815963

research-article2018
WMR0010.1177/0734242X18815963Waste Management & ResearchMahmood et al.

Original Article

Waste Management & Research

A comparison of satellite-based indices for

2019, Vol. 37(3) 219­–236
© The Author(s) 2018
Article reuse guidelines:
hazard assessment of MSW open dumps sagepub.com/journals-permissions
DOI: 10.1177/0734242X18815963
https://doi.org/10.1177/0734242X18815963

using spatial analysis journals.sagepub.com/home/wmr

Khalid Mahmood1, Zia Ul-Haq1, Fiza Faizi2 and Syeda A. Batol1

Abstract
This study compares the suitability of different satellite-based vegetation indices (VIs) for environmental hazard assessment of municipal
solid waste (MSW) open dumps. The compared VIs, as bio-indicators of vegetation health, are normalized difference vegetation index
(NDVI), soil adjusted vegetation index (SAVI), and modified soil adjusted vegetation index (MSAVI) that have been subject to spatio-
temporal analysis. The comparison has been made based on three criteria: one is the exponential moving average (EMA) bias, second is
the ease in visually finding the distance of VI curve flattening, and third is the radius of biohazardous zone in relation to the waste heap
dumped at them. NDVI has been found to work well when MSW dumps are surrounded by continuous and dense vegetation, otherwise,
MSAVI is a better option due to its ability for adjusting soil signals. The hierarchy of the goodness for least EMA bias is MSAVI>
SAVI> NDVI with average bias values of 101 m, 203 m, and 270 m, respectively. Estimations using NDVI have been found unable to
satisfy the direct relationship between waste heap and hazardous zone size and have given a false exaggeration of 374 m for relatively
smaller dump as compared to the bigger one. The same false exaggeration for SAVI and MSAVI is measured to be 86 m and -14 m,
respectively. So MSAVI is the only VI that has shown the true relation of waste heap and hazardous zone size. The best visualization
of distance-dependent vegetation health away from the dumps is also provided by MSAVI.

Keywords
Municipal solid waste, open dumps, satellite remote sensing, geographic information system, vegetation indices, hazardous zone

Received 17th July 2018, accepted 18th October 2018 by Editor in Chief Arne Ragossnig

Introduction
Municipal solid waste (MSW), whether it comes directly from
residential colonies, a material recovery facility or residue from
combustion or composting, its ultimate fate lies in its disposal by
piling it up. Moreover, the disposal of waste by piling it up is
known as the most used method and neglecting the indirect costs
is the cheapest of all the waste management techniques (Biswas
et al., 2010; Gbanie et al., 2013; Şener et al., 2010; Zhang et al.,
2011; Veverkova et al., 2013). Development of environment
friendly solutions for the disposal of MSW is a challenge as it can
contaminate soil, water and air, that ultimately disturbs ecosys-
tems around such dumps (Bellezoni et al., 2014; Hard et al., 2013;
Wang et al., 2012). The literature shows that much work has
already been done to develop criteria for addressing environmen-
tal, political, and emotional perspectives of locating such obnox-
ious facilities for the dumping of waste (Baiocchi et al., 2014;
Erkut and Moran, 1991; Geneletti et al., 2009; Gorsevski et al.,
2012; Lober and Green, 1994; Morrissey and Browne, 2004).
Still, use of the non-scientific ways of disposing the waste is very
common in non-developed and developing countries, due to their
limited resources (Ali et al., 2014; Batool and Ch, 2009; Mahmood
et al., 2015, 2016, 2017). In this way the most primitive way of
getting rid of MSW is to throw it away from the residential
colonies in open places. The waste collected from houses is just
disposed of at open places without providing any engineered
treatment to the disposal site (Ali et al., 2016; Ayub and Siddiqui,
2015; Batool and Ch, 2009; Bellezoni et al., 2014; Mahmood
et al., 2015, 2016). These kind of dumping facilities are dangerous
to the environment by a much greater degree than proper landfills
and are known as open dumps. In open dumps all the emissions
from decomposition of waste can directly damage the surrounding
environment. Two main emissions of this kind are gases and
leachate.
As a whole, waste is a noteworthy contributor to the emissions
of greenhouse gases such as carbon dioxide (CO2), methane, and
nitrous oxide into the atmosphere. This contribution comprises
about 5% of the global greenhouse gas budget that is increasing
with rise in MSW generation rate and changes with climatic
1Remote Sensing and GIS Group, Department of Space Science,
University of the Punjab, Lahore, Pakistan
2Department of Space Science, University of the Punjab, Lahore,
Pakistan
Corresponding author:
Khalid Mahmood, Remote Sensing and GIS Group, Department of
Space Science, University of the Punjab, Lahore, 54590, Pakistan.
Email: khalid.m270@yahoo.com
220
conditions (Bogner et al., 2008). For reference, a tonne of dumped
MSW has been reported as an equivalent of about 4.8 tonnes of
CO2 (Eduljee, 1995). The second counterpart, landfill leachate,
has been held responsible for the attenuation of both soil and
groundwater through many chemical, physical, and biological
processes (Ahmed and Lan, 2012; Bellezoni et al., 2014;
Mahmood et al., 2013; Veverkova et al., 2013). This contamina-
tion of soil and groundwater ultimately affects human and plant
health (Ali et al., 2014; Mahmood et al., 2016). So, collectively
both kinds of emissions affect surrounding life through direct and
indirect ways. An estimation of the radial range and severity of
the threat they pose for surrounding life is an issue needed to be
fully understood for planning any possible remedial measures.
The literature provides a very comprehensive set of waste man-
agement models and their implementation for sustainable solu-
tions (Morrissey and Browne, 2004). Still, the developing world
remains unable to plan remedial measures based on established
scientific techniques owing to their cost, and the alternative solu-
tions usually prove inappropriate over the long term. So, the
researchers have now started to develop inexpensive and easy to
adopt alternatives for scientific studies to support the decision-
making system (Mahmood et al., 2016, 2017).
The use of remotely sensed satellite data has emerged as one
of the supplementary and cost-effective substitutes of MSW
dump monitoring (Gao and Liu, 2010; Jones and Elgy, 1994; Yan
et al., 2014). One such example is Lacoboaea and Petrescu (2013)
who propose the use of this technology as a substitute for in situ
measurements that may be cumbersome for dumps monitoring,
such as determining the number of monitoring instruments, posi-
tioning of observation points, and above all time required to set
historical records.
In this regard, early researchers have used aerial photographs
instead of satellite images due to their better spatial resolution and
have successfully characterized the dumps using their spectral
response and textural details (Bagheri and Hordon, 1988; Erb
et al., 1981; Pope et al., 1996). Digital image processing and a geo-
graphic information system (GIS) almost always work together to
interpret remotely sensed satellite data (Yan et al., 2014). GIS is
popular for planning and management as it can easily link different
disciplines such as resource management, land suitability analysis,
natural hazards, forestry, transportation, geology, environmental
science, etc. (Gorsevski et al., 2012; Hannan et al., 2015; Lopez
et al., 2008). GIS not only reduces time and cost, but also provides
spatial relations that help to broaden the view of analysts. Most of
the planning and operational aspects of waste management depend
highly on spatial analysis, global positioning and mapping, etc.,
which are all key features of GIS. Recent studies have proposed
the assessment of such environmental hazards using satellite-based
bio-indicators (Mahmood et al., 2016, 2017). These indicators use
information about the health of the surrounding vegetation that can
be derived using remotely sensed satellite data.
Selection of the spatial analysis to interpret satellite data as an
indicator of MSW dumps hazards depends on geographical condi-
tions which vary due to complex land cover patterns surrounding
Waste Management & Research 37(3)
municipal solid waste open dumps (MSWODs) (Mahmood et al.,
2016, 2017). These might include urbanization patterns, incon-
sistent crop patterns, existence of other nearby sources of pollu-
tion, etc. So, in order to deal with the monitoring of MSWODs
with the help of indirect ways, all the possible methods and tech-
niques are required to be investigated for their suitability under
given natural settings around MSW dumps.
The interaction of vegetation and solar radiations is unique in
the sense that plants typically absorb in the red (around 650 nm) and
blue (around 470 nm), and reflect the green (around 510 nm) and
near infrared (700 nm to 1300 nm) portions of the electromagnetic
spectrum. Further variations in the absorption and reflection pat-
terns are controlled by plant type, pigmentation level, water con-
tent, carbon and nitrogen content, etc. In this way remote sensing of
vegetation has proved to be an attractive alternative to complex in
situ measuring devices for analyzing vegetation health (Basso et al.,
2004; Mahmood et al., 2016; Vina et al., 2011). Another advantage
of remote sensing monitoring of crops is its ability to gather infor-
mation in a non-destructive way (Basso et al., 2004).
There exists a list of VIs that have been developed for study-
ing vegetation characteristics. Use of satellite-based VIs for
studying crop variables such as percentage vegetation cover,
biomass, leaf area index, vigor, health, rate of photosynthesis,
surface energy balance and many other biophysical variables has
greatly increased in the last couple of decades (Gitelson et al.,
2002; Muneni et al., 1997; Vina et al., 2011; Wu, 2014).
Generally, two methods can be selected for the retrieval of bio-
physical information of vegetation, one is the use of VIs and the
other is model inversion (McDonald et al., 1998). VIs always
have been a preferred option over their counterpart, because of
their simplicity and relationship with various biomass parame-
ters (Silleos et al., 2006).
These indices are designed such that the influence of extrane-
ous variables such as soil reflectance, atmospheric hindrances and
varying availability of sunlight can be avoided (Moulin and Guerif,
1999). Due to limitations of spatial resolution of the used data and
varying landcovers, surrounding MSWODs, Mahmood et al.
(2016, 2017) have used many functions to simplify the results that
require a comparison of various possibilities to make the results
more reliable while passing through these simplifying functions.
Therefore, the aim of this study is to compare performance of vari-
ous available satellite-based bio-indicators, for assessing and com-
paring biohazards of two of the open dumps. In this way this study
is expected to raise the level of trust over satellite-based environ-
mental hazard assessment and remedial strategy planning.
Study area
The aim of this study is to compare bio-indicators of vegeta-
tion health that are getting popular to assess environmental
hazards of MSWODs (Mahmood et al., 2016, 2017). To make
the comparison of VIs more dynamic these have been com-
pared for two different MSW dumps situated in different geo-
graphical conditions and having different type and amount of
Mahmood et al. 221

Figure 1. Spatial association of the study area.

dumped MSW. Both the chosen sites are separated by about Table 1. Comparison of Mahmood Booti Municipal Solid
120 km of linear distance and therefore lie in similar hot and Waste Open Dump (MB-MSWOD) and Main Faisalabad
semi-arid climatological conditions. Municipal Solid Waste Open Dump (MF-MSWOD).
One of the dumps, called Mahmood Booti Municipal Solid Characteristics MB-MSWOD MF-MSWOD
Waste Open Dump (MB-MSWOD), is in Lahore. MB-MSWOD
Area of the dumping site (m2) 278,155 140,578
is a government owned dumping facility lying at 31.610°N and
Approximate height of the 20 8
74.385°E in the northern outskirts of Lahore and is in use for dumped waste heap (m)
dumping of MSW since 1997 (Mahmood et al., 2015). Lahore Depth of water table beneath 26.21 8.52
itself lies between longitudes 74.012°E to 74.641°E and between the dump (m)
latitudes 31.24°N to 31.751°N, with an average elevation of Co-efficient of permeability 42.3 30
210 m above mean sea level (MSL). With an area of 1,772 km2 averaged to water table (µm/s)
Lahore is the second largest megacity of Pakistan and the provin- Distance from nearest 210 27
residential area (m)
cial capital of Punjab (Alam et al., 2012).
Residential area within 1 km of 1,662,011 191,862
The other dump serves as the main dumping facility of the dump (m2)
Faisalabad and is called Main Faisalabad Municipal Solid Waste Population Living within 1 km of 67,379 1,748
Open Dump (MF-MSWOD). MF-MSWOD is the first govern- the dump
ment owned dumping facility in Faisalabad, lying at 31.386°N and Distance to nearest school (m) 848 3,686
73.242°E, where MSW dumping started in 2003. Faisalabad has an Distance to nearest place of 545 364
worship (m)
area of about 1,496 km2 and an approximate population of 2.86
million. It lies between longitudes 72.8°E to 73.3°E and latitudes
31.15°N to 31.63°N and has an average altitude of 186 m above
MSL. Being an industrial hub, the city of Faisalabad has about
Material and methods
3,000 small, medium and large industrial units mostly dealing with Remotely sensed satellite data of Landsat-8 from April 2013 to
the textile production. Spatial association of both the MSWODs is November 2015 has been acquired from Earth Explorer. The
shown in Figure 1 and their comparison with reference to some of used images are surface reflectance products provided by the
the physical and spatial properties is given in Table 1. United States Geological Survey (USGS). These atmospherically
222 Waste Management & Research 37(3)

Table 2. Acquisition dates of selected images over study area.

Number Date of acquisition Number Date of acquisition Number Date of acquisition

1 16 April 2013 2 2 May 2013 3 18 May 2013
4 23 September 2013 5 25 October 2013 6 10 November 2013
7 26 November 2013 8 12 December 2013 9 28 December 2013
10 29 January 2014 11 18 March 2014 12 3 April 2014
13 5 May 2014 14 21 May 2014 15 6 June 2014
16 5 August 2014 17 26 September 2014 18 12 October 2014
19 13 November 2014 20 6 January 2015 21 1 February 2015
22 17 February 2015 23 21 March 2015 24 8 May 2015
25 24 May 2015 26 9 June 2015 27 25 June 2015
28 13 September 2015 29 29 September 2015 30 15 October 2015

corrected data are generated from the Landsat Surface Reflectance
Code, which uses coastal aerosol band to perform aerosol inver-
sion tests, auxiliary climate data from the Moderate Resolution
Imaging Spectroradiometer and a unique radiative transfer
model. The index reference of the image in which both the
MSWODs lie is 038 as the row number and 149 as the path num-
ber. All the available images during the selected temporal win-
dow have been assessed for cloud overs over the study regions, as
existence of clouds can hide surface information causing errone-
ous and highly unreliable results. A total of 30 images have
passed the cloud cover scrutiny test. Table 2 shows the acquisi-
tion dates of the finally selected images. As the downloaded
images were level-1 products of Landsat-8, therefore they were
already processed for all the calibrations and atmospheric correc-
tions by the USGS to reduce any effect of atmosphere. These
atmospherically corrected data provide reflectance information
about Earth’s surface and are named as surface reflectance data.
All these images were preprocessed by performing the operations
of layer stacking, to combine their individual bands into a single
file for the onward calculations of three principal VIs.
One of these VIs is the normalized difference vegetation index
(NDVI), that is computed using equation (1), as the raster expres-
sion (Rouse et al., 1973):
RNIR − RRed
NDVI = (1)
RNIR + RRed
where, RNIR represents reflectance of a pixel in near infrared band
(band-5 of Operational Land Imager (OLI)) and RRed is the reflec-
tance of same pixel in red band (band-4 of OLI). This index has
output values ranging from -1 to +1 with greater sensitivity to
low vegetation densities. The values of the NDVI increase with
the growth of vegetation (Jackson and Huete, 1991). It is the
most widely used vegetation index as it retains the ability to nor-
malize topographic effects while producing outputs on a linear
measurement scale (Silleos et al., 2006). The NDVI assumes that
the noise effect from background soil, atmosphere, solar and
viewing effect is normalized, but it is not true (Basso et al., 2004).
The second VI considered for comparison in this study is the
soil adjusted vegetation index (SAVI). This VI has been proposed
by Huete (1988). For its development Huete (1988) introduced a
correction factor for soil adjustment to the NDVI expression given
in equation (1). After the introduction of soil adjustment factor ‘L’
to account for the background soil signal the NDVI expression
changes into equation (2) (Huete, 1988; Silleos et al., 2006).
RNIR − RRed
SAVI = (1 + L ) (2)
RNIR + RRed + L
where L may range from 0 for dense vegetation cover to 1 for
very sparse vegetation cover, whereas the typically used value of
0.5 is for intermediate vegetation cover or for a study area that
has a mixture of dense and sparse vegetation under most common
environmental conditions (Xue and Su, 2017). It is obvious that a
value of 0 for L transformed SAVI expression to NDVI.
In a study area where green cover is of 40%, the effect of noise
in the NDVI results is about 10 times greater than that of SAVI,
and corresponds to an estimation error of +/- 23% for NDVI and
+/- 2.5% for SAVI (Basso et al., 2004). Therefore, SAVI is advised
in case of sparse vegetation to compensate for the effect of exposed
soil (Xue and Su, 2017). So, SAVI is a preferred choice over NDVI
for sparse vegetation cover, but an optimization of the factor ‘L’
increases its accuracy (Qi et al., 1994; Xue and Su, 2017).
The third VI is a modification of SAVI and is called the modi-
fied soil vegetation index (MSAVI). It was developed by Qi et al.
(1994), on the basis of variation in the value of factor ‘L’ in equa-
tion (2). ‘L’ is computed to adjust its value for different soil types
and canopy cover using equation (3). Its value varies from 0 (for
sparse canopy) to 1 (for thick canopy).
(
L = 1-2 s ( NDVI ) R NIR − s ( R Red ) ) (3)
Here “s” is slope of the soil line determined using plot of RRed
versus RNIR. Putting value of L from equation (3) in to equation
(2), Qi et al. (1994) have concluded the MSAVI expression as
given in equation (4):
( R NIR + 1) − 8 ( R NIR − R Red )
2
2R NIR + 1 − (4)
MSAVI =
2
These calculated VIs were further subject to proximity analysis.
For proximity analysis multi ring neighboring zones around
both the MSWODs were created in a GIS environment. Each of
the created neighborhood zones is of 20 m radial extant away
from the dumps extended up to 1000 m for MB-MSWOD and
Mahmood et al. 223

Figure 2. Proximity zones and vegetation cover around Main Faisalabad Municipal Solid Waste Open Dump (A) and Mahmood
Booti Municipal Solid Waste Open Dump (B).

up to 1600 m for MF-MSWOD. These proximity zones have Table 3. Seasonal windows used for the study.
been used as spatial regions for averaging the value of each of
Season Starting date Ending date Number of days
the VIs within them. To avoid any possible ambiguity in the
averaging process that may arise due to non-vegetative land- Winter 6 November 10 March 116
covers, these zones have been restricted to vegetative landcover Spring 11 March 30 April 51
Dry summer 1 May 6 July 67
using spatial overlay of intersection of vegetation cover and
Monsoon 7 July 15 September 70
proximity zones. The resultant zones are shown in Figure 2
Wet summer 16 September 15 November 61
along with the zoomed detail and vegetation level and their spa-
tial distribution around both the dumps. For overcoming the dif-
ferences resulting from varying dates of sowing and watering of further averaged over temporal windows of seasons. Considered
the surrounding vegetation, the resultant averages have been seasonal windows are given in Table 3.
224
Figure 3. Vegetation indices profile comparison for both the sites before the start of municipal solid waste dumping.
Finally, the seasonal averages have been plotted against dis-
tance of corresponding proximity zone away from the source of
pollution. Complex neighboring land covers, due to existence of
multiple factors, usually result into very edgy graphs, making it
difficult to understand the variation trend. Therefore, in order to
reduce effects from sources other than the emissions from dumps
the graphs have been provided with additional function of expo-
nentially moving average (EMA) as recent studies have done, to
estimate the affected zone from emissions emanating from
MSWODs (Mahmood et al., 2016, 2017). EMA is a trend func-
tion and a momentum indicator that has ability to suppress any
insignificant anomaly in the data caused by factors other than the
main one. This function is a type of weighted moving average
that gives more significance to near values than those that are far
apart, by a weight factor, while calculating the average for a loca-
tion over graph. EMA for distance “n” from the dump is defined
using the following formula:
 2   2 
EMA ( n ) = P ( n )   + EMA ( n − 1) 1 −  (5)
 T + 1   T +1 
where P(n) and EMA(n) are the average and EMA values meas-
ured for distance n respectively, T is the number of neighbor-
hood readings incorporated while estimating EMA, which is 10
in this case. The weight provided to value at distance “n”, with
total incorporated neighboring values of 10, is 0.182 and the
Waste Management & Research 37(3)
nine neighbors share a weight of 0.818. EMA(n-1) is the meas-
ured EMA for previous distance, which is usually calculated as
a simple average of the first 10 values for an initial EMA
(EMA(0)) to kick start the onward EMA calculation. In the case
of this study true values measured for first proximity zones are
taken as EMA(0).
The study is rooted in the fact that if emissions from dumps
are responsible for any damage to surrounding vegetation, then
surely the measured vegetation health near the dumps should be
weak, and improves in going away from them (Ali et al., 2014;
Mahmood et al., 2016, 2017). At a distance where the effect of
emissions from MSWODs loses its influence, the vegetation
health curve should be flattened with almost no change in health
of plants with distance or at least the improving health trend
should vanish. Recent studies are using the same criteria by vis-
ually analyzing the distance-dependent VI curves. These studies
are using true value graphs for estimating the effective zone
radius if the MSWOD is surrounded by vegetation all around as
is the case with MF-MSWOD. If the dump is surrounded by a
variety of land covers then due to the existence of many factors
controlling VI values researchers have used EMA curves to
serve the purpose. Based on this concept, a comparison is also
needed for the bias generated by each of the VIs while shifting
from true value curves to EMA curves. Additionally, to verify
the above hypothesis of mapping the hazardous zone using
Mahmood et al. 225

Table 4. Radial extent of bio-hazard zone around municipal solid waste open dumps measured using normalized difference
vegetation index (NDVI), soil adjusted vegetation index (SAVI), and modified soil adjusted vegetation index (MSAVI).

NDVI

2013 2014 2015 Average Average TV Average EMA

EMA MF MF MB
EMA MF TV MF EMA EMA MF TV MF EMA EMA MF TV MF EMA
MB MB MB
Spring 1200 800 940 600 400 360 500 400 500 766 533 600
Dry summer 1200 900 940 900 700 340 1000 700 300 1033 766 526
Monsoon 1200 700 800 1100 700 340 900 700 160 1066 700 613
Wet summer 1200 900 ND 1100 800 320 1100 700 160 1133 800 560
Winter 1000 500 400 500 400 200 ND ND ND 750 450 300

SAVI

2013 2014 2015 Average Average TV Average EMA

EMA MF MF MB
EMA MF TV MF EMA EMA MF TV MF EMA EMA MF TV MF EMA
MB MB MB
Spring 1100 900 940 500 400 900 400 400 500 666 566 780
Dry summer 1100 900 940 1100 700 200 900 700 1000 1033 766 712
Monsoon 1100 700 760 1100 1100 300 900 800 100 1033 866 387
Wet summer 1200 1100 ND 1000 800 300 1000 700 100 1066 866 450
Winter 1100 500 400 500 400 1000 ND ND ND 750 450 700

MSAVI

2013 2014 2015 Average Average TV Average EMA

EMA MF MF MB
EMA MF TV MF EMA EMA MF TV MF EMA EMA MF TV MF EMA
MB MB MB
Spring 1200 1100 920 400 400 900 500 400 500 700 633 773
Dry summer 1100 900 960 800 700 360 900 700 1000 933 766 774
Monsoon 1100 1100 800 1100 1100 300 800 700 100 1000 966 400
Wet summer 1100 900 ND 800 700 260 800 700 100 900 766 180
Winter 700 500 360 400 400 1000 ND ND ND 550 450 680

Notes: EMA: exponential moving average; TV: true value; MF: Main Faisalabad Municipal Solid Waste Open Dump; MB: Mahmood Booti
Municipal Solid Waste Open Dump; and ND: no data.

vegetation health as a function of emissions and distance from
MSWODs the analysis has been performed to an older image of
Landsat-5, captured on 23 February 1994. In 1994 there was no
dumped MSW on both of the sides, so there should not exist the
same vegetation health improvement pattern in moving away
from the dumps.
Results and discussions
Before getting into detailed analysis of VIs curves it is impor-
tant to discuss that vegetation health improvement patterns
exist due to dumped MSW. For this Figure 3 is showing VIs
distance-dependent profiles of the same study area for 23
February 1994, when there was no MSW dumped at the sites.
Even with the use of EMA curves, no patterns of improving
vegetation health have been found around any of the MSWODs.
This way, the use of the VIs distance-dependent profile as a
bio-indicator to map MSW dumping hazards to its surrounding
is confirmed.
After confirmation of the correctness of the working hypoth-
esis, keeping in view the introduction of any possible bias in
visual analysis and EMA a comparison has been made between
VIs for their performance in measuring effective zone radius
over time. The results for comparison are given in Table 4 and
corresponding profile graphs from 2013 to 2015 are shown in
Figures 4–12. It is worth mentioning here that the results con-
tain the hazardous zone radius for MF-MSWOD estimated
using both true value curves and EMA curves whereas for
MB-MSWOD the only measurement available is the one that is
based on EMA curves. Measurement of the hazardous zone
radius for MB-MSWOD using true value curves is not possible
due to complex patterns caused by the existence of a limited
agricultural window around it.
NDVI
Although NDVI is the basic VI and the most used one, it is the
least reliable among others for studying hazardous effects of
226
Figure 4. Seasonal profile comparison of normalized difference vegetation index for 2013.
MSWODs, especially when the researcher has to depend on
EMA curves rather than true value curves due to the complexity
of surrounding geography. A comparison of both the sites from
2013 to 2015, using NDVI is shown in Figures 4–6. The EMA
bias of NDVI over true value averages out to 270 m, with a
maximum average of 320 m, found in 2013. The bias has
decreased in later years and averages to 240 m and 250 m in
2014 and 2015, respectively. Inefficiency of NDVI to handle
mixing of soil signal in sparse vegetation and the EMA-bias,
Waste Management & Research 37(3)
collectively marks inappropriateness of this VI for the compara-
tive as well as individual biohazard assessment of the MSWODs
under consideration.
The maximum range of the bio-hazardous zone of
MB-MSWOD, as measured through NDVI is found in spring and
dry summer of 2013. In the later years the distance measured for
these seasons has changed dramatically. This may be due to the
sparse cultivation pattern of vegetation, and therefore, mixing of
soil signal that NDVI is unable to account for.
Mahmood et al.
Figure 5. Seasonal profile comparison of normalized difference vegetation index for 2014.
The same decrease in the boundary extent of the hazardous
zone has also been observed for other seasons as well. Hazardous
zone radius has been decreased from 800 m in 2013 to 160 m in
2015 for monsoon, from 320 m in 2014 to 160 m in 2015 for wet
summer, and from 400 m in 2013 to 200 m in 2014 for winter. It
has happened due to absence of vegetation in the outer proximity
zones as explained by Mahmood et al. (2017). In this way, the
radial extent of hazardous zone, as measured using flattening of
NDVI curves, ranges from 300 m in winter to about 613 m in
227
monsoon, and if monsoon and wet summers are ignored then
maxima will be at about 600 m, found in spring.
SAVI
Performance of the SAVI for vegetation health measurement
around the MSWODs is ranked higher than NDVI due to its abil-
ity to handle sparse vegetation spectrally mixing with the back-
ground soil. The EMA bias of SAVI over true value averages to
228
Figure 6. Seasonal profile comparison of normalized difference vegetation index for 2015.
203 m, with maximum bias of 300 m, averages in 2013. This bias
decreased in the following years and averaged to 160 m and 150
m in 2014 and 2015, respectively. Both of these parameters rank
SAVI as a better option for this comparison of MSWODs.
The details of difference measured by SAVI between both the
compared MSWODs are shown in Figures 7–9. Radial distance
found for the vegetation hazardous zone around MB-MSWOD,
as per SAVI profiles for spring and dry summer of 2013 is the
same as that for NDVI, that is, 940 m. The peak value of this
radial distance however is 1000 m that has been found in winter
2014 and dry summer 2015. Existence of this peak radial dis-
tance of the MB-MSWOD biohazardous zone in winter is very
different behavior if compared with that of the MF-MSWOD,
where this range never exceeds 500 m for the season. A radial
Waste Management & Research 37(3)
extent of 1000 m in dry summer 2015 for the hazardous zone
around MB-MSWOD is also very different from that of the
MF-MSWOD, where it was measured to be 700 m for true value
and 900 m for EMA of SAVI.
These two measurements of 1000 m suggest that the severity
of damage to vegetation health around MB-MSWOD is greater,
but due to sparse and irregular vegetation pattern it is not very
prominent. Highest radial distance average, as measured
through SAVI, is about 800 m measured for spring, and mini-
mum is for monsoon, that is, 320 m. By ignoring the anomalous
seasons of monsoon and wet summer, the minimum radial dis-
tance measurement will fall in winter, that is, 700 m. It is inter-
esting to note here that extent of the hazardous zone varies from
700 m to 800 m around MB-MSWOD, whereas for MF-MSWOD
Mahmood et al.
Figure 7. Seasonal profile comparison of soil adjusted vegetation index for 2013.
this variation was measured to be ranging from 450 m to about
850 m, and the peak value found is in the seasons that have been
ignored for MB-MSWOD. So, the severity of damage by the
bigger MSWOD is more but is suppressed due to uneven pat-
terns of monsoon and wet summer crops. Although the scope of
this study is limited to performance evaluation of VIs as bio-
indicators of biohazards of MSWODs, existing geographical
conditions are also an important parameter that can disturb the
229
precision and hence reliability of the results of proximity
analysis.
MSAVI
Vegetation health measuring performance of MSAVI is the best
among the compared VIs due to better ability to handle mixing of
the crop–soil spectrum. A comparison of sites from 2013 to 2015,
230
Figure 8. Seasonal profile comparison of soil adjusted vegetation index for 2014.
using MSAVI is shown in Figures 10–12. The EMA bias of
MSAVI over true value averages to just 101 m, with a maximum
of 140 m, averaged for 2013. This bias has decreased and aver-
ages to 40 m and 125 m in 2014 and 2015, respectively.
All the radial distances measured for the biohazard zone of
MB-MSWOD, using MSAVI profiles, are almost the same as that
of the measure using SAVI profiles. The only difference found is
the visual clarity of the MSAVI profiles, making it easier to
Waste Management & Research 37(3)
discriminate rise of the curves from flatness. The differences
between true value and EMA observation of MSAVI for
MF-MSWOD are measured to be lowest among all three candi-
dates for measuring vegetation health. This makes the MSAVI
profile most reliable for comparison of both the MSWODs.
Ignoring anomalous seasons of monsoon and wet summer, the
effect of MF-MSWOD ranges to about 600 m on average, with
minimum of 450 m in winter and maximum of about 750 m in
Mahmood et al.
Figure 9. Seasonal profile comparison of soil adjusted vegetation index for 2015.
dry summers. Whereas, for MB-MSWOD this range averages to
about 700 m with minimum of 680 m in winter and maximum of
770 m in spring.
Considering the fact that severity of damage by the bigger
MSWOD is greater, the results from all three VIs can be tested
through another way by virtue of which the seasonal radius of the
biohazardous zone should be greater for the bigger dump of
MB-MSWOD than that of relatively smaller, at least in the sea-
sons of valid observations for MB-MSWOD. Now, if a compari-
son is made for these radial distances for spring, dry summer, and
winter the NDVI is consistently showing greater values for
MF-MSWOD, when a comparison is being made between EMA
values of both the dumps. On average the false exaggeration is of
about 374 m. For NDVI even true value results of MF-MSWOD
are higher than that of EMA value results of MB-MSWOD. The
231
same EMA comparison using SAVI, results into an average false
exaggeration of about 86 m, with no true value result for
MF-MSWOD higher than that of EMA result of MB-MSWOD.
Finally, the best method of MSAVI results into the actual situa-
tion where this false exaggeration results into a negative value of
-14 m, showing that the bigger pile of MSW poses the bigger
biohazardous zone. Overall comparison of the three VIs with
respect to all the three criteria is given in Table 5.
Conclusions
This study has successfully analyzed reliability of remotely
sensed satellite data with a comparison of optimal satellite-
based bio-indicators for monitoring MSWODs. As the remotely
sensed data are recorded through reflection of solar radiations
232
Figure 10. Seasonal profile comparison of modified soil adjusted vegetation index for 2013.
instead of direct contact with the target, so reliability of result-
ant products is always a topic of research before establishing
their use as scientific information. This study has compared
three such satellite-based bio-indicators for their suitability to
assess biohazards of two MSWODs lying under different geo-
graphical conditions.
The decision about the reliability of VIs has been made
using the EMA bias, visual clarity of their corresponding
Waste Management & Research 37(3)
distance-dependent graphs, and the difference of radial meas-
urement being produced for both the compared MSWODs in
relation to the volume of dumped waste. The maximum EMA
average bias measured is of 270 m and is found for NDVI.
For a variable that attains a maximum value of about 800 m,
the possibility of 270 m error is very high, making the
results highly unreliable. The same EMA average bias found
for SAVI is 203 m and that of MSAVI is 101 m. This way,
Mahmood et al.
Figure 11. Seasonal profile comparison of modified soil adjusted vegetation index for 2014.
results of EMA bias have ranked MSAVI highest among
the compared VIs. It is very important to mention here that
this biasing is calculated for an ideal case of MF-MSWOD,
which is surrounded by vegetation all around – the difference
could be even higher if measured for a scenario of a complex
geographical condition, that is, MB-MSWOD. The hierarchy
of the visual ease for finding the flattening point of distance-
dependent VIs is MSAVI>SAVI>NDVI. Also, MSAVI is the
233
only VI that has endorsed the fact that a bigger pile of waste
(MB-MSWOD) results in a bigger biohazardous zone that is
found to be larger by a radial size of 14 m. Whereas, SAVI has
shown a false exaggeration where mean radius of the smaller
pile is larger by 86 m than that of the bigger waste heap site.
NDVI, the least suitable VI has resulted into an even bigger
false exaggeration of 374 m. Therefore, this study concludes
that use of MSAVI as bio-indicator of vegetation health is the
234
Figure 12. Seasonal profile comparison of modified soil adjusted vegetation index for 2015.
Table 5. Performance comparison for vegetation indices (VIs).
VI/criteria Exponential moving
average bias
Normalized difference vegetation index High
Soil adjusted vegetation index Moderate
Modified soil adjusted vegetation index Low
most suitable VI for studying satellite-based hazard assess-
ment of MSWODs. It is further suggested that radius of the
hazardous zone may be assessed using a weighted average
combination of all these VIs, but obviously before that,
weighing criteria need to be established. Such an improve-
ment can make satellite data based hazard assessment more
reliable by overcoming biases in the measurements.
Waste Management & Research 37(3)
Visual clarity of Municipal solid waste open dump
VI curves size assessment
Poor Wrong, with high false exaggeration
Clear Wrong, with low false exaggeration
Clear Correct, with true exaggeration
Acknowledgements
We are grateful to the United States Geological Survey for providing
the facility of free data download via Earth Explorer. This paper has
been presented at the EurAsia Waste Management Symposium 2018.
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to
the research, authorship, and/or publication of this article.
Mahmood et al.
Funding
The authors disclosed receipt of the following financial support for
the research, authorship, and/or publication of this article:
We are thankful to the University of the Punjab, Lahore, Pakistan for
the financial support to carry out this work.
References
Ahmed FN and Lan CQ (2012) Treatment of landfill leachate using mem-
brane bioreactors: A review. Desalination 287: 41–54.
Alam K, Trautmann T, Blaschke T, et al. (2012) Aerosol optical and radiative
properties during summer and winter seasons over Lahore and Karachi.
Atmospheric Environment 50: 234–245.
Ali M, Wang W and Chaudhry N (2016) Comparing administration of hospi-
tal wastes using social network analysis. In: MATEC Web of Conferences.
EDP Sciences. pp. 14009. Available at: https://www.researchgate.net
/publication/305783332_Comparing_Administration_of_Hospital_Wastes
_Using_Social_Network_Analysis (accessed 14 June 2018).
Ali SM, Pervaiz A, Afzal B, et al. (2014) Open dumping of municipal solid
waste and its hazardous impacts on soil and vegetation diversity at
waste dumping site of Islamabad city. Journal of King Saud University
– Science 26: 59–65.
Ayub S and Siddiqui YS (2015) Municipal solid waste dumping practice and
its impact assessment. European International Journal of Science and
Technology 4: 33–53.
Bagheri S and Hordon RM (1988) Hazardous waste site identification using
aerial photography: A pilot study in Burlington County, New Jersey,
USA. Environmental Management 12: 119–125.
Baiocchi V, Lelo K, Polettini A, et al. (2014) Land suitability for waste
disposal in metropolitan areas. Waste Management & Research 32:
707–716.
Basso B, Cammarano D and Vita PD (2004) Remotely sensed vegetation
indices: Theory and applications for crop management. Rivista Italiana
di Agrometeorologia 1: 36–53.
Batool SA and Ch MN (2009) Municipal solid waste management in Lahore
city district, Pakistan. Waste Management 29: 1971–1981.
Bellezoni RA, Iwaj CK, Elis VR, et al. (2014) Small-scale landfills:
Impact on groundwater and soil. Environmental Earth Sciences 71:
2429–2439.
Biswas AK, Kumar S, Babu SS, et al. (2010) Studies of environmental quality
in and around municipal solid waste dumpsite. Resources, Conservation
and Recycling 55: 129–134.
Bogner J, Pipatti R, Hashimoto S, et al. (2008) Mitigation of global greenhouse gas
emissions from waste: Conclusions and strategies from the Intergovernmental
Panel on Climate Change (IPCC) Fourth Assessment Report. Working Group
III (Mitigation). Waste Management & Research 26: 11–32.
Eduljee GH (1995) Emissions to the atmosphere. In: Hester RE and Harrison
RM (eds) Issues in Environmental Science and Technology. Number
3. Waste Treatment and Disposal. Cambridge: The Royal Society of
Chemistry, pp.69–89.
Erb TL, Philipson WR, Teng WL, et al. (1981) Analysis of landfills with
historic air photos. Photogrammetric Engineering and Remote Sensing
47: 1363–1369.
Erkut E and Moran SR (1991) Locating obnoxious facilities in the public sec-
tor: An application of the analytic hierarchy process to municipal landfill
siting decisions. Socio-Economic Planning Sciences 25: 89–102.
Gao J and Liu Y (2010) Determination of land degradation causes in Tongyu
County, Northeast China via land cover change detection. International
Journal of Applied Earth Observations and Geoinformation 12: 9–16.
Gbanie SP, Tengbe PB, Momoh JS, et al. (2013) Modeling landfill location
using geographic information systems (GIS) and multi-criteria decision
analysis (MCDA): A case study Bo, Southern Sierra Leone. Applied
Geography 36: 3–12.
Geneletti S, Richardson S and Best N (2009) Adjusting for selection bias in
retrospective, case-control studies. Biostatistics 10: 17–31.
Gitelson AA, Kaufman YJ, Stark R, et al. (2002) Novel algorithms for remote
estimation of vegetation fraction. Remote Sensing of Environment 80:
76–87.
235
Gorsevski V, Kasischke E, Dempewolf J, et al. (2012) Analysis of the
impacts of armed conflict on the Eastern Afromontane forest region on
the South Sudan–Uganda border using multitemporal Landsat imagery.
Remote Sensing of Environment 118: 10–20.
Hannan MA, Mamun MAA, Hussain A, et al. (2015) A review of technolo-
gies and their usage in solid waste monitoring and management systems:
Issues and challenges. Waste Management 43: 509–523.
Hard M, Gamperling O and Huber-Humer M (2013) Comparison between
lab- and full-scale applications of in situ aeration of an old landfill and
assessment of long-term emission development after completion. Waste
Management 33: 2061–2073.
Huete AR (1988) A soil-adjusted vegetation index (SAVI). Remote Sensing
of Environment 25: 295–309.
Jackson RD and Huete AR (1991) Interpreting vegetation indices. Preventive
Veterinary Medicine 11: 185–200.
Jones HK and Elgy J (1994) Remote sensing to assess landfill gas migration.
Waste Management & Research 12: 357–337.
Lacoboaea C and Petrescu F (2013) Landfill monitoring using remote sens-
ing: A case study of Glina, Romania. Waste Management & Research
31: 1075–1080.
Lober DJ and Green DP (1994) NIMBY or NIABY: A logit model of oppo-
sition to solid-waste-disposal facility siting. Journal of Environmental
Management 40: 33–50.
Lopez JAR, Hernández JR, Mancilla OL, et al. (2008) Assessment of
groundwater contamination by landfill leachate: A case in México. Waste
Management 28: 33–39.
Mahmood K, Batool SA, Chaudhery MN, et al. (2015) Evaluating municipal
solid waste dumps using geographic information system. Polish Journal
of Environmental Studies 24: 879–886.
Mahmood K, Batool SA and Chaudhery MN (2016) Studying bio-thermal
effects at and around MSW dumps using satellite remote sensing and
GIS. Waste Management 55: 118–128.
Mahmood K, Batool A, Faizi F, et al. (2017) Bio-thermal effects of open
dumps on surroundings detected by remote sensing – Influence of geo-
graphical conditions. Ecological Indicators 82: 131–142.
McDonald AJ, Gemmell FM and Lewis PE (1998) Investigation of the utility
of spectral vegetation indices for determining information on coniferous
forests. Remote Sensing of Environment 66: 250–272.
Morrissey AJ and Browne J (2004) Waste management models and their appli-
cation to sustainable waste management. Waste Management 24: 297–308.
Moulin S and Guerif M (1999) Impacts of model parameter uncertainties on
crop reflectance estimates: A regional case study on wheat. International
Journal of Remote Sensing 20: 213–218.
Muneni RB, Nemani RR and Running SW (1997) Estimation of global
leaf area index and absorbed par using radiative transfer models. IEEE
Transactions on Geoscience and Remote Sensing 35: 1380–1393.
Pope P, Van-Eeckhout E and Rofer C (1996) Waste site characterization
through digital analysis of historical aerial photographs. Photogrammetric
Engineering and Remote Sensing 62: 1387–1394.
Qi J, Chehbouni A, Huete AR, et al. (1994) A modified soil adjusted vegeta-
tion index. Remote Sensing of Environment 48: 119–126.
Rouse JW, Haas RH, Schell JA, et al. (1973) Monitoring vegetation systems in
the Great Plains with ERTS. Paper presented at Third ERTS Symposium,
NASA SP-351 I, pp.309–317. Available at: https://ntrs.nasa.gov/archive
/nasa/casi.ntrs.nasa.gov/19740022614.pdf (accessed 23 March 2018).
Şener S, Şener E, Nas B, et al. (2010) Combining AHP with GIS for landfill
site selection: A case study in the Lake Beyşehir catchment area (Konya,
Turkey). Waste Management 30: 2037–2046.
Silleos NG, Alexandridis TK, Gitas IZ, et al. (2006) Vegetation indices:
Advances made in biomass estimation and vegetation monitoring in the
last 30 years. Geocarto International 21: 21–28.
Veverkova M, Toman F, Adamcova D, et al. (2013) Verifying research of
waste landfill environmental impacts using bioindicators. Polish Journal
of Environmental Studies 22: 313–317.
Vina A, Gitelson AA, Nguy-Robertson AL, et al. (2011) Comparison of dif-
ferent vegetation indices for the remote assessment of green leaf area
index of crops. Remote Sensing of Environment 115: 3468–3478.
Wang J, He J and Chen H (2012) Assessment of groundwater contamina-
tion risk using hazard quantification, a modified DRASTIC model and
236
groundwater value, Beijing Plain, China. Science of the Total Environment
432: 216–226.
Wu W (2014) The generalized difference vegetation index (GDVI) for dry-
land characterization. Remote Sensing 6: 1211–1233.
Xue J and Su B (2017) Significant remote sensing vegetation indices: A
review of developments and applications. Journal of Sensors Vol. 2017.
ID 1353691, 17 pages. Available at: https://www.hindawi.com/journals
/js/2017/1353691/ (accessed 27 April 2018).
Waste Management & Research 37(3)
Yan WY, Mahendrarajah P, Shaker A, et al. (2014) Analysis of multi-tem-
poral Landsat satellite images for monitoring land surface temperature
of municipal solid waste disposal sites. Environmental Monitoring and
Assessment 186: 1861–1873.
Zhang H, Zhang DQ, Jin TF, et al. (2011) Environmental and eco-
nomic assessment of combined bio stabilization and landfill for
municipal solid waste. Journal of Environmental Management 92:
2533–2538.