You are on page 1of 7

Vegetation indices tools for the development evaluation in

reclaimed areas
L. Halounov, P. Junek & J. Petruchov
Department of Mapping and Cartography, Faculty of Civil Engineering, Czech Technical University,
Thkurova 7, 166 29 Prague 6
halounov@fsv.cvut.cz

Keywords: remote sensing, TM multitemporal data, Ikonos, environment, monitoring,


reclamation
ABSTRACT: The northern Bohemia comprises large areas with underground mining and brown
coal open-casts. Abandoned open-casts have been reclaimed. The history of the first reclamation is
longer than thirty years. The reclamation was performed in several possible ways hydric, forestrial,
and agricultural. As a result of hydric reclamation new water basins can be found in the region.
Their main purpose is to fill in land depressions after excavation, to create new parts of the
surrounding landscape and to serve as recreational localities. Agricultural reclamation represents
the change of previous open mines into agriculturally used areas. Forestrial reclamation is a creation
of new forests whose good growth and sound stand is a measure of successfulness of the reclamation.
A list of selected vegetation indices useful for the evaluation and methods used for the evaluation
of different reclamation types from 3-year multitemporal TM satellite data and Ikonos data are
presented.
1

INTRODUCTION

Areas with open casts cover hundreds of hectares in the study area near Teplice in the northern
Bohemia. Large areas have already been excavated and have been reclaimed and large areas are
now technologically or biologically reclaimed. The reclaimed areas can be divided into two groups
mines and dumps. Hundreds of hectares of these areas were reclaimed by various procedures.
Reclamation investments reach high values. Their effectiveness has not yet been controlled in the
Czech Republic. The goal of this project is to evaluate this effectiveness. According to the Czech
Mining Law, mining organizations are responsible for the post-mining reclamation.
The reclamation process can be characterized by several phases. The first one is called technical
phase and it lasts about two years. Terrain morphology is prepared including a final surface cover
created by a fertile soil layer. The morphology is determined by resulted landforms and their slope
stabilities and by pit shapes themselves. Their final morphology and cover type are designed by
experienced specialists in individual mining organizations. Selected areas with low slope were
transformed to pastures or agriculture fields. Deep excavated areas are often transformed to water
basins. The Barbora Lake as the largest one is an example of them serving now as an already
famous recreational center. Certain areas were prepared for permanent vegetation forest or nonforest. The forests are mixed forests with mixed deciduous trees, or deciduous and coniferous
trees, and with shrubs sometimes.
The second phase is a biological one when grass and two or three-year old trees are planted out.
The biological reclamation comprises not only reforestation, but also new vegetation areas for
pasturing and preparation of agricultural fields for crop growing.

339

Mixed forests formed by deciduous trees, shrubs and conifers represent forestric reclamation.
Forestrial reclamations with deciduous trees comprise wild cherry-trees, red oak, poplar, ash-tree,
alder, maple and birch and cover more than 200 hectares. Forestrial reclamation with mixed forest
combines aforementioned deciduous trees and elm, pine, larch, and spruce. The mixed forest
covers more than 400 hectares. Reclamations with shrubs as alder, willow, juniper, hazel, smoke
plant, privet etc., can be found on more than 800 hectares.
The third phase is a temporary period when regular controls once or twice a year verify the
stability of vegetation cover in situ. In case of the stable state, the reclamation of the area is finished
and can be sold to a new owner.
The reclaimed areas are formed by individual parts with different shapes and sizes varying from
tenths of hectares to hundreds of hectares. Each part is characterized by its number, reclamation
type, year when reclamation started and year of ending. The total area of reclamation has more than
3000 hectares in the study area. The first reclamation processes were finished already in 1959 in
the earliest case, and there are many reclamation areas whose reclamation process has not been
finished yet.
Remote sensing image data evaluating objectively situations of large areas in one moment and
comparing to situations in different years seem to be a very useful tool for it. The largest reclaimed
area is covered by permanent vegetation. That is why we tested useful tools for vegetation monitoring
vegetation indices.
2

PROCESSED DATA

The following data in Table 1 were used for evaluation of reclaimed areas.
Table 1. Landsat 5 and Ikonos data used for
the processing
Sensor
Landsat 5
Landsat 5
Landsat 5
Ikonos

Date
14.8.1988
9.8.1992
10.8.1998
17.9.2003

METHODOLOGY

The evaluation of successfulness of reclaimed areas was decided to be performed by vegetation


indices. Vegetation indices are new channels calculated from measured bands. The compared
vegetation indices were mean values of individual reclaimed areas.
There are more vegetation indices offered and defined by various authors. Their goal is to
characterize the behavior of vegetation in visible and infrared bands, which according to the
spectral reflectance curve differs from other materials. The vegetation behavior in the visible band
covering 400700 nm due to pigment absorption caused by chlorophyll, xanthophyll and other
pigments represents the darkest part of the vegetation spectrum usually. The highest values in
reflectivity are in the green visible band (around 550 nm) where the pigments have a lower absorption
there. On the contrary, the spectral range 7001300 nm due to small absorption and strong scatter
offers high values being therefore very bright. Leaf water and cellulose and lignin cause absorption
in the spectral range 13002500 nm showing vegetation relatively dark.
Relations between red visible band with lowest values in visible band and near infrared band
(NIR) are used by newly calculated channels to differ vegetation from other materials.
Vegetation indices were calculated from reflectances; that is why digital values were transformed
to reflectances. The following equation shows the relation

340

L. Halounov, P. Junek & J. Petruchov

meas ( ) =

D 2 L( )
D2
=
[B( ) + G DN],
Esun ( ) cos z
Esun ( ) cos z

(1)

where D is the Earth-Sun distance (in Astronomical Units, AU), Esun() - band dependent mean
solar exo-atmospheric irradiance (W/(m2 m)), L() - spectral radiance at sensors aperture,
z - solar zenith angle, B() = Lmin(), G = (Lmax() - Lmin())/DNmax range, DN digital number.
Dark object subtraction method (Chavez, 1988) was used for atmospheric correction individually
for individual bands according to their histograms.
Nine vegetation indices were used for the evaluation: NDVI, DVI, RVI, PVI, SAVI, MSAVI,
TSAVI, EVI, WDVI.
NDVI is (the Normalized Difference Vegetation Index) is the best known and widely used
vegetation index. Its values are in the <1, 1> range. Its definition can be shown by the following
equation
NDVI = NIR red ,
NIR + red

(2)

where NIR, red are reflectances in the near infrared and red spectral bands. The vegetation index
was used by Kriegler et al. for the first time in 1969.
The Ratio Vegetation Index - RVI (Jordan, 1969) compares only single values of the NIR and
red bands. Its value can vary from 0 to infinity. Its definition is determined by the following
equation
RVI = NIR.
red

(3)

The Difference Vegetation Index mentioned in Lillesand and Kiefer (1987) as a used vegetation
index expresses the difference between NIR and red bands. It is calculated according to the 4th
equation
DVI = NIR red.

(4)

The Perpendicular Vegetation Index was first described by Richardson and Wiegand (1977). It
belongs to a group using the soil line. The soil line expresses the variation in the spectrum of bare
soil in the image. The line can be determined by choice of two or more patches of bare soil in the
image with different reflectivities and finding the best-fit line in spectral space. It can be a quite
subjective decision. This index could be considered as a generalization of the DVI, which allows
evaluation for soil lines of different slopes and passing through origin. Furthermore, the PVI is
quite sensitive to its atmospheric situation, as Qi et al. (1994) mentioned. It is important mainly in
multitemporal image processing. The PVI expresses the distance between pixel values in NIR and
red band space to soil line. Pixels with the same PVI lie at the same distance from the soil line. PVI
values are in the same range as NDVI values <1, 1>. The equation defining the PVI is
PVI = sin() NIR cos() red,

(5)

where is an angle between the soil line and the NIR axis.
Clevers (1988) introduced the Weighted Difference Vegetation Index (WDVI ). WDVI values
have an unlimited range. WDVI is also very sensitive to atmospheric conditions (Qi et al., 1994).
This index belongs to perpendicular indices, whose soil slope is arbitraryly passing through origin.
The expression for WDVI is
WDVI = NIR s.red

(6)

where s is a slope of the soil line.

Vegetation indices tools for the development evaluation in reclaimed areas

341

Huete (1988) introduced a ratio-based index: the Soil Adjusted Vegetation Index = SAVI. The
vegetation with the same amount of vegetation does not form parallel lines and these lines do not
all converge at a single point. The index was based on measurements in situ, where influence for
different soil types was eliminated by an empirically determined value L.The L-value was found to
vary between 0 for vegetation with very high densities to 1 for vegetation with very low densities.
L = 0.5 can be regarded as a default value for vegetation with intermediate vegetation density. The
soil line passes through origin and its slope is equal to 1. The SAVI values have the same range as
NDVI.
The SAVI is defined by the following equation
SAVI =

NIR red (1 + L ).
NIR + red + L

(7)

The disadvantage of SAVI resulting from knowledge of vegetation density was changed in
MSAVI - the Modified Soil Adjusted Vegetation Index and the L value was replaced by L calculated
from NDVI and WDVI according to
L = 1 s NDVI.WDVI,

(8)

where s is the soil line slope, WDVI is the Weighted Difference Vegetation Index.
Baret et al. (1989) and Baret and Guyot (1991) defined the Transformed Soil Adjusted Vegetation
Index (TSAVI ) supposing that the soil line can have arbitrary slope and intercept. The point of
convergence of lines with the same vegetation amount does not lie in origin. The TSAVI values are
in <1, 1> range.

TSAVI =

s .( NIR s . red a )
,
( a . NIR + red a . s + X .(1 + s . s ))

(9)

where a is the soil line intercept, s is the soil line slope, and X is an adjustment factor which is set
to minimize soil noise (0.08 according to authors).
Huete and Justice (1999) used another vegetation index called EVI (Enhanced Vegetation Index)
enhancing NIR band influence and compressing atmospheric influence. Its equation is

EVI = 2,5

NIR red
,
NIR + C1 red C2 blue + L

(10)

where C1 = 6, C2 = 7,5, L = 1.
Reclaimed areas of different age have different ratio between tree and grass surface cover.
Reclaimed areas with forestrial and permanent grass cover are very similar in the first years after
the third year of reclamation beginning. The reflectance influence of trees whose age coincides
usually with the age of reclamation is small if compared with influence of grass. The change occurs
for older tree stands with larger tree crowns and therefore different ratio of the tree/grass area.
Reclaimed areas were divided into 3 age classes younger than 10 years, 10 to 20 years, and
older than 20 years at the moment of measuring. The first class has a relatively small influence of
tree vegetation. We chose twenty test forest areas from stable forest near these areas. They helped
us to compare situations in reclaimed areas and natural conditions.
4

RESULTS

After calculating vegetation indices in 1988, 1992, and 1998 we created many graphs showing
vegetation indices in 1988, 1992, 1998 and 2003.
Graphs on Fig. 1 show NDVI in 1988, 1992, 1998, and 2003 for the forestrial reclamation. The
first area with low value in 1988 represents the period of the first year of biological phase. NDVIs

342

L. Halounov, P. Junek & J. Petruchov

were ranked according to NDVI in 1988. It is evident that the differences in the following years are
not alike the 1988 curve. They even have different tendencies among these three periods (19921988, 1998-1992, 2003-1998). There is also another difference between the left and the right part
of the graph caused by the real cover of reclaimed areas. The left one are only pure forests, the right
one is formed by forest parts combined by other reclamation types.
Figure 1 shows that NDVI values from 1992 are for a high number of reclaimed areas worse
than in 1988 due to the dry summer season in 1992, a similar situation occurred between 1998 and
the dry 2003 year.
NDVI for all years (Landsat 1988, 1992, 1998, Ikonos 2003
0.900000
0.800000

NDVI values

0.700000
0.600000
NDVI_MEA88
0.500000

NDVI_MEA92
NDVI_MEA98

0.400000

NDVI2003

0.300000
0.200000
0.100000
27
1
26
9a
21
7a
26
4b
23
9a
28
9
29 c
0b
25
1c
23
2
25
1b
28
5b
29
2
23
9
26
1
27
2

0.000000
Reclaimed areas

Figure 1. NDVI for all 4 processed images where the left part represents pure forest areas, the right part
are formed by forest reclamations combined also by other types of reclamation (water, agriculture, etc.).
The values are ordered by NDVI in 1988 separately for parts with the same reclamation type.

If we compare Fig. 2 with Fig. 3 showing differences between reclamation age and VIs, we find
that the highest values and ranges are at TSAVI in both cases young and old reclamation. Very
close values are at WDVI, PVI and DVI with the value range three times lower than at NDVI.
Dependency of VIs on NDVI, forest reclamations 010 year old, Landsat 1988
0.8
0.7
DVI_MEAN88

0.6

other VIs

PVI_MEAN88
0.5

WDVI_MEA88
SAVI_MEA88

0.4

TSAVI_MEA8

0.3

MSAVI_ME88

0.2

EVI_MEAN88

0.658420

0.629448

0.629213

0.602862

0.578311

0.567686

0.565322

0.480930

0.403815

0.0

0.351361

0.1

NDVI

Figure 2. Relations between NDVI and other indices for forest reclamation of 010-year old, RVI were
not included in the graph because of scale, some of its values are much higher than 1.

Vegetation indices tools for the development evaluation in reclaimed areas

343

Dependency of VIs on NDVI, forest reclamations 2030 year old,


Landsat 1988
1.0

other VIs

0.9
0.8

DVI_MEAN88

0.7

PVI_MEAN88

0.6

WDVI_MEA88

0.5

SAVI_MEA88

0.4

TSAVI_MEA8
MSAVI_ME88

0.3

EVI_MEAN88

0.2

0.667845

0.664948

0.657130

0.650344

0.630398

0.629300

0.629198

0.626472

0.619276

0.594263

0.583895

0.573708

0.569637

0.530559

0.501135

0.0

0.393229

0.1

NDVI

Figure 3. Relations between NDVI and other indices for forest reclamation of 2030-year old, RVI were
not included in the graph because of scale, some of its values are much higher than 1.

When we compared increments and decrements between years for all vegetation indices we
found out that the most positive differences were found at DVI and PVI (it was valid for 80
reclamation areas from 140, and also for 11 test areas from 20). The least number of positive
differences can be found at NDVI and RVI.
5

CONCLUSIONS

Vegetation indices describe the behavior of the land surface regarding the vegetation health state.
To study the vegetation development from multitemporal data is a very sensitive problem. We
processed three image data with nearly exactly several year differences (August 9, 10, 14). Differences
between individual VIs are not mutually linear for the selected period. The younger the forest
reclamation the higher differences between NDVI and other VIs (see Fig. 2 and Fig. 3).
Similar trends can be found at MSAVI and SAVI, then at PVI, at DVI and at WDVI (see Fig.
2 and 3). The decrease which is shown by all indices except for NDVI, is at the six-year old area
not only with forest, but also with a grass part (Fig. 2).
By comparing only time development of NDVI we find out, that it is not at all linear for
individual areas (see Fig. 1). What can be clearly discovered are dryer (1992) and wetter years
(1998, 2003) if compared to 1988.
The VIs are seriously influenced by non-homogenous situations in individual areas. Figure 1
shows differences between pure forest reclamations in the left part of the graph and forest reclamations
combined also with agricultural or hydric reclamations (the right part of the graph).
The testing of 8 VIs showed that their assessment of improvement among years is not the same.
DVI and PVI proved 60 per cent of improvement in areas (both reclaimed and tested) what was the
optimistic result in the whole study period, while NDVI and RVI showed only 30 per cent (45 from
140) of improvement for reclaimed and 20 per cent for tested areas (4 from 20).
ACKNOWLEDGEMENT
The project was financed by the Czech Grant Agency - GA CR 205/03/0218.

344

L. Halounov, P. Junek & J. Petruchov

REFERENCES
Baret, F., Guyot, G., and Major, D. 1989. TSAVI: A vegetation index which minimizes soil brightness effects
on LAI or APAR estimation, in 12th Canadian Symposium on Remote Sensing and IGARSS 1990,
Vancouver, Canada, July 10-14.
Baret, F. and Guyot, G. 1991. Potentials and limits of vegetation indices for LAI and APAR assessment,
Remote Sensing of Environment, 35, pp. 161-173.
Chavez, P. S., 1989. Radiometric calibration of Landsat Thematic Mapper multispectral images. Photogrammetric
Engineering and Remote Sensing, 55 (9), pp. 1285-1294.
Clevers, J. G. P. W. 1988. The derivation of a simplified reflectance model for the estimation of leaf area
index, Remote Sensing of Environment, 35, pp. 53-70.
Huete, A. R., Jackson, R. D., and Post, D. F., 1985. Spectral response of a plant canopy with different soil
backgrounds, Remote Sensing of Environment, 17, pp. 37-53.
Huete, A. R., 1988. A Soil-Adjusted Vegetation Index (SAVI), Remote Sensing of Environment, 25, pp. 295309.
Huete, A., Justice, C., and Liu, H., 1994. Development of vegetation and soil indices for MODIS-EOS.
Remote Sensing of Environment, 49, pp.224 234.
Jackson, R. D., 1983. Spectral indices in n-space, Remote Sensing of Environment, 13, pp. 409-421.
Jordan, C. F., 1969. Derivation of leaf area index from quality of light on the forest floor, Ecology, vol. 50,
pp. 663-666.
Kriegler, F. J., Malila, W. A., Nalepka, R. F., and Richardson, W. 1969. Preprocessing transformations and
their effects on multispectral recognition, in Proceedings of the Sixth International Symposium on Remote
Sensing of Environment, University of Michigan, Ann Arbor, MI, pp.97-131.
Lillesand, T. M. and Kiefer, R. W., 1987. Remote Sensing and Image Interpretation, 2nd edition, John Wiley
and Sons, New York, Chichester, Brisbane, Toronto, Singapore, 721 p.
Qi, J., Kerr, Y., and Chehbouni, A. 1994. External Factor Consideration in Vegetation Index Development, in
_Proc. of Physical Measurements and Signatures in Remote Sensing, ISPRS, 723-730.
Richardson, A. J. and Everitt, J. H., 1992. Using spectra vegetation indices to estimate rangeland productivity,
Geocarto International, vol. 1, pp. 63-69.
Rouse, J. W., Haas, R. H., Schell, J. A., and Deering, D. W., 1973. Monitoring vegetation systems in the great
plains with ERTS, Third ERTS Symposium, NASA SP-351, vol. 1, pp. 309-317.
Rouse, J. W., Haas, R. H., Schell, J. A., Deering, D. W., and Harlan, J. C., 1974. Monitoring the vernal
advancement and retrogradation (greenwave effect) of natural vegetation, NASA/GSFC Type III Final
Report, Greenbelt, Md. 371 p.
Terrill W. Ray, 1994. FAQ on Vegetation in Remote Sensing, kepler.gps.caltech.edu - /pub/terrill/rsvegfaq.txt,
Version 1.0: 10/13/1994.

Vegetation indices tools for the development evaluation in reclaimed areas

345

You might also like