Abstract.

The normalized difference vegetation index (NDVI) is an essential classifica-

tion method to identify the changes in land use/land cover (LULC) area and planning
for sustainable services. Machine learning and geospatial techniques are the most effec-
tive significant tools for change detection of LULC. The present study was conducted
to assess land use /land cover changes with the help of NDVI classification using Ma -
chine learning and geospatial techniques. Landsat 5 TM and 8 OLI/TIRS images are
used from 2000 to 2020 (20 years) to extract NDVI values. The NDVI values are classi-
fied into five categories, and two NDVI maps, 2000 and 2020, are generated. Five
LULC classes are identified: Water (Deep and Shallow), Built-up/ River Sand,
Fallow/Wasteland, Agricultural Land/ Crop Land, and Dense vegetation. The present
study shows that the area water areas decreased from 4.1% in 2000 to 1.9% in 2020,
and Built-up/ River sand also decreased from 8.1 % to 1.7% from 2000 to 2020, respec-
tively. The dense vegetation area was also found at 11.8% in 2020, and Agroforestry/
Sparse Vegetation areas increased from 2.1 % to 34.7% in the last 20 years.

Keywords: Machine Learning, Geospatial Techniques, NDVI; Multi-spectral

Images and Sensor, Rupnarayan River

1 Introduction

Worldwide our natural surroundings are very precious. It changes and is damaged by
nature and human activities. Our environment is precious, and it should be conserved
for future generations. Natural events such as cyclones, droughts, floods, and changes
in the Earth's surface have also affected the climate (Seneviratne et al., 2012). Several
studies (USGCRP, 2017; Friedlingstein et al., 2019) reveal that human activities are
causing global warming – a rise in average temperatures around the world over this
century. The finding of changes in the landscape of the Earth or analytical analysis
helps to gain an understanding and help to take the precaution timely (Yunfeng Hu et
al., 2018). Changing rate of LULC plays a significant role in studying global change.
Different techniques have been approached for LULC mapping, and changing pattern
has been developed throughout the globe in the last few decades (Abebe et al., 2021;
Arulbalaji et al., 2019; Nath et al., 2020, 2021). Change dynamic of Land use and
Land cover and anthropogenic activities have large effects on climate change, an in-
crease of natural hazards (flooding, drought, tropical cyclone), biodiversity loss, and
deforestation, respectively (Hassan et al., 2016; Dwivedi et al., 2005).
Presently, the potential earth observation data to extract valuable information for land
use and land covers by worldwide acceptance (Mishra et al., 2020). Earth observation
systems and Geospatial techniques are suitable for assessing land use and land cover
change analysis over large areas (Alam et al., 2020; Mishra et al., 2020; Nath et al.,
2021). It has been used for remote sensing data's spatial and spectral resolution; it
works at the micro level (Liang & Wang, 2020; Nath et al., 2022). Several methods
such as object-based classification (King et al., 2011, Koley et al., 2022), comparison
of spectral indices and principal component based (Yanan et al., 2011), cross-correla-
2

tion analysis (Jones et al., 2009), image fusion-based, post-classification comparison

(Hassan et al., 2016; Koley et al., 2022) have been used for the monitoring and
change dynamic of LULC by geospatial techniques and earth observation data (Abebe
et al., 2021).
This paper represented a variation of the vegetation index of the Rupnarayan river
basin using geospatial and machine learning techniques. Various previous researchers
have used these techniques on different river basins. This paper aims to show the spa-
tiotemporal changes in land use and land cover of the Rupnarayan river basin from
2000 to 2020. Landsat 8 provides the data over the period regarding spectral and spa-
tial resolution. This paper's prime objective is identifying and analyzing Land use
Land cover vegetation through the NDVI Indicator of the Rupnarayan basin. This
study can be significant for flood hazards and coastal management in the coastal envi-
ronment. Previous research has been done on various river basin areas of LULC
change using geospatial and machine learning techniques.

2 Study Area

The study area of this paper has been chosen as the Rupnarayan River Basin, covers
around 10797 sq. km. The Rupnarayan River Basin begins in the Chhota Nagpur
plateau foothill at Dhaleswari, near Purulia. It flows south-easterly and passes the
town of Bankura, where it is known as the Dwarakeswar River. The Shilabati River
merges it near Bandal, which takes the name Rupnarayan and is joined by the
Hooghly River at Geonkhal. The Rupnarayan River basin area experiences a typical
tropical monsoonal climate characterized (Maity & Maity, 2018). The gradient is al-
most gentle to flat in the lower basin area, composed of a lower alluvial plain and
deltaic floodplain and elevation of 10 m (Mukhopadhyay & Dasgupta, 2010). The an-
gle of the Shilabati River concerning the Rupnarayan River is 230 degrees (Das &
Bandyopadhyay, 2015). Geographically, the Rupnarayan River is a sensitive area for
environmental features; primarily Rupnarayan River flows three types of topography
Plain, Rarh, and Deltaic. Rupnarayan River flows through an area of densely popu-
lated and agricultural land; a minimal area is occupied by vegetation.
 3

Figure 1 Location of the Rupnarayan River Basin (RRB)

3 Materials and Methods

3.1 Data Collection

Earth observation data was derived from the United States Geological Survey (ht-
tps://earthexplorer.usgs.gov/) for the present study. Landsat 5 TM (30 m) data
was used for the analysis of NDVI of the year 2000, and Landsat 8/9 OLI/TIRS (30
m) was used for the analysis of NDVI of the year 2020. UTM projection and WGS84
datum were used for geo-referencing the earth observation data. Detailed information
on the data source is listed in table 1.
Table 1 Data source and used for NDVI analysis of the study area
Date Ac-
Data Path/Row Resolution Source
quired
2020-05-15 139/044 30 m
Landsat 8 -9 OLI/
15 m USGS
TIRS 2020-04-06 138/045
(PAN)
2000-01-26 138/045
Landsat 5 TM 30 m USGS
2000-09-29 139/044
4

3.2 Applied Methodology

A statistical technique has been implemented to analyze spatial and temporal changes
in vegetation geo. Satellite images do land cover changes analyses by utilizing classi-
fication methods. Landsat 8 OLI/TIRS and Landsat 5 TM images were collected from
earth explore USGS. Satellite images were collected of the Rupnarayan river basin
and combined using a mosaic method in ArcGIS. These images resolutions are 30m
images were used per pixel and classified into six areas based on digital number (DN)
values of landscape elements. The last 20 years' map was prepared by Arc-GIS 10.8
platform. NDVI classification methods did LULC classification. The present study
has considered six classes: water, built-up, wasteland, agricultural land, spare, and
dense vegetation.

Fig. 2 Flowchart of the Methodology of the study area

Data Processing and Analysis

Since 1970 satellite data has been used to monitor land cover and changes (Lillesand,
2004). Various Multi-spectral images were derived from the study area's open source
form USGS in 2000 and 2020. After collecting all data, the study area was clipped us-
ing a vector image (.shapefile). From the satellite images, Band 5 and Band 4 of
Landsat 8/9 OLI/TIRS (2020) and Band 4 and Band 3 of Landsat 5 TM (2000) were
used to understand the changes in land cover. After the extraction of suitable study ar-
eas from the selected bands, machine learning techniques were used to prepare NDVI
maps and analyze the overall changes of LULC. The present study used the multi-
spectral images of the Rupnarayan river basin to calculate the Normalized Difference
Vegetation Index values. After processing of the Landsat 8 OLI (bands 5, 4) and
Landsat 5 TM (bands 4, 3) were used for calculation of NDVI of 2000 and 2020 using
Remote sensing and GIS techniques which detect the land use and land cover changes
area. NDVI has been calculated based on the at-sensor spectral radiance of the red
 5

band 4 and infrared band 5 of Landsat 8 OLI image and the red band 3 and infrared
band 4 Landsat 5 TM images in 2000 and 2020. Theoretically, NDVI values vary be -
tween -1.0 to +1.0. NDVI represents the state of vegetation that can be implemented
with the help of a GIS tool. (Pande et al., 2021). NDVI is calculated as a ratio differ -
ence between measured canopy reflectance in the red and near-infrared bands. This
paper shows the differences between the visible red and near-infrared (NIR) bands of
Landsat 8 OLI and Landsat 5 TM images that have been used to identify areas con-
taining significant changes in the vegetation cover and other various land cover fea-
tures. The NDVI maps were calculated between 2000 and 2020 using raster calculator
spatial analysis tools. Based on the spatial ratio among both bands that are incredibly
delicate to the green biomass (Pande et al., 2021), it is demarcated as:

NIR – RED
NDVI =
NIR+ RER
Where,
‘NIR’ band indicates the near-infrared band of Landsat images.
‘RED’ band indicates visible red reflection.
The NDVI map during the years 2000 and 2020 was prepared using the Raster calcu -
lator of the spatial analysis tool. During the calculation of the NDVI values, the out-
put map was shown on a grayscale raster map with index ranges from -1 to + 1. The
thematic color was changed to NDVI per the present study's requirement. The
methodology flowchart has shown the entire step to analyze the maps of the Rup-
narayan river Basin.
NDVI is used for the analysis of vegetation that reflects in satellite bands. It detects
the land use and land cover changes in vegetation area. In analysis, it used spatial and
spectral resolution of the remote sensing data, enabling it to work at the micro level.
The main problem of this research is river-related. We have used the multi-spectral
images of the Rupnarayan river basin to calculate the NDVI values. NDVI is based on
satellite images' red band 4 and infrared band (band 5). The value of NDVI varies be -
tween -1.0 to +1.0. NDVI represents the state of vegetation that can be implemented
with the help of a GIS tool (Pande, 2021). NDVI is calculated as a ratio difference be-
tween measured canopy reflectance in the red and near-infrared bands. This paper
shows that the differences between a Landsat image's visible red and near-infrared
(NIR) bands can be used to identify areas containing significant vegetation and other
features. NDVI maps during the years 2000 and 2020. The map was designed using
the Raster calculator of the spatial analysis tool. NDVI values were calculated using
the ratio between the bands.
For Landsat 8 NDVI is calculated
NDVI= (Band 5 – Band 4) / (Band 5 + Band 4)
For Landsat 5 NDVI is calculated
NDVI = (Band 4 – Band 3) / (Band 4 + Band 3)
After analysis, a raster map was built with an index ranging from -1 to 1. As per re -
quirement, we select the color of each band. A flow chart has shown the entire step
performed to analyze the difference map of the Rupnarayan river Basin.
6

4 Result and Discussion

Agriculture is the major resource in the study area. For analysis, NDVI techniques
have been used using geospatial techniques. Satellite images have been applied for the
20 years of this present study. LULC classification was done through NDVI and clas-
sified into six categories in the study area. Generally, six NDVI classes have been
classified for the present study's analysis. In NDVI 2020, it has been observed that the
range of the NDVI value is -0.204 to 0.591. It has been classified into six LULC
classes Water (-0.204 to - 0.01), Built-up/ River Sand (-0.011 to 0.15), Fallow/Waste-
land (0.151 to 0.25), Agricultural Land/ Crop Land (0.251 to 0.35), Agroforestry/
Sparse Vegetation (0.351 to 0.45), Dense Vegetation (0.451 to 0.591). From NDVI
2000, it has been observed that the range of the NDVI value is -0.371 to 0.474. It is
also classified into six LULC classes Water (-0.371 to 0.0), Built-up/ River Sand
(0.011 to 0.15), Fallow/Wasteland (0.151 to 0.25), Agricultural Land/ Crop Land
(0.251 to 0.35), Agroforestry/Sparse Vegetation (0.351 to 0.45), Dense Vegetation
(0.451 to 0.474). From this LULC analysis, NDVI of 2000 to 2020 reveals significant
changes in the present study. The LULC map shows that the maximum area occupied
by Agricultural Land/ Crop Land (60.7%) in the year 2000; however, it reduced
rapidly from 60.7 % (2000) to 36.5 % (2020). The growth of Agroforestry/Sparse
Vegetation has rapidly increased from 2.1% to 34.7% between the years 2000 to
2020. The area of water has been reduced from 4.1 % to 1.9 % between the years
2000 to 2020.

Table 3 Area statistics for various LULC features of the study area.
NDVI Area in Area LULC Class NDVI Area in Are
Range sq. km (%) Range sq. km a
(2000) (2020) (%)
-0.371 to - 440.20 4.1 Water (Deep & Shallow) -0.204 to - 204.83 1.9
0.01 0.01
- 0.011 to 872.46 8.1 Built-up/ River Sand -0.011 to 178.39 1.7
0.15 0.15
0.151 to 0.25 2712.3 25.1 Fallow/Wasteland 0.151 to 0.25 1451.5 13.4
0.251 to 0.35 6549.46 60.7 Agricultural Land/ Crop 0.251 to 0.35 3941.55 36.5
Land
0.351 to 0.45 222.48 2.1 Agroforestry/Sparse Ve- 0.351 to 0.45 3748.93 34.7
getation
0.451 to 0.14 0.0 Dense Vegetation 0.451 to 1271.84 11.8
0.474 0.591
10797.04 100.0 Total 10797.04 100.
0

The figure portrays the total changes in percentage. It shows significant changes that
happen from the year 2000 to 2020. Shallow water level increases from 2 to 35%. It
also shows that dense vegetation parts increase from 0% to 12 %. Soil area reduces by
8 to 2%. Agriculture areas also reduce 61 to 36% and Built-up area also reduces 8.1 to
1.7 %.
 7

70.0 60.7
60.0
Area in (%)

50.0
40.0 36.5 34.7
30.0 25.1
20.0 13.4 11.8
8.1
10.0 4.1 1.9 1.7 2.1 0.0
0.0

Agricultural Land/ Crop Land

Agroforestry/Sparse Vegetation
Built-up/ River Sand

Fallow/Wasteland

Dense Vegetation
Water (Deep and Shallow)

LULC Feature
Area percentage (2000) Area percentage (2020)

Fig. 3 Distribution of the area of various LULC features of the study area
10000
5000
0
Agroforestry/Sparse Vegetation
Water (Deep and Shallow)

Agricultural Land/ Crop Land

Built-up/ River Sand

Fallow/Wasteland

Dense Vegetation
Area in Sq. km

-5000

LULC Feature
LULC, 2000 LULC, 2020 Change
( sq. Km) ( sq. Km) (sq km)

Fig. 4 Change directions of the LULC Features of the study area

Figure 4 shows the changes in the LULC area from 2000 to 2020. Figure 4 reveals
that Agroforestry/Sparse Vegetation, Dense vegetation is increased in the last two
decades. On the other hand, significant changes have been observed in Water, Built-
up/ River Sand, Fallow/Wasteland, Agricultural Land/ Crop Land areas, are drasti-
cally reduced in the last 20 year.
8

A
D

Fig. 5 Land use/land cover map of the study area A) 2000 and B) 2020
 9

5 Conclusion

The present study has been carried out on LULC through NDVI indicators using spa-
tial and machine learning applications of the Rupnarayan river basin for identifying
changes in land use dynamics in the basin area. Earth observation data has been uti-
lized to identify the LULC changes in this research. The results demonstrated that the
agricultural area had been drastically reduced from 2000 to 2020. LULC analysis also
portrayed that wasteland has rapidly increased from 2010 to 2020. Finally, the present
research study shows the changes in river basins. It prepares LULC maps which will
play a very significant role for decision-makers in identifying and protecting the flood
susceptible zones. These results will also assist in inventing better mitigation methods
for associated risk zones of the river basin.

References
Abebe, G., Getachew, D. & Ewunetu, A. (2021) Analysing land use/land cover changes and its
dynamics using remote sensing and GIS in Gubalafito district, Northeastern
Ethiopia. SN Appl. Sci. 4, 30 (2022). https://doi.org/10.1007/s42452-021-04915-8
Arulbalaji P (2019) Analysis of land use/land cover changes using geospatial techniques in
Salem district, Tamil Nadu, South India. SN Appl. Sci. 1, 462. https://doi.org/10.1007/
s42452-019-0485-5
Chauhan N (2020) Quantifying the impacts of decadal landuse change on the water balance
components using soil and water assessment tool in Ghaggar river basin. SN Applied
Sciences, 4,30. https://doi.org/10.1007/s42452-021-04915-8
Chen H (2021) Changes of the spatial and temporal characteristics of land- use landscape pat -
terns using multi-temporal Landsat satellite data: A case study of Zhoushan Island
China. Ocean Coastal Manag 213:105842. https:// doi. org/ 10. 1016/j. oceco aman.
2021. 105842
Das B, & Bandyopadhyay A (2015) Flood Risk Reduction of Rupnarayana River, towards Dis-
aster Management–A Case Study at Bandar of Ghatal Block in Gangetic Delta. Journal
of Geography & Natural Disasters, 5:1, DOI: https://doi.org/10.4172/2167-
0587.1000135
Friedlingstein, P., Jones, M. W., O’Sullivan, M., Andrew, R. M., Hauck, J., Peters, G. P., Pe-
ters, W., Pongratz, J., Sitch, S., Le Quéré, C., Bakker, D. C. E., Canadell, J. G., Ciais,
P., Jackson, R. B., Anthoni, P., Barbero, L., Bastos, A., Bastrikov, V., Becker, M., …
Zaehle, S. (2019) Global carbon budget 2019. Earth System Science Data, 11(4), 1783–
1838. https://doi.org/10.3929/ethz-b-000385668
Jones DA, Hansen AJ, Bly K (2009) Remote sensing of environment monitoring land use and
cover around parks: a conceptual approach. Remote Sens Environ 113:1346–
1356. https://doi.org/10.1016/j.rse.2008.08.018
King DJ (2011) Comparison of pixel- and object-based classification in land cover change
mapping AU—Dingle Robertson, Laura. Int J Remote Sens 32:1505–1529. https://
doi.org/10.1080/01431160903571791
Koley B, Nath A, Saraswati S, Chatterjee U, Bandyopadhyay K, Bhatta B, & Ray BC (2022)
Assessment of spatial distribution of rain-induced and earthquake-triggered landslides
10

using geospatial techniques along North Sikkim Road Corridor in Sikkim Himalayas,
India. Geo Journal. https://doi.org/10.1007/s10708-022-10585-9
Liang S, & Wang J (2020) A systematic view of remote sensing. Advanced Remote Sensing, 1–
57. doi:10.1016/b978-0-12-815826-5.00001-5
Lillesand TM, Kiefer RW, Chipman JW (2004) Remote Sensing and Image Interpretation. 5th
Edition, John Wiley, New York.
Maity SK, & Maiti R (2018). Introduction. In: Sedimentation in the Rupnarayan River.
SpringerBriefs in Earth Sciences. Springer, Cham. https://doi.org/10.1007/978-3-319-
62304-7_1
Nath A, Koley B, Saraswati S, Bhatta B, Ray BC (2021) Shoreline Change and its Impact on
Land use Pattern and Vice Versa ─ A Critical Analysis in and Around Digha Area be-
tween 2000 and 2018 using Geospatial Techniques. Pertanika Journal of Science &
Technology, 29 (1): 331 – 348.
Nath A, Koley B, Saraswati S, Ray BC (2021) Identification of the coastal hazard zone between
the areas of Rasulpur and Subarnarekha estuary, east coast of India using multi-criteria
evaluation method. Model. Earth Syst. Environ. 7, 2251–2265.
Nath A, Koley B, Saraswati S, Choudhury T, Um, JS (2022) Geospatial analysis of short term
shoreline change behavior between Subarnarekha and Rasulpur estuary, east coast of
India using intelligent techniques (DSAS). GeoJournal (2022). https://doi.org/10.1007/
s10708-022-10683-8
Pande CB, Moharir KN & Khadri SFR (2021) Assessment of land-use and land-cover changes
in Pangari watershed area (MS), India, based on the remote sensing and GIS tech-
niques. Appl Water Sci 11, 96. https://doi.org/10.1007/s13201-021-01425-1
USGCRP (2017) Climate Science Special Report: Fourth National Climate Assessment, Vol-
ume 1 [Wuebbles, D.J., D.W. Fahey, K.A. Hibbard, D.J. Dokken, B.C. Stewart, and
T.K. Maycock (eds.)]. U.S. Global Change Research Program, Washington, DC, USA,
470 pp, doi: 10.7930/J0J964J6.
Yanan L, Yuliang Q, Yue Z (2011) Dynamic monitoring and driving force analysis on rivers
and lakes in Zhuhai City using remote sensing technologies. Procedia Environ Sci
10:2677–2683. https://doi.org/10.1016/j.proenv.2011.09.416