Content uploaded by Dulan Jayasekara
Author content
All content in this area was uploaded by Dulan Jayasekara on Feb 13, 2022
Content may be subject to copyright.
WILDLANKA Vol.9, No.1, pp. 122 - 135, 2021.
Copyright 2021 Department of Wildlife Conservation, Sri Lanka.
,and
MAPPING THE VEGETATION COVER AND HABITAT
CATEGORIZATION OF MADURU OYA AND HORTON PLAINS
NATIONAL PARKS USING LANDSAT 8 (OLI) IMAGERY
TO ASSIST THE ECOLOGICAL STUDIES
ABSTRACT : Availability of accurate and detailed vegetation and habitat maps is an essential requirement
in the modern ecological studies. Geo-referenced vegetation maps of Sri Lanka’s protected areas are
scarce and only a handful of maps are available with detailed vegetation patterns. Since majority of
the current ecological research work takes place within the protected area network of the island, we
identied that preparing updated maps to assist the ecologists as well as the park management is highly
important. In this study we developed vegetation and land cover maps for Maduru Oya and Horton
Plains national parks in Sri Lanka. The procedure was based on supervised classication of Landsat 8
multispectral image data. Classication was complemented by ground truth data obtained through eld
surveys. The present study generated accurate (overall accuracy - 92-93%; Kappa – 0.89) and detailed
vegetation/land cover maps and habitat types were proposed based on vegetation patterns. The results
provide accurate information for ecologists and decision-makers to assist future research work as well as
conservation and management of protected areas concerned.
KEY WORDS:
cover
Availability of accurate and detailed
vegetation and habitat maps is an essential
requirement in the modern ecological studies.
A vegetation map includes critical information
from land management, detecting land cover
changes, understanding biodiversity patterns
and conservation planning (Dias , 2004).
(2019) vegetation
maps are based on two essential elements;
vegetation mapping group together similar plant
their arrangement pattern with spatial reference
ecological studies, the availability of vegetation
maps makes the task of the researchers easier.
Furthermore, availability of maps with detailed
vegetation patterns is scarce. As many ecological
researches are being conducted in the protected
areas of the island presently, we saw that there
is a crucial need of preparing updated maps to
assist ecologists as well as park management.
land cover mapping of forested areas of Sri
(1993) to the recent publications of Rathnayake
most of the previous work has been focused
on large scale patterns of general land cover
rather than localized rigorous assessments. We
March, 2021] MAPPING THE VEGETATION COVER AND HABITAT CATEGORIZATION 123
(2006; 2007), Abayasinghe (2014) and
Ranawana (2014) has generated vegetation
maps along with change detection for the years
(2015) have
and during the procedure several vegetation
maps have been prepared using imagery from
vegetation maps for some of the Department of
Wilpattu (Sandamali and Welikanna, 2018),
Sanctuary (Khanh and Subasinghe, 2018).
We observed two main drawbacks in most of
these past maps which limit their applicability
of high resolution digitally accessible maps, 2)
to complement the remote sensing techniques.
needs of ecological research, management and
conservation of protected areas, we developed
literature, we propose habitat categorization for
the two protected areas considered.
multispectral satellite data has enabled
the analysts to obtain spectral signature of
bodies, soil types, rocky areas, roads/other
when forested landscapes are concerned,
calculation is the most widely used method
red and near-infrared (NIR) wavelengths, from
healthy and highly photosynthetic vegetation is
We incorporated remote sensing data with
characteristics were obtained at ground level
to complement the remote sensing data.
. (2007)
to make the results understandable by a wider
of vegetation categories, including multiple
attributes of vegetation (rather than in depth
phytosociological descriptions) and building
upon the already available knowledge from
of a physiognomic nature which is based on
the basic physical structure of the vegetation
(forest, shrubland, grassland, etc.) and the main
growth forms (trees, shrubs, grasses, etc.) of
the dominant or co-dominant species in the
vegetation formation (Ichter , 2014).
which an ecologist would focus on were also
Ichter , (2014) vegetation composition
form the vegetation categories can be used to
segregate the habitat types. We believe that the
output generated through our study would be
meaningful in a remote sensing as well as in an
ecological point of view.
Study site
Horton Plains National Park
located on the southern plateau of the central
highland plateau, the altitude ranging from
WILDLANKA [Vol. 9 No. 1124
by Kirigalpoththa (2,389 m) to the west and
are respectively the second and third highest
upgraded to national park status on 11th
1988, having previously been declared a nature
in natural habitats comprises upper montane
grasslands, with a narrow ecotone of shrubs and
cause of which is uncertain but may be related
to water stress, soil conditions, air pollution
, 2009) or
natural phenomena.
FIGURE 01:
by Dulan Jayasekara, originally published in Jayasekara 2020)
Maduru Oya National Park
(588 km) lies in the districts of Ampara and
cultivation resulting in secondary forests and
vast stretches of open plains dominated by
Determination of vegetation and habitat
categories
Initially we referred to available literature
and prepared priory vegetation/habitat categories
the main vegetation types were established, on
a physiognomic basis (Dias ., 2004) with
maps available online. We established a grid
of 1×1 km plots covering the area of available
protected areas to remove some visible errors to
and study generated map areas and perimeters.
We randomly selected sampling plots and
established sampling quadrates of 10×10 m
within the larger 1×1 km plots. A total of 77
and 55 quadrates were sampled respectively
March, 2021] MAPPING THE VEGETATION COVER AND HABITAT CATEGORIZATION 125
TABLE 01: Summary of vegetation and environmental parameters and standard methods used (X
indicates whether each parameter was obtained at the selected site)
1 were obtained using standard methods. We
clusters of similar vegetation/land cover types.
Environmental Abbreviation Method followed MONP HPNP
parameter
Stem density of SD1 Average distance to the X X
plants <10cm dbh nearest woody plant with
(dbh-diameter at dbh <10cm calculated as
breast height) 1/mean area [distance]
Stem density of SD2 Average distance to the X X
plants >10cm dbh nearest woody plant with
dbh >10cm calculated as
1/mean area [distance]
(ocular estimation)
visibility (%)
cover (%) (ocular estimation)
Rock availability RA Evaluated on a scale of 1-10 X X
X
indicate June)
Pre-analysis for supervised classication
We utilized multi-spectral images from
database (http://glovis.usgs.gov/). Images
datasets with low cloud cover were considered
WILDLANKA [Vol. 9 No. 1126
and 7 were to discriminate between water and
was calculated for each image according to the
2016).
determined the number of vegetation classes
Supervised classication using Landsat 8
images
samples were selected from the data obtained
2016 to January 2021. We supplemented the
for inaccessible terrain and some known
locations.
Accuracy assessment of classied maps
Accuracy assessment was conducted to
the possibility of bias, in addition to the actual
ground survey points, we generated random
PCA clustering of vegetation and environ-
mental characteristics
shrublands, grasslands and rocky outcrops
cluster was overlapping with grasslands and
loading plot (Figure 2a) the important factors
with >10cm dbh), litter cover and litter depth.
Rocky outcrops were clearly separated based
low canopy, pigmy cloud forest, carpet grass,
tussock grass, marshes/dwarf bamboo, rocky
score plot indicated overlapping between the
two grass types and marshes/dwarf bamboo.
canopy cover, litter cover, litter depth and stem
high horizontal visibility and ground vegetation
easily categorized based on soil moisture
content (Figure 3a).
March, 2021] MAPPING THE VEGETATION COVER AND HABITAT CATEGORIZATION 127
FIGURE 02:
WILDLANKA [Vol. 9 No. 1128
FIGURE 03:
March, 2021] MAPPING THE VEGETATION COVER AND HABITAT CATEGORIZATION 129
TABLE 02:
FIGURE 04:
represent the legislated park boundary)
Spatial extents of land cover classes
with 37% accounting for 215,415ha of land
area (Figure 4). It was followed by dense
was the cloud forest (50%) covering an area of
areas repesents 25% (778ha) of park area.
Vegetation/Land cover type Percentage land cover (%) Area (ha)
MONP
Dense forest (DF) 29 170,171
WILDLANKA [Vol. 9 No. 1130
HPNP
FIGURE 05:
represent the legislated park boundary)
March, 2021] MAPPING THE VEGETATION COVER AND HABITAT CATEGORIZATION 131
TABLE 04:
TABLE 03:
Accuracy assessment of results
Class DF SL GL RO/BL R/WL Total U_Acc Kappa
Dense Forest 56 2 0 0 0 58 0.97
Shrubland 0 67 3 2 1 73 0.92
Rocky outcrops 0 3 1 17 1 22 0.77
Reservoir/Wetland 0 0 0 0 18 18 1
0.92
Kappa 0.89
Class CF CFD PCF TG CG M/DB OV RO/BL Total U_Acc Kappa
0.93
Kappa 0.89
and landcover maps for two of the prominent
remote sensing based approches are common
WILDLANKA [Vol. 9 No. 1132
provide more accurate results. It can be clearly
seen in the results of the accuracy assessments.
In our clustering process, we selected
vegetation and environmental parameters that
sophesticated phytosociological investigations.
audience and even the general management
decision making.
seperated based solely on ground survey
parameters while the shrubland habitat was
this result was actually related to the nature
of the shrublands which can be considered
a sucessional stage between the grassland
detailed and accurate vegetation and land cover
.,
plantation which currently consist of many dead/
dying teak plants. Since we did not observe any
and due to the limited distribution, it was not
highly important to conserve this area with high
priority in order to facilitate the biodiversity to
results of the present study and considering the
4. Rocky outcrops, 5. Reservoir/Wetlands.
vegetation features and landscape features along
topographical and climatic conditions of the
together a summarized and detailed account
was absent in that analysis which was focused
on identifying the cloud forest die-back. Work
cloud forest die-back areas probably due to
the low resolution and limited interpretation of
satellite images as mentioned by Abayasinghe
(2014) where they have generated
more detailed and accurate vegetation maps.
Futhermore, the images that were used in these
the present study (used 2017 satellite images)
can be considered timely to compare the change
that has occurred after about 10 years gap from
the previous analysis. We observed a slight
increase in the cloud forest cover (~3%) from
changes in the vegetation. Interestingly, our
results suggest a considerable increase (~4%) in
the carpet grass cover and a reduction in tussock
, 2002) and we posit tentatively that there
could be a possible increse in the sambar deer
. 2015) we
detailed investigations, there is the possibility
March, 2021] MAPPING THE VEGETATION COVER AND HABITAT CATEGORIZATION 133
subhabitats based on vegetation maps provided
in the present study.
comparisons with past and future vegetation/land
of vegetation classes and habitat types would
support the future ecological research work
maps generated through the present study can be
used for management and conservation decision
making. We recommend similar initiations in
use of the available technological tools to
implement modern solutions .
We appreciate the generous cooperation of
to conduct this research. We would also like to
Jayewardenepura (Department of Zoology and
to conduct this research under the university
management: a practical geoinformatics
approach.
, 2:113-131.
Ellepola, (2015). Evaluation of damage
caused by elephants (
) to the woody vegetation in
,
3(1):20-30.
and manual for standardized habitat
mapping. 7:55-74.
10.1093/obo/9780199830060-0115
Island (Azores).
, 13(8):1519-1539.
(2006).
Environment and Natural Resources,
p.40.
D.S. Karunarathna, (2015). Avifaunal
diversity in the peripheral areas of the
With conservation and management
implications.
, 8(2):121-132.
WILDLANKA [Vol. 9 No. 1134
Estimation of above ground biomass
RADAR remote sensing data.
, 26(4):608-623.
Ichter, J., D. Evans, and D. Richard, (2014).
densities and conservation assessment
of three threatened agamid species in
4(3):70-79.
forest management and monitoring in Sri
, pp.13-15.
Remote Sensing,
, 8(5):42-52.
. [online]
ianlockwood.blog/2019/09/16/
preliminary-analysis-of-land-cover-
dove-imagery/> [Accessed 14 July
2021].
Kuruppuarachchi, (2014). Detecting
, 4(1):50-58.
mapping in cropland dominated area
using information on vegetation
images.
, 1(1): 1-16.
environment for statistical computing.
(
and Environment Symposium.
Anthropogenic or natural phenomenon?.
. 13:23-45.
,
2:103-112.
series. , 9(1):1-19.
, 2(10):473:482.
March, 2021] MAPPING THE VEGETATION COVER AND HABITAT CATEGORIZATION 135
, 8(4):37-
40.
obo in Ecology. doi: 10.1093/
obo/9780199830060-0176
Received Date:
Accepted Date: