Meteor. Sot., 72, 1897-1904.Middleton, W. E., 1943: Meteorological Instruments. 2d ed. The University of Toronto Press, 226 pp.Nagurnyi, A. P., A. A. Timerev, and S. A. Egorov, 1991: Spatiotem poral variability of inversions in the lower troposphere in the Arctic. Akad. Nauk SSSR, Dokl., 319, 1110-1113.Poage, W., 1954: The dropsonde record from Alaska to the North Pole, April 1950-April 1952. Scientific Report No. 2, Contract AF 19( 122)-28, Arctic Meteorological Research, Department of Meteorology, University of California at Los Angeles, Los Angeles, 52 pp.Serreze, M. C., J. D. Kahl, and R. C. Schnell, 1992a: Low-level tem perature inversions of the Eurasian Arctic and comparisons with Soviet drifting station data. J. Climate, 5, 615-630. -, J. A. Maslanik, M. C. Rehder, R. C. Schnell, J. D. Kahl, and E. L. Andreas, 1992b: Theoretical heights of buoyant conv,,nztion above open leads in the winter Arctic pack ice cover. J. Geophys. Res., 97, 9411-9422.Skony, S. M., 1992: Thermal structure of the lower troposphere over the western Arctic Ocean. M.S. thesis, Department of Geosci ences, University of Wisconsin--Milwaukee, Milwaukee, Wis consin, 233 pp.Timerev, A. A., and S. A. Egorov, 1991: The spatial and temporal variability of surface inversions in the Arctic. Meteor. ttydrol., 50-56.Voskresensky, A. I., G. U. Karimova, M. S. Marshunova, and L. S. Petrov, 1983: Meteorological investigations at the Soviet drifting stations "North Pole." The Structure of the ,'ltmosphere over the North Polar Ocean, Vol. 381, Arctic and Antarctic Research Institute Proceedings, Gydrometeoizdat, 5-19.Vowinckel, E., and S. Orvig, 1970: The climate of the North Polar basin. World Survey of Climatology. Vol. 14, Climates of the Polar Regions, S. Orvig, Ed., Elsevier, 129-226.Zaitseva, N. A., 1990: Aerology (in Russian). State Committee for Hydrometeorology, Moscow, Russia, 328 pp.