Remote Sensing/GIS Integration
Topographic Lava Flow Output From Hawaii's East Rift
Topographic changes in an area of active volcanism are measured using remote sensing and GIS.
By Jeff Johnson, Simon Klemperer and Howard Zebker

Thirteen years of volcanic activity have altered the landscape of Hawaii's East Rift by causing extensive deforestation and producing over 80 square kilometers of new lava flows. Remote sensing and GIS have been utilized to help describe and quantify these changes. Both prove indispensable in such an area, where the topography is too volatile and transient to be efficiently mapped from the ground.

Hawaiian Volcanism
Hawaiian volcanic activity is gentle when compared with the catastrophic eruptions of volcanoes such as Krakatoa or Mount Saint Helens. Instead of culminating in a single violent event, Hawaiian volcanoes leak lava over an extended period of time. What Hawaii fails to achieve in explosivity, it compensates for in reliability. A repose period of two to three years between Hawaiian eruptions contrasts markedly with the thousands of years that characterize dormancy in other volcanic centers.
      In a typical Hawaiian eruptive episode, a fire fountain might jet as high as several hundred meters before losing volatile gasses and degenerating into a steady upwelling of low viscosity fluid. This gradual yet constant effluence of molten material may continue for years, steadily accumulating and constructing the immense, low-angle shield volcanoes typical of Hawaii. Kilauea Volcano, in the southeast corner of the Big Island of Hawaii, has been the focus of considerable activity in recent geologic history. Kilauea's juvenile East Rift (Figure 1) is heavily pockmarked with active craters, crevices, and cones. Flows which have erupted during historic times blanket the gradually sloping topography. The mantle of lava here and elsewhere in Hawaii tends to take on two distinct forms - a smooth, ropy, quickly-moving lava called pahoehoe or a more viscous, rubbly surface known as aa.

The Pu'u O'o / Kupainaha Eruptions
On Jan. 3rd, 1983, Hawaii's volcanic East Rift began spewing lava. The series of eruptions which have continued until the present are considered the most volumetric in recorded history. The outpouring of so much material has transformed what was once a benign, vegetated landscape into immense black barrens of aa and pahoehoe. The activity has also given birth to the cones of Pu'u O'o and Kupainaha which rise at the crest of the East Ridge and serve as origins for snakes of lava which enter the sea 10 kilometers to the south. Over the past decade, eruptive activity has migrated from Pu'u O'o's crater to that of Kupainaha and back again to the west. The most recent activity is manifested south of Pu'u O'o as terrestrial breakouts from lava tube skylights and tube-fed lavas which are deposited directly onto the ocean floor.

Three Different Remote Sensing Datasets
Figure 4 is a multispectral SPOT (Systeme Pour l'Observation de la Terre) high-resolution satellite image of the zone of interest. Scanned in early 1990, it depicts Kilauea's East Rift activity when the Kupainaha vent was predominantly active. Vegetated regions (red) represent strongly reflected near infrared wavelengths and are easily distinguishable from non-vegetated regions (dark). Sites of flowing lava are apparent in this image and are marked by gas clouds (white) both at Pu'u O'o summit and where the eastern flow meets the ocean producing steam.
      Figure 7 is a SIR-C/X-SAR (Space Imaging Radar - C-Band/X-Band - Synthetic Aperture Radar) image taken of the same region in October 1994. Because the shuttleborne instrument relies upon radar-wavelength backscatter from earth surfaces, textural differences are readily seen. Smooth black pahoehoe flows contrast nicely with the olive roughness of the aa. However this same characteristic makes it hard to distinguish aa from vegetation. Purple regions near the coast correspond to older vegetated pahoehoe flows. The black line is a hand-digitized outline of lava extents from all Pu'u O'o / Kupainaha eruptive episodes to date.
      Figure 3 is based upon an airborne TOPSAR (Topographic Synthetic Aperture Radar) image and a red overlay of all flows which have occurred between January 1983 and October 1994. The TOPSAR-derived DEM (digital elevation model) is used to generate artificial sunshading from a sun on the eastern horizon. The topography in the image corresponds to the August 1995 earth surface.

Estimating Lava Flow Volumes
Volume measurements and lava flow morphologies are derived from a comparison of two different temporally distinct DEMs. Rectification of these two models in x, y, and z coordinates is necessary before a differential elevation model can be derived. The resultant dataset is then manipulated with a raster processing program (ER Mapper) to produce the final results.
      The two DEMs which are compared correspond to the pre-eruption and to the present-day earth surfaces. DEM #1 contains USGS-30 meter digital topography as it appeared in 1981. DEM #2 is a TOPSAR C-Band interferometrically-derived dataset which was acquired in August of 1995. The USGS data has been digitized from topographic maps and claims 7 meter vertical accuracy and 30 meter pixel spacing. The TOPSAR data has been processed at a 7.5 meter pixel spacing and possesses 1 to 3 meters of uncertainty dependent upon slope.
      Rectification of the two DEMs in x, y space is straightforward and is accomplished by using a bilinear interpolation between 25 ground control points. Rectification in the z coordinate is more complicated because the TOPSAR elevations represent reflected backscatter from the tree canopy in vegetated regions. Because it is necessary to define which regions of the TOPSAR model correspond to radar reflections off the lava-flow surfaces and which regions correspond to tree-top reflections, the multispectral SIR-C image was utilized (see Figure 7). Boundaries of post-1983 lava flows are delineated based upon the false-color image, allowing a correlation between the DEMs.
      The resultant differential elevation model (see Figure 6) depicts topographic changes which have occurred over the past 13 years. These lava depths are systematically adjusted so that they approximate the actual flow thicknesses. It is assumed that there should be little or no topographic changes at the lava / vegetation transition - field measurements substantiate that thicknesses for single-layer, low-viscosity lavas approach zero at flow edges. An exception is where toes of lava empty into the sea; gray areas within the outlined lava boundaries in Figure 6 correspond to newly created land. The transect through the Pu'u O'o cone also reveals interesting information as it shows topographic profiles from both before and after the recent activity A vent visible in the 1981 USGS DEM matches the nadir of a depression in the present-day Pu'u O'o cone.

How Much Lava Exactly?
The calculated total-volume estimates for the Pu'u O'o / Kupainaha eruptions surpass conservative estimates for the amount of lava which has been extruded. According to the differential elevation model, approximately 1.8 cubic kilometers of lava have probably been deposited terrestrially. Another significant portion has certainly entered the sea. As expected, the eastern flow originating from Kupainaha is significantly thicker than other flows, owing to greater-sustained activity.
      The accuracy of lava volume estimates are heavily influenced by the reliability of the digital elevation models. The USGS-30 meter DEM must be questioned because it has been digitized from standard topographic contour maps. Inconsistencies between neighboring quadrangles, poor resolution, and imperfect mapping might contribute to an inflated estimate of lava production. Rectification of the DEMs to one another is another source of error. Though ground control points were established on the coast and along the East Rift, the central portions of the flows are deficient in ground control points owing to such extensive topographic changes. A last source of error is undoubtedly the lava lost to the sea. Remote sensing at this point offers no insight into the volume of material that has made its way below sea level before cooling.

Benefits of Using GIS/RS in a Topographically Unstable World
Calculating lava flow thicknesses by using a GIS to compare digital elevation models can be quick and easy when the models have been corrected to a common surface datum. As the quality of interferometry-derived elevation models increases, the limiting factor becomes the accuracy of the initial base map. For regions of active topographic change (such as Hawaii's East Rift), correlating continuously-updated DEMs will only improve the accuracy of lava thickness measurements. The TOPSAR scheduled to be flown in the summer of 1996 will be able to accurately pinpoint and measure the activity which has occurred over a one-year period. Because greater accuracy and resolution are required, new and evolving techniques must be applied to evaluate new and evolving landscapes.

A Word About TOPSAR
Radar interferometry will continue to gain precedence as a remote sensing tool. Its ability to produce accurate elevation models in a single flight pass even at night or during inclement weather is a decided advantage. The TOPSAR system which is currently generating 10 kilometer-wide swaths is intended to be a precursor for a TOPSAT (Topographic Satellite) remote sensing instrument.

About the Authors:
Jeff Johnson is a graduate student in geophysics at Stanford University. He is interested in the application of radar remote sensing to monitor active volcanoes. He will continue his studies at the University of Washington Geophysics Department in autumn of 1996. He may be reached at [email protected] Simon Klemperer is associate professor of Geophysics at Stanford University, and director of the Stanford Earth Sciences GIS Laboratory. He may be reached at 415-723-8214 (phone), 415-725-7344 (fax), or e-mail: [email protected] Howard Zebker is joint associate professor of Geophysics and Electrical Engineering at Stanford University. His research is concentrated in radar and remote sensing of the Earth and planets, specializing in the measurement and interpretation of topography and surface deformation using radar interferometric techniques. Acknowlegements The authors would like to thank ER Mapper for providing the Stanford GIS Lab with it's software under the Educational License Grant Program.

Back