Pak. J. Agri. Sci., Vol.

56(1), 187-196; 2019

ISSN (Print) 0552-9034, ISSN (Online) 2076-0906
DOI:10.21162/PAKJAS/19.7663
http://www.pakjas.com.pk

LAND COVER MAPPING AND CROP PHENOLOGY OF POTOHAR REGION,

PUNJAB, PAKISTAN
Sarah Amir1,*, Zafeer Saqib1, Amina Khan1, M. Irfan Khan1, M. Azeem Khan2 and Abdul Majid3
1
Department of Environmental Science, International Islamic University, Islamabad, Pakistan;
2
Ministry of Planning, Development and Reforms, Islamabad, Pakistan; 3ICARDA-Pakistan.
*
Corresponding author’s e.mail: [email protected]

Agriculture a major source of food and fibre affects the natural land cover and in turn is affected by climatic factors like
temperature and precipitation patterns beside other factors. The soil temperature and moisture, wind, relative humidity and
crop water requirements also affect the crop growth. Local farming practices also alter the natural landscape structure and
biodiversity in croplands. This study was conducted with the objective to find out seasonality trend and to determine the land
cover classification in Potohar region using Normalized Difference Vegetation Index (NDVI). Moreover, random points
throughout the study area were selected in order to detect the vegetation index pattern in different land cover types. For land
cover classification and acquiring seasonality trend, crop growth patterns were determined by phenological phases of a
particular crop. Land cover was classified with standard Level 1 Terrain-corrected (L1T) orthorectified images from Landsat
8 from years 2012 to 2016. Seven land cover classes within the study area were identified namely; agriculture, grasses, forest,
shrubs/tall herbs, bare soil, built-up and water bodies. Moreover, the seasonality trend over the study was found related to
different land cover classifications using 959 sample plots with 97.81% accuracy. The phase trend analysis determined the
change in vegetation cover during the years under study, which was correlated to precipitation patterns in Potohar. The NDVI
pattern was highly fluctuating in agricultural land cover due to seasonal crop growth but it remained stable throughout the year
in forest covers. Urban land cover was found to have high impact on nearby vegetation as it was related to change in land cover
type, which affected the climate of the area. Climate being a vital deriving force, affected precipitation patterns of the rain-fed
(barani) land which in turn shifted the seasonal growth of various agricultural crops.
Keywords: Agriculture, crop phenology, land cover maps, NDVI, barani (rain-fed), seasonality trend

INTRODUCTION and impacts of climate change. In this connection, the role and
behaviour of agricultural land cover is unique. It not only
The understanding of dynamics that influence the land use contribute in food provisioning but also serve as a natural sink
and land cover change (LULC) are imperative for ensuring for greenhouse gases such as carbon dioxide (CO2). It covers
food security and to address the challenges associated with more than one third of the land on earth in the form of pastures
global climatic changes. The synchronization of societal goals and croplands. Keeping in view the increasing rate of
and tripartite linkages among social, ecological and economic population, it is vital information to know that only 10% of
domains are essential to achieve the goals of sustainable increase in agricultural land is taking place which makes the
development (Engstrom et al., 2016; Hegazy and Kaloop, unavailability of food for about one billion people globally
2015). Land use and land cover are two interchangeable (Pongratz et al., 2008; Engstrom et al., 2016; Qin et al., 2015).
terminologies, where, land cover is described as the physical Land cover change can be detected by studying vegetation
characteristic of the earth’s surface and land use as the way in phenology. Phenology is the study of periodic events in the
which that characteristic is used by humans. Land cover are life cycle of living species (Vina et al., 2004; Sakamoto et al.,
classified into two broad categories i.e. natural (vegetation, 2010). An efficient method for assessment of vegetation
forests, shrubs and grasses, water bodies, bare soil etc.) and phenology is the use of satellite derived vegetation indices
man-made/human-transformed which reflect anthropogenic including Normalized Difference Vegetation Index (NDVI)
interventions in natural environment for economic activities and Enhanced Vegetation Index (EVI) (Gong et al., 2015).
include agricultural lands, settlements etc. (Rawat and Landsat 8 provides an efficient source of phenology
Kumar, 2015). assessment in the area like Pakistan having diverse crops.
The climate of an area is a product and dependent upon the Sensors of the satellite i.e., Operational Land Imager (OLI)
cumulative behaviour of land use/ land cover types and with nine bands and Thermal Infrared Sensor (TIRS) with two
human interaction in a contextual setting. The vegetative thermal bands provide the resolution of about 30 m (Jia et al.,
cover of a geographic region help us to decipher the causation 2014) which is an accurate resolution for phenological
 Amir, Saqib, Khan, Khan, Khan & Majid

assessment of large as well as small crop lands like that of July to October (Sarwar et al., 2014). The summer
Potohar region. temperature ranges between 15oC and 40oC while the range
The geographic area of Pakistan is about 80 million hectares of winter temperature is generally between 4 and 25 oC but it
(Mha), of which 18 Mha is irrigated and dry land farming is can occasionally drop. Around 994 thousand hectares area of
practiced on 12 Mha. The barani (rainfed) areas of Punjab Potohar plateau is being cultivated (GoP, 2016). About 4% of
cover about 7 Mha and are home to over 19 million people. the cultivated land is irrigated while 96% is dependent on rain
This is equivalent to about 40% of the total area of the (Majeed et al., 2010). The major crops grown in the area
Pakistan Punjab (Oweis and Ashraf, 2012). These areas, include wheat (Triticum vulgare), maize (Zea mays), barley
however, contribute less than 10% to total agricultural (Hordeum vulgare), sorghum (Sorghum bicolor), millets
production and depend solely on the rainfall. In addition to a (Panicum miliaceum), lentils (Lens culinary), gram (Cicer
food source for the citizens of the country, agriculture is also arietinum), groundnut (Arachis hypogaea) and brassica
a very important economic sector of Pakistan contributing (Brassica rapa).
about 21% in GDP. An important subsector of agriculture is
cropland constituting 39.6% of agriculture and which has a
share of about 8.3% in GDP (Khan et al., 2016). The major
share (76%) is contributed by Punjab (Rashid et al., 2014).
Kazmi and Rasul (2009) hypothesised that the underestimated
barani land (rainfed area) of Potohar plateau is capable to
significantly contribute in the economy of Pakistan because
more than 1200 kg/acre wheat is grown in the area which
shows its potential to lower import load.
Productivity of Potohar is reported to be decreased about 2.5
to 7 times due to over grazing and removal of vegetation for
purpose of obtaining fuel wood. Habitat degradation is the
obvious consequence of this event as water erosion effects the
agro ecosystems. The vegetation cover can be recovered by
increase in precipitation (Gong et al., 2015).
Land cover mapping of a region is an important task in order
to determine the ongoing changes in LULC over time. Land
cover mapping helps us determine whether a certain sector
either agriculture or urban requires proper planning and Figure 1. Location map of the study area.
management. The study of crop phenology help us to
determine changes in the climate of the area as the climate of Land cover mapping and phenology: In order to map land
the region directly affects the agricultural lands. Data about cover and trends in land surface phenology of the croplands,
crop phenology relate to food security of an area. Landsat 8 satellite data was used. All the data analyses were
The main objectives of the present study were to develop land carried out in Google Earth Engine (GEE) platform. The
cover maps for determining area under agricultural land dataset used was calibrated top-of-atmosphere (TOA)
cover, analyze the seasonal trend of Potohar region from reflectance (Collection 1 Tier 1) from Landsat 8 imagery. The
2012-2016 and to find out correlation between precipitation reflectance is calculated on the calibration coefficients
and NDVI. defined by Chander et al. (2009). The study area is covered
by World Referencing System (WRS) path 150 and row 37.
MATERIALS AND METHODS For developing land cover an image acquired 17th March 2017
was used (landsat ID:
Study area: The study area chosen for the present study was LC08_L1TP_150037_20170317_20170328_01_T1) that
Potohar plateau, Pakistan and includes major portion of contained <1% cloud cover. All the available imagery (n=77)
districts Attock, Chakwal, Islamabad, Jhelum and Rawalpindi was used for the analysis of land surface phenology trends.
(Fig. 1). The region lies between Indus River and the Jhelum The selected images from year 2012 to 2016 were processed
River and stretches from the salt range northward to the foot using Earth Engine API.
hills of Himalayas (approximately 32.5°N to 34.0°N Latitude Cloud removal: Some of the acquired data becomes
and 72°E to 74°E Longitude). The total area of this region is inaccurate due to presence of cloud cover over the study area
approximately 13,000 km2 with elevation from sea level which hinders the satellite assessment of the region. It is
fluctuating between 305 - 610 m. It has highly undulating termed as noise in the data. Such regions were masked by
topography and erratic rainfall pattern. The climate of the area removing clouds. In this study, all these bad observations
is semi-arid to humid. Out of total, about 80% rain falls during were masked using Landsat 8 quality assessment band

188
 Land cover mapping of Potohar region

(Scaramuzza et al., 2012; Roy et al., 2014). The masked amplitude 1 (peak of annual greenness) by performing
correction of bad observations is classified in four levels harmonic regression on each pixel of 5- year NDVI time
including “not determined” (Algorithm did not determine the series with temporal window of one year. Phase 0 and phase
status of this condition), “no” (0–33% confidence), “maybe” 1 were obtained in the second stage relevant to amplitude 0
(34–66% confidence), and “yes” (67–100% confidence) and amplitude 1. Values for phase image ranges from 0 to 359
(Dong et al., 2016). We used the 67–100% confidence level degrees and after every 30 degrees a change of one calendar
to exclude all the potential bad observation effects from month is represented (Eastman et al., 2009). The equation
clouds and cirrus. It helped to obtain 97.8% accuracy of utilized in script for seasonal trend analysis is expressed as
results. follows;
Spectral indices: The time series of Landsat TOA image 2πnt 2πnt
y = α0 ∑n {a n sin ( ) + bn cos ( ) } + e eq (6)
collection was used to calculate three vegetation indices, T T
Where, t is time, T is the interval of time-series, n is a
including NDVI (Normalized Difference Vegetation Index),
NDWI (Normalized Difference Water Index) and MNDWI harmonic multiplier, e is an error term, 𝛼0 is the series mean
(Modification of Normalized Difference Water Index). The and 𝑎𝑛 and 𝑏𝑛 are regression parameters.
spectral indices were calculated using following equations: Digital image classification: To classify the land cover of
B5−B4 Potohar, complex pixels of satellite image containing multiple
NDVI ( ) ;………………....…………..eq (1) spectral bands and range of millions of colours, were
B5+B4
B3−B5 classified into definite number of classes. Earth engine
NDWI ( ) ;……………………………eq (2)
B3+B5
B5−B4 B3−B7 classifier package handled the pixels using CART classifier
MNDWI ( ) , MNDWI ( )………eq (3) to generate land cover maps. It fundamentally separated the
B5+B4 B3+B7
Gaps in the original NDVI were filled using linear and forest and non-forest areas.
harmonic regression models.
Linear regression: Google Earth Engine contains a variety of
methods for performing linear regression. Linear regression
model gives the fitted values out of original NDVI values in
a linear trend. Linear model only describes the trend on the
values being increasing or decreasing over a certain period of
time. The linear trend was calculated in Earth Engine using
the following equation:
pt = β0 + β1t + et…………….eq (4)
Where, pt is NDVI at time, t is time and et is random error.
The missing data in the original NDVI were filled using this
model. This was shown as the fitted value in the graph. Linear
regression equation was used in the script in order to show the
increasing and decreasing trend for NDVI value.
Harmonic regression: In order to extract the phase and
amplitude using windowed fourier analysis for seasonal trend
analysis, a script was developed in Earth Engine Code editor
whereby the equation of harmonic regression was
incorporated along with linear regression trend. The number
of harmonic cycle taken was two. This equation of harmonic
regression is as follows;
Figure 2. Schematic flow of the methodology.
….eq (5) RESULTS AND DISCUSSION
Where Y is the NDVI, β0 is an offset, c is the trend, Ai is the
amplitude of the ith oscillation, φi is the phase component of Land cover classification: The supervised classification of
the ith oscillation, s is the fundamental frequency and T is the satellite images of Potohar distinguished seven types of land
time-dependent variable. The peak of annual greenness was cover. Figure 3 illustrates these types along with the area
represented by the amplitude while the timing of the peak covered by each land cover type. Land cover types identified
NDVI value was represented by phase. included agriculture, grasses, tree/forest, shrubs/tall herbs,
Seasonal trend analysis: Seasonal trend analysis is a two bare soil/rocks, built-up and water bodies. Some of the area
stage process to describe seasonal NDVI cycle. First stage remained unclassified due to research time and resource
dealt with obtaining amplitude 0 (mean annual greenness) and constraints.

189
 Amir, Saqib, Khan, Khan, Khan & Majid

Area that each of the land type covered was calculated

separately for five districts of Potohar namely Chakwal,
Attock, Rawalpindi, Jhelum and Islamabad. Percentage of the
area covered in each district is mentioned in Table 1 while
Figure 4 shows the same in kilometre square (km2).

Table 1. List of land cover distribution (in percentage) in

districts of Potohar in year 2017.
Land cover Islam- Attock Chak- Jhelum Rawal-
types abad wal pindi
Agriculture 37.32% 69.17% 68.65% 37.35% 38.52%
Grasses 16.42% 7.88% 7.46% 14.38% 16.13%
Trees/Forests 19.85% 16.11% 17.03% 30.78% 29.45%
Shrubs/Tall 13.81% 2.69% 3.10% 5.69% 10.94%
herbs
Bare soil/Rocks 3.01% 1.05% 1.35% 5.79% 1.58%
Built-up 8.94% 2.76% 2.25% 5.55% 3.29%
Water 0.66% 0.35% 0.17% 0.45% 0.09%

7000
Water
6000
Built-up
Figure 5. Map of Potohar region illustrating phase 1 trend
5000 Bare soils/rocks
analysis (2012 - 2016).
Shrubs/tall herbs
Area (Km²)

4000
Trees/Forest
3000 Grasses
2000 Agriculture

1000

Districts
Figure 4. Illustration of Land cover distribution in each
district of Potohar Region in kilometre square
(Km2) in year 2017.

Accuracy assessment: The overall accuracy of the land cover

map was 97.81% with value 0.971 for Kappa quotient, and
0.9903 was the maximum possible un-weighted Kappa given
the observed marginal frequencies. At 0.95 confidence
interval upper values in both methods 1 and 2 were 0.9832
and lower values were 0.9586. Both the accuracies, ‘Producer
and User’, were greater than 94% for all the land cover classes
except for class ‘grasses’ that have producer accuracy 91.61%
as shown in Table 2. Figure 6. Map of Potohar region illustrating phase 1 trend
Seasonal trend analysis: The first harmonic cycle indicates over agricultural land cover (2012 - 2016).
the first half of the year (from January to June) in terms of
phase 1 and the second harmonic cycle indicates the latter half Phase 2 trend is shown in Figure 7 & 8. The shades of the
of the year (from July to August) in terms of phase 2. Figure resulted phase maps illustrate months of the year. The colour
5 visualizes phase 1 trend analysis while Figure 6 illustrates codes for each month of phase 1 and phase 2 are mentioned
this trend over agricultural land. in Tables 3 & 4.

190
 Land cover mapping of Potohar region

Table 2. Error matrix for land cover classification showing accuracy assessment of the results in terms of user
accuracy, producer accuracy and overall accuracy with Kappa=0.971.
Land cover 1 2 3 4 5 6 7 Total Useraccuracy
1. Agriculture 340 8 0 0 1 0 0 349 97.42%
2. Grasses 6 142 1 0 0 0 0 149 95.30%
3. Trees/Forest 0 5 264 0 0 0 0 269 98.14%
4. Shrubs/Tall Herbs 0 0 0 88 0 0 0 88 100.00%
5. Bare Soil/Rocks 0 0 0 0 19 0 0 19 100.00%
6. Built-up 0 0 0 0 0 47 0 47 100.00%
7. Water 0 0 0 0 0 0 38 38 100.00%
Total possible 346 155 265 88 20 47 38
Omissions 6 13 1 0 1 0 0
Commissions 9 7 5 0 0 0 0
Correctly Classified 340 142 264 88 19 47 38
Producer Accuracy 98.27% 91.61% 99.62% 100.00% 95.00% 100.00% 100.00% Overall Accuracy
= 97.81%

Table 3. List of colour codes with their identical

description for phase 1.
Colour Identity Description
Dark Blue peak beginning of January
Moderate Blue Middle of January
Light Blue peak beginning of February
Sea Green Middle of February
Dark Green peak beginning of March
light green Middle of March
Lime Green peak beginning of April
Bright Yellow Middle of April
Orange peak beginning of May
Dark Orange Middle of May
moderate Brown peak beginning of June
Maroon Middle of June
Figure 7. Map of Potohar region illustrating phase 2 trend
analysis (2012 - 2016). Table 4. List of colour codes with their identical
description for phase 2.
Colour Identity Description
Dark Blue peak beginning of July
Moderate Blue Middle of July
Light Blue peak beginning of August
Sea Green Middle of August
Dark Green peak beginning of September
light green Middle of September
Lime Green peak beginning of October
Bright Yellow Middle of October
Orange peak beginning of November
Dark Orange Middle of November
moderate Brown peak beginning of December
Maroon Middle of December

Overall seasonality of the area is shown in Figure 9. In simple

RGB model crop phenology of different land covers in the
study area is displayed. This phenological trend is shown via
Figure 8. Map of Potohar region illustrating phase 2 trend
over agricultural land cover (2012 - 2016).

191
 Amir, Saqib, Khan, Khan, Khan & Majid

graphs in Figures 10-15 using random points throughout the

study area.

Figure 11. Illustration of phenological trend in

agricultural field in Chakwal district.

Figure 9. Map of Potohar region illustrating seasonality

trend analysis (2012 - 2016).

Precipitation correlation: The productivity of Potohar region

was correlated with precipitation as it is a barani area. Figure
16 displays the change in productivity immediately after
precipitation. In Figure 17, map of correlation lag 17 is
displayed which shows the change in NDVI 17 days after the
precipitation. Similarly in Figure 18, lag 30 map shows the Figure 12. Illustration of phenological trend in crop field
change 30 days later. of Jhelum district.

Figure 10. Illustration of phenological trend in forest Figure 13. Illustration of phenological trend in evergreen
cover near Attock district. forest at foothills of Himalayas.

192
 Land cover mapping of Potohar region

Figure 14. Illustration of phenological trend in forest near

Rawal Lake, Islamabad.
Figure 17. Map illustrating precipitation correlation at
lag 17 with changing NDVI.

Figure 15. Illustration of NDVI trend in urban built-up of

Rawalpindi district.

Figure 18. Map illustrating precipitation correlation at

lag 30 with changing NDVI.

To characterize seasonal trend and crop phenology of Potohar

region, satellite-derived time-series data was utilized. A very
interesting crop phenology trend was found to be obtained in
Figures 5 & 6. A negative NDVI value describes an early
season of the year indicated by blue colour, while, red colour
in the map having positive NDVI values indicate the peak of
the season (Fig. 5). In Figure 6, first harmonic cycle
completing in June shows the phenology of a crop which
possibly has a harvest time period in March or April as most
of the study area was represented in green which shows a
mean NDVI value. Wheat crop is an abundant crop in Potohar
which possibly describes the obtained NDVI value for March
Figure 16. Map illustrating precipitation correlation at
and April (Kazmi and Rasul, 2006). The similar trend is
lag 0 with changing NDVI.
followed in Figures 7 & 8, whereas, in July the early season

193
 Amir, Saqib, Khan, Khan, Khan & Majid

Annexure 1. Showing the codes (JAVA Script) and procedural measures adopted for data acquisition and
processing.
Processing Code
Image processing //Landsat Data
Satellite data Acquisition VarNAME=ee.imageCollection('LANDSAT/LC8_L1T_TOA')
Year: 2012-2016 .filterData('2012-01-01');
Bands Used: 1, 2, 3, 4, 5 & 7 .filter(ee.Filter.eq('WRS_PATH', 150));
Worldwide Reference System: WRS 2: 150/37 .filter(ee.Filter.eq('WRS_ROW', 37y));
//Band Selection
var bands = ['B1','B2','B3','B4','B5','B7'];
Cloud Removal var image = ee.ImageCollection('LANDSAT/LC8_L1T_TOA')
//selected year cloud remove
.sort('CLOUD_COVER')
.map(maskClouds)
Spectral Indices
B5−B4 var ndvi = image.normalizedDifference(['B4','B5']);
NDVI ( ) ;…………eq (1)
B5+B4
B3−B5 var ndwi = image.normalizedDifference(['B3','B5']);
NDWI ( ) ;………...eq (2)
B3+B5
B3−B7 var mndwi = image.normalizedDifference(['B3','B7']);
MNDWI ( )…….…eq (3)
B3+B7
Land Cover Classification var trainingImage = ee.Image([image.select('B1'),
image.select('B2'),
image.select('B3'),
image.select('B4'),
image.select('B5'),
image.select('B7'),
ndbi.rename('ndbi'),
ndvi.rename('ndvi'),
mndwi.rename('mndwi'),
Map.centerObject(roi);
Map.addLayer(trainingImage,vizParams,'True-color composite',false);
Map.addLayer(trainingImage,{'bands':['ndvi','ndbi','mndwi'],
'min':-1,'max':1},'NDXI composite',false);
/*
//Classification starts Here
print(trainingImage);
// Image classification
var predictionBands = trainingImage.bandNames();
print(predictionBands);
var trainingFeatures = agriculture //1
.merge(grasses) //2
.merge(trees) //3
.merge(shrubs) //4
.merge(bare) //5
.merge(builtup) //6
.merge(water); //8
var classifierTraining = trainingImage
.select(predictionBands)
.sampleRegions({collection: trainingFeatures,
properties: ['landcover'],
scale: 30
});
// train the classifier
var classifier = ee.Classifier.cart().train({
features: classifierTraining,
classProperty: 'landcover',
inputProperties: predictionBands
});
Accuracy Assessment on the basis of classification print(classifier.explain());
var confusionMatrix = classifier.confusionMatrix()
var accuracy = confusionMatrix.accuracy()
print(confusionMatrix,accuracy);
Linear regression formulation ee.Reducer.linearFit()
Export result for further image processing Export.image.toDrive({
image: classified,
description: 'classified_image',
scale: 30,
region: roi
});

shows negative NDVI trends indicating the sowing of a crop. Crop phenology study was conducted to determine the start
The other intermediate shades between blue and red indicate and end of seasons of various crops in Potohar and their
the presence of other minor crops. relation with the land cover and precipitation. Random points

194
 Land cover mapping of Potohar region

taken from the study area revealed the patterns of vegetation term seasonality trend helped to monitor and identify the
in various land cover types. A sub-tropical deciduous forest factors which are responsible for land cover changes in
in Attock sheds its leaves in the start of dry season so as to Potohar. The results have shown the seasonality shift based
retain water to survive in harsh weather conditions and on land cover change with the accuracy of Kappa= 0.9781.
drought. The seasonal trend didn’t show a very high change The vegetation alongwith the urban setting had shown shift in
in NDVI showing the stable nature of the forest cover. The the values of NDVI, while the forest cover away from urban
agricultural field in Chakwal shows crop phenology by area had quite stable seasonality trend. The agricultural land,
exhibiting changes in NDVI. Wheat has the harvesting period however, showed different NDVI trends for different crops
from March to May during which it shows decrease in NDVI depending upon their start of season. On the basis of these
value and then in sowing period from October to November, results, it is incurred that, vegetation in different land cover
it shows high increase in NDVI. The other minor fluctuations types is dependent upon the prevailing environment in the
show the presence of grasses in the field. The crop field of particular land cover type. Land use and land change is highly
Jhelum displays the phenologies of wheat and maize in the associated with the changing environment especially in the
area indicating high increase in NDVI in November till March context of climate change.
for wheat and from July till October or November for maize
crop. An evergreen forest on the foothills of Himalaya was REFERENCES
taken as the comparison between NDVI of deciduous and
evergreen forest. Evergreen forest indicated a very minor Chander, G., B.L. Makhram and D.N. Helder. 2009.
fluctuation in NDVI throughout the year with decreasing Summary of current radiometric calibration coefficients
trend in winters due to snow cover. Forest vegetation near for Landsat MSS, TM, ETM+, and EO-1 ALI sensors.
Rawal Lake was taken to determine the relation of water body Remote Sense Environ. 113:893-903.
with the changing NDVI. The trends in NDVI show increase, Dong, J., X. Xiao, M.A. Menarguez, G. Zhang, Y. Qin, D.
only minor, in value during monsoon and winter rains when Thau, C. Biradar and B. Moore. 2016. Mapping paddy
lake is filled by the precipitation and decrease in values in dry rice planting area in north-eastern Asia with Landsat 8
seasons. Near urban areas of Rawalpindi, the vegetation either images, phenology-based algorithm and Google Earth
consists of grasses or they may be small private crop fields Engine. Remote Sense Environ. 185:142-154.
showing negligible change in NDVI. Eastman, J.R., S. Florencia, G. Bardan, Z. Honglei, C. Hao,
In Figures 16, 17 and 18, the productivity of the crops is N. Neeti, C. Yongming, A.M. Elia and C.C.
correlated with precipitation. The areas having positive values Crema. 2009. Seasonal trend analysis of image time
indicate that the increase in precipitation increases the series. Int. J. Remote Sense 30:2721-2726.
productivity. Such areas are arid having water scarcity and Engstrom, K., M.D.A. Rounsevell, D.M. Rust, C. Hardacre,
dependant on precipitation as water source. Those having P. Alexander, X. Cui, P.I. Palmer and A. Arneth. 2016.
negative correlation show that the productivity decreases as Applying Occam's razor to global agricultural land use
the precipitation increases while those having zero values for change. Environ. Modell. Soft. 75:212-229.
correlation mean, it is not dependant on precipitation. In Gong, Z., K. Kawamura, N. Ishikawa, M. Goto, T. Wulan, D.
Figure 17, map of correlation lag 17 is displayed, which Alateng, T. Yin and Y. Ito. 2015. MODIS normalized
shows the change in NDVI 17 days after the precipitation. difference vegetation index (NDVI) and vegetation
Similarly in Figure 18, lag 30 map shows the change 30 days phenology dynamics in the Inner Mongolia grassland.
later. Lag 17 maps indicated high correlation in dense Solid Earth 6:1185-1194.
agricultural lands and forest covers near urban settings GoP. 2016. Punjab Development Statistics. 2016. Bureau of
showing human induced changes in the area. Lag 30 maps Statistics, Planning and Development Department,
indicated the high values of correlation for agricultural lands Government of the Punjab, Pakistan.
and moderate for urban settings. Correlation values for forest Hegazy, I.R. and M.R. Kaloop. 2015. Monitoring urban
covers in moist environments remained low while for arid and growth and land use change detection with GIS and
dry climatic regions like Attock, precipitation seemed to have remote sensing techniques in Daqahlia governorate
high impact. Mishra and Chauhdary (2014) also discussed in Egypt. Int. J. Sustain. Built. Environ. 4:117-124.
their study about the precipitation correlation with seasonality Jia, K., S. Liang, X. Wei, Y. Yao, Y. Su, B. Jiang and X.
analysis showing increase in brownness due to changing Wrang. 2014. Land cover classification of landsat data
precipitation patterns due to anthropogenic activities. with phenological features extracted from Time Series
Conclusion: The underlying goal of this research was to MODIS NDVI Data. Int. J. Remote Sense 6:11518-
evaluate the seasonality shift and land cover classes in 11532.
Potohar. Using the refined resolution of Landsat 8 (30m), the Kazmi, D.H. and G. Rasul. 2009. Early yield assessment of
vegetation indices were studied for an assessment of land wheat on meteorological basis for Potowar region. Pak.
cover activity based on time series trend analysis. The long J. Meteorol. 6:73-87.

195
 Amir, Saqib, Khan, Khan, Khan & Majid

Khan, A., M.C. Hansen, P. Potapov, S.V. Stehman and A.A (Rawalpindi). Pakistan Meteorological Department,
Chatta. 2016. Landsat based wheat mapping in the Islamabad. Pakistan.
heterogeneous cropping system of Punjab, Pakistan. Int. Rawat, J.S. and M. Kumar. 2015. Monitoring land use/cover
J. Remote Sense 37:1391-1410. change using remote sensing and GIS techniques: A case
Majeed, S., I. Ali, B.S. Zaman and S. Ahmad. 2010. study of Hawalbagh block, district Almora, Uttarakhand,
Productivity of mini dams in Pothwar Plateau: A India. Egypt. J. Remote Sense 18:77-84.
diagnostic analysis. Research briefings Vol (2), No (13), Roy, D.P., M.A. Wulder, T.R. Loveland, C.E. Woodcock,
Natural Resource Division, Pakistan Agricultural R.G. Allen, M.C. Anderson, D. Helder, J.R. Irons, D.M.
Research Council, Islamabad, Pakistan. Johnson and R.Kennedy. 2014. Landsat-8: Science and
Misha, N.B. and G. Chaudhary. 2015. Spatio-temporal product vision for terrestrial global change research.
analysis of trends in seasonal vegetation productivity Remote Sense Environ. 145:154-172.
across Uttarakhand, Indian Himalayas, 2000-2014. Appl. Sakamoto, T., M. Yokozawa, H. Toritani, M. Shibayama, N.
Geogr. 56:29-41. Ishitsuka and H. Ohno. 2010. A crop phenology detection
Oweis, T., M. Ashraf (eds). 2012. Assessment and options for method using time-series MODIS data. Remote Sense
improved productivity and sustainability of natural Environ. 96:366-374.
resources in Dharabi Watershed Pakistan. ICARDA, Sarwar, M., I. Hussain, M. Anwar and S.N. Mirza. 2014.
Aleppo, Syria. Baseline data on anthropogenic practices in the agro-
Pongratz, J., C. Reick, T. Raddatz and M. Claussen. 2008. A ecosystem of Potowar plateau, Pakistan. J. Anim.
reconstruction of global agricultural areas and land cover Plant Sci. 26:850-857.
for the last millennium. Global Biochem. Cycles 22:1-16. Scaramuzza, P.L., M.A. Bouchard and J.L. Dwyer. 2012.
Qin, Y., X. Xiao, J. Dong, Y. Zhou, Z. Zhu, G. Zhang, G. Du, Development of the landsat data continuity mission
C. Jin, W. Kou, J. Wang and X. Li. 2015. Mapping paddy cloud-cover assessment algorithms. IEEE. T. Geosci.
rice planting area in cold temperate climate region Remote 50:1140-1154.
through analysis of time series Landsat 8 (OLI), Landsat7 Vina, A., A.A. Gitelson, D.C. Rundquist, G. Keydan, B.
(ETM+) and MODIS imagery. Int. J. Remote Sense Leavitt and J. Schepers. 2004. Monitoring maize (Zea
105:220-233. mays L.) phenology with remote sensing. J. Agron.
Rashid, K., M. Ayaz and K. Noureen. 2014. Weather and 96:1139-1147.
maize crop development in Potowar Region Punjab

196