Introduction

Palaeochannels are found to occur at the contemporary land surface (Page et al. 1996) or in the subsurface (Harris 1980) and may be demarcated by channel form infilled by palaeochannel deposits (Clarke 2009) depending on the nature of the channel geometry and its position on the floodplain in relation to the overbank events (Charlton 2007). Therefore, it can be defined as the remnants of past streams and channel courses filled by the younger alluvium representing the fluvial system of the past (Rajawat et al. 2003; Rathore et al. 2010). Palaeochannels play important role in augmenting groundwater resources of a region (Sinha et al. 2013) as they contain highly permeable fills, which form excellent aquifers (Shirke et al. 2005). Besides, palaeochannels provide a potential reservoir of precise economic deposits (Mohammed-Aslam and Balasubramanian 2010). Nevertheless, palaeochannels also provide palaeohydrological information (Kemp and Rhodes 2010) and the location of archaeological sites (Roy and Sahu 2016).

Furthermore, the importance of palaeochannels has been stressed by various researchers in different fields such as the reconstruction of the palaeohydrological condition using palaeochannel geometry and discharge estimation (Sridhar 2007; Borisova et al. 2006; Wray 2009; Kemp and Rhodes 2010; Robertson-Rintoul and Richards 1993; Raef et al. 2016), the potentiality of groundwater recharge (Samadder et al. 2011), distribution of archaeological sites (Rajani and Rajawat 2011), reconstruction of geomorphic history (Roy and Sahu 2016), groundwater resource management (Aslam et al. 2003; Chaudhary and Aggarwal 2009; Rathore et al. 2010) and seawater and groundwater exchange (Mulligan et al. 2007).

Therefore, many enthusiastic researchers have attempted to demarcate palaeochannels based on geospatial, geophysical technique and historical evidence. Most of the previous studies were carried out using geophysical techniques all over the world. Though the geophysical methods to detect palaeochannels involve huge costs and time, they provide accurate results. With the advent of modern technology, many researchers used various advanced methods such as electrical resistivity sounding (Shirke et al. 2005; Sinha et al. 2013), ground-penetrating radar (GPR) (Weymer et al. 2014), electromagnetic induction (De Smedt et al. 2011), seismic refraction (Deen et al. 2000; Raef et al. 2016), sediment dates (Ogden et al. 2001; Borisova et al. 2006; Kemp and Rhodes 2010; Ejarque et al. 2015) and sub-surface stratigraphic records (Abbott 2004).

Remote sensing (RS) and Geographical Information System (GIS) is an efficient technique having the advantage of spatial, spectral and temporal resolution for monitoring the earth system (Mahammad and Islam 2021a; Sener et al. 2005). Based on the reflection properties of abandoned channels compared to the surroundings, the optical RS data have been used by many researchers for the delineation of palaeochannel accompanied by few image processing techniques such as image classification (Aslam et al. 2003; Khadkikar et al. 2004; Chaudhary and Aggarwal 2009; Kshetrimayum and Bajpai 2011; Rajani and Rajawat 2011; Samadder et al. 2011), contrast enhancement (Wray 2009; Mehdi et al. 2016), image transformation (Wray 2009), principal component analysis (PCA) (Rajani and Rajawat 2011; Wang et al. 2012; Mehdi et al. 2016), decorrelation stretch (DCS) (Rathore et al. 2010) and the normalized difference water index (NDWI) and normalized difference vegetation index (NDVI) (Roy and Sahu 2016).

Besides, the digital elevation model (DEM) has been used for generating the stream network and hydrological routing (Elmahdy and Mohamed 2015; Mehdi et al. 2016). High-resolution light detecting and ranging (LiDAR) data have also been employed in palaeochannel detection (Jones et al. 2007). Along with the multispectral satellite images (MSIs), radio detecting and ranging (RADAR) data of C, X and L band have also been taken by many researchers as a precise source in the field of palaeochannel investigation as these bands have more penetration capability (Elmahdy and Mohamed 2015; Islam et al. 2016; Kumar and Rajawat 2017; Abdelkareem and El-Baz 2015; Abdelkareem and El-Baz 2017; Abdelkareem et al. 2020).The present study area experienced numerous channel shifting of Damodar River especially its lower courses (Ghosh and Mistri 2012). According to De Berros (1550), the main flow of the Damodar of the sixteenth century followed the course of the present-day decayed Kana Damodar that bifurcated below Selimabad and debouched with the Hooghly at Uluberia noted by Stevenson and others (Biswas 2001). The Vanden Broucke’s map (1660) shows that the main flow of the Damodar River of the seventeenth century was maintained through Moja Damodar and found to fall into the Rupnarayan River near the present Bakshi Khal. At the same time, the larger branch of Damodar used to flow eastwardly through Barddhaman along the straight line of Gangur River which met with Hooghly River near Kalna. Subsequently, this Kalna branch was dried up according to Stevenson and others (Biswas 2001). At the same time (1660), another branch of Damodar flowing through the Amta Canal met with Hooghly River opposite of Falta. It was known as the Mondal Ghat River (Bhattacharyya 2011). The Kana Damodar became the main Damodar channel at that time and found to enter the Hooghly River near Uluberia. After 1660, a new channel opened as Kana and Kunti Nadi. This took off at the great bend of Damodar near Selimabad and flew eastward as the course of present Kana Nadi near Gopalnagar. After that, it got a turning towards the north-east along the present Kunti Nadi and entered Hooghly River near Noaserai, 4.8 km upstream of Tribeni (Bhattacharyya 2011). A chart of 1690 shows the Kana Damodar as a large stream entering into the Hooghly River near Uluberia called Jon Perdo River according to Stevenson and others (Biswas 2001). After 1770, the Kana and Kunti Nadi dried up and the main flow passed through the Amta canal and Mundeswari River (Bhattacharyya 2011). The Amta canal and Mundeswari River are now the main flow of the Damodar River and the ancient courses of the Damodar River appear as remnant channels, spill channels, abandoned channels and palaeochannels. The Khari, Behula-Gangur, Ghia-Kantal, Kana-Kunti and Kana Damodar are some notable examples.

Moreover, in recent times, Mallick and Niyogi (1972) and Niyogi (1975) studied quaternary geology and provided the geomorphological units along with the distribution of palaeochannels of the Damodar Fan Delta (DFD). Similarly, Acharyya and Shah (2007) studied arsenic contamination concerning younger alluvium and the palaeochannels. Ghosh and Mistri (2012) have identified channel and bank line shifting of Damodar River. Furthermore, Ghosh and Jana (2019) have studied the influence of historical evidence on the demarcation of palaeochannels of the DFD. They have located the historical places mentioned in the Bengali folk story to sketch the linear pattern of an ancient course of the Damodar River.

Based on the previously available literature, it appears that the palaeochannels of the study area were demarcated based on geological and photogeomorphic study, field observation and historical evidence. However, there is a lack of study for the identification of palaeochannels with the integration of geospatial, geophysical and field observation. Although there is a consciousness around the hydrological extremes including floods (Bhattacharyya 2011) and depletions of groundwater level at a rapid rate (CGWB 2006; CGWB 2017) in the study area, delineation of palaeochannels using GIScience to reduce the magnitude of the floods and droughts has not been attempted till date. Thus, the novelty of the present work lies in the mapping of palaeochannels using microwave and optical RS data coupled with field observations. Besides, this work has greater significance in the planning and management of groundwater resources of semi-critical areas like the DFD. Thus, the present study would address the following objectives:

  1. i)

    To identify the palaeochannels using optical images and radar data

  2. ii)

    To assess the effectiveness of satellite data and related image processing techniques to identify the palaeochannels

Study area

Damodar fan delta and Damodar River

The DFD consists of two alluvial fans such as Memari fan trending towards the east and Tarakeswar fan trending towards the south (Acharyya and Shah 2007; Mallick and Niyogi 1972; Niyogi 1975). It extends from 22°31′09″ N to 23°20′ N latitude and 87°49′00″ E 88°29′33″ E longitude comprising an area of ~ 3206 km2 (Fig. 1). It lies in the interfluve of Hooghly River located in the east and the Damodar River located in the west and surrounded by Kusumgram fan in the north. The DFD is a younger deltaic plain characterized by the Holocene deposit (Acharyya and Shah 2007).

Fig. 1
figure 1

Location of the study area. (a) Location of DFD in India, (b) DFD boundary on the Landsat 8 false colour composite image; source: DFD boundary was modified from Acharyya and Shah 2007; Image source: USGS Landsat 8 of 18 December 2014 and 25 December 2014)

Full size image

The Damodar River, popularly known as the ‘Sorrow of Bengal’, is an important western tributary of the Ganga River (Mahammad and Islam 2021b). It supplies huge monsoonal discharge and sediment load to the Ganga River. Originating from the Chotonagpur plateau, it traverses ~ 540 km through the states of Jharkhand and West Bengal and finally falls into the Hooghly River as the combined flow of the Rupnarayan River (Rudra 2010). The catchment area of the study river measures ~ 220069 km2 of which ~ 5250 km2 is located in West Bengal (Rudra 2010). The river is also well known for the frequent change of its lower course, which modifies the topographic configuration of the Barddhaman, Hooghly and Howrah districts (Ghosh and Mistri 2012).

The shape of the Memari fan (northern part) of the DFD is an ideal fan; however, the Tarakeswar fan (lower part) is elongated towards the south due to long evolutionary history (Mahata and Maiti 2019) (Fig. 2). The relief of the DFD has been categorized into 9 classes in which the maximum elevation (~ 37 m) of the study is found in the western part near Barddhaman town. The lowest elevation (~ 8 m) of the study area is located in the southern part near Amta. The slope is reduced towards the east and southeast direction of the study area. The longitudinal and cross profiles across the DFD have been drawn to show topographical configurations (Fig. 2(b)). The longitudinal profile (A-A1) extending from the apex part to the toe area of the DFD is nearly concave while the cross profile (B-B1) extending from the northern part to the southern boundary of the DFD is perfectly convex due to the extensive erosion at the northern boundary by the Ajay River and southern boundary by the Darakeswar River (Mahata and Maiti 2019).

Fig. 2
figure 2

(a) Elevation zone of the study area. (b) Longitudinal and cross profile of the study area; A-A1 line is a longitudinal profile that shows the reduced slope with concave shape from the apex part to the toe area of the DFD and B-B1 is the cross profile in convex shape from the north to the south direction; source: Generated from 30 m SRTM DEM

Full size image

Hydro-geomorphic attributes of Damodar fan delta

The study area, located in the alluvial plain of West Bengal, portrays a gradation in relief from ~ 8 m (near Amta) to ~ 37m (near Barddhaman town). The climate is characterized by tropical humid to sub-humid type where the maximum temperature is 31.8°C recorded in May while the minimum temperature is 19.85°C recorded in December (Bhattacharyya 2011). On average, annual rainfall amounts to 1600 mm with its maximum concentration in the monsoon period (Bhattacharyya 2011). Moreover, the study area reveals the four types of soil texture—very fine, fine, fine loamy and coarse loamy (NBSS and LUP 1992).

Geologically, the study area is a part of the Bengal basin which is a structural depression surrounded by the Chotanagpur plateau to the west, Rajmahal trap to the north and Chattagram-Tripura hills to the east (Alam et al. 2003; Rudra 2010; Sarkar et al. 2021). The evolution of the Bengal basin was initiated from the rifting of the Indian plate (Gondwana land) in the early Cretaceous period. And drifting towards the north, it stitched up with the Asian plate in the late Cretaceous period (Bandyopadhyay 2007). At the junction of Indian, Tibetan and Burma plates, the Ganga-Brahmaputra river system filled up the structural depression during ~ 150 Ma BP to the present conditions (Alam et al. 2003; Bandyopadhyay 2007; Rudra 2010). The western part of the Bengal basin consists of the four subsurface structural units extending from the west to east, i.e. (a) the shield area, (b) the basin margin zone, (c) the stable shelf zone and (d) the deeper parts of the basin (Sengupta 1966; Sengupta 1972) (Fig. 3(a)).

Fig. 3
figure 3

(a) Geological map of the Bengal basin area. (b) Cross-section along the line A-B to show the subsurface configuration and broad change in lithofacies across the basin (Source: Modified from Sengupta 1972)

Full size image

The shield area located in the western fringe of the Bengal basin is constituted by Archaean metasediments with local granitic, granophyric and doleritic intrusions (Sengupta 1966). The basin margin zone is bordered to the east by a row of enechelon fault (Sengupta 1972) (Fig. 3(b)). The stable shelf zone, located between the basin margin zone and Eocene hinge zone, is associated with a very steep slope towards the southeast covering the south-western part of Bangladesh and the southern part of West Bengal (Sengupta 1972). The stable shelf zone is occupied by Mesozoic and Tertiary formations which are increased in thickness to the southeast from ~ 914.4 m near the western margin of West Bengal to ~ 8.23 m below Kolkata (Sengupta 1972) (Fig 2(b)). The deeper parts of the basin lie beyond the hinge zone and occupy most of the eastern and south-eastern parts of Bangladesh (Sengupta 1972). The DFD is located in the stable shelf zone of the Bengal basin.

Datasets and methodology

Datasets

The palaeochannels of the DFD have been delineated using RS datasets such as Landsat 8 (L8) Operational Land Imager (OLI), Sentinel 1 (S1) A and elevation profile extracted from Google Earth images. Similarly, the validations of the probable palaeochannels have been carried out using borehole data and cross-sections from the field.

The two sets of L8 images (path/row 138/44 and 139/44, date of acquisition 18 December 2014 and 25 December 2014 respectively), with 30 m spatial resolution, were downloaded from United States Geological Survey (USGS) Global Visualization Viewer (GloVis) (https://glovis.usgs.gov/) for the present study. The L8 datasets were collected for the winter season as in this period the sky is usually cloud-free and the amount of soil moisture is low. Moreover, the S1A Interferometric Wide (IW) swath level 1 GRD data (path/frame 48/516 dated 11 November 2014) was downloaded from Alaska Satellite Facility (ASF) Vertex (https://search.asf.alaska.edu/#/) for the present study. Moreover, the four sets of SRTM DEM (~ 30 m) dated 23 September 2014 were downloaded from USGS earth explorer (https://earthexplorer.usgs.gov/) to depict the elevation zone of the study area. Besides, Google Earth Pro elevation data were used to draw the cross-sectional profiles across the palaeochannels. Apart from that, the cross-sections of the probable palaeochannels measured from the field were used to portray the surface configurations of the palaeochannels. The litholog data from the Central Groundwater Board (CGWB), Kolkata, were used for the validation of the palaeochannels.

Methodology

The study has been carried out using a robust methodological design that includes data collection, processing and validations (Fig. 4).

Fig. 4
figure 4

Methodological flowchart showing the delineation of palaeochannels using optical and radar datasets and its validations using litholog and field data

Full size image

In other words, the present research involves three distinct phase preprocessing of the RS data, delineation of palaeochannels using image processing techniques and validation of the extracted palaeochannels using the geophysical data collected and field verification.

Preprocessing of optical and radar datasets

Preprocessing attempts to minimize the distortion in satellite images produced by the sensor, atmosphere and topographic effects (Young et al. 2017). The L8 data were preprocessed using ‘Top Atmospheric Correction’ which converts the digital number (DN) value to spectral reflectance in the QGIS 3.8.1 software to remove the noises. The composite band has been generated using preprocessed bands 7 (SWIR 2), 5 (NIR) and 3 (Green) in red, green and blue (RGB) channels respectively as the 7, 5, 3 combination provides natural colour with more atmospheric removal (Butler 2013). The mosaic image has been prepared from preprocessed datasets using ArcGIS 10.4.1 software.

The S1 IW GRD data is level 1.0 multi-looked and not terrain corrected. Therefore, the preprocessing of the S1 data is required for the present study that was performed using ESA SNAP (7.0) software following some steps. At first, the ‘Apply Orbit File’ was used in the sentinel data which maximizes the geolocation quality of the data. The ‘S-1 Thermal Noise Removal’ was used to reduce the noise effects in the inter-sub swath texture by normalizing the SAR backscatter signal within the entire S1 scene. Besides, this algorithm also reduces the discontinuities between sub swaths for the scene for multi swath acquisition mode (Filipponi 2019). After that, ‘S-1 Remove GRD Border Noise’ was applied to reduce the low-intensity noise and invalid data. Furthermore, the calibration step is useful to convert the digital pixel value into the radiometrically calibrated SAR backscatters. The ‘Refined Lee’ speckle filtering was applied for the present study as it can preserve edges, line features, point targets and texture information (Lee et al. 1994). The ‘Range-Doppler Terrain Correction’ was applied using SRTM DEM 1arc second with UTM/WGS84 coordinate system to geocoded the dataset to represent the SAR images geometrically (ESA 2020). Finally, the corrected backscatter data were converted into dB format which is used in the ArcGIS software.

Delineation of palaeochannels

Normalized Difference Water Index

The present study intends to use the normalized difference water index (NDWI) method of Gao (1996) for the delineation of the open waterbodies and water-rich sediments using remotely sensed data. For the present study, NIR band (b5) and SWIR band (b7) of L8 OLI have been used to derive NDWI using the raster calculator tool of the ArcGIS software based on Eq. 1. The value of NDWI ranges from + 1 (the presence of waterbody, moist soil and marshy land) to − 1 (dry land surface or vegetation) (Lillesand et al. 2015).

$$ \mathrm{NDWI}=\uprho \mathrm{NIR}-\uprho \mathrm{SWIR}/\uprho \mathrm{NIR}+\uprho \mathrm{SWIR} $$
(1)

where ρ is the reflectance of the respective band, SWIR is the short-wave infrared band and NIR is the near-infrared band.

Normalized Difference Vegetation Index

The vegetation cover is well grown on the surface of the palaeochannels and hence the linear pattern of the vegetation cover provides a valuable signal of the presence of the palaeochannels. Following Tucker and Sellers (1986),  normalized difference vegetation index (NDVI) for the present study has been computed taking the consideration of red band (b4) and NIR band (b5) of Landsat 8 OLI using the raster calculator tool of the ArcGIS software based on Eq. 2. The NDVI ranges from + 1 (presence of vegetation) to − 1 (waterbodies).

$$ \mathrm{NDVI}=\uprho\ \mathrm{NIR}-\uprho\ \mathrm{R}/\uprho\ \mathrm{NIR}+\uprho\ \mathrm{R} $$
(2)

where ρ is the reflectance of the respective band, R is the Red band and NIR is the near-infrared band.

Principal Component Analysis

In the principal component analysis (PCA), the uncorrelated data are compacted into a fewer band (Rajani and Rajawat 2011). The blue band (b2), green band (b3), red band (b4), NIR band (b5), SWIR 1 band (b6) and SWIR2 band (b7) have been used in PCA using the ArcGIS software and produced PCs have been used in the delineation of the palaeochannels. According to Maćkiewicz and Ratajczak (1993) and Jackson and Edward (1991), the PCA can be computed as a covariance matrix as follows.

$$ y\left[\begin{array}{cc}\begin{array}{c}\begin{array}{c}y\ 1,1\\ {}y\ 2,1\end{array}\kern0.5em \begin{array}{c}y\ 1,2\\ {}y\ 2,2\end{array}\\ {}\cdots \kern0.5em \cdots \end{array}& \begin{array}{c}\begin{array}{c}y\ 1,n\\ {}y\ 2,n\end{array}\\ {}\cdots \end{array}\\ {}\begin{array}{cc}y\ m,1& y\ m,2\end{array}& y\ n,m\end{array}\right] $$
(3)

where y 1, 1 is the variance of ith variable, xi, and yij is the covariance between ith and jth variables.

Results

Identification of palaeochannels

The appearance of the palaeochannels located on both the surface and near surface can be extracted using optical and microwave satellite datasets. The delineated palaeochannels in the study area are based on the image processing techniques mentioned below.

The false colour composite (FCC) image of the current study has been prepared using the preprocessed band 7, band 5 and band 3 of the L8 in RGB channels respectively as this band combination provides the maximum atmospheric penetration with true colour (Fig. 5(a)). The composite image is used to demarcate the palaeochannels in terms of visual interpretation of vegetation cover as well as waterbodies which are seen in green and blue colours respectively. In the present study, the appearance of the palaeochannels is represented by the linear alignments of the waterbody and vegetation cover in the composite scene of L8 data.

Fig. 5
figure 5

The comparison of delineated palaeochannels based on the different image processing techniques; (a) False colour composite (7, 5, 3 in RGB), (b) sentinel 1A SAR, (c) fused image (L8 band 7, 5, 3 with Sentinel Sigma° VH dB format), (d) PCA, (e) NDWI, (f) NDVI; the red box indicates the selective window for assessing the efficiency of the palaeochannel delineation; image source: USGS L8 of 18 December 2014 and 25 December 2014 and ESA S1A of 11 November 2014

Full size image

The S1A GRD image provides the information of the near-surface features which cannot be properly visualized by the interpretation of the optical satellite images. In the present study, the moisture-rich sediment is represented by the backscatters of the S1 data. The backscatter of the processed S1 data in dB format has been categorized into five classes. The water-rich sediments are presented as a dark tone in the radar images while the dry sediments are represented as the comparatively lighter tones in the radar scene. Therefore, the class ranging from 0.12 to 0.5 with linear alignments has been recognized as the palaeochannels in the study (Fig. 5(b))

The multi-sensor fusion of the present study has been done from L8 optical and S1A GRD data using the Gaussian fusion technique which enhances the information of both the datasets. The output of the L8 OLI and S1A GRD fusion data depicts the sign of the palaeochannels in the form of green patches radiating from the apex of the present study area (Fig. 5(c)).

Moreover, a combined PCA map has been prepared using three PCs such as PC1, PC2 and PC3 in RGB. PC1 is used in the red channel which denotes the blue colour. PC2 is used in the green channel and PC3 is used on the blue channel. The palaeochannels are enriched with water. Therefore, the reflectance of the spectral signature varies significantly from the other objects surrounding it. Rajani and Rajawat (2011) have used PCA for the detection of palaeochannels. They used principal components 1, 2 and 3 as RGB. Wang et al. (2012) have used PCA with farmland masking to identify the palaeochannels. In the present study, the palaeochannels appear in blue with linear configuration and are associated with waterbodies (Fig. 5(d)). The result of PCA reveals that component 1 explains 81.41% of total variance followed by component 2 (16.68%) and component 3 (1.92%) (Table 1).

Table 1 Variance and eigenvalues derived through the principal component analysis
Full size table

In the study area, the value of NDWI ranges from − 1 to 0.73. The image has been classified into five classes using natural breaks. The first class (− 1 to − 0.42), second class (− 0.42 to − 0.26), third class (− 0.26 to − 0.11), fourth class (− 0.11 to − 0.01) and fifth class (− 0.01 to 0.73) have been categorized based on the NDWI values. The blue linear segments with the value ranging from − 1 to 0.42 associated with waterbodies are the proposed palaeochannels (Fig. 5(e)).

Similarly, the positive value of NDVI denotes the maximum intensity of vegetation cover in green colour. On the other hand, the negative value portrays the minimum intensity of the vegetation cover in blue colour. In the study area, the value of NDVI ranges from − 0.71 to 0.91. The image has been classified into five classes using natural breaks. The first class (− 0.71 to − 0.12), second class (− 0.12 to 0.24), third class (0.24 to 0.40), fourth class (0.40 to 0.57) and fifth class (0.57 to 0.91) have been categorized based on the NDVI values. In the study, the linear segments of the high NDVI value (0.57 to 0.91) in green colour are the proposed palaeochannels (Fig. 5(f)).

Based on the image processing techniques using both microwave and optical data sets coupled with the observations from topographical maps and Google earth images, the courses of six major palaeochannels are detected (Fig. 6). Apart from that, numerous river cuts and swamps have been identified. The alignment and distribution of the palaeochannels have been discussed as follows.

Fig. 6
figure 6

Distribution of the palaeochannels

Full size image

Palaeochannel I has been detected using the radar and optical data which has started from Belna and bifurcated near Saktigarth. The northern part of the palaeochannel flows towards the north up to Hatgobindapur and after that, it continues towards the east through Nabastha, Jabui and Satgachhia up to Singail. The southern part of palaeochannel I flows towards the east to meet with palaeochannel II near Gopinathpur.

Palaeochannel II has diverted from the Damodar River near Palla Colony, Barddhaman, and flows a few kilometres towards the northeast in meandering or sinuous form. After Gopinathpur, it bifurcates and the upper part is known as the Behula Nadi flowing towards the east in meandering form through Amadpur, Parhati, Debpur, Magra, Hatbaksa, Tantibaksa and Sitarambati up to Kutubpur. After Kutubpur, it takes a right angle bend and flows to the south up to Baidyapur; after that, it flows to the east through Mirhat and Senerdanga, and finally meets with the Hooghly River near Kalna and another part of the palaeochannel flows towards the south up to Somra. A segment of palaeochannel II has been traced from Debpur and flows through Shridharpur, Barwa and Isabpur to meet with the Behula Nadi at Senerdanga. From Gopinathpur, the lower part of the palaeochannel II flows to the east through Memari, Bahabpur, and Kabirpur up to Dhamas.

Palaeochannel III, probably the ancient Damodar River, has been detected from the Chanchai and flows through Ajhapur, Chokhanda, Berela, Boinchi, Simlagarh Pandua, Khanyan and Chanparai. Palaeochannel III is divided many times and forms a dendritic pattern that flows towards the east in the study area. Near Debipur, the palaeochannel leaves a channel that flows towards the northeast direction through Tola, Jamna and Bara Baharkuli up to the Behula Nadi near Natagarh. Another segment is bifurcated from the palaeochannel at Berela village and continues to the south and southeast direction through Ramnagar, Dwarbasini and Khamarpara. It is also divided near Balikukhari and the bifurcated channel flows towards the south direction. Palaeochannel III is further subdivided near Pandua town into two branches—one flows towards northeast direction through Mondolai and Depara up to Masra, and the other flows to the southeast through Namajgram, Mandaran and Dadpur up to Magra.

Palaeochannel IV has been traced from satellite data, the upper reach of which is divided into two parts. The northern part of the palaeochannel is confined to the Kantal Nadi, originating from near Masagram and flows through Nabagram, Bhastara, Kamrai, Aima and Hodla. After Hodla, the Kantal Nadi is known as Ghea Nadi. The southern part is known as the Mahindar Nadi that has started from Mahindar and flows through Chopa and Jaugram, and passes through Hatkamalpur, Dhaniakhali, Talbona, Dhanaitikari, Nalitajol and Laksmanpur and meets with the Ghea Nadi near Beraberi.

Palaeochannel V, also detected based on the above-mentioned techniques, is known as the present Kunti or Kana River that bifurcates from the Damodar River at Selimabad and flows southward, then eastward and finally northeast direction accordingly in semi-circular form and falls into Hooghly River at Noaserai.

Palaeochannel VI, also known as the Kana Damodar, has diverted from the Damodar River at Selimabad and flows towards almost the south direction through Dighir, Ballavipur, Baligari, Dwipa, Jangipara, Prasadpur and Jagatballavpur and finally meets at the outfall point of the Hooghly River near Uluberia. The palaeochannel bifurcates at Baligari and a segment continues to flow southeast and then south direction through Kaikala, Haripal, Sipai Gacchi, Furfura, Hantal and Gondalpara.

Besides, the many segments of the palaeochannel have been found in the form of river cut-offs, abandoned channels and linearly aligned waterbodies. Besides, various back swamps have also been traced based on the mentioned image processing techniques.

However, there are six major palaeochannels detected which are similar to the palaeochannels demarcated by Mallick and Niyogi (1972) and Niyogi (1975). The previous study showed the palaeochannels in the form of the ancient abandoned channels of varying dimensions mainly confined in the northern part of the study area. Besides, the palaeochannels of the present study area are different in width and length from the previous studies. The width of the palaeochannels for the present study has been verified from the surface profile derived from Google earth which depicts that the width of palaeochannels I and II is smaller than that of the palaeochannel shown in the previous studies. Palaeochannel IV is partly delineated in the previous study as a segment of the Kantal Nadi while in the present work palaeochannel IV has been demarcated as Ghia Nadi which has a longer extension. Palaeochannels V and VI are described in the previous study as small natural levees but in the present study, these are shown as Kana-Kunti Nadi and Kana Damodar River.

Efficiency of image processing techniques to identify palaeochannels

The comparison among the image processing techniques using both the radar and optical data to identify the palaeochannels has been done taking the four selected windows of the study area. This will show the efficiency of the image processing techniques.

Window 1 located in the apex part of the DFD includes all the off-take points of the palaeochannels from the Damodar River (Fig. 7(a–f)). In window 1, to assess the efficiency of the optical and radar satellite data, six locations across the palaeochannels have been selected. Near to Nabastha, the appearance of the two segments of palaeochannel I is clearly visible in dark tone by the radar satellite data, which is hardly visible in the optical L8 data. The PCA image also shows the existence of the palaeochannel which has a more clear appearance than the composite image. The analysis of the NDVI techniques provides the additional observation about the presence of the palaeochannel which depicts that the width of the palaeochannels is 470 m, while the NDVI technique shows that the width of the palaeochannel is 420 m. Besides, the transverse profile A-A′ near Nabastha provides the width of the palaeochannel ~ 470 m with 2 m depth and provides evidence of the existence of the palaeochannel. At Putunda village near Saktigarh railway station, the southern part of palaeochannel I has been shown by the L8 data in the form of linear vegetation cover while the channel has been detected by the radar satellite data in the dark tone compared to the surroundings (Fig. 7(a–f)). The PCA images show the appearance of the palaeochannel more clearly. Additionally, the analysis of NDWI also provides the observation of the palaeochannel which has a width of 280 m while the analysis of NDVI reveals that the palaeochannel has a width of 210 m. In this location, a transverse profile B-B′ has been drawn across the lower part of palaeochannel I which depicts the existence of the palaeochannel having a width of 410 m with 2 m depth. In palaeochannel II, at Sahapur, Barddhaman, a comparison has been made using optical and radar data (Fig. 7(a–f)). From the analysis, it is observed that radar data represents the palaeochannel wider than that observed from the optical data. The analysis of NDWI shows the width of the palaeochannel as 310 m while the NDVI portrays the appearance of the palaeochannel with 307 m width. A cross-section has been drawn which depicts the evidence of the presence of the palaeochannel and shows the width of the palaeochannel of ~ 550 m with 2 m depth. In palaeochannel III, near Ajhapur, the width of the palaeochannel has been proved by the radar data in a dark tone whereas the presence of the palaeochannel is observed by the optical data in terms of the presence of waterbodies and vegetation cover. It also portrays that the width of the palaeochannel detected by the radar data is wider than that observed by the optical data. The NDWI analysis provides the presence of the palaeochannel with a width of ~ 680 m whereas, the NDVI technique reveals the width of ~ 708 m. The width of palaeochannel III near Ajhapur showed by the cross-section C-C′ is 1.24 km with a depth of 3 m. For palaeochannel IV, near Nabagram, the width of the palaeochannel has been measured by the radar data as 550 m whereas the composite image and PCA reveal the width of ~ 450 m, which is comparatively lower than the radar data. Besides, the NDWI and NDVI show that the width of the palaeochannel is 430 m. The cross section D-D′ reveals that the width of the palaeochannel near Nabagram is 540 m with a depth of 2 m. Similarly, the width of the palaeochannel which is the source point of palaeochannels V and VI has also been measured at Halara, near Selimabad. The width of the channel is 290 m using the radar data while the optical data depicts the width of 585 m. The NDWI and NDVI portray that the width of the channel is 465 m and 575 m respectively. The cross-section located at Halara depicts a width of the palaeochannel of ~ 710 m with a depth of 4 m.

Fig. 7
figure 7

Window 1 shows the efficiency of different image processing techniques for the delineation of palaeochannels with respect to the cross profiles; (a) false colour composite (7, 5, 3 in RGB), (b) sentinel 1A SAR, (c) fused image (L8 band 7, 5, 3 with Sentinel Sigma° VH dB format), (d) PCA, (e) NDWI, (f) NDVI; A-A′ , B-B′, C-C′, D-D′, E-E′ and F-F′ indicate the location of palaeochannel within the cross-profiles. Image source: USGS L8 of 18 December 2014 and 25 December 2014 and ESA S1A of 11 November 2014; Elevation source: Google Earth Pro

Full size image

Window 2, located in the eastern distal part of the DFD, portrays the lower courses of the palaeochannels (Fig. 8(a–f)). From the analysis of radar data, it is observed that the appearance of palaeochannel III is clearly visible in dark tone at Belara which is 430 m in width. The optical data including RGB and PCA is unable to observe properly the existence of the palaeochannel in this location. Similarly, the appearance of the palaeochannel is also hardly visible in the NDWI and NDVI output which depicts the palaeochannel in a few metres of width. The transverse profile G-G′ at Berala indicates that the width of palaeochannel III is 430 m with a depth of 9 m which is similar to the observations of data output. The second location at Baidyapur shows the width of palaeochannel II as 650 m while the optical data including RGB and PCA is unable to detect properly the existence of the palaeochannel in this location. Similarly, the appearance of the palaeochannel is also hardly visible in the NDWI and NDVI output which depicts the palaeochannel in a few metres of width. The cross-section H-H′ near Baidyapur denotes that the width of the palaeochannel is 810 m with a depth of 4 m. At Chapta, the width of palaeochannel III has been obtained from the radar data as 650 m while the optical data reveals a width of 580 m. The NDWI and NDVI represent the width as 495 m and 310 m respectively. The cross-section I-I′ reveals that the width of palaeochannel III at chapta is 320 m with a depth of 3 m. The width of palaeochannel III near Gopalnagar has been observed by the S1A data as 750 m. The optical L8 data including RGB and PCA is unable to observe properly the existence of the palaeochannel in this location. Similarly, the appearance of the palaeochannel is also hardly visible in the NDWI and NDVI output. The cross-section J-J′ shows the width of the palaeochannel of ~ 710 m with a depth of 2 m.

Fig. 8
figure 8

Window 2 shows the efficiency of different image processing techniques for the delineation of palaeochannels with respect to the cross profiles; (a) false colour composite (7, 5, 3 in RGB), (b) sentinel 1A SAR, (c) fused image (L8 band 7, 5, 3 with Sentinel Sigma° VH dB format), (d) PCA, (e) NDWI, (f) NDVI; G-G′, H-H′, I-I′ and J-J′ indicate the location of palaeochannel within the cross-profiles. Image source: USGS L8 of 18 December 2014 and 25 December 2014 and ESA S1A of 11 November 2014; Elevation source: Google Earth Pro

Full size image

Window 3, located in Tarakeshwar fan delta, which is a newer part of the south DFD, portrays the confined courses of the palaeochannels (Fig. 9(a–f)). In the present window, the four sites have been chosen for the comparison among the image processing techniques in the present study. First, near the Bhangamora, the analysis of radar data shows that the width of the old course of the Damodar is 670 m, which is wider than the present width. The optical data including RGB and PCA can directly observe the present course of water and indicates the wider old course by the presence of vegetation cover along the banks. Besides, the NDWI and NDVI depict the presence of a wider course of the river. The measurement of the transverse profile K-K′ near Bhangamora reveals a 900 m shallow depression across the river, which indicates the existence of palaeochannel. In the second location at Dighir, the width of palaeochannel VI is clearly observed by the S1A data while the optical data including RGB and PCA show palaeochannel by a linear vegetation cover. Similarly, the appearance of the palaeochannel is also visible in the NDWI and NDVI output. The cross-section L-L′ denotes the width of the palaeochannel of ~ 475 m with a depth of 2 m. At Hatkamalpur, the width of palaeochannel IV has been obtained from the radar data as 180 m, which is hardly visible in the optical data. The NDWI and NDVI represent the width in a few meters only. The cross-section M-M′ reveals that the width of palaeochannel VI is 240 m with a depth of 3 m. The width of the palaeochannel near Bagbari has been observed by the S1A data as 100 m. The optical L8 data including RGB and PCA represents the palaeochannel with a width of 280 m. The NDWI and NDVI output depicts that the width of the palaeochannel is 130 m and 125 m respectively. The cross-section N-N′ shows that the width of the palaeochannel is ~ 295 m with a depth of 2 m.

Fig. 9
figure 9

Window 3 shows the efficiency of different image processing techniques for the delineation of palaeochannels with respect to the cross profiles; (a) false colour composite (7, 5, 3 in RGB), (b) sentinel 1A SAR, (c) fused image (L8 band 7, 5, 3 with Sentinel Sigma° VH dB format), (d) PCA, (e) NDWI, (f) NDVI; K-K′, L-L′, M-M′ and N-N′ indicate the location of palaeochannel within the cross-profiles. Image source: USGS L8 of 18 December 2014 and 25 December 2014 and ESA S1A of 11 November 2014; Elevation source: Google Earth Pro

Full size image

Window 4, located in the southeastern part of the DFD, portrays the lower courses of the palaeochannels (Fig. 10(a–f)). The four sites, like windows 2 and window 3, have been selected for the comparison among the image processing techniques in the present study. First, near Bandipur, the analysis of radar data shows that the width of the palaeochannel V is 113 m. The width of the palaeochannel observed by the L8 data is 150 m. Besides, the NDWI and NDVI depict the width of the palaeochannel of ~ 60 m and 150 m respectively. The measurement of the transverse profile O-O′ reveals that the 230 m shallow depression across the river indicates the existence of palaeochannel. In the second location, Talbona, the width of palaeochannel IV of ~ 350 m is observed by the S1A data while the optical data including RGB and PCA show a width of the palaeochannel of ~ 105 m. Similarly, the NDWI and NDVI output depicts that the width of the palaeochannel is ~ 155 m and 180 m respectively. The cross-section P-P′ reveals that the width of the palaeochannel is 330 m with a depth of 2 m. Near Dwarbasini, the width of palaeochannel III has been obtained from the radar data which is 310 m while the optical data reveals that the palaeochannel has a 135 m width. The NDWI and NDVI represent that the width is 138 m and 100 m respectively. The cross-section Q-Q′ reveals that the width of the palaeochannel is 330 m with a depth of 5 m. The width of the palaeochannel IV near Nalitajol has been observed by the radar data as 295 m; however, the optical L8 data including RGB and PCA represents a width of 100 m. The NDWI and NDVI outputs depict the width of the palaeochannel as 105 m and 115 m respectively. The cross-section R-R′ shows the width of the palaeochannel of ~ 330 m with 2 m depth.

Fig. 10
figure 10

Window 4 shows the efficiency of different image processing techniques for the delineation of palaeochannels with respect to the cross profiles; (a) false colour composite (7, 5, 3 in RGB), (b) sentinel 1A SAR, (c) fused image (L8 band 7, 5, 3 with Sentinel Sigma° VH dB format), (d) PCA, (e) NDWI, (f) NDVI; O-O′, P-P′, Q-Q′ and R-R indicate the location of palaeochannel within the cross-profiles. Image source: USGS L8 of 18 December 2014 and 25 December 2014 and ESA S1A of 11 November 2014; elevation source: Google Earth Pro

Full size image

Expressions of palaeochannels in the field

The location and dimensions of the palaeochannels as depicted have been validated with some surface expression mainly in the form of field cross-section and some sub-surface imprints derived through the study of litholog composition of the boreholes as discussed in the following sections.

Surface expression

The cross-sections across the palaeochannels at Boinchi and Simalagarh have been drawn based on the Dumpy Level survey. In the first site (Boinchi), the length of the cross-section measures 119 m (88°12′05.8″E to 88°12′09.6″E and from 23°07′09.6″N to 23°07′07.5″N) and the elevation ranges from 9.58 to 11.68 m. The elevation rapidly drops after 53 m and continues to a distance of 101 m and increases thereafter. Therefore, a valley-like depression exists amid the cross-section which is a suspected palaeochannel in this area (Fig. 11a). Similarly, the cross-section at the second site (Simlagarh) shows a length of 157.8 m (88°14′29.7″E to 88°14′29.3″E and from 23°05′09.0″N to 23°05′14.4″N). The elevation ranges from 10.75 to 12 m. A shallow depression is found amid the cross-section within the stretch extending from 39.8 to 120 m. This is also marked as a suspected palaeochannel (Fig. 11b) in this region. Moreover, the suspected palaeochannel appearing as continuous ponds and waterbodies in a linear pattern near Boinchi and Simalagarh can easily be demarcated from the Google earth image (Fig. 12a). It is noteworthy that elongated channel depressions are observed during the field survey. Some parts of the palaeochannel at Simlagarh are observed in ponding conditions and used as a local waterbody (Fig. 12b) and somewhere the palaeochannels are filled with newer sediments and transformed into the agricultural field (Fig. 12c, d). Besides, the field visits also portray that groundwater is drafted from the suspected palaeochannel area for agricultural and domestic use (Fig. 12e). However, figuring out palaeochannels using only surface expression is not possible. This evidence can only signal the probable locations of their existence. Thus, the subsurface view is essential to confirm their locations.

Fig. 11
figure 11

Cross-section across the palaeochannel shows the depression of elevation that depicts the existence of the palaeochannels at a Boinchi and b Simlagarh (source of elevation profile: Dumpy level survey)

Full size image
Fig. 12
figure 12

Google earth images and field photographs. a Surface expression of palaeochannels in Pandua and surrounding area on google earth image. This area is located in the northeastern part of the DFD shown in Figure 1. b Palaeochannel in ponding condition at Simlagarh. c, d The palaeochannels occupied by agricultural land at Boinchigram and Jabui village near Memari respectively. e Extraction of groundwater near palaeochannel at Jabui village near Memari

Full size image

Sub-surface expression

The sub-surface expressions of the palaeochannel are also helpful to identify the demarcated palaeochannels of the DFD. For this purpose, strip logs of 9 boreholes of Prasadpur (DH01), Rameswarpur (DH02), Tinna (DH03), Paira (DH04), Harit (DH05), Bitra (DH06), Shibrai (DH07), Kulupukur (DH07) and Simlagar (DH09) located on the palaeochannels (Fig. 13) are studied. The borehole comprises the different layers and portrays its characteristics concerning the environment in which they developed. The borehole DH01 comprises the layer of sand with Kankar from the depth of 25.9 to 35 m. The borehole DH02 shows the layered sand very fine occasionally mixed with clay from the depth of 18 to 34.3 m. Below this layer is sand well sorted and rounded up to the depth of 50.32 m. The borehole DH03 consists of the two layers of sand well sorted and rounded from the depth of 5.50 to 19.45 m and from 41.25 to 50.45 m. The layer sand coarse with sub-angular gravels found at the depth of 19.45 to 41.25 m and from 60 to 75.4 m and 78.6 to 90.95 m. The strip-log of the borehole DH04 shows a layer of sand well sorted and rounded found at the depth of 56.95 to 75.65 m. The borehole DH05 consists of a layer of clay with kankar at the depth of 9.14 to 12.19 m, sand medium with gravels at the depth of 12.19 to 47.24 m and sand medium to coarse grey with kankar at the depth of 51.51 to 58.52 m. Similarly, the borehole DH06 is constituted by the layer of sand coarse with gravels at the depth of 97.54 to 128.02 m. Moreover, the borehole DH07 exhibits a layer of sand mixed with clay grey at the depth of 45.72 to 54.86 m. The borehole DH08 comprises the layer of sand medium with gravel at the depth of 79.25 to 85.24 m. The borehole DH09 was constituted by the layer of sand mixed with clay grey found at the depth of 21.34 to 27.43 m, sand coarse with pebbles at the depth of 84.34 to 115.82 m.

Fig. 13
figure 13

Strip logs show the lithological compositions of sub-surface beneath the palaeochannels (Source: Generated based on CGWB litholog data)

Full size image

Discussion

Distinctiveness of surface configuration aspects and near-surface features using remote sensing

The present study examines the delineation of the palaeochannels based on their surface and near-surface configurations using integrated methods of remote sensing, geophysical investigations and field survey. The surface features have been traced out using remote sensing datasets and field survey techniques while near-surface expressions of the palaeochannels are detected using radar data and lithologs. The remote sensing techniques have investigated the palaeochannels using NDWI, NDVI, PCA, RGB and Radar datasets. In the present study (window 1, window 2, window3 and window 4), NDWI depicts the palaeochannels wider than NDVI as this technique can identify water bodies and saturated water, while NDVI focuses on biomass quantity (Szabo et al. 2016). Besides, PCA has traced out the palaeochannels in wider form compared to the RGB datasets as it transforms the original correlated remotely sensed data into the uncorrelated multispectral data containing most of the information of the original datasets (Jensen 2015). Moreover, radar datasets detect the palaeochannels in a wider form than the optical datasets as it has the penetration capability up to a few cm.

Implications of the field survey and geophysical data in the identification of palaeochannels

The field survey using dumpy level has demarcated the palaeochannels by valley-like depressions. The surface method of the palaeochannels identification depends on the surface reflectance properties. Besides, the near-surface method depends on the sub-surface lithological compositions which portray the depositional environments. The borehole lithologs along with the horizontal lithofacies provide significant information about the palaeochannels. In the study area, the lithofacies are composed of clays, sands, pebbles and gravels with different grain sizes including very fine, fine, medium and coarse and different colours such as white, black, grey, brown and yellow. The sandy lithofacies indicates the dominant channel fills whereas the finer muddy lithofacies reveals the over bank levee that formed during the floods (Pal and Mukherjee 2010). In addition, the dark grey soft clay consists of high organic matter, which represents the marshy depositional environment of the Damodar River flood plain (Pal and Mukherjee 2010). Furthermore, the layer of sand with Kankar, pebbles and gravels indicates the deposition of palaeochannels (Cao et al. 2019). The lithologs of the present study portrays the repetitive sequence of lithofacies of clay, sand layers associated with kankar, pebble and gravels as exemplified by DH01, DH02, DH03, DH04, DH05, DH06, DH07, DH08 and DH09 that indicates the riverine or near riverine deposits (Fig. 13). Besides, the well-sorted and rounded sand, found in DH02, DH03 and DH04, also indicate the fluvial depositional sequences which are similar to the findings of El Bastawesy et al. (2020). Therefore, strip log characteristics of the study area imply the presence of palaeocahnnel.

Conclusion

The integration of optical and radar data has allowed the delineation of palaeochannels and also has portrayed the significant relationship between drainage courses of the present and the past of the study area. In this study, the six major palaeochannels have been identified using various image processing techniques especially RGB, SAR, fused image, PCA, NDWI and NDVI. The present study has proved the existence of the ancient course of the Damodar River, which was used as its main course in different periods. The present study also reveals that the palaeochannels have originated due to the abrupt change of the lower course of the Damodar River.

The effectiveness study reveals that the optical data provides better information related to surface configurations, whereas the SAR data portrays the near-surface features of the study. Besides, the width of palaeochannels has been measured by the surface profile which reveals that the accuracy of the optical and radar data. The validation of the study demonstrated that the used data and methods were appropriate to demarcate the palaeochannels and understanding the past fluvial system with explanations. The demarcated palaeochannels also unfold the ancient course of the Damodar River which may play an important role in groundwater management in the semi-critical community development (C.D.) blocks of the Damodar fan delta. Moreover, the delineated palaeochannels were distributed in a radial pattern and bifurcated in its lower courses, which indicate the process of the delta formation, tectonic movements and sea-level changes as well in the study area. Further study should be taken up in this direction. Besides, as the deep buried portion of the palaeochannel is quite difficult to delineate using the remote sensing images due to its incapability to detect the features concealed at greater depth under the surface layer, the future discourse may focus on the application of intensive geophysical survey coupled with advanced geospatial techniques.