header thesis mapping - marc

thesis mapping – marc holland 2003 - RWTH Aachen

Geology of Kilauea caldera and Ka’u desert

A remote sensing based thesis mapping was done in coorperation with the University of Hawaii at Manoa in the winter 2002/2003. As part of a combined diploma thesis/mapping project a variety of different data sets were used to map a 255 km² area in 1:25.000 on the Big Island/Hawaii.

3dem with superimposed map

The excellent outcrop conditions in the proximity of the Kilauea caldera and Ka’u desert provided high quality data sets analysed prior to a 4 week lasting fieldwork period.
The processing of the raw data with the software package ENVI at the University of Hawaii at Manoa delivered preliminary maps that were checked and enhanced during the fieldwork.This page gives a short overview over the data sets and the mapping.

Data sets

The data sets of the area are of varying quality and age – mostly acquired during the 1980’s. Most sets needed to be processed and georeferenced. The content of the imagery holds data of different wavelengths offering sufficient insight in the repertoire of surface units.

DEM of mapping area

The data set shown above is a USGS DEM file (Digital Elevation Model). The data represents the elevation of the terrain. Acquisition of such data is commonly done by digitizing topographic maps or by the processing of radar. The visual information of the displayed upper image is limited but the file can be used to create contour lines or to create a 3D impression of the area. (Untreated DEM image, 17×15 km).

shaded relief derived from DEM

A mathematical operation on the DEM determines the slope of the terrain along a line of view and creates a synthetic shaded relief image. The upper image (with superimposed red contour lines) now highlights the morphology of the area with e.g. the caldera (NE) and the flanks of Mauna Loa (NW). Even the contours of major lava flows are now visible (Shaded relief image, 17
×15 km, illumination 315/25, contour interval is 100 m).

SPOT image

The SPOT data is a satellite data set. SPOT's panchromatic band (black and white) gives a resolution of 10 m/pixel.  Since the data is acquired from extreme height, limited distortion is expected. Georeferenced on the DEM the spot image was used as base image for the mapping, since it covers the entire mapping area. Visible on the upper image are younger, dark appearing lava flows as well as dark streaks, which are faults (Koa’e fault zone). (SPOT image 17
×15 km, georeferenced, panchromatic band, manually stretched)

Landsat image

The Landsat TM data set has 7 bands sampling the visual spectrum and the near infrared. It is capable of displaying information familiar to the human vision and information important to verify plant growth. The poor spacial resolution of 30 m/pixel limits the use of this data set for the mapping although it was commonly used for areas that are missing on the other data sets. (Landsat TM image, 17×15 km, georeferenced, bands 2,5,7 as RGB)

NDVI of Landsat data

The bands 3 and 4 of the Landsat data allow statements about the vegetation. The NDVI is a normalized index of the density and health of plants derived from the bands 3 and 4. The upper image shows the NDVI (on a scale from 0-100%) in the mapping area. Bright white values represent moderately to extreme vegetation cover, whereas black areas represent vegetation free areas. Note that the black area of the Ka’u desert is leeward (SW) of the Kilauea caldera. The sulphuric fumes of the crater limit plant growth downwind creating a chemical desert. The lava flows of Mauna Loa (NW) and younger craters (Kilaua Iki in the NE) are visible as well. (Indexed NDVI image, 17×15 km)

TIMS data, band 123 as RGB

The TIMS data (Thermal Infrared Multi Spectrometer) is acquired by an airborne system. The trajectory of the plane as well as its varying elevation above the ground results in distorted image files that need to be georeferenced carefully. The swaths (image coverage, see upper image) do not cover the outer parts of the mapping area but hold the most interesting central part.
TIMS samples data in the thermal infrared in 6 bands. It responds to the thermal properties of the rocks and shows similar results for the predominantly basaltic surface.
The brightness differences in the two swaths are a result of the acquisition and reflect thermal differences due to calibration and differences in the daytime (TIMS image, 17
×15 km, georeferenced, Bands 1,2,3 as RGB, manually stretched).

principal component image of TIMS

The thermal properties of the Basalts are very similar and highly correlated. This is documented by the low contrast, and dull colors on the pervious TIMS image. To stress the differences in highly correlated data, a principle component analysis can be applied. This mathematical operation recalculates the data to maximize the visual contrast to the processor. The upper image shows such an analysis. Differences within individual flow units are now highlighted. (Principle Component TIMS, 17×15 km, georeferenced, PC 1,2,3 as RGB, manually stretched)


Another airborne data set is the AIRSAR set (SAR = synthetic aperture radar). The data set uses radar wavelengths to determine the surface roughness. The sensor uses three different wavelengths (68 cm, 24 cm and 5.6 cm) to give a measure for the roughness of the surface in those scales. The three bands displayed as an RGB image allow distinguishing different surface textures. White areas are rough in all wavelengths and represent either vegetation or rugged a’a lava. Blue areas are smooth in the large scales and rough on the 5.6 cm scale. This represents roughness created by small chunks (e.g. pyroclastics)
(patched ARISAR image, 17×15 km, georeferenced, P-, L-, C-band as RGB).

field work with okkie

The major advantage of the data sets is its variety in different wavelengths. This enables good estimates of the surface units. Unfortunately the data sets were poorly georeferenced and distorted to a high extent. The poor spatial coverage as well as the differences in the important TIMS swaths required combining all data sets to provide a proper view on the surface units. However some units are not uniquely represented by the data.

A four week lasting fieldwork period was carried out to verify the results of the preliminary data processing. Small scaled observations and data retrieval for the thesis work were done as well. The lessons learned in the field allowed creating a high detail surface map of the area highlighting different surface units, relative ages and the relative roughness.

the final map
(final map, click here to enlarge)

This work was carried out with the support and help of Janos L. Urai (GED/RWTH), Stephen Martel (GG/SOEST) and Scott Rowland (HIGP/SOEST).

Marc 04-Aug-2004