PUBLICATIONS
Geochemistry, Geophysics, Geosystems
RESEARCH ARTICLE
10.1002/2016GC006501
Key Points:
 Satellite image analysis to detect
mineral occurrences in areas with
thick vegetation cover
 Identification of vegetation anomaly
zones using vegetation index and
enhancement methods
 Correspondence of vegetation
anomaly zones with geologic
data-based anomaly zones
Supporting Information:
 Supporting Information S1
Correspondence to:
A. N. H. Hede,
naftali@mining.itb.ac.id
Citation:
Hede, A. N. H., K. Koike, K. Kashiwaya,
S. Sakurai, R. Yamada, and D. A. Singer
(2017), How can satellite imagery be
used for mineral exploration in thick
vegetation areas?, Geochem. Geophys.
Geosyst., 18, 584–596, doi:10.1002/
2016GC006501.
Received 22 JUN 2016
Accepted 20 JAN 2017
Accepted article online 28 JAN 2017
Published online 15 FEB 2017
C 2017. American Geophysical Union.
V metal-induced vegetation stress [Slonecker, 2011]. However, the stress tends to appear more clearly at lon-
All Rights Reserved.
HEDE ET AL.
How can satellite imagery be used for mineral exploration in
thick vegetation areas?
Arie Naftali Hawu Hede1 , Katsuaki Koike2, Koki Kashiwaya2 , Shigeki Sakurai3, Ryoichi Yamada4,
and Donald A. Singer5
1
Earth Resources Exploration Research Group, Faculty of Mining and Petroleum Engineering, Institut Teknologi Bandung,
Bandung, Indonesia, 2Department of Urban Management, Graduate School of Engineering, Kyoto University, Kyoto, Japan,
3
Graduate School of Advanced Integrated Studies in Human Survivability, Kyoto University, Kyoto, Japan, 4Graduate School of
Environmental Studies, Tohoku University, Sendai, Miyagi, Japan, 5Consultant, Cupertino, California, United States
Abstract The Hokuroku district, northern Japan, is globally recognized for rich ore deposits (kuroko and
vein types), which have been thoroughly explored under thick vegetation cover. This situation is ideal to
evaluate the effects of ore deposits on vegetation anomalies through geobotanical remote sensing. Here
we present novel methods to detect vegetation anomalies caused by ore deposits and verify their
usefulness by comparing the anomalies with a deposit potential map produced from multiple geological
data. We use the reflectance spectra of Landsat ETM1 images acquired in summer and autumn to calculate
a vegetation index for plant physiological activity. A key variable to detect the anomalies is a variation of
vegetation index with time at each pixel. Difference in variation is enlarged by a sequence of image
enhancement methods for the detection. We find that the vegetation anomalies, defined by the large ratios,
correspond well to the high potential zones of ore deposits and known major deposits. Consequently, our
methods can extend the applicability of remote sensing-based mineral exploration to the areas covered by
thick vegetation, in addition to traditional arid and semiarid areas.
1. Introduction
The world is facing rapid increases in consumption and demand of metal resources [Patin ~o Douce, 2016],
which consequently requires innovation in mineral exploration technologies. Optical remote sensing is a
technology used in the reconnaissance survey stage because of its synoptic view which is advantageous in
detecting promising mineralized zones in less time and at a low cost [Sabins, 1999; Rokos et al., 2000; Clark
et al., 2003; Combe et al., 2006; Ciampalini et al., 2013]. However, this technology cannot be applied to
densely covered vegetation areas because the image typically expresses reflectance spectra of vegetation,
not rocks and soils below it. Image processing methods to enhance manifestations of mineralized rocks and
soils using reflectance spectra [Fraser and Green, 1987; Crosta and Rabelo, 1993; Carranza and Hale, 2002;
Guerschman et al., 2015] have been ineffective under such vegetation condition. The only useful way of min-
eral exploration under such conditions is to discriminate featured spectral patterns of vegetation anomalies
caused by the mineralization [Horler et al., 1980; Bruce and Hornsby, 1987; Rencz and Watson, 1989; Sabins,
1999; Sridhar et al., 2007]. A vegetation anomaly is an unusual physiological activity of plant species or com-
munity caused by natural or anthropogenic effects. A vegetation index (VI), a mathematical manipulation of
reflectance at multiple spectral bands, can be used for that discrimination because reflectance changes
with the plant activity and its change differs with wavelength. The Normalized Difference Vegetation Index
(NDVI) [Rouse et al., 1974] is the most representative, widely used VI.
Understanding factors affecting plant reflectance spectra [Grant, 1987] is indispensable to applying VI to
mineral exploration. Although plant growth is a complex process and the underlying physiology has not
been entirely understood, plants take up nutrients from the substrate, including metals in soils [Salt et al.,
1995; Slonecker, 2011; Lintern et al., 2013; Moffett and Gorelick, 2016]. The absorption of metals causes
changes in photosynthetic activity and chlorophyll contents. Therefore, reflectance spectra in the visible to
near-infrared (VNIR) region (350–1400 nm) and numerous VIs using this region have been marked to detect
ger wavelengths, i.e., the shortwave-infrared (SWIR) region (1400–2400 nm) by the interference of water
A NEW METHOD FOR GEOBOTANICAL RS 584
 Geochemistry, Geophysics, Geosystems 10.1002/2016GC006501

, 1999]. To take advantage of this feature,

absorption in leaves [Horler et al., 1980; Poschenrieder and Barcelo
we developed a new VI using reflectance in the VNIR and SWIR regions, the Vegetation Index considers
Greenness and Shortwave Infrared (VIGS), and has proved its high sensitivity to vegetation stress [Hede
et al., 2015]. This study aims to demonstrate how sensible VI can help identify mineralization hidden by veg-
etation by applying VIGS to multitemporal satellite image data.

2. Materials and Methods

2.1. Setting of the Study Area
2.1.1. Study Location
We selected a 40 km 3 50 km area in the Hokuroku district, Akita Prefecture, northern Japan, covering lati-
tude 40850 21.500 N to 408320 25.300 N. and longitude 1408250 0.800 E to 1408530 27.500 E. The land-cover types were
described from the Land-Use and Land-Cover (LULC) map (ver.14.02) with a 50 m grid produced by the
Earth Observation Research Center (EORC) at Japan Aerospace Exploration Agency (JAXA) [JAXA, 2014]. This
map was generated using Advanced Visible and Near Infrared Radiometer type 2 (AVNIR-2) image data of
the Advanced Land Observing Satellite (ALOS) and verified by the Biodiversity Center of Japan in the Japan
survey of vegetation [Takahashi et al., 2013]. The ASTER Global Digital Elevation Model (GDEM) of a 30 m
grid used for extracting topographic information. ASTER GDEM is a product of the Ministry of Economy,
Trade and Industry of Japan (METI) and the National Aeronautics and Space Administration (NASA), avail-
able for download from J-spacesystems ASTER GDEM site (http://gdem.ersdac.jspacesystems.or.jp).
The ASTER GDEM highlights that the Hokuroku district has irregular surface topography surrounded
by steep hills. Several urban areas such as Odate, Towada, Kosaka, and Kazuno are located on lowlands
(Figure 1a). The district is covered extensively by dense forests of deciduous and evergreen forest types
(Figure 1b) and partly by farmland, grass, and bare land. There are many ore deposits which are primarily
volcanogenic massive sulfide kuroko type, and minor fissure-filling hydrothermal vein type, that have been
thoroughly explored and studied (see geologic setting). The Hokuroku district is recognized as one of the
world’s richest kuroko deposit areas and one of the most productive mining regions in Japan for centuries.
Its total reserves and past production is estimated as 14 3 107 tons with metal concentrations 0.8–3.7%
Cu, 0.1–5% Pb, 0.8–16% Zn, and 1–8 ppm Au [Nakajima, 1989b, 1993]. The past hydrothermal activities can
be confirmed in the outcrops with wide alteration zones (Figure 1c), which supports the suitability of the
Hokuroku district as a test site of the correlation between vegetation anomalies and mineral occurrences.
2.1.2. Geologic Setting
Geological units were compiled from the seamless digital geological map of 1:200,000 by the Geological
Survey of Japan [Wakita et al., 2009]. Several references [Sato et al., 1974; Kouda and Koide, 1978; Ohmoto,
1978; Nakajima, 1989b, 1993; Sudo and Igarashi, 1997] were sourced for the location of the ore deposits, the
geological structures, and the ore generation mechanism. The residual gravity analysis used a gravity data
set by the Geological Survey of Japan [Geological Survey of Japan, 2013]. The general geologic setting can
be summarized as follows.
The Hokuroku district is chiefly composed of Neogene altered rocks and pre-Neogene basement rocks with
felsic volcanic rocks and tuffs as the predominant rock types (Figure 2a and supporting information Figure
S1). The mountainous eastern and northwestern margins of the district are overlain by Quaternary volcanic
rocks: Towada Lake and its surrounding ejecta in the northeastern edge were also formed by such recent
volcanic activity. There are several small basins in the district such as the Hanawa and Odate basins, which
are covered by terrace and alluvial deposits. The Hokuroku basin in the central district is the largest, main
basin, which is thought to be a part of volcanogenic collapses or submarine calderas formed by the under-
sea volcanic activity in Miocene Green-Tuff movement [Kouda and Koide, 1978; Ohmoto, 1978].
The residual gravity values correspond well to the basin and geologic structures (Figure 2b). The large and
small residual values are interpreted to reflect shallow depth of the basement rocks and thick acidic pyro-
clastic rocks above the basement, respectively [Kubota et al., 2004]. The residual gravity also shows the Late
Miocene basin structure in this area. Based on reconstruction of the pre-Neogene basements as the bottom
structure, the thickness of the Hokuroku basin is 2600 m at the most [Nakajima, 1993]. There are several
small basins nearly or within the Hokuroku basin formed in the Late Miocene and filled with the sediments
younger than it [Sato et al., 1974]. Large variability of residual gravity values may be caused by the rock

HEDE ET AL. A NEW METHOD FOR GEOBOTANICAL RS 585

 Geochemistry, Geophysics, Geosystems 10.1002/2016GC006501

Figure 1. Location of the Hokuroku district, Akita Prefecture, northern Japan. (a) Map of the known ore deposits overlaid upon a land cover,
generated by Japan Aerospace Exploration Agency [JAXA, 2014] and an ASTER GDEM shaded relief image (http://gdem.ersdac.jspacesystems.
or.jp) using ArcMap v10.2 (http://www.esri.com/software/arcgis/arcgis-for-desktop). Field photographs showing (b) sceneries of evergreen for-
est in which the predominant vegetation community is conifer and (c) a strongly altered outcrop of the East Kannondo in the northern Doya-
shiki deposit (marked by blue triangle in Figure 1a). Photos (b) and (c) were taken in a field survey on 26 November 2012.

facies. Because mafic rocks contain more heavy minerals than felsic rocks, their specific gravities are relative-
ly large. The large anomaly in the west-central part of the study area may suggest the presence of mafic
rocks of the early to middle Miocene, while the small residual values in the south-central part may be relat-
ed to the pre-Neogen felsic intrusive rocks (Abukuma granite).
The Hokuroku district is dominated by kuroko (‘‘black ore’’ in Japanese) deposits, a type of volcanogenic
massive sulfide deposits generated by the Miocene felsic to intermediate volcanism [Sato, 1974]. Main min-
eral constituents of kuroko are chalcopyrite, pyrite, sphalerite, galena, barite, and quartz. In addition, gyp-
sum and anhydrite are commonly abundante and form separate bodies beside sulfide-barite bodies
[Ohmoto and Skinner, 1983]. Epithermal vein-type deposits, primary sources of gold and silver, are also
abundant in dacitic or andesitic volcanic rocks and tuffs [Kouda and Koide, 1978] (see supporting informa-
tion Table S1 for the list of ore deposits in the Hokuroku district).
The generations of kuroko and vein-type deposits have been studied from several aspects of geological
and structural controls, mineralization process, and physical chemistry ore solutions [Sato, 1977; Scott, 1978;
Ohmoto, 1996; Matsuda and Koike, 2003; Glasby et al., 2004]. Subsidence of the Hokuroku basin, which was
associated with the dacite volcanism [Sato et al., 1974], caused significantly metal enrichments. Major kur-
oko deposits are embedded in the middle Miocene sedimentary sub-basins along the border of the subsid-
ing basin. Meanwhile, distribution of the vein-type deposits is scattered and most are located at high
elevation areas in the southern part [Nakajima, 1989a]. The ring structures with the 20–30 km in diameter
were identified in the Hokuroku district [Sato, 1974; Sato et al., 1974] because the kuroko deposits seem to

HEDE ET AL. A NEW METHOD FOR GEOBOTANICAL RS 586

 Geochemistry, Geophysics, Geosystems 10.1002/2016GC006501

Figure 2. Geological data-based anomaly maps. (a) Map of geological map classified into main units based on rock and facies classification
and using the seamless digital geological map at 1:200,000 [Geological Survey of Japan, 2004] for the data source. (b) Map of residual gravi-
ty categorized into five classes by m and r of the residual gravity values (see supporting information Figure S2). The maps are overlaid
upon the Miocene basins (straight and dash blue line) and ore deposits (kuroko and vein types by black and red circles, respectively). The
maps are overlaid upon an ASTER GDEM shaded relief image (http://gdem.ersdac.jspacesystems.or.jp) using ArcMap v10.2 (http://www.
esri.com/software/arcgis/arcgis-for-desktop).

be circularly distributed. The small-scale cauldrons (8–10 km in diameter) (see Figures 6a and 6b), which
may be resurgent cauldrons, are key structure to concentrate the kuroko deposits in the middle Hokuroku
basin instead of the margin of the basin [Kouda and Koide, 1978]. Post-kuroko ore veins are also related to
them. However, there are kuroko deposits unrelated to the ring structures because of the complicated sys-
tem of volcanism and the post-depositional tectonic activities. There are two major north-south striking,
vertical faults, the Ohshigenai and Hanawa faults. Although most of Neogene intrusive rocks were con-
trolled by the faults, the deposits location have little relationship with the structures [Sato et al., 1974].

2.2. Satellite Images and VI Calculation

Subscenes of the five Landsat ETM1 images acquired on 25 July 2002, 20 August 2000, 5 September 2000,
21 September 2000, and 13 October 2002 were selected because their cloud covers nearly 0% (see support-
ing information Table S2 for detailed the image acquisition dates). Selection reasons for the use of Landsat
images were their superiority over other multispectral images such as ASTER in terms of image availability
over appropriate time periods of many high-quality images which were cloud-free and low humidity, had
wide scene coverage, and had small change in the scene center location. Monthly precipitations in those four
months were 217, 45, 171, and 225 mm; average monthly temperatures were 22.78C, 25.78C, 20.68C, and
11.88C; and annual precipitations in 2000, 2001, and 2002 were 1459, 1420, and 1689 mm in the Hokuroku dis-
trict according to the Japan Meteorological Agency. Because these weather data do not indicate drought con-
dition for the target period of image analysis, drought-induced vegetation water stress can be ruled out.
These images were used for the VI analyses, with an assumption that no significant change in the land-use/
land-cover type occurred in the study area during the image acquisition period. Evidence of no change was
based on the program report of National Survey of the Natural Environment by Ministry of the Environment
of Japan that studied the change in vegetation cover from 1989 to 1999 [Ministry of the Environment, 1999].
The Landsat ETM1 images were the Level 1 Terrain-Corrected (L1T) data and downloaded from the website
of Earth Resources Observation and Science (EROS) Center, USGS (http://earthexplorer.usgs.gov/). The

HEDE ET AL. A NEW METHOD FOR GEOBOTANICAL RS 587

 Geochemistry, Geophysics, Geosystems 10.1002/2016GC006501

images were pre-processed for atmospheric corrections using the Fast Line-of-sight Atmospheric Analysis
of Spectral Hypercubes (FLAASH) module of ENVI 5.2 (Exelis VIS, Boulder, CO, USA) and for a topographic
~o et al., 2003].
correction using the C-correction method [Civco, 1989; Rian
Reflectance at each pixel was derived from the preprocessed Landsat images and used to calculate two
images for VIs, VIGS, and NDVI. VIGS is formulated as:
       
G2R N2R N2S1 N2S2
VIGS5w1 1w2 1w3 1w4 (1)
G1R N1R N1S1 N1S2

where G, R, N, S1, and S2 denote the surface reflectance in the visible green and red, near-infrared, and two
shortwave-infrared wavelengths, respectively. For a Landsat ETM1 image, G, R, N, S1, and S2 correspond to
the reflectance at bands 2, 3, 4, 5, and 7, respectively. The w1, w2, w3, and w4 are weights for each term and
a combination of w1 5 1.0, w2 5 0.5, w3 5 1.5, and w4 5 1.5 was employed in this study based on our previ-
ous result [Hede et al., 2015]. Equation of NDVI is:
N2R
NDVI5 (2)
N1R
VIGS is basically composed of the three previously normalized difference spectral indices, a green-red-
based normalized difference index [Motohka et al., 2010], an NDVI, and an SWIR-based normalized differ-
ence index [Ji et al., 2011]. The first two indices were useful to detect chlorosis-related phenomena by quan-
tifying the degree of senescence with season. The third index is sensitive to leave water contents and also
changeable with metal-induced stress [Barcelo  and Poschenrieder, 1990; Hede et al., 2015].
We demonstrated that VIGS was more sensitive to detecting metal-induced vegetation stress than NDVI
using the Landsat ETM1 images that include the Hokuroku district. Considering seasonal change of vegeta-
tion activity, temporal change in the VIGS values can be crucial to discriminate the degree of vegetation
stress. To emphasize a reflectance change caused by the stress and clarify the spatial pattern of anomalies,
a ratio of the mean (m) to the standard deviation (r) calculated on a per-pixel basis from the VIGS data set
was used. Selection of this ratio was founded on our previous studies that showed that the VIGS values
tended to be large at metal contamination zones common to the Landsat images in the growing season
[Hede et al., 2015]. The present five images were within this season. Therefore, seasonal change in the reflec-
tance of the stressed plants is expected to be small and yield a large ratio. Through the ratio, VIGS could
partly remove the phenology effect.

2.3. Enhancement Methods for the Vegetation Anomaly Detection

The following three methods were used to enhance the VIGS image and highlight the vegetation anomaly
zones.
2.3.1. Forced Invariance Method
This algorithm, originally proposed to suppress the vegetation from a remotely sensed image [Crippen and
Blom, 2001], was used to even the image contrast by reducing the elevation effect by the following two
steps. The first step is to produce a curve that approximates the relationship between the elevation and
ratio m=r (supporting information Figure S3a). This curve was defined to connect the averages of m=r in
each elevation bin. The spikes (outliers) in the curve originated from the water and cloud covers were delet-
ed using a sequence of median and mean filters, respectively. The next step is to flatten the curve to a
defined, desirable ratio by:

Revised ratio 5 original ratio 3 defined value=ratio based on the curve (3)

The total average of the original ratio was used for the defined value, which was 46 in the 8 bit integer scale
(0 2 255) in the present case. The revised ratios over 255 were changed to 255 to avoid overcorrection.
Finally, the elevation effect was confirmed to be corrected because the averages of revised ratios were con-
stant over all the elevation bins (supporting information Figure S3b).
2.3.2. Geostatistical Smoothing Approach
This approach was employed to delete the local abrupt changes of the ratios (noise). Because the ratio is a
type of regionalized random variable (Z) with a certain spatial correlation, the correlation is quantified using
the semivariogram, c (h):

HEDE ET AL. A NEW METHOD FOR GEOBOTANICAL RS 588

 Geochemistry, Geophysics, Geosystems 10.1002/2016GC006501

1
cðhÞ5 E½Z ðx1hÞ2Z ðxÞ2 (4)
2
where E, x, and h denote expectation,
the coordinate of variable in space,
and a separation distance between
two locations of a variable pair, respec-
tively. The spherical model was the
best fit to the experimental c(h) and
used for interpolating the centers of
unit cells by ordinary kriging [Zirschky,
1985] at the 30 m interval (equal to the
Landsat ETM1 image resolution) to
generate a smoothed ratio image.
2.3.3. Residual Analysis
The purpose of this analysis was to
remove regional trends and emphasize
local variation related to anomalies of
the VI values and gravity data. A
regional trend component was first
extracted by applying a moving aver-
age filter to the smoothed raster data
by the above kriging interpolation (the
ratio image and residual gravity data).
The radius for the data averaging was
set equal to be the range (the maxi-
mum distance of presence of any cor-
relation between data pairs) of the c(h)
model. A residual anomaly component
Figure 3. Forest-type distributions in each elevation range. (a) Changes in forest
was determined by subtracting the
(deciduous and evergreen types) cover rates and (b) reflectance variability by
mean (3) and standard deviation (bars above and below the mean) of VNIR and trend component from the smoothed
SWIR bands (B1–B5 and B7) of five Landsat ETM1 images for each 100 m eleva- raster data. All methodologies used in
tion range.
this research are summarized in sup-
porting information Figure S6.

3. Results and Discussion

3.1. Temporal Change in Vegetation Index
The coverage rates of deciduous and evergreen forests are almost the same, 42.19% and 42.02%, respec-
tively, between the elevations of 10 and 1100 m a.s.l. (Figure 3a). The reflectance of VNIR to SWIR bands of
five Landsat ETM1 images in July to October is correlated with the two forest types and the elevations (Fig-
ure 3b). The magnitudes of mean (m) and standard deviation (r) of the reflectance data can be interpreted
as follows. In low elevations <200 m a.s.l., mostly urban areas and paddy fields, the large r is most likely
related to anthropogenic factors that can cause large reflectance changes. The mean reflectances of the vis-
ible green and red bands (B2 and B3) increase slightly in the deciduous forest and less slightly in the ever-
green forest over 600 m a.s.l. probably due to seasonal change in chlorophyll activity. Note that the
m values of the near-infrared and SWIR bands (B4, B5, and B7), the wavelength range related to water stress,
also increase slightly with increasing the elevation. The seasonal change also affects the water stress of veg-
etation because the r values of the SWIR bands (B5 and B7) are large in both the deciduous and evergreen
forests.
Landsat-based five NDVI and VIGS images show clear seasonal changes in vegetation activity (Figure 4a).
Consistent with a previous result that spring and autumn are the best seasons to detect the plant physio-
logical response to environments [Bruce and Hornsby, 1987], both VI images are contrasted in the autumn
images. However, the differences in the VI values between deciduous and evergreen forests are more

HEDE ET AL. A NEW METHOD FOR GEOBOTANICAL RS 589

 Geochemistry, Geophysics, Geosystems 10.1002/2016GC006501

Figure 4. Temporal VI changes of study area. (a) Comparison of NDVI and VIGS of the five images. (b) Histograms of NDVI and VIGS values
from Landsat ETM1 images. The NDVI and VIGS values are individually rescaled into 8 bit integer and created in ENVI 5.2 (http://www.exe-
lisvis.com/IntelliEarthSolutions/GeospatialProducts/ENVI.aspx).

obvious by VIGS than NDVI, which demonstrates high sensitivity of VIGS to the changes in reflectance
spectra of vegetation. This sensitivity is proved by the wider range of the VI histogram of VIGS than NDVI
(Figure 4b).

3.2. Spatial Characterization of Vegetation Anomaly

The pixel-based ratio, m=r, calculated from the five VIGS images is used to quantify anomaly intensity
defined in section 2 (Figure 5a). This ratio should correct for elevation effects in the deciduous for-
ests, which have seasonal changes and relatively low VIGS values, which are dominant at the high
elevations (see Figure 3a). This effect appears bluish as low value zones in Figure 5a, but could be
reduced by the algorithm of the forced invariance method [Crippen and Blom, 2001] (Figure 5b). Fur-
thermore, a geostatistical technique and a residual analysis (see section 2) were applied to reduce
variability of the corrected image by two steps: smoothing of the image by ordinary kriging with a
unit cell size of 300 m 3 300 m (Figure 5c) and production of a residual image by subtracting the
regional trend from the kriging result (Figure 5d). The regional trend was defined by a moving aver-
age of the kriging result within 8500 m, a radius equal to the range of the semivariogram of the ratio
(supporting information Figure S4). Finally, the residual image was related to the distribution of ore
deposits as follows.
First, the residual image was grouped into three classes by thresholds using m and r of the residual values
of the deciduous and evergreen types; background (<m), low anomaly (between m and m 1 r), and high
anomaly (>m 1 r) (Figure 6a and see supporting information Figure S5). The distribution of the low and
high anomaly classes draws ring structures, which correspond well with the large basin structures that have
been interpreted as the result of submarine volcanic activity that controlled the generation of ore deposits
in the study area [Kouda and Koide, 1978; Ohmoto, 1978]. However, several large kuroko-type deposits such

HEDE ET AL. A NEW METHOD FOR GEOBOTANICAL RS 590

 Geochemistry, Geophysics, Geosystems 10.1002/2016GC006501

Figure 5. Anomaly intensity images and image enhancement. (a) Original image using the ratio (m divided by r) of VIGS values from the
five Landsat ETM1 images. (b) Flattened image through the forced invariance method to correct the elevation effect. (c) Smoothed image
using ordinary kriging with a unit cell of 300 m 3 300 m size. (d) Residual image to enhance anomalous ratios, which is Figure 5c image
minus regional trend. Values <0.2% and >99.8% in the cumulative distribution were removed as outliers. All images were rescaled into
8 bit integer and created in ENVI 5.2 (http://www.exelisvis.com/IntelliEarthSolutions/GeospatialProducts/ENVI.aspx).

as Matsuki, Matsumine, Shakanai, Kosaka, and Uchinotai appear in the background class: they are located
outside the forest areas (see Figure 1a).

3.3. Correlating Vegetation Anomaly With Ore Deposits Potential Map

Next we consider the correlation of the three classes with geology. To consider how the main geologic units
are related to ore deposits (see geological setting), the study area was divided into six major geological
units based on rock divisions and facies; sediments (s), nonalkaline felsic volcanic rocks (vf), nonalkaline
pyroclastic flow volcanic rocks (vp), nonalkaline mafic volcanic rocks (vb), accretionary complex from pre-
Neogene (b), and intrusive/plutonic rocks (p) (see Figure 2a). Because vf covers much of the study area
(45.24%), this unit occupies the largest frequency common to the three vegetation anomaly classes and its
relative dominancy increases from the background class to the high anomaly class, in particular in the ever-
green forest type (Figures 6c–6e). Accordingly, the VIGS-based vegetation anomaly is primarily distributed
in the vf area (Figure 6e). This trend is concordant with the fact that the deposits occur in the same horizon
[Sato, 1974; Kouda and Koide, 1978; Ohmoto, 1978].

HEDE ET AL. A NEW METHOD FOR GEOBOTANICAL RS 591

 Geochemistry, Geophysics, Geosystems 10.1002/2016GC006501

Figure 6. Vegetation anomaly maps and correlation with geoproperties. Classification maps based on m and r of (a) vegetation anomaly and (b) ore exploration index. Histograms of
geoproperties classified by vegetation (deciduous and evergreen types) anomaly classes: (c–e) main geological units, (f–h) residual gravity, and (i–k) ore-deposit exploration index. Ring
structures (purple line) and basin (straight and dash blue line) in Figures 6a and 6b are related to the kuroko mineralization (see geologic setting). The gray shadings in Figures 6c–6k
denote the classes of each property related to ore deposits. Figures 6a and 6b maps are overlaid upon an ASTER GDEM shaded relief image (http://gdem.ersdac.jspacesystems.or.jp)
using ArcMap v10.2 (http://www.esri.com/software/arcgis/arcgis-for-desktop).

HEDE ET AL. A NEW METHOD FOR GEOBOTANICAL RS 592

 Geochemistry, Geophysics, Geosystems 10.1002/2016GC006501

The vegetation anomalies can be related to geophysical survey results. Gravity anomalies are a useful
geophysical property for exploring basement rock structure and can be used to spatially characterize
the ore deposits in the Hokuroku district [Komazawa, 1984; Nakajima, 1993]. One example is that the
deposits are distributed near the 0 mGal contour line of the residual gravity map [Komazawa, 1984]. To
test for a correlation with the vegetation anomalies, residual gravity data were processed to extract
anomalies by the same residual analysis used in Figure 5d. The original residual gravity map was
grouped into four classes (Figure 2b) considering the shape of histogram of the residual values (support-
ing information Figure S2). The class between m and m 1 0.5r was found to best correlate with the loca-
tions of major ore deposits (see Figure 2b). Also, this class shows the highest frequency in the high
vegetation anomaly of the evergreen forest type (Figure 6h), although the difference from the subordi-
nate class is small (Figures 6f–6h).
Finally, an ore-deposit exploration index (Figure 6b) using multivariate regression analysis of seven factors
[Suzuki, 2003], i.e., positional relation to dacitic sequence and upper layer, presence of ring structures,
dacitic or rhyolite lava, andesite, vein-type deposits, alteration index, and gravity anomaly, was selected to
evaluate the vegetation anomalies (Figures 6i–6k). This index map may be the most comprehensive and
precise to express the ore deposits distribution in the Hokuroku district using more factors than used in
other studies [Singer and Kouda, 1996; Koike et al., 2002; Kubota et al., 2004]. Based on the mapped results
almost all major ore deposits were located in the range m 1 0.43r of the index values [Suzuki, 2003], the
values were classified into four classes: (1) low background (<m 2 0.5r), (2) high background (between
m 2 0.5r and m), (3) low anomaly (between m and m 1 0.5r), and (4) high anomaly (>m 1 0.5r). Although
this index map is a part of the present study area, the distribution of high anomaly zones corresponds well
to the high vegetation anomalies in Figure 6a. This correspondence is verified in that the frequency of
high anomaly zone is the largest in the high vegetation anomaly class common to the two forest types
(Figure 6k).
Our results show that the distribution of vegetation anomalies zones correspond well to the ore deposits
potential zones extracted from geoproperties and encompass the major ore deposits located in dense
forests. Another remark is that our methods can give an assessment of deposits for the zones not covered
by a geological data set, which is seen around the Hanawa and Osarizawa deposits. However, it is impor-
tant to recognize the difference in the plant communities in terms of vegetation anomaly classes that we
have defined. The background class, covering 56.05% of the study area, is composed of 40.42% deciduous
forest type and the evergreen forest type with a similar rate (33.16%), while the high anomaly class area,
covering 12.66% of the study area, is dominated by the evergreen forest type (63.29% cover rate) and the
deciduous forest type (35.12%). This different rate suggests that the deciduous plants may be affected by
external factors more strongly than the evergreen plants, which causes senescence and conceals the
effect of metal-induced vegetation stress. This concealing may explain why the Nurukawa and Towada-
ginzan mines are not located in the high anomaly class. However, our method successfully enhances
anomalies in the deciduous plants along the main ring structures (see Figures 6a and 1a). This indicates
the existence of key influence factor(s), considered below, causing vegetation anomalies around the ore
deposits.
The Hokuroku district is covered chiefly by brown forest soil, ando soil, podzol, gley soil, and gray lowland
soil [Classification Committee of Cultivated Soils, 1996] (Figure 7). The first three are mainly soil types in the
deciduous and evergreen forests areas. Brown forest soil is generally acid and low in base saturation. Ando
soil has a volcanic ash-origin and was usually formed from glass and amorphous colloidal materials. Podzols
are mostly distributed in the northern part of study area which contains interfile acidic soils. The soil acidity
increases in ascending order from ando soil, brown forest soil, to podzol with pH around 4–6 and tends to
decrease with the depth within the same soil type [Shoji et al., 1982]. Therefore, most soils in the study area
are acidic. In addition, the kuroko deposits are dominated by high sulfur content minerals such as sphaler-
ite, galena, pyrite, chalcopyrite, tetrahedrite-tennantite, and digenite [Sato, 1977; Glasby et al., 2004], which
can enhance the soil acidity by weathering. Alteration zones are in fact well developed around the deposits
with increasing sulfur contents in top soils [Izawa et al., 1978]. Such low pH of the soil strongly affects plant
physiological activity. Because the major ore deposits (all are currently closed mines), e.g., the Ezuri, Fuka-
zawa, and Osarizawa, are concordantly located in the high anomaly class of vegetation, soil acidity can be
regarded as the most plausible factor for the vegetation stress.

HEDE ET AL. A NEW METHOD FOR GEOBOTANICAL RS 593

 Geochemistry, Geophysics, Geosystems 10.1002/2016GC006501

Figure 7. Map of soil types [Classification Committee of Cultivated Soils, 1996], overlaid upon ring structures (purple line), basin (solid and
dash blue line), an ASTER GDEM shaded relief image which is explained in the caption of Figure 6.

Several methods have been proposed to spatially map ore deposits in the Hokuroku district using multiple
data sets of geological, geophysical, geochemical, and drilling surveys with spatial techniques such as geo-
graphical information systems, neural networks, and multivariate regression [Singer and Kouda, 1996; Koike
et al., 2002; Suzuki, 2003; Kubota et al., 2004]. However, spatial mapping still requires improvements because
of lack of outcrops, limited drilling density and depth, complicated generation mechanisms and structures
of deposits. Although we use only a set of reflectance spectra images of vegetation, the vegetation anoma-
lies distribution was compatible with the high potential zones identified in earlier research. Therefore, our
proposed method can give a preliminary spatial map of possible deposit locations by emphasizing subtle
vegetation stress in thick vegetation covered areas and can be combined with existing exploration data-
bases. As a next step, ground-truth studies are needed to verify our results by investigating soil geochemis-
try and plant reflectance spectra over the Hokuroku region.

4. Conclusions
The effectiveness of the vegetation index (VI) for identifying mineral deposits was demonstrated by the
case study of the kuroko deposits in the Hokuroku district, northern Japan. Vegetation anomalies were
detectable though a sequence of image enhancement methods with the VI analyses. Superiority of the pre-
sent methods was demonstrated by the correspondence of the vegetation anomaly zones with the geolog-
ic anomaly zones by a multivariate analysis of geological and geophysical data and ore deposits exploration
index. Correlation of the geologic anomalies with the mineralization was confirmed in the preceding stud-
ies. It is also noted that the vegetation anomaly zones detected appear overlapped with the ring structures
that controlled the distributions of ore deposits, and most major ore deposits in the forest areas were in
agreement with the vegetation anomalies. Some deposits did not appear as vegetation anomalies, which
are a limitation of the proposed methods, because they were not located in natural vegetation areas. Some
plants may be affected by anthropogenic, external factors, and by seasonal changes that induce senescence
of deciduous plants at high elevations. These can conceal the effect of metal-induced vegetation stress. Our
next step is to enhance the detection accuracy of vegetation anomaly by considering these effects for an
advanced remotely sensed mineral resource exploration in thick vegetation areas together with appropriate
ground truthing.

HEDE ET AL. A NEW METHOD FOR GEOBOTANICAL RS 594

 Geochemistry, Geophysics, Geosystems
Acknowledgments
The authors express their sincere
gratitude to Mr. Ryoichi Kouda of the
Geological Survey of Japan, AIST and
Mr. Hidekaze Kato of the International
Institute for Mining Technology
(MINETEC) for their valuable
instructions regarding the kuroko
deposits in the Hokuroku district, and
to two anonymous reviewers for their
valuable comments and suggestions
that helped improve the clarity of the
manuscript. This research was partially
supported by the Japan Science and
Technology Agency (JST) and the
Japan International Cooperation
Agency (JICA) through Science and
Technology Research Partnership for
Sustainable Development (SATREPS).
All data for this paper are properly
cited and referred to in the reference
list.
HEDE ET AL.
10.1002/2016GC006501
References
Barcel o, J., and C. Poschenrieder (1990), Plant water relations as affected by heavy metal stress: A review, J. Plant Nutr., 13(1), 1–37, doi:
10.1080/01904169009364057.
Bruce, B., and J. K. Hornsby (1987), A Canadian perspective on the application of satellite remote sensing to regional geobotany, Geocarto
Int., 2(3), 53–59, doi:10.1080/10106048709354108.
Carranza, E. J. M., and M. Hale (2002), Mineral imaging with Landsat Thematic Mapper data for hydrothermal alteration mapping in heavily
vegetated terrane, Int. J. Remote Sens., 23(22), 4827–4852, doi:10.1080/01431160110115014.
Ciampalini, A., F. Garfagnoli, C. Del Ventisette, and S. Moretti (2013), Potential use of remote sensing techniques for exploration of iron
deposits in western Sahara and southwest of Algeria, Nat. Resour. Res., 22(3), 179–190, doi:10.1007/s11053-013-9209-5.
Civco, D. L. (1989), Topographic normalization of Landsat Thematic Mapper digital imagery, Photogramm. Eng. Remote Sens., 55(9),
1303–1309.
Clark, R. N., G. A. Swayze, K. E. Livo, R. F. Kokaly, S. J. Sutley, J. B. Dalton, R. R. McDougal, and C. A. Gent (2003), Imaging spectroscopy: Earth
and planetary remote sensing with the USGS Tetracorder and expert systems, J. Geophys. Res., 108(E12), 5131, doi:10.1029/
2002JE001847.
Classification Committee of Cultivated Soils (1996), Classification of Cultivated Soils in Japan (3rd Approximation), Natl. Inst. of Agro-Environ.
Sci., Tsukuba, Japan.
Combe, J.-P., P. Launeau, P. Pinet, D. Despan, E. Harris, G. Ceuleneer, and C. Sotin (2006), Mapping of an ophiolite complex by high-
resolution visible-infrared spectrometry, Geochem. Geophys. Geosyst., 7, Q08001, doi:10.1029/2005GC001214.
Crippen, R. E., and R. G. Blom (2001), Unveiling the lithology of vegetated terrains in remotely sensed imagery, Photogramm. Eng. Remote
Sens., 67(8), 935–943.
Crosta, A. P., and A. Rabelo (1993), Assessing Landsat/TM for hydrothermal alteration mapping in central-western Brazil, in Proceedings of
the Ninth Thematic Conference on Geologic Remote Sensing: Exploration, Environment, and Engineering, pp. 1053–1061, Environ. Res. Inst.
of Mich., Ann Arbor, Mich.
Fraser, S. J., and A. A. Green (1987), A software defoliant for geological analysis of band ratios, Int. J. Remote Sens., 8(3), 525–532, doi:
10.1080/01431168708948659.
Geological Survey of Japan (2004), Geological Map of Japan 1:20,000, CD-ROM Version, Geol. Surv. of Jpn., AIST, Tsukuba, Japan.
Geological Survey of Japan (2013), Gravity Database of Japan, DVD Edition, Digital Geoscience Map P-2, Geol. Surv. of Jpn., AIST, Tsukuba,
Japan.
Glasby, G. P., T. Yamanaka, J. Yamamoto, H. Sato, and K. Notsu (2004), Kuroko and hydrocarbon deposits from Northern Honshu, Japan:
A possible common hydrothermal/magmatic origin?, Resour. Geol., 54(4), 413–424, doi:10.1111/j.1751-3928.2004.tb00217.x.
Grant, L. (1987), Diffuse and specular characteristics of leaf reflectance, Remote Sens. Environ., 22(2), 309–322, doi:10.1016/0034-
4257(87)90064-2.
Guerschman, J. P., P. F. Scarth, T. R. McVicar, L. J. Renzullo, T. J. Malthus, J. B. Stewart, J. E. Rickards, and R. Trevithick (2015), Assessing the
effects of site heterogeneity and soil properties when unmixing photosynthetic vegetation, non-photosynthetic vegetation and bare
soil fractions from Landsat and MODIS data, Remote Sens. Environ., 161, 12–26, doi:10.1016/j.rse.2015.01.021.
Hede, A. N. H., K. Kashiwaya, K. Koike, and S. Sakurai (2015), A new vegetation index for detecting vegetation anomalies due to mineral
deposits with application to a tropical forest area, Remote Sens. Environ., 171, 83–97, doi:10.1016/j.rse.2015.10.006.
Horler, D. N., J. Barber, and A. R. Barringer (1980), Effects of heavy metals on the absorbance and reflectance spectra of plants., Int. J.
Remote Sens., 1(2), 121–136, doi:10.1080/01431168008547550.
Izawa, E., T. Yoshida, and R. Saito (1978), Geochemical characteristics of hydrothermal alteration around the Fukazawa Kuroko deposit, Aki-
ta, Min. Geol., 28(151), 325–335, doi:10.11456/shigenchishitsu1951.28.325.
JAXA (2014), High resolution land use land cover map, Japan Aerospace Exploration Agency (JAXA) Earth Observation Research Center
(EORC), Tsukuba, Ibaraki, Japan. [Available at http://www.eorc.jaxa.jp/ALOS/lulc/lulc_jindex.htm.]
Ji, L., L. Zhang, B. K. Wylie, and J. Rover (2011), On the terminology of the spectral vegetation index (NIR 2 SWIR)/(NIR 1 SWIR), Int. J.
Remote Sens., 32(21), 6901–6909, doi:10.1080/01431161.2010.510811.
Koike, K., S. Matsuda, T. Suzuki, and M. Ohmi (2002), Neural network-based estimation of principal metal contents in the Hokuroku district,
northern Japan, for exploring kuroko-type deposits, Nat. Resour. Res., 11(2), 135–156.
Komazawa, M. (1984), On the quantitative gravimetric analysis in the Hokuroku district, Geophys. Explor., 37(3), 123–134.
Kouda, R., and H. Koide (1978), Ring structures, resurgent cauldron, and ore deposits in the Hokuroku volcanic field, northern Akita, Japan,
Min. Geol., 28(150), 233–244, doi:10.11456/shigenchishitsu1951.28.233.
Kubota, H., T. Urabe, R. Yamada, and S. Tanimura (2004), Exploration indices and mineral potential map of the kuroko deposits in northeast
Japan, Resour. Geol., 54(4), 387–397.
Lintern, M., R. Anand, C. Ryan, and D. Paterson (2013), Natural gold particles in Eucalyptus leaves and their relevance to exploration for bur-
ied gold deposits, Nat. Commun., 4, 2274.
Matsuda, S., and K. Koike (2003), Sensitivity analysis of a feedforward neural network for considering genetic mechanisms of kuroko depos-
its, Nat. Resour. Res., 12(4), 291–301.
Ministry of the Environment (1999), National survey on the natural environment, Ministry of the Environment, Fujiyoshida, Yamanashi,
Japan. [Available at http://www.biodic.go.jp/english/kiso/vg/vg_kiso_e.html.]
Moffett, K. B., and S. M. Gorelick (2016), Relating salt marsh pore water geochemistry patterns to vegetation zones and hydrologic influen-
ces, Water Resour. Res., 52, 1729–1745, doi:10.1002/2015WR017406.
Motohka, T., K. N. Nasahara, H. Oguma, and S. Tsuchida (2010), Applicability of Green-Red Vegetation Index for remote sensing of vegeta-
tion phenology, Remote Sens., 2(10), 2369–2387, doi:10.3390/rs2102369.
Nakajima, T. (1989a), Explanation Text of the Geological Map for Mineral Resources Assessment of the Hokuroku District, Scale 1:50,000, Miscel-
laneous Map Series 27, Geol. Surv. of Jpn., Tsukuba, Ibaraki, Japan.
Nakajima, T. (1989b), Geological Map for Mineral Resources Assessment of the Hokuroku District, Scale 1:50,000, Miscellaneous Map Series 27,
Geol. Surv. of Jpn., Tsukuba, Ibaraki, Japan.
Nakajima, T. (1993), Reconstruction of the depositional circumstances of the Kuroko deposits in the Hokuroku basin, Bull. Geol. Surv. Jpn.,
44(2/3/4), 251–282.
Ohmoto, H. (1978), Submarine calderas: A key to the formation of volcanogenic massive sulfide deposits?, Min. Geol., 28(150), 219–231, doi:
10.11456/shigenchishitsu1951.28.219.
Ohmoto, H. (1996), Formation of volcanogenic massive sulfide deposits: The kuroko perspective, Ore Geol. Rev., 10(3–6), 135–177.
A NEW METHOD FOR GEOBOTANICAL RS 595
 Geochemistry, Geophysics, Geosystems 10.1002/2016GC006501

Ohmoto, H., and B. J. Skinner (Eds.) (1983), The Kuroko and Related Volcanogenic Massive Sulfide Deposits, Soc. of Econ. Geol., Littleton, Colo.
Pati~
no Douce, A. E. (2016), Metallic mineral resources in the twenty-first century. I. Historical extraction trends and expected demand, Nat.
Resour. Res., 25(1), 71–90, doi:10.1007/s11053-015-9266-z.
Poschenrieder, C., and J. Barcel o (1999), Water relations in heavy metal stressed plants, in Heavy Metal Stress in Plants SE-10, edited by
M. N. V. Prasad and J. Hagemeyer, pp. 207–229, Springer, Berlin, Heidelberg.
Rencz, A. N., and G. P. Watson (1989), Biogeochemistry and LANDSAT TM data: Application to gold exploration in northern New Brunswick,
J. Geochem. Explor., 34(3), 271–284, doi:10.1016/0375-6742(89)90117-9.
Ria~
no, D., E. Chuvieco, J. Salas, and I. Aguado (2003), Assessment of different topographic corrections in Landsat-TM data for mapping veg-
etation types (2003), IEEE Trans. Geosci. Remote Sens., 41(5 Part 1), 1056–1061, doi:10.1109/TGRS.2003.811693.
Rokos, D., D. Argialas, R. Mavrantza, K. St.-Seymour, C. Vamvoukakis, M. Kouli, S. Lamera, H. Paraskevas, I. Karfakis, and G. Denes (2000),
Structural analysis for gold mineralization using remote sensing and geochemical techniques in a GIS environment: Island of Lesvos,
Hellas, Nat. Resour. Res., 9(4), 277–293, doi:10.1023/A:1011505326148.
Rouse, J. W., R. H. Haas, J. A. Schell, and D. W. Deering (1974), Monitoring vegetation systems in the Great Plains with ERTS, in Proceedings
of the Third Earth Resources Technology Satellite-1 Symposium-Volume I: Technical Presentations, NASA Spec. Publ. 351, pp. 309–317, NASA,
Washington, D. C.
Sabins, F. F. (1999), Remote sensing for mineral exploration, Ore Geol. Rev., 14(3–4), 157–183, doi:10.1016/S0169-1368(99)00007-4.
Salt, D. E., M. Blaylock, N. P. B. A. Kumar, V. Dushenkov, B. D. Ensley, I. Chet, and I. Raskin (1995), Phytoremediation: A novel strategy for the
removal of toxic metals from the environment using plants, Nat. Biotechnol., 13(5), 468–474, doi:10.1038/nbt0595-468.
Sato, T. (1974), Distribution and geological setting of the kuroko deposits, in Geology of Kuroko Deposits. Mining Geology Special Issue 6,
edited by T. Ishihara, pp. 1–10, Soc. of Min. Geol. of Jpn., Tokyo, Japan.
Sato, T. (1977), Kuroko deposits: Their geology, geochemistry and origin, Spec. Publ. Geol. Soc. London, 7(1), 153–161, doi:10.1144/
GSL.SP.1977.007.01.18.
Sato, T., S. Tanimura, and T. Ohtagaki (1974), Geology and ore deposits of the Hokuroku district, Akita prefecture, in Geology of Kuroko
Deposits. Mining Geology Special Issue 6, edited by T. Ishihara, pp. 11–18, Soc. of Min. Geol. of Jpn., Tokyo, Japan.
Scott, S. D. (1978), Structural control of the Kuroko deposits of the Hokuroku district, Japan, Min. Geol., 28(151), 301–311, doi:10.11456/
shigenchishitsu1951.28.301.
Shoji, S., Y. Fujiwara, I. Yamada, and M. Saigusa (1982), Chemistry and clay mineralogy of ando soils, brown forest soils, and podzolic soils
formed from recent Towada ashes, northeastern Japan, Soil Sci., 133(2), 69–86.
Singer, D., and R. Kouda (1996), Application of a feedforward neural network in the search for kuroko deposits in the Hokuroku district,
Japan, Math. Geol., 28(8), 1017–1023, doi:10.1007/BF02068587.
Slonecker, E. T. (2011), Analysis of the effects of heavy metals on vegetation hyperspectral reflectance properties, in Hyperspectral Remote
Sensing of Vegetation, edited by P. S. Thenkabail, J. G. Lyon, and A. Huete, pp. 561–578, CRC Press, Boca Raton, Fla.
Sridhar, B. B. M., F. X. Han, S. V Diehl, D. L. Monts, and Y. Su (2007), Spectral reflectance and leaf internal structure changes of barley plants
due to phytoextraction of zinc and cadmium, Int. J. Remote Sens., 28(5), 1041–1054, doi:10.1080/01431160500075832.
Sudo, S., and T. Igarashi (1997), Mineral Resources Map of Tohoku (1:500,000), Geol. Surv. of Jpn., Tsukuba, Ibaraki, Japan.
Suzuki, T. (2003), Application of multiple regression analysis to extract promising areas of kuroko deposits in the Hokuroku area, Akita
pref., Japan, Shigen-to-Sozai, 119(4,5), 149–154, doi:10.2473/shigentosozai.119.149.
Takahashi, M., K. N. Nasahara, T. Tadono, T. Watanabe, M. Dotsu, T. Sugimura, and N. Tomiyama (2013), JAXA high resolution land-use and
land-cover map of Japan, in Geoscience and Remote Sensing Symposium (IGARSS), 2013 IEEE International, pp. 2384–2387, The Institute of
Electrical and Electronics Engineers, Inc., Piscataway, N. J.
Wakita, K., T. Igawa, and S. Takarada (2009), Seamless Geological Map of Japan at a Scale of 1:200,000 DVD Edition, Digital Geoscience Map G-
16, Geol. Surv. of Jpn., AIST, Tsukuba, Japan.
Zirschky, J. (1985), Geostatistics for environmental monitoring and survey design, Environ. Int., 11(6), 515–524, doi:10.1016/0160-
4120(85)90187-4.

HEDE ET AL. A NEW METHOD FOR GEOBOTANICAL RS 596