Jeanette Harlow – Detecting Caves with Thermal Imagery Within Ka’a’awa Valley


[ hide ]

    Detecting Caves with Thermal

    Imagery Within Ka’a’awa Valley

    California State University Long Beach &

    National Science Foundation – Research Experiences for Undergraduates

    Jeanette Harlow,  June 2013

    Advisors  –  Dr. Suzanne Wechsler, Dr. Carl Lipo,  Dr. Chris Lee,  Dr. Matt Becker


    Legends abound of sacred burial caves within and around Ka’a’awa Valley.  Should these caves actually exist, is would be important to document, preserve and protect them.  Due to the extreme terrain and vegetation of this valley, locating caves that may have significance is very difficult.  3-D imaging with Agisoft PhotoScan and Thermal imaging, utilizing UAV’s, are a way to locate caves that have a difference in temperature from their surroundings.  Within the two weeks that the project took place, I was not able to locate caves from their surroundings using data acquired was.  The caves and surrounding vegetation were too similar in the thermal imaging scale.  Further testing in ideal conditions, experimental calibrations with the thermal camera, and multiple control caves with known parameters could be a way to successfully detect caves with thermal imaging.


    Ka’a’awa Valley is an archaeologically unexplored paradise. Archaeological structures are found on the ground and legend states that there are over 400 burial caves (1) in the Kualoa ranch Valley walls. These caves may vary in size and depth. Observed in the field, often with the aide of binoculars, one can see small erosional caves with varying depths and opening diameters. Legend states that one of the sacred Pohukaina cave entrances is on the Kualoa Ranch. This lava tube is believed to stretch through Oahu and be a sacred burial sanctuary. Finding these caves, documenting and especially preserving those with archaeologically significant sites would be a benefit to future generations as well as learning about cultural practices of burial or other findings.


    Caves tend to have a relatively stable temperature compared to the surface of which they exit. Infrared thermographic remote sensing may indicate these temperature differences that can then be correlated to caves.  The control location of a known cave, called the “David Morgan” (DM) cave, was used to test the hypothesis of cave identification with thermal imagery. It is located on the south and east side of the valley with a nearly northern facing entrance. This cave is inaccessible, approximately 557m up a nearly sheer wall. Depth is unknown and diameter of the opening is difficult to determine due to the inaccessibility and distance.

    The DM cave is said to contain bones, and was guarded by two long sticks in the entrance in the shape of an X. One of those sticks has since fallen, so only one stick is visible. Vegetation is found at random positions along the wall, mostly in areas where the steepness changes to something less than vertical, but also has sparse vegetation growth on the steep walls.

    Fig 1. "David Morgan"cave, focus area.

    Using a DJI Phantom quadcopter with a Pentax Optio WG2-GPS camera, systematic transects of the S.E canyon walls were flown in an up/down fashion. Photos were taken every 10 seconds and transects were flown slowly to create overlap of photos. Different angles toward the walls were photograph to allow PhotoScan to create a 3-D model of all gullies and fins in the area.


    Fig 2. Phantom with Pentax Optio

    Using the Phantom quadcopter, a FLIR 640 Tau thermal camera was flown with the same fashion of transects specifically focusing on the wall with the DM cave and the adjoining gully and wall to the east of the cave. The thermal camera was set to take photos whenever it encountered a change in scenery. All transects were flown in the late morning to mid-afternoon over several days, under partly cloudy to cloudy conditions.

    Fig 3. Phantom with thermal camera

    An Instant Eye UAV with a GoPRO3 camera was used to fly as close to the cave as possible to acquire imaging of the contents of the cave. A Nikon COOLPIX P510 with a telephoto lens was used at base camp to focus in on the cave for alternative photos.

    Using photos taken with the Optio, and the thermal images, careful correlations were made based on general shapes to determine locations and features such as vegetation and rock outcrops. With this general principle and using many different image comparisons, the DM cave was analyzed.


    A 3-D Model of the target area was successfully created by PhotoScan. It was determined that to make a high resolution model, multiple transects were needed in the same general area. Phantom flying, as well as the Instant Eye, in conditions needed to get high resolution images to capture the nooks and crannies of the caves as well as the small scale valley walls, was unpredictable. The wall proximity needed to get the high resolution meant battling updrafts and drift. This may have compromised the quality of the images due to blurriness from jolty movements, however increasing the amount of pictures compensated for the low quality photos. A distance of ~100m or less from the wall was ideal for higher resolution images using the Optio or GoPro3.

    Fig 4.

    Fig 4. 3-D Model including area of interest

    Thermal imagery carefully compared to true color image of the 3-D model did show differences in temperature of the rock wall outcrops from other features. These other features included caves, vegetation, and relatively cooler spots of outcrop that may be a result of differences in physical properties of the rock, or moss cover. Thermal imagery did not successfully differentiate the DM cave temperature from vegetative temperature with the current techniques used.

    The task of taking the 3-D model and creating an inverted DEM with the Y plane on the X axis is a work in progress. Should the inversion be successful, we will be able to examine the contours where dimples in the x plane that may overlap with thermal imagery “cold” spots. These areas may be indicative of caves and can be further examined.

    Fig 5. Thermal Image of walls of interest

    Fig 7. 3-D model image with correlative patterns used to compare thermal images

    Fig 6. Thermal with correlative patterns of vegetation. Cool areas are light, hot areas are dark



    Images of the DM cave taken to determine archaeological significance gave insight into the quality of the cameras used. The GoPRO3, despite its closeness to flying the cave, between 50-100m, gave grainy resolution. The Optio, flying between 50 and 100 m to the cave, also gave grainy resolution but the resolution was better. The ultimate photos taken were with the Nikon telephoto camera taken from base camp ~550m away from the cave. It is possible that a GOPRO mounted on the Phantom may give smoother pictures because of the shock mounts for camera connection; this would need to be tested.

    Fig 10. Nikon Close up from 550 m

    Fig 8. GoPro Closeup from ~ 50m

    Fig 9. Optio Close up from ~50m


    In the time provided, climactic conditions encountered, limited knowledge of physical dimensions of caves, and lack of familiarity and experience with thermal imaging, thermal imagery was not able to locate the DM cave in the Ka’a’awa Valley from the surrounding vegetation.

    Uncertainty and Future Work

    Longer exposures of direct sun on the valley walls could show a difference in the amount of thermal radiation and could separate vegetation from caves. Having direct access to caves to measure depth and temperature could be a source of controlled experimentation in a systematic way to explore gradations in thermal signatures. Deeper caves may have more consistent temperatures, and shallow caves may have some indirect absorption of the heat from walls.  With this knowledge, the deeper caves would be an ideal control for the experiment performed.

    A source of random error could be due to the wind and drafts along the valley walls. Cool air may be getting sucked out and quickly blending with the surrounding warm air. This can be tested by locating caves, possible at ground level and/or near western side of the valley, that are not as affected by winds.

    Finally, it is possible that the small, yet strong, spots of vegetation found on the valley walls seen via thermal and affirmed by true color could be small caves that are a conduit of a small water source or bowl shape that collects rainwater that allows a small but dense growth of vegetation to cover the cave. It is possible that, as mentioned previously, the overlying vegetation is protecting the escaping cave temperature from quick dispersion by wind, hence the strong thermal signatures. This would need to be tested by physical examination.


    This project was made possible through funding by the National Science Foundation Research Experience for Undergraduates program. Award #1005258 Geospatial Research and Mapping (GRAM). I would like to thank Dr. Suzanne Wechsler, Dr. Carl Lipo, Dr, Chris Lee, Dr. Matt Becker, Paul Nesbit, Scott Winslow , Mike Ferris, Michelle Baroldi, Briton Voorhees, Emily Allen, and especially Ted Ralston for all his help and encouragement and the Kualoa Ranch for this memorable and enriching opportunity!


    • Fischer, John. “Exploring Kualoa Ranch and the Ka’a’awa Valley of Oahu, Hawaii.” . Web. 28 Jun 2013. <>
    • Howarth, Francis. “Hawaiian Lava Tubes- A Preliminary Report.” A Special Session of the 29th Annual Convention of the National Speleological Society. Ed. William R. Halliday Western Speleological Survey, 1976. 32-34. Web. 25 Jun. 2013. <>.
    • Thompson, Jim, and Jim Thompson. “Experimental Research on the Use of Thermography to Locate Heat Signatures from Caves.”2005 National Cave and Karst Management Symposium. . Web. 28 Jun. 2013. <>.