SOIL MOISTURE DETERMINATION USING REMOTE SENSING DATA FOR

THE PROPERTY PROTECTION AND INCREASE OF AGRICULTURE

PRODUCTION

Ivan Potić1,2, Marko Bugarski1,2, Jelena Matić-Varenica1

1
Sector for Topography and Cartography, 2Department for Topography-Unit for
Photogrammetry and Remote Sensing, Republic Geodetic Authority, Serbia
[email protected]

Paper prepared for presentation at the

“2017 WORLD BANK CONFERENCE ON LAND AND POVERTY”
The World Bank - Washington DC, March 20-24, 2017

Copyright 2017 by author(s). All rights reserved. Readers may make verbatim copies of this
document for non-commercial purposes by any means, provided that this copyright notice
appears on all such copies.
Abstract
Today we are witnesses of global climate changes confirmed by the temperature increase in past 100
years which cause more frequent natural disasters across the globe. In the past six years, the
environment in Serbia had suffered great changes when several natural disasters occurred (forest fires,
drought, flooding’s). In 2012 the entire territory was affected by extreme drought, and according to
the unofficial data, the Government allocated about €150 million to redress its effects. According to
the Serbia floods 2014 Recovery Assessment Report, the total amount of the damage and effects is
over €1,700 million. Altogether the damage amounts to approximately €1.9 billion. To provide
geospatial data that enables generation of adequate information related to floods and droughts, we
applied the remote sensing method that relies on the use of soil moisture index (SMI) which in its
algorithm uses the data obtained from satellite sensors. As presented by Hunt et al., the index is based
on the actual content of water (Θ), water capacity and wilting point. Multispectral satellite images
from visible (red band) and infrared bands (near infrared and thermal bands) are essential for the
calculation of the index.
Key Words: Climate change, Soil moisture index (SMI), Remote sensing, Flood, Drought
Introduction
Global climate changes are confirmed by the temperature increase in past 100 years. According to
EPA1 data, today temperature is 1°C increased about the same year in a previous century, whereas the
predictions for the next 100 years are up to 8°C, 3-5°C by the end of the XXI century. Floods,
droughts, forest fires and other natural disasters are more frequent across the globe. The struggle with
the nature of these disasters is comprehensive and requires the involvement of a vast number of
people to repair and rebuild the damage and conduct the adequate protection. Climate change that
causes natural disasters, directly affect the quality of human lives and has a significant impact on
agricultural production. Previous climatologic data that are still in use in agricultural research changed
dramatically, and it is of great significance their revision and update. To facilitate the struggle with
changes in nature and speed up the remediation of consequences, use of modern methods and
techniques is not only desirable, but it is rather necessary. These technologies and methods include
GIS analysis and processing of the data obtained by remote sensing technology (LiDAR, Radar,
Satellite images and others).
In the past six years, the environment in Serbia had suffered great changes when several natural
disasters occurred (forest fires, drought, flooding’s, etc.). Due to the climate change, a significant
number of landslides have occurred, many houses have been destroyed, agricultural yields has been
reduced, the amount of toxic/carcinogenic substances increased in agricultural products. The
sustainable development as planned and as we apply begins to lose its basis, and critical revision is
necessary if we want to be ready for the upcoming future.
According to the Recovery Assessment Report „Serbia Floods 2014“, 24 municipalities were
characterized as the most affected by the floods that occurred during the third week in May 2014,
where the precipitation was more than 200mm per sq m during one week. The damage (total effect)
which emerged after the floods in these municipalities was estimated at €1,525 million, of which 57%
relates to the value of the destroyed properties, while the remaining 43% relates to the losses in the
agriculture production (Serbia Floods, 2014). According to data from the same document, the total
amount of the damage and effects caused by these floods is over €1,700 million for all affected
municipalities in the Republic of Serbia.
In contrast to these floods in 2012, the entire territory of the Republic of Serbia was under the
influence of drought which was characterized as extreme in most of the country. In the period from
June 17th to July 16th, 2012, the average daily temperature ranged from 18 to 26°C, which is 3.5-6°C
higher temperature compared to the average for that time of the year. The highest recorded
temperature was in the city Loznica on August 24th, 2012 with T-max 41.0°C. Precipitation for the
given period was only about 20% of the average precipitation for that period of the year
1
US Environmental Protection Agency
(Meteorological Annual 1, 2013). Since the irrigated areas in the Republic of Serbia are among the
smallest in Europe (Bogdanov, 2012), the consequences of such weather conditions were catastrophic.
Direct drought consequences were visible in 2013 when the level of aflatoxin was increased in corn
and livestock food (type B1(AFB1)) and cow’s milk (type M1 (AFM1)) (Spirić et al., 2014).
According to the unofficial data, the Government of the Republic of Serbia allocated about €150
million to redress the effects of 2012 draught, which altogether with the damage caused by the 2014
floods amounts to approximately €1.9 billion.
Aiming to support sustainable development through the provision of geospatial data and services,
Republic Geodetic Authority (RGA) realized development project between 2010 and 2015 and
implemented a technical infrastructure that comprises flexible, fully scalable and interoperable
components well suited for the needs of public
authorities and business in Serbia.
Through the usage of a comprehensive multi-source
data, and developing of capacities in the remote
sensing application RGA enabled a variety of thematic
maps to be produced: Generic Land Cover, Agriculture
Land Cover, Natural Habitat, Landscape
Fragmentation, Natural Habitat Change, Ecological
Network, Natural Hazard (droughts, floods,
landslides), Biophysical Parameters and other. Using
of national capacities and developing of the Soil
Moisture Index Map of the state on the efficient and
economical way was seen as an excellent opportunity
of helping the public and private sector in the field of
agriculture and water management.
Figure 1 Area of Interest For the area of interest two locations covering 441,87
sq km were chosen (Fig. 1). Area of interest for the drought-
affected year 2012 is located in Vojvodina, covering the area north of the Danube River while the area
of interest for the flood-affected year 2014 is located in the area south of the Sava River. Both areas
were affected by the nature disasters in the corresponding years.
Materials and methods
Applications of the remote sensing method provide geospatial data that enables generation of
adequate information related to floods and droughts when the human settlements and agricultural
areas are directly affected. The method applied in this paper relies on the use of soil moisture index
(SMI) which in its algorithm uses the data obtained from satellite sensors. The index is based on the
actual content of water (Θ), water capacity and wilting point (Hunt et al., 2008). Multispectral satellite
images from visible (red band) and infrared bands (near infrared and thermal bands) are essential for
the calculation of the index. Only a few of today’s satellite platforms carry sensors that can collect the
data from that particular spectrum. Among these sensors, and free of charge to use, are Aster and
Landsat TM, ETM+, and OLI&TIRS.
Two different satellite scenes were downloaded using USGS web application Earth Explorer. Path
and row of the scenes are 186 029 from the Landsat 7 and Landsat 8 satellite. The images are dating
from 30.08.2012 and 09.06.2014. Necessary bands from the scenes for the calculation are Red and
Near Infrared (NIR) for the Normalized Difference Vegetation Index (NDVI) calculation and Thermal
Infrared (TIR) band(s) for the Land Surface Temperature (LST) calculation. The spatial resolution of
the used bands is 30 m for both scenes while the radiometric resolution is 8bit for Landsat 7 data and
16bit for Landsat 8 data and 16 days as a temporal resolution for both satellites. Map projection of the
scenes is UTM Zone 34N with WGS84 datum. Selected areas of interest are cloud free. In the
procedure of Landsat data preprocessing, Dark Object Subtraction (DOS1) atmospheric correction
and radiometric correction were performed using QGIS SCP plugin software.
Soil Moisture Index
The algorithm that applies to the calculation of SMI function is based on the utilization of the NDVI
and LST which are calculated using multispectral satellite imagery for each pixel. There are different
presenting modes for SMI. Michigan Tech Research Institute (2009) presents SMI values from 0 to 1
where higher values close to 1 represent higher estimated soil moisture levels. Hunt et al. (2008)
present SMI continuous data by values from -5 to 5, where Θ=5 accounts for the water capacity of the
field while Θ=-5 is the wilting point. Further, these data can be used to create soil moisture map that
can be used to define necessities for monitoring irrigation or drainage of the land. Parida et al. (2008)
classify SMI values > 0.3 as no drought or favorable soil moisture conditions, whereas SMI values <
0.3 were classified into three categories of drought, namely severe, moderate, and slight drought.
In presented work, continuous SMI data ranged between 0 and 1 best describe the level of soil
moisture. Therefore, MTRI (2009) method used to present mode of SMI.
Soil Moisture Index calculation
Soil moisture index is based on empirical parameterization of the relationship between land surface
temperature (LST) and normalized difference vegetation index (NDVI) and calculated using
following Eq. 1 (Zeng et al. 2004, Parida et al. 2008, Wang et al. 2009):
SMI = (LSTmax – LST) / (LSTmax – LSTmin) (1)
where LSTmax and LSTmin are the maximum and minimum surface temperature for a given NDVI and
LST is Land Surface Temperature, the surface temperature of a pixel for a given NDVI derived using
remotely sensed data.
LSTmax and LSTmin are calculated using following Eq. 2 and 3, respectively (Zhan et al. 2004, Parida et
al. 2008):
LSTmax = a1 * NDVI + b1 (2)
LSTmin = a2 * NDVI + b2 (3)
where a1, a2, b1 and b2 are the empirical parameters obtained by the linear regression (a present slope
and b present intercept) defining both dry and wet (warm and cold) edges of the data (Fig. 2).

Figure 2 Definition of SMI. Scatterplot in LST-NDVI space (Parida et al. 2008)

First step in SMI calculation is the conversion of digital number (DN) to spectral radiance (L –
W/m2/sr/µm) using following Eq. 4 (Lwin 2010):
L = Lmin + (((Lmax – Lmin) / (QCALMAX – QCALMIN)) * (DN – QCALMIN)) (4)
where Lmin and Lmax are spectral radiance calibration constants (Table 1), QCALMAX and
QCALMIN are the highest and lowest quantized calibration pixel values (Table 2) and DN is the
Digital Number.
Table 1. Landsat spectral radiance (Lmin and Lmax) values for thermal bands
Landsat 5 Band 6 Landsat 8 Band 10 and 11
Radiance minimum Radiance maximum Radiance minimum Radiance maximum
1.238 15.303 0.10033 22.00180
Source: NASA (2013), USGS (2015)

Table 2. Quantized calibration pixel (Qcal min and Qcal max) values
Landsat 5 (Band 6) Landsat 8 (Bands 10 and 11)
Radiance minimum Radiance maximum Radiance minimum Radiance maximum
1 255 1 65535
Source: NASA (2013), USGS (2015)

Two inputs must be calculated (LST and NDVI) to be able to calculate LSTmax and LSTmin.
LST (K) is calculated using Landsat 5 and Landsat 8 Thermal bands using Eq. 5 (Weng et al. 2004):
LST = Tb / [1 + (λ * Tb / c2) * ln (ε)] (5)
where: Tb (Eq. 6) is At-Satellite Brightness Temperature, λ is wavelength of emitted radiance,
c2 = 1.4388*10-2 m K and it is presented with Eq. 7 and ε – emissivity (typically 0.95).
Tb = (K2 / (ln(K1 * ε / L + 1))) (6)
where K1 – sensor dependent calibration constant 1 and K2 – sensor dependent calibration constant 2
(Table 3), ε – emissivity (typically 0.95) and L – spectral radiance (Lwin 2010).
c2 = h ∗ c / s (7)
−34
where (Weng et al. 2004) h is Planck’s constant = 6.626∗10 J s, c is velocity of light = 2.998∗108
m/s and s is Boltzmann constant = 1.38 ∗ 10−23 J/K.

Table 3. Landsat 5 and 8 TIRS Thermal constants

Landsat 5 (Band 6) Landsat 8 (Bands 10 / 11)
K1 K2 K1 K2
607.76 1260.56 774.89 / 480.89 1321.08 / 1201.14
Source: Lwin (2010), NASA (2013), USGS (2015)

The ratio of the reflectivity differences for the NIR and the Red band to their sum (NDVI) is
calculated using following Eq. 8 (Rouse et al. 1974):
NDVI = (NIR – Red) / (NIR + Red) (8)
The final step in the data collecting is the determination of empirical parameters by linear regression.
To do so, statistical software was developed being able to process the data for the same pixel from
two raster sets, LST and NDVI, and present the distribution of the data in the scatterplot (Fig. 5).
Linear regression values range from 0 at the “dry edge’’ (limited water availability) to 1 at the ‘‘wet
edge’’. Pixels close to
the dry edge are drier relative to the wet edge which are wetter (maximum
um evapotranspiration -
unlimited water access).. The scatter plot position of a pixel defines its moisture condition. The
parameters are implemented in the Equations 2 and 3.

Results
NDVI (Fig. 3) and LST (Fig. 4) are calculated
calculate as necessary data to acquire SMI calculation.
calculat NDVI
values are within range -11 to 1 where negative values present lack of vegetation or diseased
vegetation, while positive values show healthy and dense vegetation. LST values are the surface
temperature in °C.

Figure 3 NDVI for areas of interest for 2012 and 2014

Figure 4 Land surface temperature (LST) for areas of interest

The final result from Eq. 1 is presented with the values range from 0 to 1, where values near 1 are the
regions with a low amount of vegetation and surface temperature and present a higher level of soil
moisture. The
he values near 0 are the areas with a significant amount of vegetation and surface
temperature and present the low level of soil moisture (Fig. 6).
The calculation gave us following
ing results: LST values for 2012 and 2014 are within 13.4°
13.4 and 39.9°C
and 23.1° to 36.7°C, respectively. The scatterplot is created only with the pixels with positive
p NDVI
values and corresponding LST values. The LST pixels values range from 19.3°° to 39.9°C for 2012
and a full range of 2014 values (Fig. 3, 4 and 5 ).

Figure 5 Scatterplots for corresponding areas of interest

Figure 6 Soil Moisture Index maps

Fig. 6 present tree SMI maps for years 2011, 2012 and 2014. 2011 was the drought-affected
drought year with
mean summer precipitation (Climatic
Climatic characteristics, 2012) and therefore is presented as a contrast to
the natural disaster characterized drought-affected 2012 and flood-affected
affected 2014. As we can see from
the maps in Fig 6,, 2012 has a much larger drought affected area than 2011 and huge water bodies
which have occurred
red after the flooding in 2014. Histograms in Fig. 7 present the SMI data
distribution.

Figure 7 SMI Histograms for selected years

The most SMI values for 2012 are within 0.1 and 0.6 with mean value 0.4 and present dry state of
o the
soil. For 2014, most of the values are within range 0.1-0.2
0.1 with a max value of 0.46 and mean value
0.16. A significant
ignificant part of the 2014 area is covered by thin layer of river sediments which are dry
since the flood occurrence
rence was one month earlier than the footage was taken.. As can be seen from
Fig. 6 and 7, SMI can derive fast data for the affected areas. Higher values of the index present the
higher water/moisture presence and lower values present dry areas.

Conclusion
Obtained results present the distribution of soil moisture. The data is in accordance with the soil
condition from the corresponding periods, but more explicit confirmation of the data punctuality is
necessary to verify the accuracy of the data. Field data collection would be the best option for the
accuracy verification. For the flood-affected
flood affected 2014, SMI has accurately depicted the surface moisture
one month after the flood. As can be seen
seen from Fig. 6 (2014), there are three major flood-affected
flood
areas compared to the 2011 map.
map In the south-eastern part of the map, we can see emerged lake at the
site of the open pit lignite mine that did not exist in the period before the flood. In the northwest part
of the map, we have the similar situation with the emerged large water bodies. The alluvial planes are
presented with higher value of the index (drier areas) due to the river sediments rapid draining.
draining The
SMI values for drought-affected
affected 2012 are also in accordance with the condition for the given period.
As can be seen from the comparison to 2011 map, the entire research area has a lack of moisture in
the soil. This data can be used as an excellent indicator of the soil moisture condition and can assist in
the multi-criteria decision-making (MCDM) strategy for resolving problems with the emerged
environmental stresses.
The advantages of multispectral satellite bands (visible, near and thermal infrared) in soil moisture
analysis are reflected in following:
• Fine spatial resolution,
• Broad coverage,
• Satellite data free of charge,
• Low-Cost methodology,
• Good temporal resolution (USGS/NASA Landsat courtesy, 16 days),
• Ease of access,
• The possibility for the data to be improved with ground/field data.
The limitations are:
• Satellite sensors limitation (Landsat, Aster, IRMSS-2, MTI with spatial resolution up to
120m),
• Land cover dependence,
• Limited surface penetration (reflectance data),
• Cloud contamination dependence (the footage should be cloud-free),
• Verification of the results is tough without field data or complementary data (e.g.,
meteorological stations data).
Nowadays, when the temporal resolution of Landsat satellite images is 16 days during the year
(NASA, 2013), it is possible to provide quality, accurate and up-to-date data. The data can be used for
smart agricultural production planning, but also for spatial and construction planning. In the case of
floods and droughts, data can be used for prevention activities related to them. The advantage of
remote sensing, beside continuously derived data, is in the continuously monitoring of large areas
(Shafian and Maas, 2015). SMI is potentially a very effective indicator of soil moisture with the
application in irrigation scheduling and water resource monitoring.

The main agricultural problem in Serbia is the existence of a large number of small-scale farmers who
are very sensitive to natural hazards. One of the priorities of the European Commission (EC) Rural
Development Programmes (RDP) 2014-2020 (The Common Agricultural Policy after 2013) is
promoting resource efficiency and supporting the shift towards a low-carbon and climate-resilient
economy in the agriculture, food and forestry sectors. The method of this paper is consistent with the
EU's Common Agriculture Policy (CAP) and helps the implementation of its integral parts: Climate-
smart Agriculture (CSA), Climate-change Adaptation (CCA), Climate-change Mitigation (CCM) and
Disaster Risk Reduction (DRR). Three priority action CAP areas to protect and enhance rural heritage
are:

• Biodiversity and the preservation and development of 'natural' farming and forestry systems,
and traditional agricultural landscapes;
• water management and use;
 • dealing with climate change (The Common Agricultural Policy after 2013).

Furthermore, this paper is in accordance with the Draft of the Country Action Programme for Serbia
for the year 2017 regarding European Integration Facility through IPA II (2014-2020) which beside
other, has focus on disaster risk reduction in agriculture. This work is a specific attempt to help, assist
and improve the resilience of agricultural activities to climate change in Serbia. It should help in
development of an early warning system and early response on national and local level; timely
distribution of information which would minimize production loss; in improving prevention; and
help in decision-making.

The information on natural resources and their changes are necessary for managing the policy of
sustainable development. According to the UN 2030, Agenda for Sustainable Development, among 17
Sustainable Development Goals (SDGs) are: reduce poverty, zero hunger and life on land. The
significance of geospatial information in the achievement of these goals is doubtless. In the future
period the course of RGA capacity improvement will not be changed, and dealing with contemporary
remote sensing methods will continue. Besides that, the Republic Geodetic Authority of Serbia will
continue to provide up to date geospatial data and consultancy to all interested parties, both
governmental and local institutions as well as regional and to cooperate with respectable organizations
from all over the world.

References:
Bogdanov N. (2012). Poljoprivreda u uslovima suše: razmere štete, očekivane posledice i izazovi.
Makroekonomske analize i trendovi 213-214, 19 – 22 [in Serbian].
Climate characteristics, Significant climate events in Serbia during 2011. (2012). Republic
Hydrometeorological Service of Serbia.
EPA - US Environmental Protection Agency. www3.epa.gov [last accessed October 7th, 2016].
Hunt E.D., Hubbard K.G., Wilhite D.A., Arkebauer T.J. and Dutcher A.L. (2008). The development
and evaluation of a soil moisture index. Int. J. Climatol. 29: 747 – 759. doi: 10.1002/joc.1749.
Landsat 8 (2016). http://landsat.usgs.gov/landsat8.php [last accessed October 7th, 2016].
Lwin K. K. (2010). Estimation of Landsat TM Surface Temperature Using ERDAS Imagine Spatial
Modeler. SIS Tutorial Series, Division of Spatial Information Science.
Meteorological Annual 1, (2013). Climatological data 2012. Republic Hydrometeorological Service
of Serbia.
Michigan Tech Research Institute (2009). Inputs to the Multi-Criteria Decision Model. National
Consortium for Remote Sensing in Transportation.
NASA (2013). Landsat 7 Science Data Users Handbook. Available at
[landsathandbook.gsfc.nasa.gov].
Parida B. R., Collado W. B., Borah R., Hazarika M. K., and Samarakoon L. (2008). Detecting
Drought-Prone Areas of Rice Agriculture Using a MODIS-Derived Soil Moisture Index. GIScience &
Remote Sensing, 45, No. 1, p. 109 – 129. doi: 10.2747/1548-1603.45.1.109.
Rouse J. W., Haas R. H., Schell J. A. and Deering D. W. (1974). Monitoring vegetation systems in the
Great Plains with ERTS. Third ERTS-1 Symposium, 10–14 Dec. 1973, NASA SP-351, Washington
D.C. NASA, pp. 309–317.
Serbia Floods 2014. Recovery Assessment Report (2014). United Nations Serbia, EU Delegation in
Serbia, World Bank Group and Government of the Republic of Serbia.
Shafian S. and Maas S. (2015). Index of Soil Moisture Using Raw Landsat Image Digital Count Data
in Texas High Plains. Remote Sens., 7, 2352 – 2372; doi:10.3390/rs70302352.
Spirić D.M., Stefanovic S.M., Radicevic T.M., Djinovic Stojanovic J.M., Jankovic V.V., Velebit B.M.
and Jankovic S.D. (2014). Study of Aflatoxins Incidence in Cow Feed and Milk in Serbia During
2013. Hem. ind. 69 (6), 651 – 656 [in Serbian]. doi: 10.2298/HEMIND140619087S.
Transforming our world: the 2030 Agenda for Sustainable Development (2015). The General
Assembly of the UN.
USGS (2015). Landsat 8 (L8) Data Users Handbook. Version 1.0. Department of the Interior, U.S.
Geological Survey.
Wang L. and Qu J. (2009). Satellite remote sensing applications for surface soil moisture monitoring:
A review. Front. Earth Sci. China 3: 237 – 247. doi:10.1007/s11707-009-0023-7.
Weng Q., Lu D. and Schubring J. (2004). Estimation of land surface temperature–vegetation
abundance relationship for urban heat island studies. Remote Sensing of Environment, Elsevier
Science Inc., Box 882 New York NY 10159 USA, 89, 467 – 483.
Zeng Y., Feng Z. and Xiang N. (2004). Assessment of soil moisture using Landsat ETM+
temperature/vegetation index in semiarid environment. Geoscience and Remote Sensing Symposium,
IGARSS '04. Proceedings. 2004 IEEE International, Volume: 6, 4306 – 4309. doi:
10.1109/IGARSS.2004.1370089.
Zhan Z., Qin Q. and Wang X. (2004). The Application of LST/NDVI Index for Monitoring Land
Surface Moisture in Semiarid Area. Geoscience and Remote Sensing, IGARSS ‘04. Proceedings 2004
IEEE International, 3:1551 – 1555.
-, The Common Agricultural Policy after 2013. Available at: [ec.europa.eu/agriculture/cap-post-
2013_en].