the keweenaw current and ice rafting: use of satellite...

J. Great Lakes Res. 25(4):642–662Internat. Assoc. Great Lakes Res., 1999

The Keweenaw Current and Ice Rafting: Use of Satellite Imagery to Investigate Copper-rich Particle Dispersal

Judith Budd*, W. Charles Kerfoot, Andrew Pilant1, and Lisa M. Jipping2

Remote Sensing And Ecosystem Science InstituteMichigan Technological University

Houghton, Michigan 49931

ABSTRACT. The immense surface area and large volume of Lake Superior causes thermal characteris-tics to resemble marine waters, yet the completely bounded shoreline and low flushing rate introduceunique features. Previously, shoreline inputs were considered minor, as annual river discharges accountfor only 0.36% of the total hydrologic volume of the lake. However, thermal bar formation and windshear from prevailing westerlies impound warm waters along the southern coastline, creating a coastalexposure corridor with strong counterclockwise circulation known as the Keweenaw Current. Dischargesfrom rivers and industrial sources are confined, then entrained. Here infrared AVHRR (Advanced VeryHigh Resolution Radiometer) satellite imagery was utilized, verified by NDBC (NOAA National DataBuoy Center) buoy surface data, to document thermal features of offshore waters and the coastal zone.

Five stamp mills at Freda/Redridge discharged over 45 million metric tons of stamp sands between1895 and 1922. The coarse fraction forms beach sands that now extend 23 kilometers north from theirsources and that blanket shallow-water sandy sediments. The finer fractions disperse much farther thanthe coarse fractions, moving along the primary track of the Keweenaw Current. SPOT and TM (ThematicMapper) imagery were used to document how Ontonagon clays and Freda/Redridge stamp sand particlesare entrained by the Keweenaw Current. The two particle types have distinctive reflective spectra.

An additional transport mechanism, revealed by RADARSAT ScanSAR (Synthetic Aperature Radar)imagery, is ice rafting. Nearshore ice incorporates large amounts of coastal sands and deeper-water sed-iments. Spring break-up of coastal ice results in large drifting ice packs that are pushed by prevailingwesterlies and currents around the tip of the Keweenaw Peninsula into the Caribou Basin.

INDEX WORDS: Keweenaw Current, ice rafting, AVHRR, copper, water temperature, Lake Superior.

642

INTRODUCTION

The immense size of Lake Superior, its boundedwaters and latitudinal position generate intriguingtemperature and water circulation patterns. The lakeis the largest freshwater body in the world from anaerial extent, covering 82,100 km2, with a length of563 km and a breadth of 259 km. As in many of theLaurentian Great Lakes, the watershed is character-ized by a relatively high water to terrestrial surfacearea ratio (39.1% water surface, 61.9% terrestrial).Lake dominance of the hydrologic cycle is furtherenhanced by evapotranspiration. Although the ter-

restrial area constitutes 61.9% of the total surfacearea, evapotranspiration reduces the surface water-stream inputs to only 46% of the annual total. Thusmost (54%) of the vast stored waters in the lakeoriginally fell as direct precipitation onto its watersurface, explaining why the surface waters are bothvery dilute and simultaneously sensitive to long-distance atmospheric contaminants (IJC 1977, EC& USEPA 1987).

The volume of water stored is so large relative todischarge, there is a tendency to discount the yearlyshoreline inputs and to emphasize the atmosphericcomponent. The best available estimates are that thelake discharges only 0.56% of its volume annuallythrough the St. Marys River at Sault St.Marie, lead-ing to a long water residence time (191 yrs). Thetotal discharged annually into the lake from river,stream, and surface runoff constitutes only 0.36%

1Present address: Environmental Protection Agency, Office of Researchand Development, Landscape Characterization Branch, Mail Drop 56,Reasearch Triangle Park, NC 277112Present address: U.S. Army Corps of Engineers, Detroit Office, Detroit,MI*Corresponding author. E-mail: [email protected]

Satellite Imagery of the Keweenaw Current and Ice Rafting 643

constrained by limited spatial or temporal informa-tion (LeDrew and Franklin 1985). Moreover, giventhe immense size of Lake Superior, the thermalbudget and structure are an interrelated product ofregional meteorology and local influence upon cli-mate (Feit and Goldenberg 1976, Bennett 1978,Phillips 1978, Assel 1985). The establishment ofNDBC moored buoys and fixed monitoring stationsin conjunction with nearly synoptic coverage by in-frared Advanced Very High Resolution Radiometer(AVHRR) and RADARSAT ScanSAR satellite plat-forms opens exciting possibilities for studies of lakethermal properties, water circulation patterns, andice rafting.

Based on the long renewal time and the relativelyrapid seasonal mixing, there is a tendency to dis-count shoreline dynamics and to emphasize the sta-bility of offshore characteristics. This is furtherunderscored by the observation that 95% of the pol-lutants that enter the lake ecosystem do so throughatmospheric inputs (EC & USEPA 1987). Whereasthat assessment may be correct for certain organiccontaminants, it can not be generalized to many in-organic contaminants. Here it is illustrated how theseasonal development of thermal regimes combinewith prevailing westerly winds to create strongcounter-clockwise coastal currents. The coastalregime constitutes a corridor that transports dis-solved and suspended particulate materials alongshorelines and into deeper waters.

Various industrial operations historically dis-charged materials into the coastal zone of Lake Su-perior. At the turn of the century, large-volumestamp mills from copper mining operations openedalong the shorelines of the Keweenaw Peninsula(Kerfoot et al. 1994, Kerfoot and Robbins 1999).On the western coast near Freda and Redridge, 46 ×106 metric tons of stamp sands were discharged byfive high-volume stamp mills between 1895 and1922, with 2.4 × 106 metric tons reworked by recla-mation efforts between 1956 and 1964. Along theeastern shoreline near Gay, two large mills dis-charged 38 × 106 tons (Fig. 1). The coarse-grainedfraction of the stamp sands discharged at Freda andRedridge are plastered 22.6 km northeastward asbeach sands or are present underwater as vast fieldspeppering shallow-water sands (Kraft 1979),whereas the discharges at Gay formed black sandbeaches that stretch 8.1 km south to the TraverseRiver.

The circumstantial evidence for long-distancetransport of shoreline-discharged particles is verysuggestive. If reported surface copper concentra-

of the volume of the lake (IJC 1977, EC & USEPA1987).

Shoreline discharges can be important and are in-timately tied to coastal thermal structure and circu-lation patterns. Early knowledge of Lake Superiorwarming cycles (Beeton et al. 1959, Phillips 1978),thermal fronts and bar formation (Smith 1974,Spain et al. 1975), currents (Harrington 1895,Adams 1970, Smith 1972, Sloss and Saylor 1976,Green and Terrell 1980, Pickett 1980), and zones ofupwelling and downwelling (Ragotzkie 1974,Niebauer 1976) relied almost exclusively on ship-board measurements. Early data collection methodsemphasized ship surveys (bucket temperature, ba-thythermograph, or thermister measurements; Bee-ton et al. 1959, IJC 1977). Unfortunately,ship-board methods were severely constrained bytransit time, hence tradeoffs in spatial and temporalcoverage departed from truly synoptic monitoringand led to interpolation of values between recordingsites. The resulting contours of surface thermal pat-terns depicted idealized or stylized monthly temper-ature isoclines (Webb 1975, Phillips 1978).

Attempts to enhance temporal and spatial cover-age included improvements in both aerial surveysand establishment of continuously monitoring buoyand fixed recording stations. In 1966 the Hydrome-teorology Division of the Atmospheric Environ-mental Service (AES) of Canada established aprogram of airborne radiation thermometer (ART)surveys of the Great Lakes that bordered Canada.The aim was to improve the accuracy and coverageof surface temperature measurements provided byother means (Irbe 1972a,b). The surveys were con-ducted at approximately monthly intervals duringthe ice-free season. These surveys represented avast improvement over previous studies, but still in-corporated some liabilities. For example, the timespan between surveys, 1 month, meant that manythermal phenomena could not be tracked reliably(thermal bar and eddy formation, periodic up-welling). Although presented as synoptic, the datacollected from ART surveys also were not truly si-multaneous because of the time required to com-plete flight paths.

During the last few decades, ocean surface tem-peratures have been estimated by infrared radiome-ters aboard earth-orbiting satellites (Bernstein 1982,Schluessel et al. 1987, Barton et al. 1989, Minnett1990). Until recently, only a few attempts have ex-tended this technology to freshwater bodies. This isunfortunate, because satellite remotely sensed datacan provide new insights into problems previously

644 Budd et al.

tions across eastern Lake Superior are divided bypresettlement concentrations to produce an enrich-ment index, surface plots reveal high values thatfollow the path of the Keweenaw Current (Kerfootet al. 1994). The pattern suggests that copper from mining activities was transported a consider-able distance from the original Keweenaw Penin-sula sources, extending well into the CaribouBasin, despite low mass sedimentation rates in theinterior basin (Fig. 2). Yet the exact track of theKeweenaw Current, the amounts of sediment trans-ported, or other transport mechanisms are poorlyunderstood.

Here AVHRR satellite imagery is used to de-scribe the seasonal development of thermal struc-ture that, with prevailing westerlies, leads to theformation of the Keweenaw Current. In addition,TM and SPOT imagery are used to document off-shore movement of shoreline stamp sand particlesand their entrainment in the Keweenaw Current.RADARSAT ScanSAR satellite imagery reveals anadditional seasonal transportation mechanism, i.e.,

ice rafting, which also moves sediments off the tipof the Keweenaw Peninsula into the Caribou Basinand which contributes to observed patterns ofcoarse and fine copper-rich particle dispersal.

METHODS AND MATERIALS

NDBC Buoy Data

To obtain values for long-term water surface tem-perature trends, NDBC (National Data Buoy Cen-ter) data for 1984 to 1989 were purchased.Additional records were obtained for the detailed1993 to 1994 AVHRR studies. As part of the Na-tional Weather Service (NWS), NDBC operates au-tomated data acquisition systems from mooredbuoys and fixed monitoring stations throughout theGreat Lakes. The location of the three open-lakeNOMAD buoys in Lake Superior (Isle RoyaleBasin, buoy 45001, Latitude 48.N Longitude87.6W; Caribou Basin, buoy 45004 47.2N 86.5W;Chefswet Basin, buoy 45006 47.3N 90.0W) and

FIG. 1. Locations of buoys and shoreline sites shows the three open-lakeNDBC (National Data Buoy Center) buoys in Lake Superior (45001, Isle RoyaleBasin; 45004, Caribou Basin; 45006, Chefswet Basin) and four other fixedCMAN monitoring stations (PILM4 and ROAM4, both off Isle Royale; DISW3,off the Apostle Islands; and STDM4, at Stannard Rock); main figure Peninsulalocalities discussed in the text (circles, mines; triangles, stamp mills).


four other fixed monitoring CMAN stations(PILM4 48.2N 88.4W, ROAM4 47.9N 87.2W, bothoff Isle Royale; DISW3 47.1N 90.7W, off the Apos-tle Islands; and STDM4, 47.2N 87.2W at StannardRock) are shown in Figure 1. Temperature recordsfor the fixed stations extend throughout the year,whereas buoys are removed from Decemberthrough March.

The weather data were available on 9 track mag-netic tape in a 152 character-string format. Imbed-ded in this character-string was information on thestation, time of day, and weather parameters. Theweather data consisted of hourly climatologicalreadings. Programs were written within the statisti-cal package Statistical Analysis System (SAS) tocalculate and extract daily averages of six variables:1) air temperature, 2) surface temperature, 3) windspeed, 4) wind direction, 5) wave height, and 6)wave period. After the daily averages were ex-tracted, they were imported into the databaseDBASE IV, which served as a link to the imageprocessing software, and as a means of archivingthe information. Here the air temperature and bulksurface temperature are used directly, while refer-ring in text to the other variables. While long-termaverages are useful, it should be pointed out that

time-averaging across years may obscure certainseasonal events, such as abrupt warming followingcollapse of thermal bars (rate of early season warm-ing).

AVHRR Sensor and CoastWatch Images

AVHRR sensors are mounted on polar-orbiting,sun-synchronous NOAA Television and InfraredObservation Satellite (TIROS-N, currently NOAA-10 and NOAA-11) weather platforms. Two satel-lites orbit the earth every 102 minutes at an altitudeof 833 km. One satellite images Lake Superior atabout 0230 UTC (Universal Coordinate Time, oth-erwise known as Greenwich Mean Time) and 1430UTC, whereas the other images at about 0730 UTCand 1930 UTC. The devices record a continuousimage that covers a 2,400 km swath width with abasic spatial resolution (pixel size) of 1.1 km atnadir. The sensors are five-band multispectral scan-ners that detect radiation from the red to thermal in-frared regions of the electromagnetic spectrum(Table 1). Kidwell (1991) summarizes sensor speci-fications and Planet (1988) lists prelaunch calibra-tions for channels 1 to 5.

AVHRR has real-time radiometric calibration

FIG. 2. Copper enrichment in surface sediments off the tipof the Keweenaw Peninsula (after Kerfoot et al. 1994). Piediagrams indicate % enrichment, where 0% = background(i.e., pre-1870).

646 Budd et al.

Image processing (IP) utilized ERDAS SoftwarePrograms (Table 2). The > 90% cloud free imagesfor portions of Lake Superior were based on visualanalysis of near infrared (channel 2), aerosol-cor-rected reflectance (channel 1 minus 2), and channel4 for SST images. Upon receipt, the data were man-ually adjusted to a base map overlay provided byCoastWatch. Because of a clock error on the satel-lite, navigation has an error. In the most extremeexamples, the images were misregistered by asmuch as 14 pixels or 19 km. Without the adjust-ment procedure, the high variation between imageswould have precluded comparisons at the pixellevel (for the purposes of regression of in situ waterquality parameters against MCSSTs). Pixels withchannel 2 albedos that exceeded a threshold valuewere flagged as clouds or land and masked (Stumpf1992). This procedure provided a highly accurateoverlay of the shoreline, given the coarse spatialresolution of the images.

Lake Surface Temperature Data

There is a distinction between skin surface andbulk temperatures. Thermal infrared remote sensorsdetect radiation emitted from the first 50 nm of thesurface or skin of the lake and the intervening at-mosphere. Lake skin temperatures are generallyseveral tenths of a degree colder than the lowerwater strata due to evaporative cooling at the air-water interface (Saunders 1967, Grassl 1976, Paul-son and Simpson 1981). Bulk temperature, avariable usually measured by shipboard or buoy in-strumentation, measures temperature in the upper0.5 to 2.0 m of the water column. Understandingand quantifying the relationship between bulk andsatellite-derived sea surface temperatures (SST) hasoccupied two decades of research (McMillin 1975,McMillin and Crosby 1984, McClain et al. 1985a,Minnet 1990, Schluessel et al. 1990, Walton et al.1990, Wick and Emery 1992). SST algorithms re-late the black body temperature detected by thesatellite to buoy bulk temperature. The thermal in-frared radiation detected in bands 4 and 5 originatesfrom the surface skin of the water, whereas the co-efficients for the SST equations are found from re-gression of skin temperatures against bulk watertemperatures (0.5 to 2 m) of buoys. The equationsare designed to reduce water vapor effects on tem-perature (Walton et al. 1990), as the relation be-tween skin and bulk temperature is presumed to beconstant. Different equations are used for daytimeand nighttime images, assuring that the skin/bulk

TABLE 1. Channel characteristics of theAVHRR sensor (from Kidwell 1991).

Channel Wavelength (µm) Characteristics

1 0.58–0.68 reflected solar energy; (red) clouds, coastlines, and

vegetation

2 0.72–1.00 reflected solar energy;(near-infrared) clouds, coastlines, and

vegetation

3 3.55–3.93 reflected solar energy;(mid-infrared) thermal emission;

clouds and sea surfacetemperature

4 10.3–11.3 thermal emission;(thermal infrared) clouds and sea surface

temperature

5 11.5–12.5 thermal emission;(thermal infrared) clouds and sea surface

temperature

using an on-board black-body target and a view ofcloud cover. The temperature of the black-body,which is close to terrestrial temperatures, is moni-tored by four platinum thermisters. Thus, in princi-ple, the AVHRR sensor should provide uniformwater surface temperature measurements orbit afterorbit. The noise level in the two thermal infraredchannels is low, reported as 0.12 K for a 300 K tar-get, but in reality appears to be much lower (Min-nett 1990, Kidwell 1986). The data from eachchannel are collected in 10-bit resolution (1,024grey levels).

The AVHRR images came from the NOAACoastal Active Access System (NCAAS), an on-line archiving service provided through the NOAACoastWatch Program. As part of CoastWatch pre-processing, the images were georectified to a Mer-cator projection with pixel width off the KeweenawPeninsula of 1.34 km2, converted to degrees Celsiususing one of three multichannel sea surface temper-ature (MCSST) algorithms, and subset into 512 ×512 pixel datasets (Leshkevich et al. 1993). DEC-CON, a freeware program available through theCoastWatch Program, removed header informationcontained in the files and converted the images to aformat suitable for importing into the image pro-cessing software (Townsley and Stumpf 1996).


relationship is maintained, which allows for analy-sis of diurnal variations of surface bulk temperature(Walton et al. 1998, applications for CoastWatchimages, Pichel et al. 1995, Leshkevich et al. 1993).Only one of the three daytime algorithms pertain tothis study, i.e., OCNMAP, a linear algorithm fordaytime images and a nonlinear procedure that in-corporates improved atmospheric corrections fornighttime images (Leshkevich et al. 1993).

A primary source of in situ data allowed SST val-ues in the CoastWatch imagery to be comparedagainst bulk temperatures: NOAA National Data Buoy Center (NDBC) buoys. Seasonal differ-ences between AVHRR SST values and buoy bulktemperatures were investigated over the 1993 sea-son. Of 72 images obtained for study, twenty rela-tively “cloud free” scenes were selected to illustratea typical yearly cycle. The original and processedimages are on file with the Lake Superior Ecosys-tem Research Center, Michigan Technological University.

SPOT-HRV, TM, and RADARSAT ScanSAR Images

The High Resolution Visible (HRV) sensors ofSPOT-1 offer high spatial resolution (10 mpanchromatic, 20 m multispectral), but containbands only in the green, red, and near infraredwavelengths. However, the images are useful fordetecting and mapping distinct water masses andturbidity (Vande Castle et al. 1988). Bands 1 and 2(green and red wavelengths) generally are excellentfor picking up turbidity patterns. The Landsat The-matic Mapper (TM) sensor has 30 m pixels and 7spectral bands, capable of differentiating biotic(Chlorophyll a) and sediment signatures.

For imaging ice packs, RADARSAT-1 ScanSARdata were acquired from the Alaska SAR Facilityand the NOAA Satellite Active Archive. Nominalimage characteristics are a 500 km swath width, 50or 100 m pixel size, and 8 bit quantization ofbackscatter intensity. Because ScanSAR imagery isconstructed from multiple narrow beam modes, the

TABLE 2. IP procedure for temperature contour maps.

PRELIMINARY IMAGE PROCESSING:

NCAAS Retrieve NOAA CoastWatch images.DECCON Shareware program provided by NOAA that strips off header information and converts data

from 11-bit to 8-bit format.

ERDAS IMAGE PROCESSING:Step 1. Import AVHRR imagery.

LDDATA Import AVHRR image.BSTATS Generate simple statistics.

Step 2. Image Georeferencing.

CLASOVR Adjust image to CoastWatch base map.FIXHED Repeat steps A-C with adjustments.

Step 3. Isolating the Water Signature.

RECODE Use visible channels 1-2 reflectance data to create land and cloud mask. GISEDIT Edit out inland lakes and rivers.MASK Overlay mask on georeferenced image.

Step 4. SST Verification

DATATAB Select data points that match in situ data.

Step 5. Convert 8-bit satellite data to IDL format (FORTRAN program).

ERDTORAW Converts ERDAS binary datasets to ASCII datasets.

Step 6. Run IDL software to create contoured temperature maps.

.RUN TEMP IDL is a command language image presentation software program. A special program withinIDL was written to produce false gray-scale and color composites of AVHRR thermal data

648 Budd et al.

data are as yet uncalibrated. Hence, backscatter isrepresented as 8-bit digital numbers between 0 and255, rather than the more usual form of calibratedbackscatter expressed as a backscattering coeffi-cient, sigma naught. Image processing consisted ofinteractive contrast enhancement and speckle re-duction (image smoothing) using a 3 × 3 medianfilter.

Reflectance Spectra for Ontonagon Clays andStamp Sand Fine Fractions

Sediment refectance methods followed Bhargavaand Mariam (1991). Samples of kaolinite clay(Sigma) and sieved fractions of both OntonagonRiver bank clays and stamp sand samples were sus-pended in a large aquarium. The walls of the 50-gallon aquarium were spray-painted black tominimize internal reflectance. A 1,000 watt tung-sten halogen lamp, suspended overhead at a 45%angle to the water surface, provided illumination. Ahand-held narrow-slit (10°) spectroradiometer (An-alytical Spectral Devices, Boulder, Co.; model Per-sonal Radiometer) was held 1 meter above thewater surface, reading reflectance over a 16 cm di-ameter circle (202 cm2) on the surface of the slurry.Original amygdaloidal stamp sand tailing materialwas obtained from the Isle Royale piles and sievedinto particle size fractions. Kaolinite clay (control),Ontonagon clay, and stamp sand fine particle frac-tions were introduced in concentrations that rangedbetween 0 mg/L (control, background water spec-tra) to 1,280 mg/L. Readings were made for se-lected size fractions of Ontonagon clay (< 53,53–63, 63–75, 75–90 µm) and stamp sand(125–150, 150–180 µm). Clearly, further spectralwork is required, as iron oxides act in a size-spe-cific fashion, making very fine fractions of stampsands (so-called slime fraction) more reddish(amygdaloid) or purplish (conglomerate). Ηere re-sults from the < 53 µm Ontonagon clay and the 125to 150 µm stamp sand fractions are reported.

RESULTS

Seasonal Patterns At Buoy Stations

Data from NDBC buoys illustrate some generalcharacteristics of Lake Superior thermal patterns.Because of its vast size and depth, the mean tem-perature of Lake Superior lags behind the cyclicalvariation of both air and surface water temperature(Phillips 1978). Although its more northern latitudehas some influence, relative to the other Laurentian

Great Lakes, Superior warms and cools much moreslowly, attaining maximum surface temperaturesmuch later than other smaller, regional lakes, LakeMichigan, and especially the much shallower LakeErie (Fig. 3a).

The thermal inertia of Lake Superior is very evi-dent in the seasonal air temperature versus bulk sur-face temperature trends. Figures 3b and 3c show thedaily mean temperature trends for air temperatureand bulk surface temperature at an open-water buoystation in Lake Superior (buoy 45006; ChefswetBasin, off the Apostle Islands) and at a fixed shore-line station (ROAM4, Rock of Ages Lighthouse,Isle Royale). The curves are hourly water surfacetemperatures averaged during the interval from1984 through 1989. Heat storage lags in fall andspring are influenced by two factors: 1) large quan-tities of stored heat during summer, carrying overinto the fall, and 2) the formation and thawing ofice. Each acts to moderate air and surface tempera-ture cycles. The minimum surface water tempera-ture of less than 0.5°C occurs over a 6-week periodfrom late February through early April), whereasthe maximum surface water temperature is reachedbetween late July and early September, dependingupon the site. In open waters, the surface tempera-ture is lower than the mean lake temperature frommid-December through early May (Phillips 1978).With the onset of spring warming, the surface tem-perature of open water rises slowly from the end ofMarch to about 3 to 5°C in mid-June.

Serving as an introduction to the spatial hetero-geneity characteristic of seasonal warming, Figure4 compares warming and cooling trends at the threeoffshore buoys during a single season (1993). Theshallower western basin station (45006, Chefswet)shows more rapid spring warming, early (lateJuly/early August) and high maximum temperature,and relatively early decline of temperature in thefall. In contrast, the central, deepwater station(45001, Central) exhibits a delay in spring warmingthrough July, a low and late maximum temperatureand late fall cooling. After attaining surface max-ima, temperature decreases rapidly during Septem-ber and October at all stations, reaching 4°C byearly November. The deep offshore (45001, Cen-tral) station cools more slowly during Novemberand December than the shallower, nearshore sta-tions, again related to the heat storage effects. How-ever, Buoy 45004 (Stannard Rock, KeweenawBay), shows a curious pattern of early warming andtemperature irregularities, despite being in one ofthe deeper basins.


Correspondence Between Buoy And AVHRR Data Sets

It was found that the 1993 AVHRR SST tempera-tures reproduce the buoy surface bulk temperaturesclosely (Fig. 5a; r = 0.971) with only a slight nega-tive bias (< 1°C) at low temperatures. Seasonalplots of AVHRR MCSST temperatures versusNDBC buoy bulk temperatures (Fig. 5b) illustratethe very close absolute agreement between values.

Coastal Spatial Thermal Patterns (AVHRR)

The spatial pattern of seasonal warming andcooling for Lake Superior in 1993, created fromprocessed AVHRR imagery, is shown in the twoFigure 6 panels. Cooler temperatures are toward theblue, whereas warmer are toward the red. Land ismasked as black, while clouds are masked as white.Although the processed pixels are ca. 1.1 km square(1.34 km2 area), the resolution is more than ade-quate for resolving broad-scale patterns of surfaceskin temperatures, since the lake surface area isover 82,000 km2. For example, large-scale eddiesfrom the Keweenaw Current are evident in summerscenes, curling off the tip of the Peninsula, often asfar as Stannard Rock. For purposes of discussion,the temperature contour maps illustrate a typicalseasonal cycle that passes through six stages:

Mid-Winter Cooling Phase: Ice Maximum

From January to early March, surface tempera-tures are everywhere below 4°C. Freezing startsfrom the shallow and wind-protected margins andepisodically progresses lakeward. Ice cover reachesits maximum from February to early April (peaked

FIG. 3. Comparison of long-term temperatureaverages between different lakes and stations: a)contrast in seasonal daily bulk surface tempera-tures between open-water Lake Superior (45006,Chefswet Basin), Lake Michigan (45007 42.7N87.1W), and Lake Erie (45005 41.7N 82.4W)buoys, b) comparisons of mean daily air tempera-tures (+) with bulk surface (-) temperatures at twoLake Superior monitoring sites (Buoy 45006,Chefswet Basin; ROAM4, southern Isle Royale).Values averaged over 1984–1989.

FIG. 4. Bulk surface temperatures at NDBCLake Superior buoys (45001, Isle Royale Basin;45004, Caribou Basin; 45006, Chefswet Basin)during 1993.

650 Budd et al.

ca. 1 April 1993). Central upwelling (IJC 1977)keeps large areas ice-free most years, particularly inthe eastern basin.

Initial Stages of the Spring Warming Phase

During early March to April, the whole lake isstill below 4°C and portions remain covered withfloating ice (see RADARSAT ScanSAR section)during characteristic inverse stratification of thewater column. Ice lies along the periphery, particu-larly in the Apostle Islands, Thunder Bay, Ke-weenaw Bay, Whitefish Bay, and smallerembayments along the southern shoreline, keepingtemperatures near freezing. Warmer water (1.3°C)is located both in the mid-lake and along the north-west coast. By early April, the lake water column isessentially isothermal with a temperature of approx-imately 2°C (Jipping 1991). At the end of this pe-riod, the land responds more quickly to increasedinsolation and the warming of the atmosphere. Es-pecially along the warmer southern margin, riversand streams discharge large volumes of warmer wa-ters during the spring snow-melt period.

Thermal Bar Formation Stage of Warming Phase

During the initial stages of the warming phase,water along the shoreline heats faster than offshorewaters and also receives warmer river discharges.Along the southern shoreline, this phenomenon isespecially pronounced in shallow embayments (Du-luth basin, Chequamegon Bay, Keweenaw Bay,Whitefish Bay). Temperatures start to move above4°C in bay regions, whereas the middle portion ofthe lake remains below 4°C (5/25 to 6/28 scenes,Fig. 6)

Later June scenes (6/10 to 6/28 scenes, Fig. 6)show a definite temperature gradient developingalong the south shore. The average surface tempera-ture of the main body of the lake at that time is ap-proximately 2 to 3°C. Along the southern coastalmargins and the western arm of the lake, surfacetemperatures range from 5 to 11°C. Again the riseis pronounced in bays. The boundary between theinshore warm region and the isothermal offshorewaters is termed the “thermal bar” (Huang 1972,Ullman et al. 1998). The thermal bar position iseasily determined by the location of the 4°Cisotherm. The thermal bar acts to confine shorelinedischarges along the coast. This isotherm marks apredominately vertical boundary between watersfrom different sources, near the temperature of

FIG. 5. Verification of surface temperature datain Lake Superior: a) regression of 1993 AVHRRMCSST skin surface temperatures against NDBCbuoy bulk surface temperatures. See Methods sec-tion for discussion of procedures. Fit regressionequation and R2 values given in insert. b) Com-parison of 1993 AVHRR MCSST temperatureswith NDBC buoy bulk surface temperatures at thethree open-water buoy sites, plotted through time.


maximum density. The density causes a sinkingmotion along the offshore edge of the 4°C isotherm,and a type of isolation of flows on either side of the4°C isotherm. Waters of different color, marked dif-ferences in phytoplankton populations, and differ-ent flow patterns are commonly observed along the4°C isotherm (Ullman et al. 1998).

Sinking along the 4°C isotherm causes surfaceconvergence. The convergence at the thermal bar isthought to result in an upwelling in the middle por-tion of the lake during this period of time(Ragotzkie and Bratnick 1965, IJC 1977, Bennett1978). The upwelling to the surface of the offshorebasin provides water colder than the temperature ofmaximum density (4°C) until the disappearance ofthe thermal bar. The spring thermal bar period usu-ally lasts for 4 to 6 weeks. The shallow nearshorewater gradually warms toward the deeper offshoreareas during the course of summer.

During May to early July the horizontal tempera-ture difference between the edges of the lake and themiddle portion increases several fold. Shoreline wa-ters become much warmer than offshore waters andcoalesce to form a continuous southern corridor.Under the influence of prevailing westerlies associ-ated with barometric highs (Niebauer l976), coastalwaters form a major counterclockwise coastal cur-rent along the western shoreline of the KeweenawPeninsula. The Keweenaw Current (Ragotzkie l966,Smith l972, Smith and Ragotzkie 1970) compressesnear the tip of the peninsula, then jets with obviouseddy patterns into the Caribou Basin (6/28 to 7/11scenes, Fig. 6). SPOT and TM images capturecoastal sediments becoming entrained in the Ke-weenaw Current (Figs. 7a–b). These sediments are amixture of Ontonagon clays from the OntonagonRiver discharge, sandy beach sediments from weath-ering shoreline rock formations (Freda Sandstone,Copper Harbor Conglomerate), and stamp sand ma-terial broadcast along the shoreline from stampmills. From the Freda and Redridge shorelines northfor 23 km, waves and longshore currents work thedark sand beaches (stamp sands), suspending parti-cles that are entrained offshore into the KeweenawCurrent (Fig. 7b).

Summer Warming Phase: Keweenaw Current andOffshore Upwelling Followed by Thermocline Period

After the disappearance of the thermal bar, thelake surface warms above 4°C. During this time, thecentral and eastern basins are characterized by a rel-atively cool, central upwelling region that is charac-

teristic of Lake Superior. Along the shoreline,coastal upwelling regions may result from shifts inprevailing wind patterns, as barometric highs arefollowed by lows. Strong barometric lows can leadto prevailing southeasterly winds, causing upwellingalong the western coastline of the peninsula(Niebauer 1976). By late June or early July, offshorewaters in the western basin are much more homoge-neous and a weak thermocline begins to grow.

The July scenes (7/11 to 7/30) show that as theheat content increases in summer, the surface tem-perature gradient becomes stronger, ranging from15 to 19°C along the shoreline and from 4 to 7°C inthe central eastern basin. During this time, there isvery little change in the central part of the easternbasin, due to strong water density differences re-stricting mixing and to the development of mid-lakeupwelling.

Mid- to late-August (8/07 to 8/24) shows amajor increase in central, offshore surface tempera-tures. By late August (8/20 to 8/24), the tempera-ture ranges from 14°C at the central portion of thelake to 20°C toward the shoreline. The early to lateAugust scenes capture the transformation of thelarge isothermal midlake zone into a stratified re-gion. By early August through mid-September, thelake becomes strongly thermally stratified. That is,there is a sharply defined vertical temperature gra-dient everywhere across the lake. The nearly ho-mogeneous epilimnion is isolated from the coldhypolimnion.

The thermal maximum occurs in the first twoweeks of September. By mid- to late September(9/15 to 9/23 scenes), surface temperatures rapidlydecline. The average surface temperature of themain body of the lake is approximately 11°C. Up-welling is evident along the western coastline of theKeweenaw Peninsula on the September 15 image.

Early Fall Cooling Period

The lake begins to cool in late October to No-vember over its entire surface (Feit and Goldenberg1976, Assel 1985, 1986). By December, the watersbecome nearly homogeneous again, as the surfacewaters are above 4°C, but the edges are cooler thanthe center.

Late Fall/Early Winter Cooling Period: InvertedStratification

During this final stage, shoreline waters cool to4°C, then below 4°C. Rapid cooling in embayments

652 Budd et al.

FIG. 6. Seasonal development of the Lake Superior coastal zone as illustrated by AVHRRimages. Scenes illustrate the thermal bar interval in May and subsequent development of a strongKeweenaw Current by late June to early July. Note changes in temperature scale through thesequence of images.

654 Budd et al.

to 4°C leads to profile-bound density currents, asthe denser waters sink toward deeper portions of thebasin. Once below 4°C, ribbons of low-density,cool waters ring the shoreline (Smith 1973).Whereas inverted temperature regimes should leadto formation of an autumn thermal bar, the phenom-enon is not as pronounced as its spring counterpartlargely due to great disturbance by early winterstorms. The temperature difference between theedges and the main body of water is only 4°C,much smaller than the gradient during the springwarming phase, and sometimes barely recognizable.Soon ice formation begins along certain (shallow,protected) shorelines.

Full winter coverage of Lake Superior byAVHRR imagery is extremely difficult to obtain be-cause of extensive cloud cover. However,RADARSAT ScanSAR uses 5.6 cm wavelengthsthat can penetrate through cloud cover and darknessto reveal drifting ice and open water regions. Maxi-mum ice cover does not occur until mid-March(Phillips 1978).

Reflectance Spectra for Ontonagon Clays and Stamp Sands

SPOT and TM imagery permits zooming in oncoastal Keweenaw waters to check for suspendedparticles. Entrained particles seen in SPOT and TMimages are a mixture of particles from varioussource materials: clay and silt-sized particles dis-charged from the Ontonagon River, which erodesprevious lake-bed deposits along its banks, re-worked clay, silt, and sand from weathering shore-line rock outcrops (Freda Sandstone, CopperHarbor Conglomerate) and reworked stamp sands.The stamp sands were discharged from five largemills at Freda and Redridge, plus smaller mills be-tween Eagle River and Eagle Harbor. In this high-energy environment, silt-sized and clay-sizedparticles are frequently seen during spring riverhigh-discharge periods, after storms, and in the fallhigh-wave period.

Reflectance spectra for pure kaolinite clay, On-tonagon clay, and stamp sands are shown in Figure8. The different spectra suggest that Ontonagonand stamp sand components may be separated fromeach other by use of multi-band satellites such asLandsat TM or the Sea-viewing Wide Field-of-View Sensor (SeaWiFS), potentially allowing esti-mates of suspended solids for each type of sourcematerial.

FIG. 7. Movement of stamp sands and entrain-ment by the Keweenaw Current. SPOT-HRVpanchromatic image acquired on September 24,1992 illustrating a) drifting of beach sands formedfrom stamp sands and suspended particles fromthe Freda to Redridge northeastward to the NorthEntry region; b) suspended sediments along theentire Keweenaw coastline.

A.

B.


RADARSAT ScanSAR Images of Ice Rafting

The Keweenaw Current is not the only transportoption operating along the Keweenaw Peninsula.Ground studies and aerial observations of Lake Su-perior and other Great Lakes show that nearshoreice may incorporate large sediment loads (primarilyshoreline sand, with subordinate pebbles, cobbles,but also organic-rich deeper sediments). The sedi-ments are concentrated in nearshore ice facies by avariety of mechanisms, principally becausenearshore regions are most directly exposed to sedi-ment supply and entrainment (Barnes et al. 1994).

Sediments are subsequently transported to midlakeareas via ice rafting. Here ground-based and remotesensing studies were used to document sedimententrainment mechanisms around the KeweenawPeninsula. A combination of ground photos andaerial photos show fine details, whereas syntheticaperature radar (SAR) images illustrate large-scaleice pack features.

Those who venture out on the lake during winterknow that shoreline ice is “dirty.” A typicalnearshore ice complex (NIC) consists of agrounded ice foot, one or more shore-parallelridges (with coalescent ice-volcanoes), and inter-vening shore-parallel ice lagoons (Fig. 9). The fea-tures of a typical NIC are described in detail byBarnes et al. (1994), O’hara and Ayers (1972), andMarsh et al. (1973). The NIC remains shorefast formuch of the winter, although it may become activewhenever directly exposed to waves. The outeredge of the NIC is a boundary with mobile ice fa-cies, notably 1) a belt of brash and slush ice, and 2)a band of drifting pack ice. NIC zones form as ashore-parallel grounded ridge of ice constructedthrough advection of slush ice, water, and sedi-ments, primarily during winter storms. The icefootis a product of episodic pulses under conditions ofhigh-energy waves and concurrent snowfall thatproduces copious amounts of slush. Initial growthis an early season phenomenon, prior to thenearshore zone becoming inaccessible to large,deepwater waves. If snow is present in sufficientamounts, a belt of slush floats lakeward of the ice-foot edge. Under stormy conditions, large waveschurn the water into a turbid brown, due to sus-pended sediments and lumps of slush ice. Wavesincorporate sediment into the floating slush, whichin turn provides a mechanism for lifting sedimentsonto the ice foot. The high surface to volume ratioof slush increases its effectiveness as a sand andpebble lifting agent.

Sediment entrainment mechanisms include thefollowing:

1. Waves entrain sediments into the water col-umn and floating slush;

2. Waves deposit entrained sediments (directlyfrom water or as slush) on top of solid ice(floes and ice foot);

3. Ice keels scour and entrain sediments (diversusing SCUBA report that ice keels may reach20 to 30 m depth);

4. Sediment-laden anchor ice detaches from lakebottom;

FIG. 8. Comparison of reflectance spectra frompure kaolinite, Ontonagon, and stamp sand finefractions.

656 Budd et al.

5. Sediment-laden river ice enters lake;6. Sediment-laden river water flows onto river or

lake ice (aufeis);7. Cliff/bluff sediments fall onto ice surface;8. Wind-blown (eolian) deposition.

Based on our field experience in Lake Superior, thefirst four mechanisms are probably the most signifi-cant in terms of bulk mass of sediments incorpo-rated into NIC zones.

Three main environmental factors facilitate thesediment entrainment-transport process: 1) shallowbathymetry (2 m or greater waves are common inwinter storms), 2) abundant source materials (un-consolidated sediments), and 3) strong onshorewinds and waves. The Keweenaw coast immedi-ately off McLain State Park, at the North Entry to

the Keweenaw Waterway, exemplifies these condi-tions (Fig. 10). A mixture of natural beach sandsand stamp sands coat the melting ice foot in earlyApril, 1997. Sandy layers incorporated into the icefoot are exposed in broken cross sections.

Late winter RADARSAT ScanSAR images ofLake Superior from March and April, 1997, showthe NIC as a bright band along the coast from On-tonagon to Keweenaw Point (Fig 11). Pack ice isoriginally consolidated, then surges back and forth,and drifts past the tip of the peninsula under the in-fluence of strong westerly winds. Images on 11March 1997 document the pack ice drifting pastCopper Harbor and Manitou Island (Fig. 12).Clearly, particle transport via ice rafting is real, yetthe mass transport details require further research.

DISCUSSION

If used in isolation, NDBC buoy and CMANfixed stations provide detailed temporal tempera-ture records at scattered sites, but little informationon spatial patterns. However, when verified by thetemporal point data, AVHRR spatial imagery allowsreconstruction of large-scale Lake Superior surfacethermal patterns. The spatial results are striking intheir ability to capture large-scale details of thecoastal zone. Moreover, the data are truly synopticand the errors are relatively small.

Satellite imagery has been available for decadesin the U.S., yet relatively few individuals and insti-tutions have utilized this source for analysis of sur-face water temperatures in the Great Lakes. Theprimary difficulty has been the expense of purchas-ing scenes and the software necessary for process-ing satellite imagery. Exceptions include Sabatini’s(1971) use of data from Nimbus 1, 2, and 3,Strong’s (l974) use of data from the NOAA 2 satel-lite, Lathrop and Lillesand’s (1986, 1987) use ofThematic Mapper (TM) data, and Lathrop et al.’s(1990) use of NOAA 9 infrared data. Vande Castleet al. (1988) strongly advocated use of TM, SPOT-1, and AVHRR in water quality studies. Other in-vestigations not dealing primarily withtemperature, but utilizing this information, includeStrong and Eadie’s (1978) use of imagery fromLandsat, the NOAA polar orbiting satellites, andthe Skylab Earth Terrain Camera system to investi-gate lake surface temperature and calcium precipi-tation patterns, plus Mortimer’s (1988) use ofCoastal Zone Color Scanner (CZCS) data to test avariety of hypotheses about chlorophyll a andupwellings.

FIG. 9. Oblique aerial photograph of NIC(nearshore ice complex) and sediment-covered iceridges. As ice ablates in spring, sediments are con-centrated at the ice surface. Each ridge representsan episode of ice formation through advection ofsand-bearing slush ice and spray.


FIG. 11. RADARSAT ScanSAR image of Lake Superior (April 1, 1997). Nearshoreice on windward coasts is the dominant sediment accumulation facies. Offshore icefacies are less likely to entrain sediments. The NIC appears as a narrow bright bandalong the coast from Ontonagon to Keweenaw Point. ScanSAR Wide B image, pixelsize = 100 m, swath width = 500 km. Copyright Canadian Space Agency, 1997.

FIG. 10. In situ photos of sediments inicefoot: A) sand ablation cones on surfaceof eroding icefoot, west McLain State Park,MI, April 1997, and B) beach sand/stampsand mixture in vertical cross-section oficefoot (matchbook for scale). Cone struc-tures form through differential ice ablation.

658 Budd et al.

In the regression analysis of AVHRR SST versusbuoy bulk temperature, errors of a fraction of a de-gree are not unreasonable. In fact, the global SSThas a standard error of estimate about 0.5 to 0.7 de-gree. The Lake Superior comparison in this studycame close to this, except for the bias at low tem-perature. The errors can be attributed to atmos-pheric variability (0.1 to 0.2 K; Kidwell 1986),noise in the radiometric satellite measurements(0.12 K; Kidwell 1986), errors in the buoy data (0.1K; Verity 1988), the difference between bulk andskin temperatures (0.1 to 0.5 K; McClain etal.1985a,b; Schluessel et al. 1987; Schluessel et al.1990; Minnett 1990), and cooling or warming ef-fects associated with diurnal variation. Anotherpossible source of error is the uncertainty intro-

duced by spatial and temporal variability (Schlues-sel et al. 1987, Minnett 1990). That is, the satellitemeasurement is a near-instantaneous spatial aver-age, whereas the buoy measurement is a temporalaverage at a given sample location. The error in-duced by this aspect probably is not very largesince the AVHRR surface temperature seldomshows gradients of 0.5°C per pixel (Schluessel etal. 1987).

One possibility for the low bias, suggested by R.Stumpf (personal communication), is that the globalanalysis is made over the open ocean, where airtemperatures are almost certainly higher when thewater is below 5°C. The water vapor correctionmay be over-compensating for consistently lowerwater vapor content over Lake Superior.

FIG. 12. ScanSAR imageof A) Keweenaw Peninsularegion, (March 11, 1997),with close-up B) showingice drifting from NW to SEpast Copper Harbor (CH)and Manitou Island (MI).ScanSAR Wide B image,pixel size = 50 m. CopyrightCanadian Space Agency,1997. A


Current limitations are primarily the 1 km2 reso-lution of the AVHRR imagery and the fact thatAVHRR senses only the surface temperature (notrevealing internal structure). True variations on adiurnal basis produce about 0.5 K differences be-tween day and night temperatures (Bernstein 1982).At present, given minimal cloud coverage, satellitemonitoring could potentially produce four scenesper day. Buoy data, taken at hourly intervals, couldbe utilized for general interpolation of diurnal cycles.

Temperature contouring reveals several specialfeatures of Lake Superior. The first is the long lagtime in summer warming, the second is the thermalbar and Keweenaw Current phase, and the third isthe large central upwelling region and other more-localized coastal upwelling zones associated withstrong westerlies or southeastern winds. Thenearshore coastal features constrain and distributewarm waters. On one hand, offshore waters remaincooler longer than in other Laurentian Great Lakesand smaller regional lakes, whereas the KeweenawCurrent is also responsible for broadcasting largevolumes of coastal waters and entrained materialsacross some of the deepest regions in Lake Superior(Caribou Basin).

Computer contouring derived from correctedAVHRR imagery gives a more instantaneous andcompletely objective view of spatial temperatureisoclines than previous methods. The general pat-terns of surface water temperature based onAVHRR data are very similar to those foundthrough ship (Webb 1975) and ART surveys, yetdetails of thermal gradients and large-scale eddystructure are much more evident in the AVHRR-based isoclines. For example, summer surface watercirculation patterns are dominated by a steepnearshore-offshore temperature gradient and acounter-clockwise nearshore current system(Adams 1970, Smith and Ragotzkie 1970, Sloss andSaylor 1976). In the AVHRR-derived images, thesteep shoreline temperature gradients are very evi-dent and irregularities in contours capture large-scale eddy structure at several locations (off the tipof the Keweenaw Penninsula). Planned studies ofthe Keweenaw Current using drogues and drifters(NOAA/NSF KITES Project) will help quantifycurrents off the tip of the Peninsula and across theCaribou Basin.

Determining the amount of sediment transportedin this system will be challenging, requiring a com-bination of current meters, sediment traps, and

water sampling. Wind-driven, shore-parallel coastaljets which lie within 10 kilometers of the shorelinehave been documented in all of the Great Lakes(Murthy and Dunbar 1981). However, direction re-versals, peculiar to closed basins, have not been ob-served thus far in Lake Superior, presumablybecause the lake circumference is so large (Viek-man and Wimbush 1993). Maximum current veloci-ties are reached in the summer stratified periodwhen stratification hinders the ability of wind tomix the water column. However, high sedimentloads come in the spring and late fall. For example,the Ontonagon River contributes nearly 25% of thetotal riverborne sediment load to the lake (Kemp etal. 1978). Yet surface transport of clay-sized parti-cles from this source can best be seen in May, whenthe warmer river waters are so buoyant that the sed-iment-laden strata move over cooler coastal waters.In mid-summer, when temperatures of river andcoastal waters are similar, the clay-laden watersplunge below shoreline strata (Strong 1995).

Estimating the amount of sediment transportedby ice rafting will be equally challenging. Althoughthe preliminary observations reported here docu-ment substantial sediment loads, the amount and di-rection of ice transport has not been systematicallydocumented, nor the sediment loads and ice wastingprocess. A combination of ice coring and ice packdrift tracing by RADARSAT imagery might accom-plish this task.

AVHRR imagery, supplemented by higher resolu-tion SPOT-HRV, LANDSAT TM, or RADARSATscenes, supplies the amount of spatial coverage andreal-time detail necessary to monitor dynamicchanges associated with many meso-scale lake phe-nomena. In the future, application of AVHRR andthe recently launched SeaWiFS instrument to docu-ment thermal bar formation, coastal jet and large-scale eddy mixing, and upwelling events holdsexciting promise.

ACKNOWLEDGMENTS

Image acquisition and processing was supportedby grants from the Office of Sea Grant, NationalOceanic and Atmospheric Administration (NOAA),U.S. Department of Commerce, SG-1993-08, To JBand WCK through the Cooperative Institute forLimnology Ecosystems Research (CILER), Michi-gan Sea Grant College Program, Fund RA0547-MICHI; and the National Aeronautics and SpaceAdministration (Grant #NRA 94-OSS-16). Addi-tionally, we thank Michigan Great Lakes Protection

660 Budd et al.

Fund grants #910417 and #090192 to WCK andJ.A.Robbins, and separate portions of NSF OCE97-12827 to JB and WCK. Research ExcellenceFunds (REF), provided by the state to the Lake Su-perior Ecosystem Research Center (LaSER), sup-ported hardware, software, and image storagecapability. Radar and ice studies were supported byNASA Grant NAG5-7051 to DP under theRADARSAT ADRO program. Additional supportcame from an Apostle Islands National Seashoregrant to WCK. A Research Excellence Fund (REF)grant from the State of Michigan to MTU supportedLJ throughout her Master’s project. The NDBCbuoy and CMAN fixed station data also were pur-chased by LaSER. George Leshkevitch and DavidSchwab (NOAA/GLERL) aided us with NOAACoastWatch program connections, whereas R.Stumpf offered valuable suggestions. We thankSarah Green (MTU) and R. Stumpf (USGS,NOAA) for editorial comments. Radarsat SAR im-agery and support for ice field work were providedby NASA Radarsat ADRO Grant #364.

REFERENCESAdams, C.E. Jr. l970. Summer circulation in western

Lake Superior. In Proc. 13th Conf. Great Lakes Res.,pp. 862–879. Intern. Assoc. Great Lakes Res.

Assel, R.A. l985. Lake Superior cooling season tempera-ture climatology. NOAA TM ERL GLERL-58. GreatLakes Environmental Research Laboratory, 2300Washtenaw, Ann Arbor, MI 48104.

———. l986. Fall and winter thermal structure of LakeSuperior. J. Great Lakes Res. 12(4):251–262.

Barnes, P.W., Kempema, E.W., Reimnitz, E., andMcCormick, M. 1994. The influence of ice on south-ern Lake Michigan coastal erosion. J. Great LakesRes. 20(1):179–195.

Barton, I.J., Zavody, A.M., O’Brien, D.M., Cutten, D.R.,Saunders, R.W., and Lluwellyn-Jones, D.T. 1989.Theoretical algorithms for satellite-derived sea sur-face temperatures. J. Geophys. Res. 94:3365–3375.

Beeton, A.M., Johnson, J.H., and Smith, S.H. 1959. LakeSuperior limnological data 1951–1957. United StatesFish and Wildlife Service, Special Scientific Report,Fisheries No. 297.

Bennett, E.B. l978. Characteristics of the thermal regimeof Lake Superior. J. Great Lakes Res. 4:310–319.

Bernstein, R.L. 1982. Sea surface temperature estimationusing the NOAA-6 satellite advanced very high reso-lution radiometer. J. Geophys. Res. 87:9455–9465.

Bhargava, D.S., and Mariam, D.W. 1991. Effects of sus-pended particle size and concentration on reflectancemeasurements. Photogramm. Eng. Remote Sens.57(5):519–529.

Environment Canada and U.S. Environmental ProtectionAgency (EC & USEPA). 1987. The Great Lakes: AnEnvironmental Atlas and Resource Book. ChicagoUSEPA Great Lakes National Program Office, andToronto: Environment Canada, Ontario Region.

Feit, D.M., and Goldenberg, D.S. l976. Climatology ofsurface temperatures of Lakes Superior, Huron, Erie,and Ontario. TDL Office Note 76-16, NationalWeather Service, Systems DEvelopment Office,Techniques Development Laboratory, Silver Spring,MD 20233.

Grassl, H. 1976. The dependence of the measured coolskin of the ocean on wind stress and total heat flux.Boundary Layer Meteorol. 10:465–474.

Green, T., and Terrell, R. E. 1980. The surface tempera-ture structure associated with the Keweenaw Currentin Lake Superior. J. Geophys. Res. 83:419–426.

Harrington, M. S. 1895. Currents in the Great Lakes.U.S. Weather Bur., Bull. B, Washington, D.C.

Huang, J.C.K. 1972. The thermal bar. Geophysical FluidDynamics 3:83–107.

International Joint Commission (IJC). 1977. Waters ofLake Huron and Lake Superior, vol. III part B: LakeSuperior. ed. Upper Lakes Reference Group. pp.249–275. Windsor, Ontario.

Irbe, J.G. l972a. Aerial surveys of Great Lakes watertemperatures, April 1968 to March 1970. ClimatologyDivision, Atmospheric Environment Service,Downsview, ON.

———. 1972b. An operational program for measuringsurface water temperature by airborn radiation ther-mometer [ART] survey. In Proceedings, First Cana-dian Symposium on Remote Sensing, pp. 183–200.Atmospheric Environment Service, Department of theEnvironment, Downsview, Ontario.

Jipping, L. M. 1991. Advanced very high resolutionradiometer (AVHRR) imagery applied to Lake Supe-rior thermal structure. Ms. thesis, Michigan Techno-logical University,

Kemp, A.L., Williams, J.D.H., Thomas, R.L., and Gre-gory, M.L. 1978. Impact of man’s activities on thechemical composition of the sediments of LakesSuperior and Huron. Water, Air, and Soil Pollution10:381–402.

Kerfoot, W.C., and Robbins, J.A. 1999. Nearshoreregions of Lake Superior: Multi-element signatures ofmining discharges and a test of Pb-210 depositionunder conditions of variable sediment mass flux. J.Great Lakes Res. 25:697–720.

———, Lauster, G., and Robbins, J.A. 1994. Paleolim-nological study of copper mining around Lake Supe-rior: Artificial varves from Portage Lake provide ahigh resolution record. Limnol. Oceanogr.39(3):649–669.

Kidwell, K.B. 1986. NOAA Polar Orbiter Data UsersGuide. National Oceanic and Atmospheric Adminis-tration, Washington, D.C.


———. 1991. NOAA Polar Orbiter Data Users Guide.National Oceanic and Atmospheric Administration,Washington, D.C.

Kraft, K.J. 1979. Pontoporeia distribution along theKeweenaw shore of Lake Superior affected by coppertailings. J. Great Lakes Res. 5:28–35.

Lathrop, R.G., and Lillesand, T.M. 1986. Use of The-matic Mapper Data to assess water quality in GreenBay and Central Lake Michigan. Photogrammet. Eng.Rem. Sens. 52 (5):671–680.

———, and Lillesand, T. M. 1987. Calibration of the-matic mapper thermal data for water surface tempera-ture mapping: case study on the Great Lakes. RemoteSensing Environ. 22:297–307.

———, Vande Castle, J.R., and Lillesand, T.M. 1990.Monitoring river plume transport and mesoscale cir-culation in Green Bay, Lake Michigan through satel-lite remote sensing. J. Great Lakes Res. 16:471–484.

LeDrew, E.F., and Franklin, S.E. 1985. The use of ther-mal infrared imagery in surface current analysis of asmall lake. Photogrammet. Eng. Rem. Sens. 51(5):565–573.

Leshkevich, G.A., Schwab, D.J., and Muhr, G.C. 1993.Satellite environmental monitoring of the GreatLakes: A review of NOAA’s Great Lakes CoastWatchprogram. Photogrammet. Eng. Rem. Sens. 59(3)371–380.

Marsh, W.M., Marsh, B.D., and Dozier, J. 1973. Forma-tion, structure, and geomorphic influence of LakeSuperior icefoots. Amer. J. Sci. 273:38–64.

McClain, E. P., Pichel, W.G., and Walton, C.C. 1985a.Comparative performance of AVHRR-based multi-channel sea surface temperature. J. Geophys. Res.90:11,587–11,601.

———, Pichel, W.G., and Walton, C.C.. 1985b. Multipleatmospheric-window techniques for satellite-derivedsea surface temperatures. J. Geophys. Res (90)C6:11,587–11,601.

McMillin, L.M. 1975. Estimation of sea surface tempera-tures from two infrared window measurements withdifferent absorption. J. Geophys. Res. 80(36):5113–5517.

———, and Crosby, D.S. 1984. Theory and validation ofthe multiple window sea surface temperature tech-nique. J. Geophys. Res. 89(C3):3655–3661.

Minnett, P.J. 1990. The regional optimization of infraredmeasurements of sea surface temperature from space.J. Geophys. Res. 95:13,497–13,510.

Mortimer, C.H. 1988. Discoveries and testable hypothe-ses arising from Coastal Zone Color Scanner imageryof southern Lake Michigan. Limnol. Oceanogr.33:203–226.

Murthy, C.R., and Dunbar, D.S. 1981. Structure of theflow within the Coastal Boundary Layer of the GreatLakes. J. Great Lakes Res. 11:1567–1577.

Niebauer, H.J. l976. Wind driven coastal upwelling in

Lake Superior. Ph.D. thesis, University of Wisconsin-Madison.

O’Hara, N.M., and Ayers, J.C. 1972. Stages of shore icedevelopment. In Proc. 15th Conf. Great Lakes Res.,pp. 521–535. Internat. Assoc. Great Lakes Res.

Paulson, C.A., and J.J. Simpson. 1981. The temperaturedifference across the cool skin of the ocean. J. Geo-phys. Res. 86(C11):11,044–11,054

Phillips, D.W. 1978. Environmental climatology of LakeSuperior. J. Great Lakes Res. 4(3–4):288–309.

Pichel, W., Clemente-Colon, P., Sapper, J., Duda, C.,and Krasnopolsky,V. 1995. Validation and qualitycontrol of CoastWatch satellite-derived water-surface-temperature imagery. In Proceedings of the ThirdThematic Conference on Remote Sensing for Marineand Coastal Environments. Seattle WA, ERIM, Sep-tember 18–20.

Pickett, R.L. 1980. Observed and predicted Great Lakeswinter circulations. J. Phys. Oceanogr. 10:1140–1145.

Planet, W.G. 1988. Data Extraction and Calibration ofTIROS-N/NOAA Radiometers. NOAA Tech. MemoNESS 107-Rev. 1, Washington, D.C.

Ragotzkie, R.A. l966. The Keweenaw Current, a regularfeature of summer circulation of Lake Superior. Tech.Rept. No. 29. Univ. of Wisconsin-Madison, Dept. ofMeteorology.

———. l974. Vertical motions along the northern shoreof Lake Superior. In Proc. 17th Conf. Great LakesRes., pp. 456–461. Internat. Assoc. Great Lakes Res.

———, and Bratnick, M. l965. Infrared temperature pat-terns on Lake Superior and inferred vertical motions.In Proc. Eighth Conf. Great Lakes Res., pp. 349–357.Internat. Assoc. Great Lakes Res.

Sabatinti, R.R. 1971. The feasibility of using NimbusHRIR and THIR measurements for obtaining surfacetemperatures of Lake Michigan. In Proc. 14th Conf.Great Lakes Res., pp. 487–494. Internat. Assoc. GreatLakes Res.

Saunders, P.M. 1967. The temperature at the ocean-airinterface. J. Atmos. Sci. 24:269–273.

Schluessel, P., Shin, H.Y., Emery, W.J., and Grassl, H.l987. Comparison of satellite-derived sea surface tem-perature with in situ skin measurements. J. Geophys.Res. 92:2859–2874.

———, Emery W.J., Grassl, H., and Mammen, T. 1990.On the bulk-skin temperature difference and itsimpact on satellite remote sensing of sea surface tem-perature. J. Geophys. Res. 95:13,341–13,356.

Sloss, P.W., and Saylor, J. H. 1976. Large-scale currentmeasurements in Lake Superior. Techn. Rep.N.O.A.A. ERL 363-GLERL 8.

Smith, N. l972. Summertime temperatures and circula-tion patterns in Lake Superior. Ph.D. thesis, Univ. ofWisconsin-Madison.

———. 1973. The fall overturn and winter cooling inLake Superior. Limnol. Oceanogr. 18(3):483–487.

662 Budd et al.

———. l974. A note on spring warming in Lake Supe-rior. Water Res. Bull. American Water ResourcesAssoc. Vol. 10. No. 5.

———, and Ragotzkie, R.A.. 1970. A comparison ofcomputed and measured currents in Lake Superior. InProc. 13 Conf. Great Lakes Res., pp. 670–675. Inter-nat. Assoc. Great Lakes Res.

Spain, J.D., Wernert, G.M., and Hubbard, D.W. l975.The structure of the spring thermal bar in Lake Supe-rior, II. 18 Conf. Great Lakes Res., Albany.

Strong, A.E. 1974. Great Lakes temperature maps bysatellite (IFYGL). In Proc. 17th Conf. Great LakesRes., pp. 321–333. Internat. Assoc. Great Lakes Res.

———, and Eadie, B.J. 1978. Satellite observations ofcalcium carbonate precipitations in the Great Lakes.Limnol. Oceanogr. 23(5):877–887.

Strong, M.L. 1995. Use of HIS transformation withLANDSAT Thematic Mapper (TM) imagery fordetection of suspended sediments and sediment load-ing in freshwater systems. Ms. thesis, Forestry andWood Products, Michigan Technological University,Houghton, MI.

Stumpf, R.P. 1992. Remote sensing of water clarity andsuspended sediments in coastal waters. In Proc. 1stThematic Conf. on Rem. Sens. for Marine and CoastalEnviron., pp. 293–305. New Orleans, LA.

Townsley, B., and Stumpf, R.P. 1996. DECCON Soft-ware to Decompress NOAA IMGMAP CompressedImagery and Convert to Common Image FileFormats. U.S. Geological Survey, Center for CoastalGeology, St. Petersburg, FL.

Ullman, D., Brown, J., Cornillion, P., and Mavor, T.1998. Surface temperature fronts in the Great Lakes.J. Great Lakes Res. 24(4):753–775.

Vande Castle, J.R., Lathrop Jr., R.G., and Lillesand,T.M. 1988. The significance of GIS and remote sens-ing technology in Great Lakes monitoring andresource management. In The Great Lakes: Livingwith North America’s Inland Waters, ed. D.H. Hick-cox, pp. 155–161. American Water Resources Associ-ation, Bethesda, Maryland.

Verity, C.M. 1988. Coastal-marine automated network(C-MAN) users guide. National Data Buoy Center,Stennis Space Center, MS.

Viekman, B.E., and Wimbush, M. 1993. Observations ofthe vertical structure of the Keweenaw Current, LakeSuperior. J. Great Lakes Res. 19:470–479.

Walton, C.C., McClain, E.P., and Sapper, J.F. 1990.Recent changes in satellite-based multi-channel seasurface temperature algorithms. Presented at Scienceand Technology for and New Ocean Decade. MTS 90.Mar. Technol. Soc., Washington, D.C.

———, Sullivan, J.T., Rao, C.R.N., and Weinreb. 1998.Corrections for detector nonlinearities and calibrationinconsistencies of the infrared channels of theadvanced very high resolution radiometer. J. Geo-phys. Res. 103(C2):3323–3337.

Webb, M.S. 1975. A surface water temperature climatology of the Laurentian Great Lakes. Climato-logical Studies No. 27, Atmospheric EnvironmentService.

Wick, G.A., and Emery, W.J. 1992. A comprehensivecomparison between satellite-measured skin and mul-tichannel sea surface temperature. J. Geophys. Res. 97(C4):5569–5595.

Submitted: 20 July 1998Accepted: 28 August 1999

the keweenaw current and ice rafting: use of satellite...

Documents