JACK N. RINKER

U. S. Army Engineer Topographic LaboratoriesFort Belvoir, VA 22060

Airborne Infrared ThermalDetection of Caves andCrevasses*Under certain conditions crevasses in Greenland and caveopenings in Puerto Rico were detected with airborne infraredthermal scanners.

INTRODUCTION

A s with many ventures, the thermal de­tection of subsurface voids is easier in

principle than it is in practice. In order to useairborne infrared thermal sensing techniqueson other than a trial and error basis, it isnecessary to have information about the fol­lowing items: the physical basis for the exis-

The results reported here are portions of tworesearch projects: crevasse detection andcave detection.

The efforts in crevasse detection weredone in the late 1950's and early 60's whenthe author was on the staff of the U. S. ArmySnow, Ice and Permafrost Research Estab­lishment (USASIPRE), Wilmette, Illinois. In

ABSTRACT: Airborne infrared thermal scanners can be used to detectcrevasses and cave openings, but only under certain conditions. First,the temperature inside the void must be significantly different fromexternal conditions; and second, some mechanism must exist tobring this thermal difference to the surface where it can be detectedby a scanner. Furthermore, it must be determined if other eventsinfluence this mechanism. In the case of crevasses, conduction andconvection both playa role in altering the surface temperature of thesnow bridge over a crevasse. For caves, convection is the mechanismthat brings about the temperature alteration. Convection is linked tothe breathing cycle which, in turn, is caused by changes in atmos­pheric pressure. From ground measurements ofinternal temperature,external temperature, and atmospheric pressure a flight time can bepicked that will provide the most favorable circumstances. The cavesignal is more of a problem because it is frequently surrounded bysimilar looking signals caused by other events. Results are given for acrevasse field in Greenland and for a cave system in Puerto Rico.

tence of a "detectable signal", its magnitudein the geometric sense as well as in terms ofradiant energy, the dependence of this signalon other natural phenomena, and the varia­tion of the signal with the passage of time.

* Presented at the Fall Meeting, AmericanSociety of Photogrammetry, Washington, DC,September 1974.

the early 60's, this laboratory became part ofthe U. S. Army Cold Regions Research andEngineering Laboratory (USACRREL),Hanover, New Hampshire. The crevassestudies and airborne thermal sensing tookplace on the Greenland ice cap and were ajoint effort between the group atUSASIPRE/USACRREL and the InfraredLaboratory, Institute of Science and

PHOTOGRAMMETRIC ENGINEERING AND REMOTE SENSING,

Vol. 41, No. 11, November 1975, pp. 1391-1400.1391

1392 PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING, 1975

Technology, The University of Michigan.These early efforts in the use airborne in­frared thermal sensing techniques for cre­vasse detection were first reported in theclassified literature. Since then, the imageryhas been declassified and is presented hereas illustrative material associated with thegeneral problem of detecting subsurfacevoids.

The cave detection results are from a por­tion of a research program concerned withremote sensing and environmental analysiswithin a study area in the Commonwealth ofPuerto Rico. The field work and data collec­tion were done in 1970, when the author wason the staff of USACRREL, Hanover, NewHampshire. t The airborne thermal missionswere done by the U. S. Air Force, Rome AirDevelopment Center, under th~ jurisdictionof Mr. A. Stringham, Chief of the Reconnais­sance and Mapping Branch. Although thesetwo efforts may seem unrelated, they are,from the physical point of view, similar prob­lems and are therefore combined in this pre­sentation.

BASIS FOR DETECTION

All matter at a temperature above absolutezero (00 K or -2730 C) emits electromagneticenergy. Ifthe matter is hot enough, e.g., a redhot stove pipe, a tungsten light bulb, the sun,etc., it emits sufficient energy to influencethe eye or a photographic emulsion. How­ever, at normal earth surface temperatures,Le., -500 C to +500 C, the amount of energyemitted is far below the threshold levels ofeither a photo emulsion or an eye. Thus, todetect such energy, special materials areused that are not only sensitive to infraredradiation but which have some property, suchas electrical resistance, that changes rapidlyand significantly with minute variations inthe incoming signal. When working withenergy levels associated with the tempera­ture range of -500 C to +500 C, a detectormost sensitive to infrared energy withwavelengths of8 to 14 micra is normally used.This is so for two reasons: at such tempera­tures materials emit more energy at thesewavelengths than at any others, and the at­mosphere is relatively transparent to radia­tion of these wavelengths. The signal fromthe detector is amplified and then displayedon a cathode ray tube as well as recorded onmagnetic tape, or directly on film via a mod­ulated glow tube or a cathode ray tube trace.By convention, the gray tones are printed so

t This work was supported by the U. S. ArmyExploratory Development Program on Barriersand Counterbarriers.

that light tones indicate higher radiationlevels, or warmer areas, and darker tones rep­resent cooler areas.

In order to detect a given object, it musthave a radiation characteristic that differsfrom its surroundings, Le., a different tem­perature or a different emissivity, or somecombination of the two. The term emissivitydescribes the level at which a given materialabsorbs and emits energy at any selectedtemperature for the wavelength of interest.Materials emit energy at those wavelengthsat which they absorb. A material with anemissivity of unity is a perfect absorber, re­flecting none of the radiation striking its sur­face; it emits more energy than a materialwith a lesser emissivity at the same physicaltemperature. Emissivity is a function of themolecular species, its temperature, and theportion of the electromagnetic spectrumbeing considered. For example, snow ap­pears very bright in the visible part of thespectrum which means that it has a high re­flectivity and a low emissivity. However, forinhared radiation in the 8 to 14 micra range, itis nearly a perfect absorber, reflecting butlittle of such energy striking its surface.

In order to decide how to detect any givenobject, it must first be determined if the ob­ject ever shows a thermal difference from itsbackground. A subsurface void, such as a cavein the ground or a crevasse in an ice field, canbe detected only if an opening can be foundthat has a temperature different from that ofthe adjacent background, or if thermal con­duction, or convection, in the overburden hasbrought the surface above the void to a tem­perature different from those of adjacent sur­faces. The latter, although of applicability increvasse detection, has little practical bear­ing on cave studies.

A crevasse is a crack in the ice of the upperportion ofa glacier and is formed in responseto stress induced by movement of the glacierover different slopes of bedrock. Because ofthe weight of the overburden, the bottom ofthe glacier is plastic and it conforms to thesurfaces over which the glacier moves. Thetop portion of ice is brittle rather than plastic,and consequently can snap and form a crackas the glacier bends. This crack, which canget wider with time, usually bridges overwith snow during the winter months, whicheffectively obscures the underlying crack.Figure 1 is an interior view of a crevasse.During the winter, the internal temperatureof the crevasse can be many degrees warmerthan the outside air and exposed snow sur­faces. In fact, if a hole develops in the bridgeover the crevasse, the relatively warm inter-

AIRBORNE INFRARED THERMAL DETECTION OF CAVES & CREVASSES 1393

nal air can rise as a plume of steam visible forsome distance. In many places the bridge ispermeable to air and if the atmospheric pres­sure drops, the relatively warm internal aircan warm the bridge as it passes through, thusbringing the surface of the bridge to a tem-

perature different from the surface of the adja­cent firn. The bridge can also alter its temper­attire by conduction. Field measurements re­vealed these differences and set the basis fortrying thermal detection techniques.

With respect to caves, the problem is onlyslightly different, for the mantle of material(soil and rock) over a cave is nonpermeable,and is usually ofsuch a thickness that thermalconduction processes cannot establish a reli­able and persistent surface thermal pattern.For detection purposes, this leaves the open­ings or, more correctly, the mass of air in theopenings, as well as the material around theopenings or lining them. As do crevasses,caves breathe in and out as the atmospheric

FIG. 1. Interior view of a crevasse in the ice capofnorthwest Greenland not far from the Thule AirBase region. The rope ladder in the background ishanging through a hole dug in the snowbridgeover the crevasse. The bridge was 3 to 4 metersthick and the distance from the underside of thebridge to the working level near the bottom of thecrevasse was about 20 meters. The relatively flatsurface near the bottom was formed by the freez­ing of accumulated meltwater that trickled downduring previous summer thaws. The cyclic trickleand dripping of surface melt gradually builds upthe varied collection of icicles and other iceshapes so characteristic of arctic crevasses. Thiscrevasse was part of the crevasse field shown inFigure 3.

pressure changes, and if the temperature ofthe air inside the cave differs enough fromthat of the outside air, soil, and vegetation,there is the possibility that it will thermallyalter the immediate environment in andaround the opening as it passes over it duringthe "exhaling" part of the "breathing cycle".In order to determine the sequence and ex­tent of such events, it is necessary to monitorthem for several days by placing suitable re­cording instruments in the cave as well as onthe ground surface near a cave opening. Inaddition, a variety of day and night groundmeasurements should be made during thestudy period. Primarily these would be radia­tion and contact temperature measurementsof surfaces of interest, e.g., soil, rock, roads,vegetation, etc, and frequent incoming radia­tion measurements of the sky at zenith. Withthese data in hand, one can determine if it ispossible to detect the opening and the propertime at which to do so.

RESULTSCREVASSES

In addition to measurements on crevassesin the Greenland ice sheet, an undersnowtarget was set up at the SIPRE field stationnear Houghton, Michigan, to serve as ameans for evaluating airborne techniques.The target was a refrigerated device in theshape of a large letter Z. The surface of thesnow over the cold target was kept at a some­what lower temperature than the surface ofthe snow adjacent, which was warmed by theflow ofheat from the unfrozen mass ofgroundbelow. Figure 2 shows the results of a flightover the Z target and over a compacted snowtarget. The compacted snow, having a greaterdensity, conducted heat from the ground tothe surface at a higher rate than the uncom­pacted snow. Consequently, that surface isportrayed as a relatively warm area. Fieldmeasurements in Greenland showed that in­deed the surface of the snow bridge over acrevasse could have a temperature differentfrom the adjacent firn and that this differencearose from both conductive and convectiveprocesses.

Figure 3 is a comparison of photo imageryto infrared thermal imagery over a crevassefield on the Greenland ice cap, not far fromThule Air Base. The light streaks in the ther­mal image, which indicate relatively warmsignals, mark the locations of crevasses. Spe­cifically, they mark portions of the bridge por­ous enough, or thin enough, to be markedlyaltered by the volume of warmth below. Theair temperature inside one of these crevasseswas in the range of -80 C to -120 C while the

1394 PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING, 1975

FIG. 2. Airborne nighttime infrared thermal images of specialtargets for studying thermal patterns in snow. The undersnow target,designed to simulate a crevasse (left image), was made by connectinga series of hollow panels together to form a large letter Z (265 feet onthe diagonal and 136 feet along each end) which rested on an insulat­ing layer at ground level. Chilled trichloroethylene was circulatedthrough the panels, and by conduction the surface of the snow abovethe target could be maintained at a temperature lower than that of theadjacent snow surface. In this image the temperature difference be­tween surficial snow over the target and the snow adjacent was 30 C.By convection, thermal imagery is usually printed so that lightertones indicate relatively warmer signals. Thus, the colder snow overthe Z target shows as a dark streak. To the right is the image of acompacted snow area (100 feet by 100 feet) created by driving trackedvehicles over the surface. The denser, compacted snow (d = 0.452),can conduct heat from the ground up to the surface faster than theloose, uncompacted snow, and thus shows as a relatively hot signal (3°C warmer in this case). In the daytime when the sun shines on thesurfaces, the compacted area can conduct surficial heat downward ata greater rate than the uncompacted snow, and thereby maintains itssurface at a lower temperature. Thus, in daytime this signal can bereversed, and the compacted snow show as a darker gray tone. Asurface wind can degrade, and even erase these signals.

outside air was closer to -350 C. The interiorview shown in Figure 1 is of one of the cre­vasses in this group.

Figure 5 shows thermal imagery ofanothercrevasse field in the same general region ofthe ice cap. In addition to the warm traces ofthin areas in the bridges over the crevasses,the imagery also depicts the compacted snowtrails and the campsite. The emissivity ofsnow is relatively high, 0.90 to 0.95 for freshsnow, which means that the material is a verygood absorber. The thermal energy detectedby the scanner comes not from any depth, butfrom the surface. Since this signal is primarilya surface phenomenon, it is influenced byatmospheric events from above as well asconductive and convective events from be­low. A wind blowing over the snow surfacecan erase the signal imposed by subsurfacevents, and inscribe thereon the marks of itsown passing. Figure 6, of the same area as inFigure 4 (upper), shows the influence ofwindon the surface thermal pattern.

CAVES

A more familiar form of subsurface void is acave and, with the exception of conductiveevents, the physical basis for detecting such afeature is analogous to that associated withthe crevasses. Figure 7 shows the approxi­mate location of the study area within theisland of Puerto Rico.

This particular cave system, which exists inmassive beds of heavily fractured and se­verely eroded limestone, is associated withthe underground passage of the CamuyRiver. The river flows north in a deep narrowvalley until it disappears at the base ofa steepface. From this point, it continues north viaunderground channels for several kilometersbefore emerging to resume its course as asurface stream. The location of much of theunderground system was unknown, and theinfrared thermal scanner flights were made inthe hope of finding additional sinkholes thatwere directly connected to the undergroundchannels and voids.

AIRBORNE I FRARED THERMAL DETECTION OF CAVES & CREVASSES 1395

FIG. 3. Comparative photo and thermal imag­ery of a crevasse field in northwest Greenland.The upper image is a conventional aerialphotograph (original scale 1:20,000) and thelower two are infrared thermal displays (8 to 14micra bandpass). In all images the main trail,which is labeled in the photograph, is visible asare the remains ofan old meteorological stationand associated snow drifts just above the maintrail. Although traces of the crevasses could notbe seen by the eye or in the photograph, theywere recorded by the thermal scanner. Thelight streaks trending up and down, which indi­cate relatively warm signals, mark the crevasselocations. More specifically, they mark thoseportions of the snow bridges that are suffi­ciently thin or permeable to develop a surfacethermal anomaly as described in the text. Thecrevasse shown in Figure I is one of this group.

Limestone is slightly soluble in acidicground water and, over long periods of time,slowly dissolves along fractures and cracks,constantly enlarging them. Systems of theseinterconnected and enlarged fractures caneventually capture a surface stream and causeit to flow through the underground channels,further increasing the rate of dissolution anderosion. Eventually some of the voids andchannels become so large that the overlyingmaterial is too weak to support the span and itfails, falling to the bottom. Figure 8 is a stereopair ofair photos that portrays the ruggednessof this karst terrain and shows some of thesinkholes associated with the cave. Collapsestructures are prominent in this image andtwo such features have been labeled, i.e., thesinkholes Tres Pueblos and Empalme whichconnect directly to the cave system. Figure 9is a set of oblique aerial photos of some of thesinks and cave openings. Tres Pueblos Sink isabout 600 feet in diameter and some 450 feetdeep. The Camuy River flows along the bot­tom of this sink and is easily seen from thesurface. Empalme Sink also descends di­rectly to the river416 feet below the opening.

Recording hygrothermographs wereplaced in parts of the cave system and at asurface station, along with a recording baro­graph. Within the cave the air temperature wasfairly constant, about 70° F, while the outsideair showed a typical diurnal variation rangingfrom the high 60' s to the low 80's. A portion ofthis information is shown on the graph in

11 MARCH 196'2NI;., S1o!lon !I-5 Bldg'jNIP 55

Scol, I· 20,000

PANCHROMATIC FILM(.e..a'~l a"

flOg.., All '000 ft

INFRARED SCANNER STRIPC.......n ..> ,nd,(O~ b, ....hPl. .....,

FIG. 4. An infrared thermal image of the same area shown inFigure 3. The light-toned vertical line running through theimage is the main trail. The light-toned horizontal line extend­ing to the right of the main trail is an access trail to the crevasseentrance ofthe crevasse shown in Figure 1. The light streaks thattrend diagonally up from left to right are crevasses. The light­toned square is a compacted snow area and the light tone signalto the right of the main trail, which extends downward to theright, is wind streaking. The wind is from the upper left andsurfaces in the lee of the expedition vehicles, old camp build­ings, and snow drifts show as warmer signals. This imagery wastaken with an InSb detector at an altitude of about 2,000 feet inMarch of 1962.

1396 PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING, 1975

FIG. 5. Infrared thermal imagery of a crevasse field innorthwest Greenland (1420 hrs., March 1962). This area is aglacial feed to the Moltke Glacier that flows intoWolstenholm Fiord. The well-defined light streaks are por­tions of crevasse bridges that are relatively thin. The darkstreaks associated with them are massive portions of thebridge composed more of ice than of snow. Where the bridgehas iced up, it is relatively impermeable to air and maintainsa cooler temperature. The large light area in the upper leftpart of the image is a campsite. Leading out from it, as well asin other parts ofthe images, are vehicle trails which show as awarmer signal. Since this is winter, the snow surface is rela­tively cold. Each centimeter below the surface the tempera­ture is a little warmer until it reaches a maximum that repre­sents the previous summer; with further depth, the tempera­ture decreases to a minimum established by the previouswinter. This cyclic variation of temperature repeats itselfwith each cycle becoming less evident until, at some depth, arelatively stable temperature is reached. At this time of year(late winter) the compacted snow of the trails conducts thestored heat to the surface and thus shows as a warm signal, ascompared to the adjacent uncompacted snow. This imagerywas taken with an InSb detector at an altitude ofabout 2,000feet.

Figure 10 which combines data from bothtemperature recorders. In the early morninghours, the interior of the cave is warmer thanthe outside ground air by as much as four orfive degrees, and this was coincident with asmall decrease in atmospheric pressure.Thus, at this point on the cycle, the caveshould be "exhaling" slightly, and the warminterior air might be detectable either as amass inside the sink or as a result of warmingthe adjacent soil and vegetation surfaces. Formost types of infrared thermal missions it ispreferable to fly at night to avoid the com­plicating factor of intense solar heating withits high levels of radiation and large radiationdifferences. Although flights were frequentlyscheduled over the Camuy region during theairborne program, weather and clouds usu­ally prevented them.

Some flights were made during the earlymorning of 11 June 1970, under a relativelyclear sky with little surface wind. Increasing

cloud cover prevented additional passes.Some of these results are shown in Figures 11and 12.

The only two sinkholes or openings knownto connect with the cave that were consis­tently registered in the image were TresPueblos and Empalme. Also these are theonly ones that open directly, in a verticalsense, to the river. Other sinks connect to theinterior, but do so by inclined, somewhattwisted, and longer paths. Careful compari­son to stereo aerial photography did not re­veal additional "detected openings."

In Figure 12, Run 6, there is a row of threesmall hot spots labeled as firepots. These arepails ofburning charcoal used to mark knownlocations. In regions that are well endowedwith distinct and easily seen ground patterns,i.e., roads, rivers, ponds, structures, plowedfields, etc., it is not difficult to verify the pres­ence or absence of some desired signal as­sociated with a known cause. However, in

AIRBORNE INFRARED THERMAL DETECTION OF CAVES & CREVASSES

IR THERNAL - CREVASSES - GREENLAND

FIG. 6. Effect of wind on the surface thermal pattern. Inthese infrared thermal images of the same area as in theupper illustration of Figure 4, a surface wind is cooling thesnow surface. Areas protected from the wind by vehicles,buildings, snowdrifts, etc., were radiating at a higher energylevel. Although the wind streaks are evident in the lowerimage, the wind not strong enough to erase the thermalpattern of the crevasses. In the upper image, the wind speedwas such that it greatly subdued the surface thermal patternestablished by the crevasses although the compacted trailsshow. This is usually the result when winds across the sur­face exceed 8 to 10 knots.

1397

unknown regions, or where corroborativepattern detail does not exist and the lookedfor signal cannot be found, one does not knowif the signal was not recorded because it didnot exist, or because the aircraft did not flyover the proper spot at the proper time. As anaid to this problem, small firepots can beplaced in a well defined pattern at or aroundthe area of interest. Ifthe image records these

PUERTO RICO

ATlANTIC OCEAN

D CAMUYCAVIAlliA

• utUAOO

....OJUNIA~

CARIBBEAN SEA

markers, at least it is known that the aircraftpassed over the site in question.

A second problem associated with manydetection procedures is that of sorting out thesignal being sought from amidst a clutter ofsimilar looking signals. It is relatively simple,in the sense of physics, to determine if it ispossible to detect a given signal and underwhat conditions, but in real situations, it is

.C"'c.u ...~

FIG. 7. The general location ofthe cave study is shown on thisoutline map of Puerto Rico. The Camuy Cave is located inmassive beds of limestone that dip gently to the north. TheCamuy River flows through this region as a subsurface streamover 400 feet below the terrain envelope.

1398 PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING, 1975

~.~

FIG.8. Portions of a stereo pair ofair photos over the study area (USGS photography, 1951,scale 1: 15,000). Collapse structures are very evident. Of these, the large sinkholes, TresPueblos and Empalme, connect directly to the underground channel of the Camuy River.Tres Pueblos Sink is about 600 feet in diameter and 450 feet deep.

FIG. 9. A cIoserview ofthe collapse structures marked in Figure 7. Empalme Sink, to the left,provides a drop of416 feet to the surface ofthe Camuy River. Tres Pueblos Sink, to the right, isabout 600 feet across and some 450 feet deep. The Camuy River flows along the bottom at theright hand edge. Photos by Dave Atwood of the U. S. Army Cold Regions Research andEngineering Laboratory, Hanover, New Hampshire.

AIRBORNE INFRARED THERMAL DETECTION OF CAVES & CREVASSES 1399

frequently difficult to identify a particularsignal even though it was recorded. Forexample, consider the imagery of any singlerun in Figure 11. It is known from ground

experience that Empalme Sink is located atthe intersection of two roads, that these roadshave a unique pattern, and that they will bevisible in the thermal image. The imagery

ATMOSPHERIC PRESSURE

FIG. 10. The upper graph shows the combined temperature records from instruments on thesurface and within Espiral Sink, not far from the Empalme site. The lower graph shows thevariation in atmospheric pressure during the same period of time. The mission was flown inthe early hours of 11 June 1970, at a time when the cave was warmer than the outside and"exhaling" (note small dash and arrow above pressure trace).

, .a.TRES

FIG. 11. Infrared thermal images of the Camuy Cave area. The twosinkhole connections that were consistently detected, Empalme andTres Pueblos, are marked. In the flyover at 0117 hours (Run 3), the cavewas in the early phase of its exhaling cycle and Empalme Sink is barelydiscernible as a warmish mass. By 0129 hours (Run 6), Empalme Sink hasdeveloped into a very strong signal. Figure 12 shows portions of two ofthese runs at an enlarged scale. Note the background confusion ofsignalsthat make it difficult without auxiliary information to identify the par­ticular signal being sought. The other signals are caused by small ponds,bare rock outcrops, houses, etc.

1400 PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING, 1975

FIG. 12. Enlargements of portions of the thermal images from Runs 3 and 6,shown in Figure 11. In Run 3 the opening of Empalme Sink is not discernible,although there is a warmish area in the general vicinity of the opening thatmarks the beginning of the warming trend brought about by exhalation of thewarm interior air. Twelve minutes later (Run 6) the trend has proceeded to thestage where the sink opening itself can be detected. The charcoal firepots usedfor site identification are apparent in Run 6. They were not ignited until justafter the flyover of Run 3.

can then be examined for the pattern of the cord any thermal variation that could be as­proper road and road intersection to see if sociated with cave openings? It cannot bethere is some sort of signal that might repre- concluded from this alone that infrared ther­sent the hole that is known to be there. How- mal techniques cannot detect cave openings,ever, if one knew nothing of this region, was for perhaps the flights were made at thehanded imagery of one of these runs, and wrong time of day or under the wrongasked to comment on the presence or absence meteorological conditions. Thus, it is impor­of cave openings, silence would probably en- tant to examine the physical basis of a prob­sue, for there are too many similar looking lem and to collect ground data associatedsignals arising from other causes such as with the important parameters of that prob­small ponds, bare rock outcrops, etc. If the lem before coming to firm conclusions aboutthermal image could be compared to stereo choice of detector, times for flying, altitude,aerial photography or if, as in this case, a and weather restrictions. At times, factorssequence of thermal images was taken to pick beyond the investigators control determineup changes in a given signal, there would be a the choice of sensor system and times for fly­basis for noting areas of suspicion. The ab- ing. Even so, this basic information is of helpsence of the desired signal can also pose a in analyzing whatever imagery does resultproblem. Suppose this imagery did not re- from the missions.