This article was downloaded by: [Dalhousie University]

On: 02 September 2012, At: 12:03

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number:
1072954 Registered office: Mortimer House, 37-41 Mortimer Street,
London W1T 3JH, UK

International Journal of
Remote Sensing
Publication details, including instructions for
authors and subscription information:
http://www.tandfonline.com/loi/tres20

HyMap hyperspectral
remote sensing to detect
hydrocarbons
B. Hörig, F. Kühn, F. Oschütz & F. Lehmann

Version of record first published: 25 Nov 2010

To cite this article: B. Hörig, F. Kühn, F. Oschütz & F. Lehmann (2001): HyMap
hyperspectral remote sensing to detect hydrocarbons, International Journal of
Remote Sensing, 22:8, 1413-1422

To link to this article: http://dx.doi.org/10.1080/01431160120909

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/

terms-and-conditions

This article may be used for research, teaching, and private study
purposes. Any substantial or systematic reproduction, redistribution,
reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any
representation that the contents will be complete or accurate or up to
date. The accuracy of any instructions, formulae, and drug doses should
be independently verified with primary sources. The publisher shall not
be liable for any loss, actions, claims, proceedings, demand, or costs or
damages whatsoever or howsoever caused arising directly or indirectly in
connection with or arising out of the use of this material.
 int. j. remote sensing, 2001, vol. 22, no. 8, 1413 –1422

HyMap hyperspectral remote sensing to detect hydrocarbons

B. HÖRIG, F. KÜHN
Federal Institute for Geosciences and Natural Resources (BGR), Wilhelmstrasse
25-30, 13593 Berlin, Germany; e-mail: [email protected]

F. OSCHÜTZ and F. LEHMANN

German Aerospace Centre (DLR), Rudower Chaussee 5, 12489 Berlin,
Germany
Downloaded by [Dalhousie University] at 12:03 02 September 2012

(Received 5 January 2000; in Ž nal form 24 August 2000)

Abstract. The ability of airborne hyperspectral remote sensing methods to detect

hydrocarbons was investigated by the Federal Institute of Geosciences and
Natural Resources. Reference areas of deŽ ned geometry and chemical properties
were prepared, e.g. sandy soil, oil-contaminated soil, grass, plastic tarpaulins. The
aim of the study was to collect hyperspectral data from these areas and simultan-
eously determine their spectra with the infrared intelligent spectroradiometer
GER Mark V IRIS. The data corrections and further processing were based on
data provided by a Ž eld spectrometer.
This study showed that airborne hyperspectral remote sensing can be used to
detect hydrocarbons eYciently. Hydrocarbon-bearing substances are character-
ized by typical absorption features in the spectra. The availability of the high
signal-to-noise-ratio HyMap hyperspectral imaging system permits these features
to be recognized in the pixel spectra even if they are not very pronounced. Oil-
contaminated soil and other materials containing hydrocarbons can be dete-
cted and located directly and unambiguously by image processing focused on
the spectral characteristics of hydrocarbons. By this procedure, atmospheric
correction of the HyMap data is not necessary.

1. Introduction
Several remote sensing projects of the Federal Institute for Geosciences and
Natural Resources (BGR) have focused on the detection of hydrocarbons using
multispectral remote sensing. In 1993, Landsat Thematic Mapper (TM ) data and
spectroscopic Ž eld data were evaluated to determine the potential for the detection
of outcropping oil shale in Pakistan (Bannert et al. 1994). A complex study on
satellite and airborne remote sensing to detect oil-contaminated ground in abandoned
military training areas was carried out at the beginning of the 1990s (Kühn and
Hörig 1995). In both cases, the spectral resolution of the Landsat TM and Daedalus
scanner bands was too poor to recognize hydrocarbons using subtle absorption
features that can be detected with Ž eld spectroscopy.
The availability of newest hyperspectral airborne scanners oVered the possibility
to successfully continue studies focusing on direct detection of hydrocarbon-bearin g
rocks and soils. The 1998 Pro Smart Experiment, organized by the German
Internationa l Journal of Remote Sensing
ISSN 0143-116 1 print/ISSN 1366-590 1 online © 2001 Taylor & Francis Ltd
http://www.tandf.co.uk/journals
DOI: 10.1080/01431160010013450
 1414 B. Hörig et al.

Aerospace Centre (DLR), provided an opportunity to test the hyperspectral HyMap

scanner designed and operated by the Australian company Integrated Spectronics
Ltd. Several European remote sensing organizations have used this experiment to
evaluate the capability of this technology to clarify speciŽ c questions of geology,
mining, land-use, hydrology, etc.
The BGR has investigated whether this new scanner can solve problems which
had arisen during the above-mentione d studies in Pakistan and in abandoned military
training areas. We used the large space available on the BGR premises in Berlin to
prepare reference areas. The reference areas were of deŽ ned geometry and chemical
properties (e.g. sandy soil, oil-contaminated soil, grass, plastic tarpaulin). The aim of
the study was to collect HyMap data for these areas and to compare it with the
spectra obtained simultaneously with the GER Mark V IRIS Infrared Intelligent
Spectroradiometer. We expected to obtain hyperspectral data of spatially and spec-
Downloaded by [Dalhousie University] at 12:03 02 September 2012

trally well-deŽ ned objects. The HyMap data and the data provided by the Mark V
IRIS were then used to develop an optimum data correction and processing proced-
ure. The purpose of the processing was to unambiguousl y identify the hydrocarbon-
bearing reference areas on the BGR premises in the HyMap data. Moreover, this
procedure was to be applied to detect other hydrocarbon-bearin g materials in the
HyMap data taken for other parts of Berlin. The general aim of the study was to
provide a simple methodology for the routine detection and mapping of hydrocar-
bons. Another aim was to improve our knowledge of the spectral properties of
hydrocarbons. Here, we report on the methods applied and the general results of
the experiment.

2. Methods and sensors

2.1. Test Ž eld
The test Ž eld was prepared on the BGR premises in the Spandau district of
Berlin. This BGR branch oYce occupies former barracks of German and allied
troops. A large parking lot formerly used by military vehicles provided ideal condi-
tions for the reference areas. The ground used for the HyMap test is shown in
Ž gure 1. Figure 1 shows an image taken with the DLR’s Airborne High Resolution
Stereo Camera (HRSC-A), taken simultaneously with the HyMap scanner images.
The reference areas prepared for the experiment are characterized in table 1.
The concrete surface of a parking lot (15), lawn (18), single trees (e.g. 19) and a
gravel-paved area (20) were used as additional reference objects. The reference areas
were used to deŽ ne the spectral properties of the reference objects, to evaluate the
eVect of undersampling in the case of small targets, and to deŽ ne the limits of the
method in the case of slightly contaminate d targets with low-intensity spectral
features. Figure 2 shows reference areas on the BGR premises.

2.2. Field spectroscopy

The spectral properties of the reference areas were determined with a GER Mark V
IRIS Infrared Intelligent Spectroradiometer. The Mark V was operated with a 7°
by 3° dual Ž eld of view in the spectral range from 0.385 to 2.548 mm (849 bands).
Its spectral band widths varied between 2 and 6 nm. The spectroradiometer was
mounted on a cart so that it could be quickly moved between the reference areas
during the HyMap  ights. The spectroscopic data were recorded as ratio spectra
(percent re ectance) and as radiance spectra (mW/(cm2 nm sr)).
 Hyperspectral remote sensing to detect hydrocarbons 1415
Downloaded by [Dalhousie University] at 12:03 02 September 2012

Figure 1. HRSC-A image showing the reference sites on the BGR premises in Spandau,
Berlin (see §2.1 for explanation).

2.3. HyMap and HRSC-A airborne scanner

2.3.1. HyMap scanner
Selected speciŽ cations of the HyMap scanner are explained in table 2. The
HyMap  ights were carried out on 2 and 3 September 1998. To obtain optimal
spectral and spatial resolution we tried to  y the HyMap scanner at a relatively low
altitude. Our original objective to obtain HyMap datasets of 1 m GIFOV (ground
instantaneous Ž eld of view) could not be achieved because the aircraft’s lowest
operating speed was too high. If there are to be no gaps between the scan lines (i.e.
no undersampling), the minimum (i.e. best) GIFOV was approximatel y 4 m with a
 ight altitude of 2000 m. Therefore, we decided on two separate  ights at altitudes
 1416 B. Hörig et al.

Table 1. Reference areas.

1—rooŽ ng felt (4 m×4 m) 10—sand, uncontaminated (4 m×4 m)

2—sand mixed with CaCO3 (4 m×4 m) 11—sand, uncontaminated (6 m×6 m)
3—sand, slightly oil-contaminated† (4 m×4 m) 12—sewage sludge (4 m×4 m)
4—sand, highly oil-contaminated‡ (4 m×4 m) 13—sewage sludge (4 m×4 m)
5—sand, highly oil-contaminated‡ (2 m×2 m) 14—sewage sludge (4 m×4 m)
6—sand, highly oil-contaminated‡ (1 m×1 m) 16—plastic tarpaulin, orange (6 m×8 m)
7—sand, uncontaminated, moist (4 m×4 m) 17—plastic tarpaulin, black (4 m×8 m)
8—sand, uncontaminated (1 m×1 m) 21—heap of transparent plastic sheet
9—sand, uncontaminated (2 m×2 m)

† 25 ml lubricating oil (API SF/CC) per kg sand.

‡ 100 ml lubricating oil (API SF/CC) per kg sand.
Downloaded by [Dalhousie University] at 12:03 02 September 2012

Figure 2. Reference areas on the BGR premises.

Table 2. Selected speciŽ cations of the HyMap scanner (from http://www.intspec.com).

Spectral range: 0.440 –2.543 mm

Spectral bands: 128
Spectral band widths: 10–20 nm
IFOV: 2.5 mrad (along track)
2.0 mrad (across track)
GIFOV: 5 m at 2500 m operating altitude
Signal to noise ratio: >500:1

of 2200 m and 1137 m (table 3). The 2200 m  ight provided data with 4.4 m across-
track (5.5 m along-track) GIFOV and full coverage of the ground. The 1137 m  ight
was a compromise. Spatial resolution was better at this altitude (2.27 m across-track
GIFOV, 2.84 m along-track GIFOV), but because no further reduction of the air-
craft’s speed was possible and the integration time of the scanning system cannot be
changed, gaps occur between the scan lines. This undersampling results from the
2.84 m along-track GIFOV (i.e. width of each scan line) and the 4.67 m between the
centres of the pixels of each scan line.
 Hyperspectral remote sensing to detect hydrocarbons 1417

Table 3. Selected speciŽ cations of the HyMap  ights during the BGR Oil Experiment, 2 and
3 September 1998.

2 September 1998 3 September 1998

Operating time: 15:00 CET 11:00 CET

GIFOV: 2.84 m (along track) 5.50 m (along track)
2.27 m (across track) 4.40 m (across track)
Flight direction: SW–NE SW–NE
Flight altitude: 1137 m 2200 m
Swath width: 1.364 km 2.640 km
Additional equipment: HRSC-A camera
Aircraft: Cessna Caravan 208 Cessna Caravan 208
Downloaded by [Dalhousie University] at 12:03 02 September 2012

2.3.2. HRSC-A camera

The DLR’s HRSC-A camera was operated simultaneously with the HyMap
scanner. The HRSC-A camera provided high-resolution images of the area covered
by HyMap (Ž gure 1). It was used to verify the ground objects in the HyMap data
and to rectify the HyMap images. The HRSC-A camera is the airborne version of
the High Resolution Stereo Camera (HRSC) originally designed for space applica-
tions. The HRSC-A has a CCD system with 5272 pixels per line, 7 mm pixel size
(i.e. 10 cm at 2500 m altitude) and 8 bit radiometric resolution. For more reference
information, see Scholten et al. (1999 ).

3. Results and discussion

In most cases, visible to short-wave infrared spectra are recorded by a Ž eld or
laboratory spectrometer as percentage re ectance of the incident radiation.
Consequently, most spectra available from spectral databases are percentage
re ectance spectra. The general advantage of percentage re ectance spectra is that
they can be compared even if they were recorded at diVerent times, at diVerent
geographic locations, and with diVerent spectrometers.
In this study, the HyMap pixel spectra were evaluated by comparison with the
Mark V radiance spectra, which were made at the same time as the HyMap  ights.
This was done because the HyMap and Mark V instruments were calibrated in a
similar way.
Radiometric corrections and correction for systematic errors were applied to the
radiance values of the HyMap data by the DLR. Consequently, the digital number
of each HyMap pixel is proportional to the absolute value of the radiance re ected
by the ground and can be compared directly with the Mark V spectra of the
respective reference area. Atmospheric corrections to the HyMap data were not
necessary.
The upper plot of Ž gure 3 shows Mark V radiance spectra of selected reference
areas. As already demonstrate d by Clutis (1989), the hydrocarbon-bearin g reference
objects are characterized by absorption maxima at wavelengths of 1730 and 2310 nm.
Our experience is that plastic objects (e.g. plastic tarpaulin) show a relatively sharp
absorption maximum at 1730 nm, whereas this peak is less prominent in the spectra
of oil-bearing soils and rocks. The question was whether this less prominent absorp-
tion peak (or radiance minimum) can be recognized in the HyMap pixel spectra,
despite noise produced by the atmosphere between the scanner and the ground.
The lower plot in Ž gure 3 shows the HyMap pixel spectra of the same reference
 1418 B. Hörig et al.

(a)
Downloaded by [Dalhousie University] at 12:03 02 September 2012

(b)

Figure 3. GER IRIS Mark V radiance spectra (a) and HyMap spectra (b).

areas as in the upper plot. The spectra were calculated using ENVI image processing.
It can be seen that the same maxima/minima that are characteristic of hydrocarbons
are present in both plots. Although less prominent, they are signiŽ cant enough for
hydrocarbon-bearin g materials to be detected when the pixel spectra are evaluated.
However, eYcient mapping of the locations of hydrocarbons requires image pro-
cessing capable of accentuating all pixels with such absorption maxima. All pixel
spectra and images are from data obtained at 1137 m  ight altitude.
The best results were obtained for the 1730 nm maximum/minimum deŽ ned by
 Hyperspectral remote sensing to detect hydrocarbons 1419

the HyMap SWIR-1 bands 21 (1668.22 nm), 26 (1729.31 nm) and 31 (1788.98 nm).
These bands were used for FCC images based on SWIR-1 bands 21/red, 26/green
and 31/blue combined with linear contrast stretching. Figure 4 shows an extremely
enlarged HyMap image processed as described above (2.27 m pixel). The reference
areas on the BGR premises can be seen here (see Ž gure 1). The HyMap SWIR-1
bands 21, 26 and 31 are located in a narrow portion (100 nm wide) of the electromag-
netic spectrum. Consequently, most of the surface materials appear grey because
they show similar spectral characteristics in this portion of the spectrum. Only the
hydrocarbons appear coloured due to their signiŽ cant absorption features within
the same narrow portion. This procedure combined with linear stretching lead to
greyscale images with coloured depiction of all hydrocarbon-bearin g materials.
Consequently, the oil-contaminated soil (3, slightly contaminated; and 4, heavily
contaminate d), the plastic tarpaulin (16) and the plastic sheet (21) appear pinkish.
Downloaded by [Dalhousie University] at 12:03 02 September 2012

The intensity of the colour of the large oil-contaminated reference areas (3 and 4) is
proportional to the oil content. The percentage of the smaller areas covered by pixels
could not be determined. However, the 2.27 m GIFOVs covering or touching the
small oil-contaminated reference areas (2 m×2 m and 1 m×1 m) appear slightly
pinkish (5 and 6). Additionally, it must be kept in mind that the lower limit on the
aircraft speed caused undersampling.
This procedure applied to all of the HyMap data leads to images like in Ž gure 5.
All objects containing hydrocarbons appear pinkish. The hydrocarbon feature is not
visible or has a lower intensity if the surface is covered, for example, by dirt. Most

Figure 4. Reference areas on HyMap images (SWIR-1 bands 21/R, 26/G, 31/B with linear
stretching).
 1420 B. Hörig et al.
Downloaded by [Dalhousie University] at 12:03 02 September 2012

Figure 5. HyMap image processed as in Ž gure 4, showing hydrocarbon-bearing materials

at the ground surface in Spandau, Berlin (BGR premises and the oil-contaminated
reference areas, A, artiŽ cial grass; T, race track; B, plastic roofs).

of the pinkish areas are roofs of buildings (e.g. B), artiŽ cial grass (A), or light athletics
race tracks (T). One site within an industrial area in Spandau, Berlin showed
indications of oil-contaminated ground (outside the area of Ž gure 5 ).
Oil-contaminated soil and other materials made of hydrocarbons could be disting-
uished in colour composite images based on HyMap bands in the visible and the
 Hyperspectral remote sensing to detect hydrocarbons 1421

near-infrared portions of the spectra. In these parts of the spectrum, the colours of
the objects help distinguish, for example, between plastic, artiŽ cial grass, rooŽ ng
felt, and oil-contaminated soil (Ž gure 6). The VIS/NIR HyMap spectra of hydrocar-
bon-bearing materials diVer depending on the colour of the material. Nevertheless,
it is impossible to diVerentiate between hydrocarbons and non-hydrocarbon s on the
basis of the VIS/NIR part of the spectrum.
Downloaded by [Dalhousie University] at 12:03 02 September 2012

Figure 6. HyMap colour composite image in natural-like colours based on VIS bands 16/R,
10/G and 3/B.
 1422 Hyperspectral remote sensing to detect hydrocarbons

4. Conclusion
This study showed that airborne hyperspectral remote sensing can be used
to eYciently detect hydrocarbons on the ground surface. In radiance spectra,
hydrocarbon-bearin g substances are characterized by typical absorption maxima at
about 1730 nm and 2310 nm. High signal-to-noise-rati o hyperspectral imaging system
makes it possible for these to be recognized within the pixel spectra although they
are less pronounced. The existence and the location of hydrocarbon materials and
oil-contaminated soil can be detected and located directly and unambiguousl y using
a simple data correction and processing procedure, it is not necessary to apply an
atmospheric correction. We found that the SWIR part of the spectrum can be used
to detect hydrocarbons, and the VIS/NIR part to distinguish between diVerent
hydrocarbon-bearin g materials.
A prerequisite for eYcient mapping of hydrocarbons was the availability of a
Downloaded by [Dalhousie University] at 12:03 02 September 2012

hyperspectral sensor with a high signal-to-noise ratio (HyMap) and simultaneous

Ž eld spectroscopy data for well-deŽ ned reference areas so that the spectral signatures
of the target objects could be deŽ ned. Using this approach it should be possible to
detect and map any objects (rocks, minerals, soils, etc.) directly in the same simple
way, even if they are characterized by low-intensity radiance spectrum features.

Acknowledgments
The authors thank the German Aerospace Centre (DLR) for the opportunity to
participate in the Pro-Smart Experiment. Also, we would like to thank J. Flohrer,
M. Brand and T. Bucher (all DLR) for their collaboration and advice during the
planning and operation of the HyMap–HRSC-A Campaign. The HyMap scanner
was operated by P. Cocks (Integrated Spectronics Ltd, Australia); the HRSC-A data
were processed and rectiŽ ed by DLR. We are also indebted to Dr H. Raschka
(former director of the BGR Berlin branch oYce) for his support in preparing the
BGR premises for the experiment. Advisory support regarding chemical questions
was provided by A. Krenz (BGR).

References
Bannert, D., Kühn, F., and Hörig, B., 1994, Ergebnissse spektroradiometrischer Unter-
suchungen an Ölschieferaufschlüssen in Pakistan. 14. Wissenschaftlich-Technische
Jahrestagung der DGPF vom 5–10 Oktober 1994 in Dresden, pp. 171–178.
Clutis, E. A., 1989, Spectral re ectance properties of hydrocarbons: remote sensing implica-
tions. Science, 245, 165–168.
Kühn, F., and Hörig, B., 1995, Environmental remote sensing for military exercise places.
Remote Sensing and GIS for Site Characterizations: Applications and Standards,
ASTM STP 1279, American Society for Testing and Materials, pp. 5–16.
Scholten, S., Sujew, S., Wewel, F., Flohrer, J., Jaumann, R., Lehmann, F. Pischel, P., and
Neukum, G., 1999, The high resolution stereo camera HRSC-A—digital 3-D image
acquisition; photogrammetric processing and data evaluation. Proceedings, Joint
Workshop Sensors and Mapping from Space 1999, Institut für Photogrammetrie und
Ingenieurvermessung, Universität Hannover, Hannover, No. 18.