The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLVIII-1/W2-2023

ISPRS Geospatial Week 2023, 2–7 September 2023, Cairo, Egypt

SATELLITE REMOTE SENSING FOR ASSESSING THE SPATIOTEMPORAL

CHANGES OF THE ECOLOGICAL STATE OF THE AGRICULTURAL LANDS IN
ARMENIA

Grigor Ayvazyan and Shushanik Asmaryan

Centre for Ecological-Noosphere Studies, National Academy of Sciences, Abovyan Street 68,
Yerevan 0025, Armenia, ([Link], [Link])@[Link]

KEYWORDS: Remote sensing, LANDSAT, Google Earth Engine, Grassland, Pasture, NDVI, Mann-Kendall test.

ABSTRACT

Proper management and monitoring of the ecological condition of natural feeders are crucial for both the economy and natural
environments. Armenia is no exception, and addressing this issue requires comprehensive studies combining traditional methods with
remote sensing technologies. Remote sensing in agriculture utilizes various sensors and technologies to collect data about crops, soil
conditions, and agricultural parameters from a distance. This study focuses on the utilization of remote sensing, including satellite
imagery, to assess the ecological state of grasslands and pastures in the Syunik administrative district of Armenia. Over a 22-year
period (2000-2021), the study utilized Landsat series (5,7,8) multispectral optical images processed through Google Earth Engine
(GEE) to analyze changes in the Normalized Difference Vegetation Index (NDVI) as an indicator of vegetation health. The results
reveal favorable trends in vegetation health across the study regions, as indicated by consistently positive Z-values in the Seasonal
Mann-Kendall test. While no significant trends were observed in some areas, others showed improvements in biomass conditions. This
study provides valuable insights into the resilience and health of grasslands and pastures in the Sisian and Goris regions, offering a
basis for targeted interventions and strategies to preserve these ecosystems for future generations.

1. INTRODUCTION The infrared range closest to the visible is near-infrared (NIR).

Both visible and infrared ranges are used for remote sensing of
Unplanned, irregular use of grasslands and pastures without agricultural land (Roy et al., 2016). The continuous
observing the basic norms of grazing leads to thinning of development of remote sensing methods and geographic
vegetation, disturbance of interrelation between vegetation information systems (GIS) made them important tools that
elements, and reduction of pasture productivity, which give an opportunity to observe the spatio-temporal changes for
afterward to degradation (Tovmasyan, 2019). Thus, studying, large territory, assess paces of change, and arrange appropriate
monitoring and proper management of the ecological state of managing (assessing the spatial-temporal changes) of
the mountainous grasslands is becoming imperative because agricultural lands. In the context of remote sensing, only visual
of the threat of degradation. Today, pastures and grasslands are processing of satellite images and mapping was carried out for
under threat of degradation all over the world, which leads to pastures and grasslands. Also, for the first time, Garegin
disturbance of natural balance, a decrease of soil productivity, Tepanosyan developed and evaluated the method of assessing
and consequently to decrease of plant and animal productivity. the degradation of grasslands through spectral analysis of
Therefore, it should be noted that the study, monitoring, and satellite images (Tepanosyan, 2019). This study is the first
correct management of the ecological condition of pastures attempt to assess the spatiotemporal changes of the ecological
and grasslands are in the interests of both the economy and state of the biomass of grasslands and pastures of some rural
natural ecosystems (Tovmasyan, 2019). It is one of the key communities of the Syunik administrative district (marz) of the
issues in Armenia as well and it requires large-scale studies RA based on satellite remote sensing data and technologies.
combining traditional methods supplemented by remote
sensing (RS) technologies (Muradyan et al., 2022). 2. MATERIALS AND METHODS
Remote sensing in agriculture refers to the use of various
sensors and technologies to collect data from a distance about 2.1 Study area
crops, soil conditions, and other agricultural parameters. These
data are crucial for making informed decisions and optimizing The Republic of Armenia is a mountainous country. It covers
agricultural practices. Some common remote sensing an area of 29.8 thousand square kilometers and is located at an
techniques used in agriculture include satellite imagery, aerial average height of 1700 m above sea level. 9.9% of the territory
photography, drones, and ground-based sensors. Remote of the republic lies at an altitude of up to 1000m above sea
sensing is used to identify stressed areas in pastures by level, 76.6% at an altitude of 1000-2500m, and 13.5% at an
capturing the spectral signatures of healthy plants that differ altitude of more than 2500m (Tovmasyan, 2019). Such
from those of damaged and stressed plants (Metternicht et al., variations in height complicate the development and
2010). The wavelengths of the electromagnetic spectrum used management of areas, especially agricultural lands. According
for remote sensing in agriculture cover only a small part of the to the data of the Real Estate Cadastre Committee under the
spectrum. The visible range of the electromagnetic spectrum RA Government (Tepanosyan, 2019), more than 68% of the
is from about 400 nm to about 700 nm. The color green, republic's land fund is agricultural land, more than half of
associated with plant health/activity, has a wavelength which, 57%, are natural fodder, pastures, and grasslands,
centered around 500 nm. Wavelengths in the infrared (IR) which have a great environmental and socio-economic
range are longer (up to about 25 µm) than in the visible range. importance. The study area includes the pastures and

This contribution has been peer-reviewed.

[Link] | © Author(s) 2023. CC BY 4.0 License. 1325
 The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLVIII-1/W2-2023
ISPRS Geospatial Week 2023, 2–7 September 2023, Cairo, Egypt

grasslands of Tsghuk, Spandaryan, Sarnakunk, and Akner,

Khnatsakh, Khndzoresk, Khoznavar, Vaghatur, Verishen,
Tatev, Harzhis rural settlements accordingly in Sisian and
Goris administrative regions of Syunik district in 2000-2021,
the most mountainous district of RA. These regions are located
in the south of Armenia and are characterized by dry
continental climate conditions (Tovmasyan, 2019).

2.2 Data

According to G. Metternich found that the Landsat series of

multispectral optical images (MSS, TM, ETM +, OLI) are the
most common and used data source; and 49% of articles report
their utilization (Metternicht et al., 2010). Time series of
LANDSAT images for 2000-2021 were processed via Google
Earth Engine (GEE) open-source data cube. LANDSAT is the
most used database and is the largest component of the GEE
data portal. A harmonization of LANDSAT 5, LANDSAT 7,
and LANDSAT 8 satellite data was implemented and a series
of fragments with less than 30% cloud cover were selected for
processing. The harmonization process was carried out by
applying surface reflectance sensor transformation functions
(ETM+ to OLI and OLI to ETM +) and coefficients proposed
by Roy. This enables the transformation of ETM+ surface
reflectance to OLI surface reflectance or vice versa (Roy et al.,
2016). In total, 1087 fragments were processed via calculating,
during which we encountered a problem related to some
distorted fragments. Since June 2003, the sensor has acquired
and delivered data with gaps caused by the Scan Line a)
Corrector (SLC) failure. However, this fact does not force
ejecting the distorted fragments, since it may result in missing
valuable data. Various vegetation indices are used to
determine the viability of agricultural sites, especially pastures
and grasslands. The most common and used index is the
Normalized Difference Vegetation Index (NDVI): a
standardized way to measure the health of vegetation
(biomass) NDVI imagery allows farmers to analyze vegetation
conditions for all seasons from NDVI maps, curves, and
values, as well as predict productivity and potential yield.
NDVI shows the vigor level of the crop and it is calculated as
the ratio between the difference and the sum of the refracted
radiations in the near-infrared and the red, that is as (Roy et
al., 2016) ,
𝑁𝐼𝑅−𝑅𝐸𝐷
𝑁𝐷𝑉𝐼 = (1)
𝑁𝐼𝑅+𝑅𝐸𝐷

where NIR = Near-infrared

RED = Red

2.3 Image processing and testing

Time series of NDVI values were calculated in the GEE

platform using JavaScript scripting, for the April-October
season of the 2000-2021 time period and a set of maps for
2000, 2005, 2010, 2015, and 2021 years (Figure 1, Figure 2)
were produced via ArcGIS 10.6 which we will show the state
of the biomass for each fifth year in the study period. The b)
NDVI values were classified according to the scheme
presented in Table 1. (“Agricolus,” n.d.)

This contribution has been peer-reviewed.

[Link] | © Author(s) 2023. CC BY 4.0 License. 1326
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLVIII-1/W2-2023
ISPRS Geospatial Week 2023, 2–7 September 2023, Cairo, Egypt

c) e)

Figure1. The NDVI values of the pastures and grassland of

the rural communities (Tsghuk, Sarnakunk, Spandaryan) of
Sisian region: a) 2000, b) 2005, c) 2010, d) 2015, e) 2021.

a)

d)

This contribution has been peer-reviewed.

[Link] | © Author(s) 2023. CC BY 4.0 License. 1327
 The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLVIII-1/W2-2023
ISPRS Geospatial Week 2023, 2–7 September 2023, Cairo, Egypt

b) e)

Figure2. The NDVI values of the pastures and grassland of

the rural communities (Tatev, Harzhis, Akner, Verishen,
Vaghatur, Khnatsakh, Khoznavar, Khndzoresk) of Goris
region: a) 2000, b) 2005, c) 2010, d) 2015, e) 2021.

The NDVI values were classified according to the scheme

presented in Table 1. (“Agricolus,” n.d.)

NDVI INTERPRETATION
< 0.1 Bare soil
0.1 – 0.2 Almost absent canopy cover
0.2 – 0.3 Very low canopy cover
Low canopy cover, low vigour or very low
0.3 – 0.4
canopy cover, high vigour
Mid-low canopy cover, low vigour or low
0.4 – 0.5
canopy cover, high vigour
c) Average canopy cover, low vigour or mid-
0.5 – 0.6
low canopy cover, high vigour
Mid-high canopy cover, low vigour or
0.6 – 0.7
average canopy cover, high vigour
0.7 – 0.8 High canopy cover, high vigour
0.8 – 0.9 Very high canopy cover, very high vigour
0.9 – 1.0 Total canopy cover, very high vigour

Table 1. Classification of NDVI values of agricultural plots

Afterward, trend analysis was performed on the Jupiter

Netbook platform via the Seasonal Mann-Kendall statistical
test (Mavromatis and Stathis, 2011), using Python coding. The
Mann-Kendall calibration method was chosen, which is
widely used for the analysis of trends in climatic and
hydrological time series (Shadmani et al., 2012). There are two
advantages to using this test, it is a non-parametric test and
d)
does not require a normal distribution of the data, and the test
has low sensitivity to sharp time discontinuities due to
inhomogeneous time series (Shadmani et al., 2012). Moreover,
it allows us to take seasonality into account as well. Any data
reported as non-detects are included by assigning them a total
value that is less than the smallest measured value in the data
set (Mavromatis and Stathis, 2011). Under this test, the null
hypothesis H0 assumes that there is no trend and this is tested
against the alternative hypothesis H, which assumes that there
is a trend. A null hypothesis was proposed assuming that there
is no trend between variables. T, H, P, α (0.05), and Z are
descriptive parameters, which show the state of, alternative
trend, value of the significance, significance level, and
standardized statistic test accordingly. The positive Z indicates

This contribution has been peer-reviewed.

[Link] | © Author(s) 2023. CC BY 4.0 License. 1328
 The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLVIII-1/W2-2023
ISPRS Geospatial Week 2023, 2–7 September 2023, Cairo, Egypt

an increasing trend in the time series, while the negative Z is a Tren

decreasing trend (Buhairi, 2010). H P Z α
d
Pastur 3.06 0.0
3. RESULTS Yes True 0.002
e 0 5
Tatev
The research reveals crucial insights into the state of biomass Grassl 2.24 0.0
Yes True 0.024
in the grasslands and pastures of the Sisian and Goris regions and 4 5
over 22 years, from 2000 to 2021. By utilizing the NDVI Harz Grassl 2.45 0.0
Yes True 0.014
values as a key indicator, the study sheds light on the overall his and 1 5
health and trends of vegetation in these areas with high Pastur 0.000 3.63 0.0
agricultural importance. Figure 1 shows that according to the Yes True
Akne e 2 5 5
NDVI values, the state of the biomass in grasslands and
pastures of the rural communities of the Sisian region in 2010 r Grassl 2.73 0.0
Yes True 0.006
and 2015 is much better than in 2000, 2005, and 2021. Figure and 9 5
2 demonstrates that in the Goris region, the situation is a little Pastur 2.54 0.0
bit different. The state of the biomass is better in 2005, 2010, Yes True 0.010
Veris e 9 5
and 2015 than in 2000 and 2021. While, the time series hen
analysis through the Seasonal Mann-Kendal test shows that no Grassl 2.08 0.0
Yes True 0.037
trend is observed neither in grasslands nor in pastures of the and 2 5
Sisian region (Tsghuk, Sarnakunk, and Spandaryan). So, the Pastur 1.48 0.0
No False 0.138
null hypothesis is confirmed, due to P and Z values (Table 2): Vagh e 3 5
P > α, and |Z| > 1,96 (critical value of Z, if α=0,05) (Shadmani atur Grassl 1.51 0.0
et al., 2012)․ No False 0.130
and 2 5
Pastur 3.03 0.0
Tre Yes True 0.002
H P Z α Khna e 5 5
nd
tsakh Grassl 1.42 0.0
Fals 0.23 1.17 No False 0.154
Pasture No 0.05 and 2 5
e 8 9
Tsghuk Pastur 1.74 0.0
Grasslan Fals 0.39 0.85 No False 0.080
No 0.05 Khoz e 5 5
d e 2 4
navar Grassl 1.79 0.0
Fals 0.20 1.26 No False 0.072
Pasture No 0.05 and 4 5
Sarnak e 6 2
unk Grasslan Fals 0.19 1.30 Pastur 3.21 0.0
No 0.05 Khnd Yes True 0.001
d e 3 0 e 0 5
zores
Fals 0.13 1.50 k Grassl 2.96 0.0
Pasture No 0.05 Yes True 0.002
Spanda e 1 7 and 8 5
ryan Grasslan Fals 0.05 1.89
No 0.05 Table3. The descriptives of the Seasonal Mann-Kendall test
d e 7 7
of the rural communities (Tatev, Harzhis, Akner, Verishen,
Table2. The descriptives of the Seasonal Mann-Kendall test Vaghatur, Khnatsakh, Khoznavar, Khndzoresk) of the Goris
of the rural communities (Tsghuk, Sarnakunk, Spandaryan) of region.
the Sisian region.
It is significant that, for the Sisian and Goris regions, the
According to the Seasonal Mann-Kendal statistical test in the calculated Z-values were found to be positive in all cases,
Goris region, no trend is observed for both pastures and indicating a consistently favorable state of vegetation over the
entire 22-year time series. This uniformity in positive values
grasslands of Vaghatur and Khoznavar rural communities and
across both regions suggests that, on the whole, the vegetation
for grasslands of Khnatsakh rural community where P > α, and has been thriving and resilient over the long period, despite
|Z| > 1,96. However, a positive trend is observed for both any localized fluctuations in biomass health.
pastures and grasslands of Tatev, Harzhis, Akner, Verishen,
and Khndzoresk communities, and for the grasslands of 4. CONCLUSIONS
Khnatsakh village according to the P and Z values:P < α and
|Z| > 1,96 (Table 3). Seasonal Mann-Kendall statistical testing enables to
validatation of the results of the time series analysis to identify
the trend and trend behavior of spatio-temporal changes. No
trend of NDVI values in the time series 2000-2021 was
observed in pastures and grasslands of Tsguk, Sarnakunk, and
Spandaryan rural communities of the Sisian region, and hence
no change in biomass condition was observed. While in the
pastures of the Khnatsakh community of Goris region and in
pastures of Vaghatur and Khoznavar no change in the biomass
condition was observed, while in the grasslands of Khnatsakh
village and in both grasslands and pastures of Akner,
Khndzoresk, Verishen, Tatev, Harzhis rural communities an
improvement in the biomass condition was noted. So, no

This contribution has been peer-reviewed.

[Link] | © Author(s) 2023. CC BY 4.0 License. 1329
 The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLVIII-1/W2-2023
ISPRS Geospatial Week 2023, 2–7 September 2023, Cairo, Egypt

degradation of the biomass condition was observed in the REFERENCES

studied clusters of the rural communities in the Goris and
Sisian regions, but improvement was observed only for those Agricolus [WWW Document], n.d. URL
of the Goris region. By recognizing the positive trends and [Link]
resilient state of vegetation, policymakers and local ndvi-ndmi/
communities can work together to develop strategies that Buhairi, M.H. Al, 2010. Analysis of Monthly, Seasonal, and
protect and enhance the vital grasslands and pastures in Sisian Annual Air Temperature Variability and Trends in Taiz
and Goris, ensuring the preservation of these valuable City - Republic of Yemen. J. Environ. Prot. (Irvine,
ecosystems for future generations. The identification of Calif). 01, 401–409.
specific communities with positive trends in vegetation health [Link]
can serve as a basis for implementing targeted interventions to Mavromatis, T., Stathis, D., 2011. Response of the water
promote further improvements. balance in Greece to temperature and precipitation
trends. Theor. Appl. Climatol. 104, 13–24.
ACKNOWLEDGMENT [Link]
Metternicht, G., Zinck, J.A., Blanco, P.D., del Valle, H.F.,
The work was supported by the Science Committee of RA, in 2010. Remote Sensing of Land Degradation:
the frames of the research project № 22AA-1E021. The Experiences from Latin America and the Caribbean. J.
authors also thank the Center for Ecological-Noosphere Environ. Qual. 39, 42–61.
Studies NAS RA for dedicated support during the research. [Link]
Muradyan, V., Asmaryan, S., Ayvazyan, G., Dell’Acqua, F.,
2022. Multidecadal Trend Analysis of Armenian
Mountainous Grassland and Its Relationship to Climate
Change Using Multi-Sensor NDVI Time-Series.
Geosci. 12.
[Link]
Roy, D.P., Kovalskyy, V., Zhang, H.K., Vermote, E.F., Yan,
L., Kumar, S.S., Egorov, A., 2016. Characterization of
Landsat-7 to Landsat-8 reflective wavelength and
normalized difference vegetation index continuity.
Remote Sens. Environ.
[Link]
Shadmani, M., Marofi, S., Roknian, M., 2012. Trend Analysis
in Reference Evapotranspiration Using Mann-Kendall
and Spearman’s Rho Tests in Arid Regions of Iran.
Water Resour. Manag. 26, 211–224.
[Link]
Tepanosyan, G., 2019. Applicability of linear spectral
unmixing in delineating potential erosion areas in
highland pastures ( a case study of Nerkin Sasoonashen
rural community ). Agroscience, 2016, No. 1-2 44–49.
Tovmasyan Gagik, 2019. Guidelines for improving degraded
natural forage lands (pastures and hayfields). (In
Armenian)

This contribution has been peer-reviewed.

[Link] | © Author(s) 2023. CC BY 4.0 License. 1330