Open Geosci.

2016; 8:302–309

Research Article Open Access

Marzieh Mokarram* and Dinesh Sathyamoorthy

Relationship between landform classification and

vegetation (case study: southwest of Fars
province, Iran)
DOI 10.1515/geo-2016-0027 of natural resources in maintaining sustainable vegetation
Received Mar 29, 2015; accepted Nov 02, 2015 management for assessment of land use capabilities. Fur-
thermore, the study on the relationship between vegeta-
Abstract: This study is aimed at investigating the relation-
tion and landform classification is important because the
ship between landform classification and vegetation in the
distribution of vegetation based on the analysis of land-
southwest of Fars province, Iran. First, topographic posi-
form characteristics is an important aspect in the process
tion index (TPI) is used to perform landform classification
of understanding ecology [1–3]. In addition, the existence
using a Shuttle Radar Topography Mission (SRTM) digital
of landforms such as ridges indicate flood frequency. On
elevation model (DEM) with resolution of 30 m. The classi-
the other hand, increasing vegetation decreases floods in
fication has ten classes; high ridges, midslope ridges, up-
the area [4, 5].
land drainage, upper slopes, open slopes, plains, valleys,
Debelis et al. [6] used soil characteristics and vegeta-
local ridges, midslope drainage and streams. Visual inter-
tion as a function of landform position to develop a system
pretation indicates that for the local, midslope and high
that allows for extrapolations to be made at the landscape
ridge landforms, normalized difference vegetation index
scale on the relationship with vegetation. Loučková [7] in-
(NDVI) values and tree heights are higher as compared
vestigated the association between landforms and vege-
to the other landforms. In addition, it is found that there
tation. The results obtained suggest that recently created
are positive and significant correlations between NDVI and
landform geomorphic forms are key environmental deter-
tree height (r = 0.923), and landform and NDVI (r =
minants of riparian vegetation distribution patterns. In the
0.640). This shows that landform classification and NDVI
research of Cremon et al. [8], the main goal was to inves-
can be used to predict tree height in the area. High correla-
tigate the relationship between paleolandforms and veg-
tion of determination (R2 ) 0.909 is obtained for the predic-
etation classes mapped based on the integration of opti-
tion of tree height using landform classification and NDVI.
cal and synthetic aperture radar (SAR) data using the de-
Keywords: Landform classification; topographic position cision tree analysis. The results obtained indicated that the
index; normalized difference vegetation index (NDVI); tree former was useful for separating between forest and open
heights; correlation coefficients vegetation classes. Ahmad Zawawi et al. [1] studied the re-
lationship between landform classes and normalized dif-
ference vegetation index (NDVI). They found that NDVI de-
creases with elevation, and increases with tree height [9].
1 Introduction
This study is aimed at investigating the relationship
between landform classification and vegetation in the
Information on terrain characteristics is very important to
southwest of the Fars province, Iran. It will be demon-
explain geographical constraints and map the variability
strated using multiple regressions that landform classifi-
cation and NDVI can be used to accurately predict tree
height.
*Corresponding Author: Marzieh Mokarram: Department of
Range and Watershed Management, College of Agriculture and
Natural Resources of Darab, Shiraz University, Darab, Iran; Email:
m.mokarram @shirazu.ac.ir; +98-917-8020115; Address: Darab, 2 Study Area
Shiraz university, Iran; Postal Code: 71946-84471
Dinesh Sathyamoorthy: Science & Technology Research Insti- This study was carried out in west of Fars Province, which
tute for Defence (STRIDE), Ministry of Defence, Malaysia; E-mail:
is located in southern Iran and has an area of about
dinesh.sathyamoorthy@stride.gov.my

© 2016 M. Mokarram and D. Sathyamoorthy, published by De Gruyter Open.

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivs 3.0 License.
Unauthenticated
Download Date | 3/29/18 6:41 AM
 Relationship between landform classification and vegetation |
Figure 1: SRTM DEM of the study area.
122,400 km2 and is located at longitude of N 29∘ 24′ –29∘ 35′
and latitude of E 50∘ 24′ –50∘ 65′ (Figure 1). The altitude of
the study area ranges from the lowest of 1,538 m to the
highest of 2,830 m. There are three distinct climatic regions
in the Fars Province: 1) The mountainous area of the north
and northwest with moderate cold winters and mild sum-
mers. 2) The central regions with relatively rainy mild win-
ters and hot dry summers. 3) The region is located in the
south and southeast, has cold winters with hot summers.
The average temperature for the area is 16.8 ∘ C, ranging
between 4.7∘ C and 29.2∘ C [10].
The study area is a biodiversity of rich mountains,
relief and lithology, and other geological characteristics
such as sedimentary basin and elevated reliefs [10]. The
major land use categories of the area are agriculture, range
land, farming and forests. Range lands are found in large
parts of the north and south of the study area; forests lands
form a belt in the center of the study area; while wood
lands are located in small parts of the north and south of
the study area (Figure 2).
As the study area is located in a semi-arid region, its
river floods in the parts of years. The dominant causes of
gully formation in the area are rangeland destruction, land
use change from rangeland to dryland, misdesign and con-
struction of road culverts, road construction in sensitive
areas, improper irrigation, and destruction of channels for
flood conveyance [10].
303
3 Materials and methods
3.1 Landform classification
In this study, the topographic position index (TPI)
method [11] was used to perform landform classification
from a Shuttle Radar Topography Mission (SRTM) digital
elevation model (DEM) of the study area, which was down-
loaded from http://srtm.csi.cgiar.org. TPI is the difference
between the elevation at a cell and the average elevation
in a neighborhood surrounding that cell. Negative values
indicate the cell is lower than its neighbors, while positive
values indicate that the cell is higher than its neighbors.
TPI values provide a powerful means to classify the land-
scape into morphological classes [12].
TPI values can be calculated from two neighborhood
sizes. A negative small-neighborhood TPI value and a pos-
itive large-neighborhood TPI value is likely to represent
a small valley on a larger hilltop. Such a feature may be
reasonably classified as an upland drainage. Conversely, a
point with a positive small-neighborhood TPI value and a
negative large-neighborhood TPI value likely represents a
small hill or ridge in a larger valley [13] (Table 1 and Fig-
ure 3).
Unauthenticated
Download Date | 3/29/18 6:41 AM
304 | M. Mokarram and D. Sathyamoorthy

Figure 2: Land use map of the study area.

Figure 3: Landform classification map of the study area.

Unauthenticated
Download Date | 3/29/18 6:41 AM
 Relationship between landform classification and vegetation | 305

Table 1: Definitions of landform classes using TPI [13].

Classes Description
Canyons, deeply incised streams Small Neighborhood TPI: TPI ≤ −1
Large Neighborhood TPI: TPI ≤ −1
Midslope drainages, shallow valleys Small Neighborhood TPI: TPI ≤ −1
Large Neighborhood TPI: −1 < TPI < 1
upland drainages, headwaters Small Neighborhood TPI: TPI ≤ −1
Large Neighborhood TPI: TPI ≥ 1
U-shaped valleys Small Neighborhood TPI: −1 < TPI < 1
Large Neighborhood TPI: TPI ≤ −1
Plains small Neighborhood TPI: −1 < TPI < 1
Large Neighborhood TPI: −1 < TPI < 1
Slope ≤ 5∘
Open slopes Small Neighborhood TPI: −1 < TPI < 1
Large Neighborhood TPI: −1 < TPI < 1
Slope > 5∘
Upper slopes, mesas Small Neighborhood TPI: −1 < TPI < 1
Large Neighborhood TPI: TPI ≥ 1
Local ridges/hills in valleys Small Neighborhood TPI: TPI ≥ 1
Large Neighborhood TPI: TPI ≤ −1
Midslope ridges, small hills in plains Small Neighborhood TPI: TPI ≥ 1
Large Neighborhood TPI: −1 < TPI < 1
Mountain tops, high ridges Small Neighborhood TPI: TPI ≥ 1
Large Neighborhood TPI: TPI ≥ 1

Table 2: Characteristics of NDVI signatures [1, 9]. where Red and NIR stand for the spectral reflectance mea-
surements acquired in the visible (red) and near-infrared
NDVI Dominant cover regions respectively. NDVI values vary between −1 to +1,
< 0.1 Water, pond and streams with low NDVI values indicating sparse or unhealthy vege-
0.1 to 0.2 Bare areas, soil and rock tation, and high values indicating greener plants (Table 2).
0.2 to 0.3 Shrubs, grassland, agriculture areas and dry Water typically has an NDVI value less than 0, bare soils
forests between 0 and 0.1, and vegetation over 0.1 [9, 19].
0.3 to 0.6 Dense vegetation
0.6 to +1.0 Very dense vegetation and tropical rainforest
3.3 Multiple regressions
3.2 Vegetation cover classification using The relationships between NDVI and different parameters
NDVI analysis that characterize vegetation, such as leaf area, percentage
of plant fraction and plant biomass, have been highlighted
Vegetation cover is an important factor as it has a strong by several authors [20–22]. For this study, the relationships
relation to root strength that represents site quality and between NDVI, tree height and landform classes were de-
land use suitability [1]. One of most important vegetation termined using multiple regressions [21]. The general form
indices is NDVI, which gives a measure of the amount of of the regression equations is according to Eq. 2 [23]:
vegetation in the study area, differentiating vigorous from
less vigorous vegetation [1, 14, 15]. In this study, NDVI was Y = A0 + A1 X1 + A2 X2 + . . . + b n X n (2)
computed from a Landsat ETM+ satellite image (May 2010)
using the following equation [16–18]: where Y is the dependent variable, A0 is the intercept,
A1 . . . bn are regression coefficients, and X1 –X n are inde-
NDVI = (NIR − Red) / (NIR + Red) (1) pendent variables referring to basic soil properties.

Unauthenticated
Download Date | 3/29/18 6:41 AM
306 | M. Mokarram and D. Sathyamoorthy

Figure 4: Areas of the landform classes.

Table 3: Correlation values (r) showing the relationships between 4.2 Vegetation cover classification using
the landform classes and vegetation variables analyzed in this
study.
NDVI analysis

As shown in Figure 6(a), the NDVI values for the study

Parameters Landform NDVI Tree height
area range between −0.58 and 0.69. According to Table 2,
Landform 1 .640* .408
NDVI ranging from 0.1 to 0.2, which represents rock, soil
NDVI 1 .923**
and bare areas, is found at the upper slope, open slope
Tree height 1
*
and midslope drainage classes. Flat areas and high ridges
Correlation is significant at the 0.05 level (2-tailed).
**
have higher NDVI values, ranging from 0.2 to 0.3, covering
Correlation is significant at the 0.01 level (2-tailed).
lower tree heights. Higher NDVI values of more than 0.3
are found concentrated at valleys and streams, where tree
height is higher (more than 14 m) [1]. The dominant covers
4 Results and Discussion based on the NDVI values for the study area are shown in
Figure 6(b). It was found that for the local, midslope and
4.1 Landform classification high ridge landforms, NDVI values are higher as compared
to the other landforms. Furthermore, it was found that the
Using TPI, the landform classification map of the study
locations with the highest NDVI has the most vegetation
area was generated. The classification has ten classes;
(tree height).
high ridges, midslope ridges, upland drainage, upper
slopes, open slopes, plains, valleys, local ridges, midslope
drainage and streams (Figure 3). The areas of the landform
4.3 Multiple regressions
classes are shown in Figure 4. It is observed that the largest
landform is streams, while the smallest is plains. The calculated simple linear correlation coefficients (r) be-
By comparing the landform classification and tree tween landform classes, tree height and NDVI are summa-
height (Figure 5) maps of the study area, it was found rized in Table 3. It was found that there is positive and sig-
that high ridges, midslope ridges, upland drainage, upper nificant correlations between NDVI and tree height (r =
slopes, open slopes, plains and valleys consist of vegeta- 0.923), and landform and NDVI (r = 0.640). This indi-
tion heights of > 20, 14–20, 11–14, 8–11, 4–8, 2–4 and < 2 m cates that landform classification and NDVI can be used
respectively. Based on this, it can be concluded that ridge to predict tree height.
landforms have more vegetation than the other landforms. In to Table 4, R2 for prediction of tree height through
This is as in ridges, the climate is suitable for growth of veg- landform and NDVI is 0.909 in the model 1 that as is shown
etation [21]. good correlation model. Standard Errors refers to the stan-

Unauthenticated
Download Date | 3/29/18 6:41 AM
 Relationship between landform classification and vegetation | 307

Figure 5: The tree height map for the study area.

Table 4: MLR model summary for the tree height prediction. predict tree height using landform and NDVI is as follows:

Model R R2 Adjusted R2 Tree height = −35.117 + 98.040 × NDVI (3)

1 0.935a 0.909 0.883 − 0.209 × landform
a
Predictors: (Constant), NDVI, landform
According to Table 5, the highest standardized coef-
Table 5: Performance indices and coeflcients of variables for differ- ficients (β) for prediction of variables for tree height pre-
ent MLR models for the tree height prediction. diction using NDVI and landform was obtained in NDVI
(98.04).
Model Unstandardized Coeflcients t
B Std. Error
Constant −35.117 5.174 −6.787
1 NDVI 98.040 12.965 7.562 5 Conclusion
Landform −.209 .100 −2.096
a. Dependent Variable: tree height
The landform classes obtained in the southwest of the Fars
province. Landform classifications using TPI show that
the landform classification map of the study area was ten
dard errors of the regression coefficients, which can be classes; high ridges, midslope ridges, upland drainage,
used for hypothesis testing and constructing confidence upper slopes, open slopes, plains, valleys, local ridges,
intervals. The standardized coefficient (B) is what the re- midslope drainage and streams. From the analysis, NDVI
gression coefficients would be if the model were fitted to values for the study site ranges between −0.58 and 0.69.
standardized data, that is, if from each observation, the The relationship between landform classification and
sample mean is subtracted. In addition, the t statistic tests vegetation was investigated. The results obtained showed
the hypothesis that a population regression coefficient β is that the high ridges, midslope ridges, upland drainage, up-
0, that is, H0 : β = 0. It is the ratio of B to its standard er- per slopes, open slopes, plains and valleys classes consist
ror. Based on this, it was determined that the equation for of tree heights of > 20, 14–20, 11–14, 8–11, 4–8, 2–4 and

Unauthenticated
Download Date | 3/29/18 6:41 AM
308 | M. Mokarram and D. Sathyamoorthy

(a)

(b)

Figure 6: (a) NDVI values obtained for the study area (b) Dominant covers map.

Unauthenticated
Download Date | 3/29/18 6:41 AM
 Relationship between landform classification and vegetation |
< 2 m respectively. Ridge landforms (high, midslope and
local ridges) were found to have highest tree heights and
NDVI values. It was found that there are positive and sig-
nificant correlations between NDVI and tree height (r =
0.923), and landform and NDVI (r = 0.640). This shows
that landform classification and NDVI can be used to pre-
dict tree height in the area, with high value of R2 of 0.909
obtained this prediction. Using deep understanding of the
surface terrain characteristics could be detected potential
and specific constraints of the tree.
References
[1] Ahmad Zawawi A., Shiba M., Janatun Naim Jemali N., Land-
form Classification for Site Evaluation and Forest Planning: In-
tegration between Scientific Approach and Traditional Concept.
Sains Malaysiana. 2014, 43(3), 349–358.
[2] Hoersch B., Braun G., Schmidt U., Relation between land-
form and vegetation in alpine regions of Wallis, Switzerland.
A multiscale remote sensing and GIS approach. Computers
Environment and Urban Systems, 2003, doi: 10.1016/S0198-
9715(01)00039-4.
[3] Pfeffer K. E., Pebesma J., Burroug P. A., Mapping alpine vegeta-
tion using vegetation observations and topographic attributes.
Landscape Ecology 2003, 18, 759–776.
[4] Burr D., Baker V. R., Carling P., Megaflooding on earth and mars.
Cambridge, UK; New York: Cambridge University Press. 2009.
[5] Chalmers A. C., Erskine W. D., Keene A. F., Bush R.T., Relation-
ship between vegetation, hydrology and fluvial landforms on
an unregulated sand-bed stream in the Hunter Valley, Australia.
Austral Ecology, 2012, 37(1), 193–203.
[6] Diallo O., Diouf A., Hanan N.P., Ndiaye A. Y., Prevost Y., AVHRR
monitoring of savana primary production in Senegal, West
Africa: 1987-1988. International Journal of Remote Sensing.
1991, 12 (6), 1259–1279.
[7] Loučková B., Vegetation–landform assemblages along selected
rivers in the czech republic, a decade after a 500-year flood
event. River Research and Applications. 2012, 28, 1275–1288.
[8] Henrique C. E., Fátima Rossetti D., Zani H., Classification of Veg-
etation over a Residual Megafan Landform in the Amazonian
Lowland Based on Optical and SAR Imagery. Remote Sens. 2014,
6, 10931–10946.
[9] Zawawi A. A., Shiba M., Jemal N. J. N., Landform Classification for
Site Evaluation and Forest Planning: Integration between Scien-
tific Approach and Traditional Concept. Sains Malaysiana, 2014,
43(3), 349–358.
309
[10] Soufi M., Morpho-climatic classification of gullies in Fars
province, southwest of Iran. Isco 2004 – 13th International Soil
Conservation Organisation Conference – Brisbane, 2004.
[11] Weiss A. D., Topographic Position and Landforms Analysis.
Ecoregional Data Management Team the Nature Conservancy,
2005, 343–4345.
[12] Jenness J., Topographic Position Index (tpi_jen.avx) exten-
sion for ArcView 3.x”, v. 1.3a. Jenness Enterprises, 2010,
http://www.jennessent.com/arcview/tpi.htm.
[13] Tagil S., Jenness J., GIS- based automated landform classifi-
cation and topographic, Landcover, and geologic attributes of
landforms around the Yazoren Polje, Turkey. Journal of Applied
Sciences, 2008, 8(6), 910–921.
[14] Miguel A., Alloza G. J. A., Mayor A. J., Detection and mapping
of burnt areas from time series of MODIS-derived NDVI data
in a Mediterranean region. Central European Journal of Geo-
sciences. 2014, 6(1), 112–120.
[15] Rouse J. W., Haas R. H., Schell J. A., Deering D. W., Monitoring
vegetation systems in the great plains with ERTS. Third Earth
Resources Technology Satelite-1 Symposium, 1973, 309–317.
[16] Jackson R. D., Slater P. N., Pinter P. J., Discrimination of growth
and water stress in wheat by various vegetation indices through
clear and turbid atmospheres. Remote Sensing Environ, 1983,
13, 187–208.
[17] Silveira C. M. S., Tchepel O. A., Influence of the spatial resolu-
tion of satellite-derived vegetation parameters on the biogenic
Volatile Organic Compounds (VOC) emission modeling. Central
European Journal of Geosciences, 2014, 6(1), 104–111.
[18] Tucker C. J., Red and Photographic Infrared Linear Combina-
tions for Monitoring Vegetation, Remote Sensing of Environ-
ment, 1979, 8(2), 127–150.
[19] Calera A., Martínez C., Melia J., A procedure for obtaining green
plant cover: Relation to NDVI in a case study for barley. Interna-
tional Journal of Remote Sensing, 2001, 22 (17), 3357–3362.
[20] Carlson T. N. Y., Ripley D. A., On the relation between NDVI, frac-
tional vegetation cover, and leaf area index. Remote Sensing of
Environment. 1997, 62, 241–252.
[21] Franklin S. E., Hall R. J., Moskal L. M., Maudie A. J., Lavigne M. B.,
Incorporating texture into classification of forest species com-
position from airborne multispectral images. International Jour-
nal of Remote Sensing, 2000, 21, 61–79.
[22] Wylie B. K., Meyer D. J., Tieszen L. L. Y., Mannel S., Satellite map-
ping of surface biophysical paameters at the biome scale over
the North American grassland. A case study. Remote Sensing of
Environment, 2002, 79, 266–278.
[23] Gunst R. F., Mason R. L., Regression Analysis and Its Application.
Marcel Dekker Inc., New York, 1980.
Unauthenticated
Download Date | 3/29/18 6:41 AM