The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XL-8, 2014

ISPRS Technical Commission VIII Symposium, 09 – 12 December 2014, Hyderabad, India

Time-Series analysis of MODIS NDVI data along with ancillary data for Land use/Land
cover mapping of Uttarakhand
Sandeep Kumar Patakamuria*, Shefali Agrawalb, and M. Krishnavenia

a Centre for Water Resources, Anna University, Guindy, Chennai, India-600 025
b IndianInstitute of Remote Sensing, 4-Kalidas Road, Dehradun, India-248001
*[email protected]

KEY WORDS: Land use/land cover, NDVI, MODIS, Phenology, MOD13Q1.

ABSTRACT:

Land use and land cover plays an important role in biogeochemical cycles, global climate and seasonal changes. Mapping land use
and land cover at various spatial and temporal scales is thus required. Reliable and up to date land use/land cover data is of prime
importance for Uttarakhand, which houses twelve national parks and wildlife sanctuaries and also has a vast potential in tourism
sector. The research is aimed at mapping the land use/land cover for Uttarakhand state of India using Moderate Resolution Imaging
Spectroradiometer (MODIS) data for the year 2010. The study also incorporated smoothening of time-series plots using filtering
techniques, which helped in identifying phenological characteristics of various land cover types. Multi temporal Normalized
Difference Vegetation Index (NDVI) data for the year 2010 was used for mapping the Land use/land cover at 250m coarse
resolution. A total of 23 images covering a single year were layer stacked and 150 clusters were generated using unsupervised
classification (ISODATA) on the yearly composite. To identify different types of land cover classes, the temporal pattern (or)
phenological information observed from the MODIS (MOD13Q1) NDVI, elevation data from Shuttle Radar Topography Mission
(SRTM), MODIS water mask (MOD44W), Nighttime Lights Time Series data from Defense Meteorological Satellite Program
(DMSP) and Indian Remote Sensing (IRS) Advanced Wide Field Sensor (AWiFS) data were used. Final map product is generated
by adopting hybrid classification approach, which resulted in detailed and accurate land use and land cover map.

1. INTRODUCTION Various Global land cover data sets were prepared using multi
temporal AVHRR NDVI data sets (De Fries et al., 1998; IGBP
Land Use and Land Cover (LULC) play a vital role in DISCover, Loveland et al., 2000; UMD GLCC, Hansen et al.,
understanding the complex earth’s ecosystem processes, 2000). Major disadvantages of using AVHRR data is its poor
biogeochemical cycles (Achard et al., 2004; Liu et al., 2008; geometric accuracy and limited radiometric calibration (Cihlar
Sellers et al., 1997), climate change studies (Liu et al., 2004) et al., 1997; Meyer et al., 1995).
and deforestation studies(Hansen et al., 2008; Skole and
Tucker, 1993; Steininger et al., 2002). Land use and land cover Global Land Cover (GLC) 2000 maps were produced making
information is extensively used by various governmental and use of SPOT VEGETATION data (Agrawal et al., 2003;
non-governmental organizations and thus placing a great Bartholomé and Belward, 2005; Han, 2004; Mayaux et al.,
demand on detailed, reliable and up to date LULC maps (Perera 2006; Stibig et al., 2007). NDVI data derived from VGT sensor
and Tsuchiya, 2009). International Geosphere Biosphere is also utilized for national level mapping of vegetation cover
Program (IGBP 1990), Intergovernmental Panel on Climate (Vancutsem et al., 2009; Xiao et al., 2002).
Change (IPCC 1990), National Research Council (NRC 2005),
and Global Terrestrial Observation systems (GTOS 2009) etc., Land use land cover mapping is also possible using multi
have identified land use and land cover as an important driver resolution data obtained from various satellite platforms like
for climate change. Land use/land cover also impacts the socio- Landsat (Bakr et al., 2010; Huang et al., 2009; Olthof and
economic conditions of the people (Hietel et al., 2005; Lo and Fraser, 2007; Steininger et al., 2002), Indian Remote Sensing
Yang, 2002; Ribeiro and Lovett, 2009; Swetnam et al., 2011) IRS-1C satellite based Wide Field Sensor (WiFS) data with 189
and hence detailed scientific study is required in this field for meter resolution (Joshi et al., 2001, 2006) and Indian Remote
better environmental planning, management and conservation. Sensing IRS-P6 satellite based Advanced Wide Field Sensor
(AWiFS) data with 56 meter resolution (Kandrika and Roy,
Before the era of satellite remote sensing, mapping of land 2008). However, these data sets suffer limitations in terms of
resources started with manual interpretation of aerial spatial coverage, temporal resolution and availability of cloud
photographs (Colwell, 1960). These techniques were tedious free data (Asner, 2001; Hansen et al., 2008; Song et al., 2001).
and time consuming. Global land cover data sets were first
prepared using various published maps, atlases, national sources Several researchers demonstrated the use of MODIS vegetation
and ground level surveys (Matthews, 1983; Olson et al., 1983). Indices to produce land use land cover maps (Clark et al., 2010;
The global land cover maps generated by this method suffered García-Mora et al., 2012; Klein et al., 2012; Thenkabail et al.,
lack of consistency in classification schema used, differences in 2005; Wardlow et al., 2007; Zhang et al., 2008). In the present
spatial resolutions and variables in measurement study, MOD13Q1 vegetation indices product at 250m
techniques(Running et al., 1994; Townshend et al., 1991). resolution was used.

Land use land cover mapping of larger areas using satellite Present study has the advantage of (i) Using temporal plots of
remote sensing started with National Oceanic and Atmospheric various land cover classes to understand the status of the
Administration (NOAA) Advanced Very High Resolution vegetation, (ii) Smoothening the time-series curves to eliminate
Radiometer (AVHRR) data. (Nemani and Running, 1997). the erroneous interpretations, (iii) Use of ancillary data to

_____________________________
*Corresponding Author
This contribution has been peer-reviewed.
doi:10.5194/isprsarchives-XL-8-1491-2014 1491
 The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XL-8, 2014
ISPRS Technical Commission VIII Symposium, 09 – 12 December 2014, Hyderabad, India

support classification decision making and (iv) Use of land use (MOD44W) product derived from MODIS and SRTM data
land cover maps derived from AWiFS data alongside with high (Carroll et al., 2009) downloaded from LP DAAC, Defense
resolution Google earth imagery as reference data for Meteorological Satellite Program (DMSP) Operational
classification accuracy assessment. The approach presented in Linescan System (OLS) nighttime lights time series data
this study is expected to have better accuracy and is adoptable downloaded from National Geophysical Data Center
for land use land cover mapping at different geographical (http://ngdc.noaa.gov/eog/dmsp/downloadV4composites.html),
locations. Advanced Wide Field Sensor (AWiFS) imagery downloaded
from Bhuvan portal developed by National Remote Sensing
2. STUDY AREA AND DATA DESCRIPTION Centre Open Earth Observation Data Archive (NOEDA) to
visualize and disseminate earth observation data (http://bhuvan-
Uttarakhand is a state located in the northern region of India noeda.nrsc.gov.in/download/download/download.php), and
with a geographical extent of 53,483 km2. The state is situated Land use land cover maps at 1:250,000 scale derived using
between 28o 43' to 31o 27' N latitudes and 77o 34' to 81o 02' E AWiFS, obtained from National Remote Sensing Centre,
longitudes (Figure. 1). It is bordered by the Tibet autonomous Department of Space, Government of India, Hyderabad.
region on the north, Nepal on the east and the Indian states of
Uttar Pradesh to the south, Haryana to the west and Himachal
Pradesh to the North West. Climate of the state falls into
temperate zone and tropical zone with temperatures varying
from sub-zero to 43o C. Elevation values ranges from 200 m to
8000m above mean sea level and supports a variety of flora and
fauna (Forest Survey of India, 2011; Ministry of Economics and
Statistics, 2010). Topography and climate plays a major role in
ecology, socio economic status and culture of the people in this
region (SoE, 2004). Agriculture is the major sector in this area
and many industries are forest dependent. There are over fifteen
important rivers and over a dozen glaciers in the state.
Uttarakhand has over two hundred medium to large sized
hydro-projects and these projects highly influence the
environment and living conditions of the people (Rana et al.,
2007; WSMD, 2009).

The state is housing six national parks, six wild life sanctuaries
and two conservation reserves. The state is one of the major
contributors of the tourism sector in India, attracting several
domestic and international tourists. The state also has a rich and
ancient cultural heritage and is also called as Dev-Bhoomi
which means land of Gods. Uttarakhand state is sensitive to
natural calamities like flash floods; landslides and earth quakes
and requires well planned environmental management system.
In September 2010 Uttarakhand has faced floods due to heavy
rain fall resulting in loss of lives and property (Sharma,
2012).This study can help planners and decision makers to
Figure 1. Location map of the study area- Uttarakhand
better manage and protect the land resources of Uttarakhand.

In the present study, MODIS 250m resolution vegetation

indices product MOD13Q1 is used for mapping the land use 3. METHODS
land cover of Uttarakhand. The product contains 16-day
The methods adopted in this study can be broadly divided into
composite of Normalized Difference Vegetation Index (NDVI),
five stages: (i) Data preparation, (ii) Image classification, (iii)
Enhanced Vegetation Index (EVI), blue, red, near infrared
Generating time-series plots and smoothening, (iv) Grouping
(NIR), mid-infrared (MIR) and pixel reliability (Huete et al.,
the classes based on similarity and (v) Accuracy assessment.
2002, 1999; Solano et al., 2010). Entire study area is covered by
The flow chart depicted in Figure. 2 represents the systematic
two MODIS tiles numbered H24V05 and H24V06. “Data
approach adopted in the present study.
required for the present study was obtained through the online
Data Pool at the NASA Land Processes Distributed Active
Archive Center (LP DAAC), USGS/Earth Resources 3.1 Data Preparation
Observation and Science (EROS) Center, Sioux Falls, South A total of 23 tiles for H24V05 and 23 tiles of H24V06 covering
Dakota (https://lpdaac.usgs.gov/get_data)" with the help of the study area were downloaded for the year 2010 starting from
United States Geological Survey (USGS) Earth Explorer (EE) January to December. These images were then mosaicked
tool (http://earthexplorer.usgs.gov/). together to form 23 images, and later layer stacked to form
annual composite image.
Ancillary data collected for the study includes Digital Elevation
Model (DEM) data derived from Shuttle Radar Topographic
3.2 Image classification
Mission (SRTM) (Farr et al., 2007; Jarvis et al., 2008)
downloaded from Consortium for Spatial Information website In the present study ISODATA clustering technique was used to
(http://www.cgiar-csi.org/data/srtm-90m-digital-elevation- exploit its ability to further split and merge the clusters (Jensen,
database-v4-1#download). MODIS land water mask

This contribution has been peer-reviewed.

doi:10.5194/isprsarchives-XL-8-1491-2014 1492
 The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XL-8, 2014
ISPRS Technical Commission VIII Symposium, 09 – 12 December 2014, Hyderabad, India

1996; Richards and Jia, 1999). MODIS data was grouped to 150 S. No. Class label Criteria
clusters with 25 iterations and a convergence threshold of 0.95. 1 Snow/Ice Snow/Negative NDVI values,
The classification resulted in 105 useful clusters with non-zero Higher altitude zones.
values. 2 Barren land Waste lands do not support any
vegetation. Very low NDVI values
3.3 Generating time-series plots and smoothening (<0.2), no snow cover or
vegetation observed during image
Once the classification process was finished, a set of four interpretation.
random samples were collected from each of the 105 clusters 3 Crop land Single crop and double/triple crop
and multi-temporal NDVI plots were generated for each of the areas and fallow lands were
420 sample locations. identified based on the growth
cycles observed in an annual year.
There are several factors such as atmospheric conditions, sensor 4 Scrub Scrub lands/degraded forests
viewing geometry, cloud cover, aerosol concentrations, off- lands/Degraded contain shrubs and woody
nadir viewing and low sun zenith angles that causes noise in forest vegetation and were identified by
NDVI time series (Gutman, 1991; Holben and Fraser, 1984; visual interpretation and NDVI
Kobayashi and Dye, 2005; Li and Strahler, 1992). Savitzky- plots showed non-uniform pattern.
Golay filtering technique was used in this study to filter the 5 Grass land High NDVI values observed
noise in NDVI response. It is a simplified least squares fit during small portion of the year.
convolution for smoothing and computing derivatives of a High resolution image
spectrum (Savitzky and Golay, 1964). interpretation, reference maps and
expert discussions helped in
3.4 Grouping the classes based on similarity delineating grasslands.
The clusters with similar temporal characteristics were grouped 6 Water bodies Rivers, lakes and reservoirs etc.,
together based on the theory that similar land cover classes are classified as water bodies.
exhibit similar temporal patterns. Time-series patterns of NDVI values are low throughout
sample points were extracted from each cluster, elevation data, the year. MOD44W water mask
image interpretation of AWiFS false color composites and product helped in delineating water
Google earth visual interpretation helped in proper bodies.
identification of land use/land cover classes. Table.1 below 7 Built-up land Built-up land consisted of roads
shows the criteria followed in attributing suitable land use/land and buildings. NDVI values were
cover class to each cluster. uniform throughout the year.
Urban mask derived from DMSP-
3.5 Accuracy assessment OLS nighttime lights data helped
in identifying built-up land.
Accuracy assessment is a key stage in thematic mapping using 8 Evergreen High NDVI values, perennial
remotely sensed data. Classification accuracy assessment forest nature was observed in NDVI time
indicates the qualitative aspects of the land cover map. It series.
determines whether the generated map serves the purpose of the 9 Deciduous High NDVI values were observed
users or not. Accuracy assessment is a complex process forest and during autumn season, NDVI
involving several problems of consideration (Foody, 2008, values started decreasing in
2002). autumn season due to shredding of
leaves.
In this study, accuracy assessment is carried out using error Table 1. Criteria followed in identification and attribution of
matrix. It accounts for commission errors, as well as omission land use/ land cover class names to each cluster
errors and kappa statistics which reflects the difference between
actual agreement and the agreement expected by chance 4. RESULTS AND DISCUSSIONS
(Lillesand and Kiefer, 2008).
4.1 Plotting NDVI time-series and smoothening
Reference data for accuracy assessment is collected with the
Different land cover classes portray dissimilar temporal
help of National land use and land cover maps derived by using
characteristics at various stages of the year. This helps in
multi-temporal AWiFS for the year 2009-2010 and Google
identification and classification of land use/land cover using
Earth (GE) image interpretation. Stratified random sampling is
multi-temporal remotely sensed satellite imagery. A matlab
used with a minimum of 25 samples per class, and a total of 300
routine was coded to generate the plots and to smoothen their
sample points were generated. Grids were generated with 250 x
responses. Time series plots were generated for a set of four
250 m size for each of the sample point to match the pixel size
randomly selected points from each of the 105 clusters.
of the MODIS NDVI data and exported to Google Earth for
Savitzky-Golay filtering was applied on the plots to obtain
visualization.
smoothened, noise free values to ease the interpretation process
as this method is found to be suitable for smoothening the time
series without much disturbing the seasonality parameters
(Chen et al., 2004; Jonsson and Eklundh, 2002). Time-series
plots of various land use/land cover classes and their
corresponding smoothened plots were shown in the Figure. 2
(a)-(h).

This contribution has been peer-reviewed.

doi:10.5194/isprsarchives-XL-8-1491-2014 1493
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XL-8, 2014
ISPRS Technical Commission VIII Symposium, 09 – 12 December 2014, Hyderabad, India

(f)
(a)

(g)
(b)

(h)
(c)
Figure 2 (a)-(h). NDVI plots for various land use/land cover
classes namely (a).Snow/Ice, (b).Barren land, (c).Grass
land, (d).Shrub land/Degraded forest, (e).Crop land, (f).Built-up
land, (g).Deciduous forest, and (h). Evergreen forest, and their
smoothened response to Savitzky-Golay filer.

Time-series NDVI profiles for the Snow/ice areas were

observed to be negative (<0) throughout the year (Figure. 2 (a)).
Barren lands support minimal to no vegetation, thus NDVI plots
contained low (<0.1) and fairly uniform values. In areas where
(d) the barren lands contained sparse vegetation, NDVI values are
observed to be increasing with the growth of the vegetation
(Figure. 2 (b)). Vegetation indices plots for grass lands samples
showed lower values (<0.4) in the present observations (Figure.
2 (c)). At some sample locations, NDVI plots of the grasslands
were confused with the temporal pattern of seasonal crops. In
such cases, the grass lands were differentiated based on its
lower NDVI values and shorter growth periods. Shrub
lands/Degraded forest is a mixed land cover class, whose NDVI
plots showed a decreasing trend and shorter growth periods
(Figure. 2 (d)). Crop lands in the study area were identified with
steep, high amplitude and sudden decrement in the vegetation
indices plots attributed to the respective growth, maturity and
(e) harvest of the agricultural practices. In the study area, fallow
lands, single crop pattern, double crop pattern and slope
agriculture patterns were identified and grouped as crop lands

This contribution has been peer-reviewed.

doi:10.5194/isprsarchives-XL-8-1491-2014 1494
 The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XL-8, 2014
ISPRS Technical Commission VIII Symposium, 09 – 12 December 2014, Hyderabad, India

(Figure. 2 (e)). Built-up lands consist of buildings, roads and a

small portion of urban vegetation. NDVI values are of lower
amplitude (<0.4) and plots followed fairly uniform pattern
(Figure. 2 (f)). Decreasing pattern of the NDVI plots in the
deciduous class is associated with shredding of leaves during
autumn season (Figure. 2 (g)). Evergreen forests showed high
NDVI values throughout the year and minute amplitude
disturbances are in response to the canopy moisture content
during wet seasons (Figure. 2 (h)).
(g) (h)
4.2 Ancillary reference collection
Figure 3 (a)-(h). Various land use/land cover classes as
Present study made use of several ancillary reference datasets visualized in Google Earth (a).Snow/Ice, (b).Barren lands,
like SRTM DEM, MODIS Water mask, DMSP-OLS nighttime (c).Grass lands, (d).Shrub lands/Degraded forest, (e).Crop
lights time series data, AWiFS imagery and Google Earth© lands, (f).Built-up land, (g).Deciduous forest, (h).Evergreen
high resolution imagery. All these data sets were re-projected to forest.
the spatial resolution of MODIS imagery. SRTM DEM data
was used to identify elevation values which helped in
differentiating land cover classes. Water mask and nighttime Reference grids were created to match the spatial resolution of
lights data helped in classification of water bodies and built-up MODIS pixels. Many researchers have used google earth
lands respectively. imagery for classification and validation of land use/land cover
maps (Clark et al., 2010; Colditz et al., 2012). When a mixed
For collecting Google earth reference data, rectangular grids class is observed inside the grid, the majority land cover type
were generated with side length of 250 meters around each of (>70%) present in the grid is allocated for the sample. Snow/Ice
the 420 sample points that were collected during time-series class was observed at higher altitudes and fresh snow appeared
plots generation. Various land cover classes of the present brighter than permanent snow. Barren lands support minimum
study were viewed in Google earth as shown in Figure. 3. to no vegetation cover and these samples were easily
differentiated from other classes based on texture information.
Shrub lands/degraded forest is a land cover class consisting of
trees and shrubs. Crop lands were identified by rectilinear
shapes and plow lines and slope agriculture was identified with
the help of elevation information. Identification of grass lands
was a difficult task experienced in the study. Grass lands were
mixed with crop lands and shrub land classes, and
differentiating these classes was found to be a challenging task.
Built-up lands contained man made buildings and road
networks. Evergreen forests were differentiated from deciduous
frosts with the help of elevation information. Deciduous forests
(a) (b)
are located at the lower altitude zones compared to evergreen
forests as deciduous forest could not adapt to the cold and dry
weather at high altitudes. No shedding of leaves was observed
in evergreen forests during autumn season.

4.3 Final Classification and map compilation

All the clusters belonging to a particular land use/land cover

class were recoded at the final stage of classification Urban
mask derived from DMSP-OLS Night time lights data and
water mask derived from MOD44W product were appended to
(c) (d) the final land use/land cover map of Uttarakhand for the year
2010. Map generated in the present study is presented on a
reduced scale in Figure. 4.

Snow/ice cover was evident in the higher altitudes and

contributed about 12.39% of the study area. Ever green forests
are evident throughout the state and constituted about 24.08%
of the area. Deciduous forests covered about 13.07% of the
area. Single crop lands, double/triple crop lands and fallow
lands together are labeled as crop lands accounted for 20.22%
of the entire area. Shrub/degraded forests covered 12.5% of the
area. Water bodies contributed 0.43% of the area and built-up
(e) (f)
lands occupied about 0.16% of the area. Grass lands and barren
lands accounted for 9.56% and 7.55% of Uttarakhand area
respectively.

This contribution has been peer-reviewed.

doi:10.5194/isprsarchives-XL-8-1491-2014 1495
 The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XL-8, 2014
ISPRS Technical Commission VIII Symposium, 09 – 12 December 2014, Hyderabad, India

Qualitative aspects of the thematic mapping effort were

validated using error matrix (Lillesand and Kiefer, 2008)
generated with the help of three hundred sample points obtained
by stratified random sampling method. NRSC-ISRO land
use/land cover map for the year 2009-2010 was used as
reference image for the purpose of validation. Rectangular
reference grids were generated for all three hundred points with
dimensions of 250m x 250m for the purpose of visualization on
Google Earth. Spatial distortions occurred in the google earth
reference will be negligible compared to the pixel size of the
imagery used in the study (Clark et al., 2010). Table. 3
represent the error matrix, producers accuracy, users accuracy
and kappa statistics of the accuracy assessment process.

Error matrix obtained in the present study clearly highlights the

complex nature of accuracy assessment. Present study
incorporated careful examination of sample locations on
Figure 4. Land use/land cover map of Uttarakhand for the year multiple source like published reference maps, google earth as
2010 on a reduced scale well as AWiFS imagery for assuring the quality of
classification. The result showed an overall accuracy of 78.67%
with over all kappa statistics of 0.7600.
4.4 Evaluation of mapping results
5. CONCLUSION
Results obtained in the present study were evaluated in both
qualitative and quantitative aspects. Reference map classes During the land use and land cover mapping of Uttarakhand,
were aggregated to match the classification schema of the barren lands, dry river beds, fallow lands and portions of the
present study for quantitative validation. Area statistics for all built-up land exhibited similar temporal pattern at some of the
the classes were calculated for the present map and compared locations. In order to better delineate Built-up lands and Water
with the area extent of reference map obtained from Bhuvan bodies, the study made use of urban mask derived from DMSP-
portal (Aggarwal et al., 2012) as shown in Table. 2. Results OLS nighttime lights time series data and water mask derived
were also compared with Forest Survey of India publication from MOD44W product. Identification of grass lands and shrub
(FSI) statistics published in India State of Forest report 2011. lands/mixed forest classes is observed to be a difficult task as
the temporal patterns were difficult to interpret and differentiate
Area statistics in these classes. It is thereby suggested to consider for more
from NRSC number of sample locations at these clusters. Elevation
Area statistics land use/land information obtained from SRTM DEM data helped in
Land use/land from the present cover map differentiating various vegetation classes. Prior knowledge
cover class study (sq.km) (sq.km) about the study area and suitable reference datasets also plays a
Crop lands 10815 11982 major role in improving the quality of classification. It is of
Evergreen prime importance to have a clear definition and classification
forest 12881 12902 criteria for each land use/land cover type used in a particular
Deciduous mapping exercise.
forest 6993 6274
Shrub/Degrade Time series plots generated in the study may contain
d forest 6689 6470 disturbances due to sensor geometry, atmospheric errors and
Water bodies 86 62 sensor calibrations. Savitzky-Golay filtering was adopted in the
Built-up lands 231 204 present study to smoothen the time-series curves and there by
Grasslands and avoid erroneous interpretations. The study incorporated the
other waste visual interpretation of google earth imagery in classification
lands 15788 15589 and accuracy assessment phases. With recent advancement in
web based GIS it would be a great opportunity to deploy crowd
Total area 53483 53483
sourcing in which experts from the field, research collaborators
Table 2. Area statistics of land use/land cover map derived in
and volunteers can collectively work at various stages of land
the present study compared with reference map
cover classification and validation.
It is observed that the results obtained are in close agreement
Land use/land cover map generated in the present study was
with the area statistics of the reference data. Minor over
evaluated in both qualitative and quantitative aspects. Over all
estimation in water bodies and built-up lands are attributed to
accuracy of the present map is found to be 78.67% with over all
the use of urban mask and water mask in delineating these land
kappa coefficient of 0.7600. The methods and techniques used
cover classes. Differences in the remaining land cover classes
in the study demonstrated the ability of multi-temporal NDVI
may be attributed to the moderate resolution of the imagery
data in mapping land use/land cover. Various governmental,
used in the study. According to Forest Survey of India report,
non-governmental and research organizations can make use of
the total forest area in Uttarakhand is 24,766 sq.km and non-
the results (available
forest area is 28,717 sq. km (Forest Survey of India, 2011).
athttp://doi.pangaea.de/10.1594/PANGAEA.816654) obtained
When aggregated, the present study results showed a total forest
in the present study for sustainable ecological, economical and
area of 26,563 sq. km and non-forest area of 26,920 sq. km.
environmental development planning and management.

This contribution has been peer-reviewed.

doi:10.5194/isprsarchives-XL-8-1491-2014 1496
 The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XL-8, 2014
ISPRS Technical Commission VIII Symposium, 09 – 12 December 2014, Hyderabad, India

Classifie SH/ Row Producers Users

d Data S/I GL BL EF DF CL DF WB BU Total Accuracy Accuracy Kappa
S/I 27 0 7 0 0 0 0 0 0 34 87.10% 79.41% 0.7704
GL 0 25 2 1 3 0 0 0 0 31 73.53% 80.65% 0.7817
BL 4 3 26 0 0 0 0 1 0 34 74.29% 76.47% 0.7336
EF 0 4 0 25 2 1 1 0 0 33 69.44% 75.76% 0.7245
SH/DF 0 0 0 4 26 3 1 0 0 34 66.67% 76.47% 0.7295
CL 0 2 0 1 2 27 0 0 1 33 71.05% 81.82% 0.7918
DF 0 0 0 5 0 3 26 0 0 34 89.66% 76.47% 0.7395
WB 0 0 0 0 2 4 0 28 0 34 90.32% 82.35% 0.8032
BU 0 0 0 0 4 0 1 2 26 33 96.30% 78.79% 0.7669
Overall Kappa
Column Overall Accuracy = statistics=0.760
Total 31 34 35 36 39 38 29 31 27 300 78.67% 0
S/I: Snow/Ice, GL: Grass lands, BL: Barren lands, EF: Evergreen forest, SH/DF: Shrub/degraded forest, CL: crop lands, DF: Deciduous forest, WB:
Water bodies and BU: Built-up lands.
Table 3. Error matrix for the land use/land cover map of Uttarakhand

ACKNOWLEDGEMENTS
Carroll, M.L., Townshend, J.R., DiMiceli, C.M., Noojipady, P.,
The authors would like to thank Indian Institute of Remote Sohlberg, R. a., 2009. A new global raster water mask at 250 m
Sensing, Dehradun and Anna University, Chennai for providing resolution. Int. J. Digit. Earth 2, 291–308.
an opportunity to carry out this work. MODIS data products doi:10.1080/17538940902951401
were obtained at the NASA Land Processes Distributed Active
Chen, J., Jönsson, P., Tamura, M., Gu, Z., Matsushita, B.,
Archive Center (LPDAAC), USGS/EROS center. National land
Eklundh, L., 2004. A simple method for reconstructing a high-
use/land cover map was obtained from National Remote
quality NDVI time-series data set based on the Savitzky–Golay
Sensing Centre/ISRO.
filter. Remote Sens. Environ. 91, 332–344.
doi:10.1016/j.rse.2004.03.014
REFERENCES
Chen, X., Mithal, V., Vangala, S.R., Brugere, I., Boriah, S.,
Achard, F., Eva, H.D., Mayaux, P., Stibig, H.-J., Belward, A., Kumar, V., 2011. A study of time series noise reduction
2004. Improved estimates of net carbon emissions from land techniques in the context of land cover change detection. Dep.
cover change in the tropics for the 1990s. Global Biogeochem. Comput. Sci. Eng. Reports 11–016.
Cycles 18, 1–11. doi:10.1029/2003GB002142
Cihlar, J., Ly, H., Li, Z., Pokrant, H., 1997. Multitemporal ,
Aggarwal, S., Shadab, S., Karnatak, H.C., 2012. Thematic Data Multichannel AVHRR Data Sets for Land Biosphere Studies
Dissemination on Bhuvan Gateway to Indian Earth Artifacts and Corrections. Remote Sens. Environ. 60, 35–57.
Observation, in: Open Source Geospatial Resources to doi:10.1016/S0034-4257(96)00137-X
Spearhead Development and Growth. pp. 1–5.
Clark, M.L., Aide, T.M., Grau, H.R., Riner, G., 2010. A
Agrawal, S., Joshi, P.K., Shukla, Y., Roy, P.S., 2003. SPOT scalable approach to mapping annual land cover at 250 m using
VEGETATION multi temporal data for classifying vegetation MODIS time series data: A case study in the Dry Chaco
in south central Asia. Mater. Chem. 84, 157–163. ecoregion of South America. Remote Sens. Environ. 114,
2816–2832. doi:10.1016/j.rse.2010.07.001
Asner, G.P., 2001. Cloud cover in Landsat observations of the
Brazilian Amazon. Int. J. Appl. Earth Obs. Geoinf. 22, 3855– Colditz, R.R., López Saldaña, G., Maeda, P., Espinoza, J.A.,
3862. Tovar, C.M., Hernández, A.V., Benítez, C.Z., Cruz López, I.,
Ressl, R., 2012. Generation and analysis of the 2005 land cover
Bakr, N., Weindorf, D.C., Bahnassy, M.H., Marei, S.M., El- map for Mexico using 250m MODIS data. Remote Sens.
Badawi, M.M., 2010. Monitoring land cover changes in a Environ. 123, 541–552. doi:10.1016/j.rse.2012.04.021
newly reclaimed area of Egypt using multi-temporal Landsat
data. Appl. Geogr. 30, 592–605. Colwell, R.N., 1960. Manual of photographic interpretation.
doi:10.1016/j.apgeog.2009.10.008 The American society of photogrammetry.

Bartholomé, E., Belward, a. S., 2005. GLC2000: a new De Fries, R.S., Hansen, M., Townshend, J.R.G., Sohlberg, R.,
approach to global land cover mapping from Earth observation 1998. Global land cover classifications at 8 km spatial
data. Int. J. Remote Sens. 26, 1959–1977. resolution: The use of training data derived from Landsat
doi:10.1080/01431160412331291297 imagery in decision tree classifiers. Int. J. Remote Sens. 19,
3141–3168. doi:10.1080/014311698214235

This contribution has been peer-reviewed.

doi:10.5194/isprsarchives-XL-8-1491-2014 1497
 The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XL-8, 2014
ISPRS Technical Commission VIII Symposium, 09 – 12 December 2014, Hyderabad, India

Dwyer, M., Schmidt, G., 2006. The MODIS reprojection tool. Holben, B., Fraser, R., 1984. Red and near-infrared sensor
Earth Sci. Satell. Remote Sens. 162–177. response to off-nadiir viewing. Int. J. Remote Sens. 5, 145–160.
doi:10.1080/01431168408948795
Farr, T., Rosen, P., Caro, E., 2007. The shuttle radar
topography mission. Rev. Geophys. 15, 1–33. Huang, C., Kim, S., Song, K., Townshend, J.R.G., Davis, P.,
doi:10.1029/2005RG000183.1.INTRODUCTION Altstatt, A., Rodas, O., Yanosky, A., Clay, R., Tucker, C.J.,
Musinsky, J., 2009. Assessment of Paraguay’s forest cover
Foody, G.M., 2002. Status of land cover classification accuracy change using Landsat observations. Glob. Planet. Change 67,
assessment. Remote Sens. Environ. 80, 185–201. 1–12. doi:10.1016/j.gloplacha.2008.12.009
doi:10.1016/S0034-4257(01)00295-4
Huete, a, Didan, K., Miura, T., Rodriguez, E.., Gao, X.,
Foody, G.M., 2008. Harshness in image classification accuracy Ferreira, L.., 2002. Overview of the radiometric and
assessment. Int. J. Remote Sens. 29, 3137–3158. biophysical performance of the MODIS vegetation indices.
doi:10.1080/01431160701442120 Remote Sens. Environ. 83, 195–213. doi:10.1016/S0034-
4257(02)00096-2
Forest Survey of India, 2011. India State of Forest Report 2011,
in: India State of Forest Report 2011. Forest Survey of India, Huete, A., Justice, C., Leeuwen, W. Van, 1999. MODIS
Dehradun, pp. 236–240. Vegetation index alogrithm theoritical basis document,
doi:http://www.fsi.org.in/cover_2011/uttarakhand.pdf Algorithm theoretical basis ….

Fritz, S., McCallum, I., Schill, C., Perger, C., See, L., Jarvis, A., Reuter, H.I., Nelson, A., Guevara, E., 2008. Hole-
Schepaschenko, D., van der Velde, M., Kraxner, F., filled SRTM for the globe, Version 4. CGIAR-CSI SRTM 90m
Obersteiner, M., 2012. Geo-Wiki: An online platform for Database. Int. Cent. Trop. Agric. Cali, Columbia. http//srtm.
improving global land cover. Environ. Model. Softw. 31, 110– csi. cgiar. org.
123. doi:10.1016/j.envsoft.2011.11.015
Jensen, J.R., 1996. Introductory digital image processing.
García-Mora, T.J., Mas, J.-F., Hinkley, E.A., 2012. Land cover Prentice Hall New Jersey, New Jersey.
mapping applications with MODIS: a literature review. Int. J.
Digit. Earth 5, 63–87. doi:10.1080/17538947.2011.565080 Jones, K.B., Survey, U.S.G., Discipline, B., 2006. Importance
of Land Cover and Biophysical Data in Landscape-Based
Gutman, G., 1991. Vegetation indices from AVHRR: An Environmental Assessments, in: North America Land Cover
update and future prospects. Remote Sens. Environ. 35, 121– Summit. pp. 215–250.
136.
Jonsson, P., Eklundh, L., 2002. Seasonality extraction by
Han, K., 2004. A land cover classification product over France function fitting to time-series of satellite sensor data. Geosci.
at 1 km resolution using SPOT4/VEGETATION data. Remote Remote Sens. 40, 1824–1832.
Sens. Environ. 92, 52–66. doi:10.1016/j.rse.2004.05.005
Joshi, P.K., Singh, S., Agarwal, S., 2001. Land cover
Hansen, M.C., Defries, R.S., Townshend, J.R.G., Sohlberg, R., assessment in Jammu & Kashmir using phenology as
2000. Global land cover classification at 1 km spatial resolution discriminant – An approach of wide swath satellite ( IRS –
using a classification tree approach. Int. J. Remote Sens. 21, WiFS ). Curr. Sci. 81, 392–399.
1331–1364. doi:10.1080/014311600210209
Joshi, P.K.K., Roy, P.S., Singh, S., Agrawal, S., Yadav, D.,
Hansen, M.C., Stehman, S. V, Potapov, P. V, Loveland, T.R., 2006. Vegetation cover mapping in India using multi-temporal
Townshend, J.R.G., DeFries, R.S., Pittman, K.W., Arunarwati, IRS Wide Field Sensor (WiFS) data. Remote Sens. Environ.
B., Stolle, F., Steininger, M.K., Carroll, M., DiMiceli, C., 2008. 103, 190–202. doi:10.1016/j.rse.2006.04.010
Humid tropical forest clearing from 2000 to 2005 quantified by
using multitemporal and multiresolution remotely sensed data. Justice, C.O., Townshend, J.R.G., Holben, B.N., Tucker, C.J.,
Proc. Natl. Acad. Sci. U. S. A. 105, 9439–9444. 1985. Analysis of the phenology of global vegetation using
meteorological satellite data. Int. J. Remote Sens. 6, 1271–
Herold, M., Mayaux, P., Woodcock, C.E., Baccini, a., 1318. doi:10.1080/01431168508948281
Schmullius, C., 2008. Some challenges in global land cover
mapping: An assessment of agreement and accuracy in existing Kandrika, S., Roy, P.S., 2008. Land use land cover
1 km datasets. Remote Sens. Environ. 112, 2538–2556. classification of Orissa using multi-temporal IRS-P6 awifs data:
doi:10.1016/j.rse.2007.11.013 A decision tree approach. Int. J. Appl. Earth Obs. Geoinf. 10,
186–193. doi:10.1016/j.jag.2007.10.003
Hietel, E., Waldhardt, R., Otte, A., 2005. Linking socio-
economic factors, environment and land cover in the German Klein, I., Gessner, U., Kuenzer, C., 2012. Regional land cover
Highlands, 1945–1999. J. Environ. Manage. 75, 133–143. mapping and change detection in Central Asia using MODIS
doi:10.1016/j.jenvman.2004.11.022 time-series. Appl. Geogr. 35, 219–234.
doi:10.1016/j.apgeog.2012.06.016
Hird, J.N., McDermid, G.J., 2009. Noise reduction of NDVI
time series: An empirical comparison of selected techniques. Kobayashi, H., Dye, D.G., 2005. Atmospheric conditions for
Remote Sens. Environ. 113, 248–258. monitoring the long-term vegetation dynamics in the Amazon
doi:10.1016/j.rse.2008.09.003 using normalized difference vegetation index. Remote Sens.
Environ. 97, 519–525. doi:10.1016/j.rse.2005.06.007

This contribution has been peer-reviewed.

doi:10.5194/isprsarchives-XL-8-1491-2014 1498
 The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XL-8, 2014
ISPRS Technical Commission VIII Symposium, 09 – 12 December 2014, Hyderabad, India

Li, X., Strahler, A., 1992. Geometric-optical bidirectional Rana, N., Sati, S.P., Sundriyal, Y.P., Doval, M., Juyal, N.,
reflectance modeling of the discrete crown vegetation canopy: 2007. Socio-economic and environmental implications of the
Effect of crown shape and mutual shadowing. IEEE Trans. Geo hydroelectric projects in Uttarakhand Himalaya, India. J. Mt.
Sci. Remote Sens. 30, 276–292. Sci. 4, 344–353. doi:10.1007/s11629-007-0344-5

Lillesand, T.M., Kiefer, R.W., 2008. Remote sensing and image Ribeiro, S.C., Lovett, A., 2009. Associations between forest
interpretation, Distribution. Wiley. characteristics and socio-economic development: a case study
from Portugal. J. Environ. Manage. 90, 2873–81.
Liu, J., Liu, S., Loveland, T.R., Tieszen, L.L., 2008. Integrating doi:10.1016/j.jenvman.2008.02.014
remotely sensed land cover observations and a biogeochemical
model for estimating forest ecosystem carbon dynamics. Ecol. Richards, J. a., Jia, X., 1999. Remote Sensing Digital Image
Modell. 219, 361–372. doi:10.1016/j.ecolmodel.2008.04.019 Analysis. Springer Berlin Heidelberg, Berlin, Heidelberg.
doi:10.1007/978-3-662-03978-6
Liu, S., Kairé, M., Wood, E., Diallo, O., Tieszen, L.L., 2004.
Impacts of land use and climate change on carbon dynamics in Rouse, J., Haas, R., Schell, J., Deering, D., 1973. Monitoring
south-central Senegal. J. Arid Environ. 59, 583–604. vegetation systems in the Great Plains with ERTS., in: Third
doi:10.1016/j.jaridenv.2004.03.023 ERTS Symposium, Vol. 1. pp. 309–317.

Lo, C.P., Yang, X., 2002. Driver of Land-Use/Cover Changes Running, S.W., Loveland, T.R., Pierce, L.L., 1994. A
and Dynamic Modeling for the Atlanta, Georgia Metropolitan Vegetation Classification Logic-Based on Remote-Sensing for
Area. Photogramm. Eng. Remote Sensing 68, 1073–1082. Use in Global Biogeochemical Models. Ambio 23, 77–81.

Lu, D., Weng, Q., 2007. A survey of image classification Savitzky, A., Golay, M.J.E., 1964. Smoothing and
methods and techniques for improving classification Differentiation of Data by Simplified Least Squares
performance. Int. J. Remote Sens. 28, 823–870. Procedures. Anal. Chem. 36, 1627–1639.
doi:10.1080/01431160600746456 doi:10.1021/ac60214a047

Matthews, E., 1983. Global vegetation and land use: New high- Sellers, P.J., Dickinson, R.E., Randall, D.A., Betts, A.K., Hall,
resolution data bases for climate studies. J. Clim. Appl. F.G., Berry, J.A., Collatz, G.J., Denning, A.S., Mooney, H.A.,
Meteorol. 22, 474–487. Nobre, C.A., Sato, N., Field, C.B., Henderson-Sellers, A., 1997.
Modeling the Exchanges of Energy, Water, and Carbon
Mayaux, P., Eva, H., Gallego, J., Strahler, A.H., Herold, M., Between Continents and the Atmosphere. Science (80-. ). 275,
Agrawal, S., Naumov, S., Miranda, E.E. De, Bella, C.M. Di, 502–209.
Ordoyne, C., Kopin, Y., Roy, P.S., 2006. Validation of the
global land cover 2000 map. IEEE Trans. Geosci. Remote Sens. Sharma, S., 2012. Catastrophic hydrological event of 18 and 19
44, 1728–1739. doi:10.1109/TGRS.2006.864370 September 2010 in Uttarakhand, Indian Central Himalaya—an
analysis of rainfall and slope failure. Curr. Sci. 102, 327–332.
Meyer, D., Verstraete, M., Pinty, B., 1995. The effect of
surface anisotropy and viewing geometry on the estimation of Skole, D., Tucker, C., 1993. Tropical deforestation and habitat
NDVI from AVHRR. Remote Sens. Rev. 12, 3–27. fragmentation in the Amazon: satellite data from 1978 to 1988.
doi:10.1080/02757259509532272 Science 260, 1905–10. doi:10.1126/science.260.5116.1905

Ministry of Economics and Statistics, 2010. Uttarakhand at a SoE, U., 2004. Uttaranchal Status of Environment, in:
Glance 2010-11. Dehradun. Environment Protection and Pollution Control
board,Uttarakhand.
Morisette, J.T., Privette, J.L., Justice, C.O., 2002. A framework
for the validation of MODIS Land products. Remote Sens. Solano, R., Didan, K., Jacobson, A., Huete, A., 2010. MODIS
Environ. 83, 77–96. doi:10.1016/S0034-4257(02)00088-3 Vegetation Index User ’ s Guide ( MOD13 Series ).

Nemani, R., Running, S., 1997. Land cover characterization Song, C., Woodcock, C.E., Seto, K.C., Lenney, M.P.,
using multitemporal red, near-IR, and thermal-IR data from Macomber, S.A., 2001. Classification and Change Detection
NOAA/AVHRR. Ecol. Appl. 7, 79–90. doi:10.1890/1051- Using Landsat TM Data When and How to Correct
0761(1997)007[0079:LCCUMR]2.0.CO;2 Atmospheric Effects? Remote Sens. Environ. 75, 230–244.
doi:10.1016/S0034-4257(00)00169-3
Olson, J.S., Watts, J.A., Allison, L.J., 1983. Carbon in live
vegetation of major world ecosystems. ORNL. Steininger, M.K., Tucker, C.J., Townshend, J.R.G., Killeen,
T.J., Desch, A., Bell, V., Ersts, P., 2002. Tropical deforestation
Olthof, I., Fraser, R., 2007. Mapping northern land cover in the Bolivian Amazon. Environ. Conserv. 28, 127–134.
fractions using Landsat ETM+. Remote Sens. Environ. 107, doi:10.1017/S0376892901000133
496–509. doi:10.1016/j.rse.2006.10.009
Stibig, H.-J., Belward, a. S., Roy, P.S., Rosalina-Wasrin, U.,
Perera, K., Tsuchiya, K., 2009. Experiment for mapping land Agrawal, S., Joshi, P.K., Beuchle, R., Fritz, S., Mubareka, S.,
cover and it’s change in southeastern Sri Lanka utilizing 250m Giri, C., 2007. A land-cover map for South and Southeast Asia
resolution MODIS imageries. Adv. Sp. Res. 43, 1349–1355. derived from SPOT-VEGETATION data. J. Biogeogr. 34, 625–
doi:10.1016/j.asr.2008.12.016 637. doi:10.1111/j.1365-2699.2006.01637.x

This contribution has been peer-reviewed.

doi:10.5194/isprsarchives-XL-8-1491-2014 1499
 The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XL-8, 2014
ISPRS Technical Commission VIII Symposium, 09 – 12 December 2014, Hyderabad, India

Swetnam, R.D., Fisher, B., Mbilinyi, B.P., Munishi, P.K.T.,

Willcock, S., Ricketts, T., Mwakalila, S., Balmford, A.,
Burgess, N.D., Marshall, A.R., Lewis, S.L., 2011. Mapping
socio-economic scenarios of land cover change : A GIS method
to enable ecosystem service modelling. J. Environ. Manage. 92,
563–574. doi:10.1016/j.jenvman.2010.09.007

Thenkabail, P.S., Schull, M., Turral, H., 2005. Ganges and

Indus river basin land use/land cover (LULC) and irrigated area
mapping using continuous streams of MODIS data. Remote
Sens. Environ. 95, 317–341. doi:10.1016/j.rse.2004.12.018

Townshend, J., Justice, C., Li, W., Gurney, C., McManus, J.,
1991. Global land cover classification by remote sensing:
present capabilities and future possibilities. Remote Sens.
Environ. 35, 243–255. doi:10.1016/0034-4257(91)90016-Y

Vancutsem, C., Pekel, J.-F., Evrard, C., Malaisse, F., Defourny,

P., 2009. Mapping and characterizing the vegetation types of
the Democratic Republic of Congo using SPOT
VEGETATION time series. Int. J. Appl. Earth Obs. Geoinf. 11,
62–76. doi:10.1016/j.jag.2008.08.001

Wardlow, B., Egbert, S., Kastens, J., 2007. Analysis of time-

series MODIS 250 m vegetation index data for crop
classification in the U.S. Central Great Plains. Remote Sens.
Environ. 108, 290–310. doi:10.1016/j.rse.2006.11.021

White, M. a., 2005. A global framework for monitoring

phenological responses to climate change. Geophys. Res. Lett.
32, L04705. doi:10.1029/2004GL021961

WSMD, 2009. Uttarakhand State Perspective and Strategic

Plan. Watershed Management Directorate, Dehradun, pp. 105–
108.

Xiao, X., Boles, S., Liu, J., Zhuang, D., Liu, M., 2002.
Characterization of forest types in Northeastern China, using
multi-temporal SPOT-4 VEGETATION sensor data. Remote
Sens. Environ. 82, 335–348. doi:10.1016/S0034-
4257(02)00051-2
Zhang, X., Sun, R., Zhang, B., Tong, Q., 2008. Land cover
classification of the North China Plain using MODIS_EVI time
series. ISPRS J. Photogramm. Remote Sens. 63, 476–484.
doi:10.1016/j.isprsjprs.2008.02.005

This contribution has been peer-reviewed.

doi:10.5194/isprsarchives-XL-8-1491-2014 1500