The above table shows that in coastal regions specially where there is
regular accumulation of sea water in high tides, only few exotic plant
species with high salinity tolerance levels can thrive well.
Increased soil salinity not only causes less water intake in plants but
also leads to acute nutrient imbalances (Hu & Schmidhalter, 2005). This
often causes toxins detrimental for plant growth to accumulate and
reduction in water infiltration in the event of high sodium ion
(Na+) concentration(Qadir & Schubert, 2002).
Statistics reveal that globally about 1 billion hectares, close to 7%
of earth’s continental crust is salinity affected (Lewis & Maslin,
2015). Soil salinity can occur in any soil type anywhere on earth.
However, semi-arid and arid regions are the worst affected (Jordán, et
al., 2004). Soil salinity depends upon (Jolly, McEwan, & Holland,
2008)-
Salt could be found along with parent material like rock or salt
layers accumulated over time.
Leaching and weathering of parent rock materials causes the free ion
accumulation in soil thereby increasing soil salinity.
During erosion by wind or water, soil salinity can occur by materials
brought by erosive forces from one area to another.
In case when the underground water becomes saline due to prolonged
leaching, the places where the water table is near the surface soil
becomes saline after long spells of dryness and evaporation from
surface.
Most common cause of soil salinity is however the unscientific flood
irrigation. The irrigation water from tube-wells or canals if
containing dissolved salts often leads to soil salinity or increase in
the salinity levels of soil.
Remote sensing is a non-evasive and time saving tool which can be
applied effectively for monitoring of soil salinity levels and mapping
of salinity affected areas (Katsaros, Vachon, Liu, & Black, 2002).
Since many decades and ever since the inception of remote sensing as
an advanced surveying tool with launch of first remote sensing
satellite- Landsat-1 in early 1970s; spaceborne remote sensing found
extensive use in agriculture (Zhang et al., 2002). However, there was
still paucity of dedicated use of spaceborne remote sensing for soil
health studies (Pampaloni & Calvet, 2007; Wagner, Lemoine, & Rott,
1996). Initial soil health studies were done in conjunction with
remote sensing research for precision agriculture using optical
multispectral satellite datasets (Viterbo & Betts, 1999; Y. Xie, Sha,
& Yu, 2008).
Since the optical, thermal, and hyperspectral datasets have their own
constraints owing to their non availability every time and in all
seasons due to dense cloud covers specially in monsoon months, a need
for all weather data was felt (Kim & Hong, 2007). It was due to this
that microwave or Synthetic Aperture RADAR (SAR) remote sensing came
into use (Hong & Pan, 2000). It has a penetration ability and being
an active sensor, it has high temporal resolution with 24-hour data
availability (Holmes, 1959; Liu et al., 2011). Apart from this, owing
to its several modelling and decomposition approaches, it becomes the
most versatile branch of remote sensing research (Engman, 1991; Larson
et al., 2010).
But spaceborne remote sensing has its own limitations (Pachepsky,
Guber, & Jacques, 2005). High resolution optical datasets are mostly
available on commercial and paid basis and almost all currently active
SAR Datasets are available at very high costs (Delaney, 1974; Walker,
Houser, & Willgoose, 2004). This often causes budgetary constraints
on advanced research. Hence a cost-effective study is needed to carry
out research that is feasible and at par with those conducted using
commercial data. Moreover, SAR data has lots of speckle noise which is
absent in optical data (Chong & T, 2005). Keeping all the limitations
and constraints in view, this study was conducted using synergy of
both optical and SAR remotely sensed and freely available high
temporal resolution satellite data for soil salinity estimation and
modelling.
This study uses backscatter coefficients generated after calibrating
the SAR data from Sentinel-1 checks their sensitivities for soil
moisture and electrical conductivity values collected from ground
along with NDSI calculated from Sentinel-2 optical multispectral data
of the same dates (David R . Anderson, 2000; Pope & Webster, 1972).
Also, surface temperature data in Fahrenheit was also used using a
thermal imager on ground during field data collection (Congalton,
Fenstermaker, & Mcgwire, 1991; Richardson & Hollinger, 2005). This
study is unique in the sense that it uses a multisensor remote sensing
approach along with a synergy of field data to be fed into developing
a Ordinary Least Squares (OLS) model for estimation of soil salinity
(Dekker, 1998; Sowter et al., 2016).1.1 Short Literature reviewRemote sensing is defined as the science and art of data collection,
processing, interpretation, and analysis of data from a distance
without coming in actual tangible contact with the target object
(Guzha, 2004; Metternicht & Zinck, 2003). In optical remote sensing,
the target is illuminated by the Sun’s rays and the reflectance is
captured by the sensors of the satellite operating in narrow band
ranges of the electromagnetic spectrum and capture imagery in several
wavelength band ranges (Herrick, 2000; Stenberg, Rossel, Mouazen, &
Wetterlind, 2010; K. T. (eds). 2013 Wymann von Dach S, Romeo R, Vita
A, Wurzinger M, 2014).
In microwave remote sensing, the sensor being an active sensor,
illuminates the target with emitted electromagnetic waves in the
microwave region (Raina, Joseph, & Haribabu, 2010; K. T. (eds).
Wymann von Dach S, Romeo R, Vita A, Wurzinger M, n.d.). The waves are
received back after interaction with the target and are received on
the receiver of the sensor platform which generates an image of the
target object (Braidwood, 1960; Muir, Pretty, Robinson, Thomas, &
Toulmin, 2010). Remote sensing data is collected from a variety of
sensors like visible and infrared sensors, optical and thermal imaging
sensors, hyperspectral sensors and microwave sensors and based on the
respective behavioural properties of salinity effected areas with the
incident radiation, mapping of soil salinity is done (Schmugge et al.,
2002). In remote sensing applied to land resource surveys, wavelengths
between 0.4-1.5mm are most used (Aslan et al., 2016).
Landsat TM (Thematic Mapper) and SPOT were the satellites that found
usage widespread for natural resource mapping for landscapes spread
over hectares of land (Akshar Tripathi, 2018). The image type depended
upon not only the purpose but also on the sensor used and the number
of spectral bands it offered (Lobell et al., 2015), with Landsat
providing much greater number of spectral bands than SPOT. Landsat
images were used for classification, both supervised and unsupervised
(Dengsheng Lu, Tian, Zhou, & Ge, 2008). Multispectral bands 3,4,5 is
used along with TM bands 7 for the proper mapping of salinity effected
soils as described by (Hari Shanker Srivastava et al., 2008).
Davidson & Finlayson, 2007; Mahlke, 1996, used synergy of
thermal and microwave remotely sensed data and used RADAR backscatter
for fresh and saline water and surface temperature as parameters to
model for soil salinity. (Leckie, 1984) found that Landsat bands 1-5
and 7 are sensitive for soil minerals and are good for mapping when
salinity causing minerals are dominant in soil also soil salinity
affects the thermal properties of the soil. (Anderson & Croft, 2009)
used Landsat MSS data to produce maps for calcareous, gypsiferous and
clayey soils and also found the TM bands helpful when used with aerial
photographs for arid and semi-arid regions (Mougenot, Pouget, &
Epema, 1993). Sharma, Saxena, & Verma, 2000 used topographic survey
maps and standard False colour composite from Landsat MSS imagery to
map saline and non-salinity affected areas. Saha, Kudrat, & Bhan,
1990, used digital image classification using Landsat data
and successfully mapped saline, non-saline and moderately saline areas
which were waterlogged for a long period of time, with 96% accuracy
in classification. Similar classification study was done by D Lu &
Weng, 2007.
Calvão & Palmeirim, 2004 used band rationing and proved that in
Middle to Near Infrared bands from Landsat were useful in mapping
chlorosis affected soils. Mougenot et al., 1993, used thermal
infrared band and found that the hygroscopic properties of soil can be
analysed and the reflectance from leaves of plants depends upon the
chemical composition of the dissolved salts in the up taken water and
morphology of plant. Hansen, Dubayah, & Defries, 1996, found
classification tree was useful when Normalised Differential Vegetation
Index (NDVI) is used along with brightness index for mapping of soil
salinity affected areas in Morocco and Pakistan Respectively. Ahmed &
Luis, 2010, used multiple regression analysis using electrical
conductivity values from field and based on that generated a soil
salinity map for entire Mexico. Mayaux et al., 2004 used NDVI and
Surface Energy Balance Algorithm for Land (SABAL) to map and classify
the soils based on the various salinity levels.
It is clear from the studies above that most of the soil salinity
studies conducted were of qualitative classification and mapping
based. There are only few dedicated studies for soil salinity
estimation and modelling using multiple remote sensing sensors in a
quantitative way.
Most of the studies conducted for soil salinity estimation using
remotely sensed data have been done at surface level since remote
sensing is a surface phenomenon. But after going through literature,
it was found that it is the sub-surface soil salinity that affects the
plant growth as the salts underneath get trapped in root nodules and
prevent further moisture intake by plants. This study is one of the
few studies which estimate soil salinity in terms of electrical
conductivity using remotely sensed SAR and optical data in synergy
with field data. Apart from this, it was also aimed to provide a
simple and robust soil salinity estimation approach, in this study.
Early stage of wheat crop growth was chosen owing to the presence of
more exposed area of soil to the satellite sensor during it pass.
Moreover, C-band satellite data from Sentinel-1 cannot penetrate the
vegetation cover once the crop matures. Hence study of soil becomes
difficult.1.2 Electrical Conductivity as Soil Salinity IndicatorThe amount of electrical current that a material allows to pass
through it, or the current carrying capacity of a material is defined
as its electrical conductivity(Frackowiak, 2001). Electrical
conductivity is also known as specific conductance and is measured in
mS.m-1 (milli Siemens per meter)(Riffat & Ma,
2003). Electrical conductivity of soil correlates with soil properties
like soil texture, cation exchange capacity, salinity, as a
measurement of current conducting capacity of soil (Landauer, 1978).
Since soil salinity refers to the concentration of ions in the soil
pore water that make the further water up take difficult, the
laboratory determination of soil salinity is a cumbersome process
which is time taking(Corwin & Lesch, 2003). Electrical conductivity
measurement as an indicator of soil salinity is an effective and time
saving method for soil salinity estimation(Rhoades, 1993). The cause
of the electrical conductivity is the ions present in the soil which
become more loosely bound to the soil pores in presence of moisture
owing to the high dielectric constant of water (Corwin & Lesch,
2005). The ions get aligned once the electric current is applied to
soil, which are otherwise in random state in the soil(Rhoades &
Miyamoto, 1990).1.3 SAR Backscatter coefficientsBackscatter is defined as the RADAR signal that returns to the
receiver of the antenna after interaction with target under
observation(Allbed & Kumar, 2013). The coefficient of scatter of
RADAR signal in RADAR direction is called backscatter coefficient,
represented by sigma nought σ0(Verhoest et al.,
2008). It is the RADAR backscatter per unit area of the distributed
target with which the incident signal interacts. It is measured in
degree decibels -dB ( Tripathi, Maithani & Kumar, 2018; Shashi, 2019;
Tripathi & Maithani, 2018).
Beta nought (β0) is the brightness coefficient of
the RADAR. It is a dimensionless quantity. Gamma nought (γ0) is the RADAR backscatter coefficient dependent
upon RADAR Brightness and incidence angle and is suitable for volume
Scatterers(Saha, 2011; Tripathi and Tiwari, 2019).1.4 Problem StatementThe prime objective of the study is to estimate the soil salinity in
early stage of wheat crop growth at sub-surface level in a simple and
robust way with freely available satellite datasets while maintaining
the accuracy good enough.2. Study Area and Datasets2.1 Study AreaRupnagar district of Punjab state in India is the study area for this
study. Punjab has been the cradle of India’s first green revolution
and has been a primarily agriculture dependent state(Kumar, Kumar, &
Mittal, 2004). Though in terms of annual rainfall, it lies in the
semi-arid zone but owing to the nutrient rich alluvial soil brought
down from the Himalayas by the five rivers and extensive canal and
tube-well network for irrigation, Punjab leads the country in wheat
production. Along with Haryana, it is the wheat bowl of the country.
Rupnagar lies to the north of India’s first planned city after
independence- Chandigarh. Rupnagar is a district and headquarters of
Rupnagar division of Punjab(Levinson et al., 2004).
Rupnagar located on the banks of river Sutluj is one of the prominent
sites of the erstwhile Indus Valley Civilization. Located between
30.970 N and 76.530 E, Rupnagar’s
average elevation is 260m above mean sea level and is bordered by the
Shivalik hills of the mighty Himalayas in North and North-East(Kumar
et al., 2004; Levinson et al., 2004).2.2 DatasetsSAR/Microwave remote sensing data from Sentinel-1 satellite of the
European Space Agency (ESA), Optical multispectral remote sensing data
from Sentinel-2 satellite of the ESA. The datasets were acquired from
the Alaska Satellite Facility (ASF) which is a freely available
datasets for download from this portal. The data used were of
20th November and 27th December
2019 and 20th January 2020, which were the days of
satellite pass over Rupnagar. It was decided to acquire the field data
of volumetric soil moisture and Electrical Conductivity (EC)
simultaneously on the same dates at 60 cm depth. The following table
(Table 2) gives details of Datasets used-Table 2. DATASETS USED