Groundwater potential zones identification and validation in Peninsular India

ABSTRACT Groundwater mapping is essential for meeting the water requirement of people. Identification of groundwater potential zone was attempted for a watershed located in Kanchipuram district, Tamil Nadu, India. The Landsat 8 and Landsat 5 data were used for land use/land cover analysis. For delineating groundwater potential zone, total seven thematic layers namely drainage density, slope, geology, soil, geomorphology, rainfall, land use/land cover and observed groundwater levels were considered during the analysis. Afterwards thematic layers were converted into raster using GIS platform. Further, after assigning weights and ratings to each thematic layer, overlay analysis was applied and total five zones were delineated as very good, good, moderate, poor, and very poor. The majority of area has moderate groundwater potential zone. The potential zone map was also validated using ROC method. The map of 2005 shows a highest accuracy compared to rest year maps.


Introduction
Groundwater is an important natural resource used in different sectors, namely, domestic, agriculture and industrial (Singh et al., 2009).In India, majority of population depends on groundwater (Singh et al., 2010) and also in the world because groundwater act as alternative source to surface water (Arabameri et al., 2019).Therefore, mapping and monitoring of groundwater resources are very important in semi-arid and arid areas (Subba Rao & Prathap Reddy, 2004;Subba Rao et al., 2013).
Studies reported that for delineation GWPZ mainly required input data such as on land use/land cover (LULC), relief, slope, drainage density, lithology, geological structures, soil, geomorphology, water level, rainfall, and others (Murmu et al., 2019;Pande et al., 2021;Singh et al., 2010).The GWPZs can be assisted by surveys, namely, geophysical, hydrogeological, geological, drilling and satellite data, global positioning system data and GIS (Subba Rao, 2006, 2016;Kumar et al., 2008;Mondal et al., 2008;Subba Rao et al., 2016).However, direct methods are time-consuming, costintensive and required a skilled manpower (Fetter, 1994).The combination of earth observation satellite data and GIS gives an information about the different effective components of groundwater existence and its movement corresponding to the geology, soils, geomorphology, LULC, drainage density, geology, and lineaments (Jha et al., 2007;Subba Rao, 2011Yadav et al., 2018Kumar et al., 2018Kumar et al., 2020).These factors can play an important role in identification and evaluation of GWPZs (Thomas et al., 1999).Overlay analysis of thematic layers in GIS environment help to identify the occurrence of groundwater (Kumar et al., 2008;Nagarajan & Singh, 2009).Therefore, our objective of present study was to delineate the GWPZs using integrated method and to validate the output using Receptor Operating Curve (ROC) method for a watershed located in Kanchipuram, Tamil Nadu, India.

Study area
The district Kanchipuram is located in the southern part of India (Figure 1).The Kanchipuram has a total geographical area of 1909.58 km 2 .The land is flat and aspect is towards the south and east with average elevation is 83.2 m above mean sea level, whereas in east it is bound by Bay of Bengal.Agriculture is dominant class and 47% of population engaged in agriculture activities (Rawat et al., 2019).The climate is humid and hot and region receives the rain under the north-east and south-west monsoons.Relative humidity in November to January varies from 83 to 84%; however, in June, it is around 58%.The minimum and maximum temperature is 20°C and 37°C, respectively.

Physiography
The major portion is on undulating topography with countless depressions and has many natural tanks which are used for irrigation purpose.Based on Geological Survey of India, the region has three major geomorphic units namely erosional (Chingleput-Tirukkalukkunram surface), fluvial (Palar surface), and marina (Mahabalipuram surface).The drainage pattern is radial and drainage network is sub-dendritic.

Data sets and sources
Total seven types of dataset were collected and processed.Landsat 8 (4 February 2020, 14-Octo -2015), and Landsat 5 (16-Octo-2010, 11 May 2005) data sets have downloaded from web portal: https://earthexplorer.usgs.gov/and pre-processed before further analysis.Rainfall data in grid format were downloaded from (https://crudata.uea.ac.uk/ cru/data/hrg/) CRU TS 2.1 Global Climate Database (The CRU TS2.1 Climatic Research Unit (CRU) of University of East Anglia has produced climate dataset).The SRTM 30 m digital elevation model data were downloaded from web portal: https://earthexplorer.usgs.gov/and used for slope and relief analysis.Geology and geomorphology of study area was collected from Bukhosh portal of Geological Survey of India (https://bhukosh.gsi.gov.in/Bhukosh/Public), and soil map was downloaded from Food and Agricultural Organization (FAO) web portal: https://www.fao.org/soils-portal/datahub/soil-maps-and-databases/en/.The observed groundwater level data for validation were downloaded from Central Ground Water Board (CGWB), India web portal: http://cgwb.gov.in/GWdata-access.html.Global Positioning System (GPS, eTrex Legend® HCx, with ± 15 m accuracy) was selected for collection of co-ordinates of all training samples for supervised classification of satellite data.

Methodology
Land use/land cover (LULC) of different years was prepared using Landsat 5 and 8 satellite data.Geomorphology, and geology maps were prepared from resource atlas and soil map from Food and Agriculture Organization (FAO) soils portal (www.fao.org/soils-portal/data-hub/soil-maps-anddatabases/).The drainage network and slope map were prepared from SRTM 30 m DEM and with the help of drainage map, and drainage density map was generated in GIS platform using density analysis tool.The total annual rainfall distribution map of different years were prepared using inverse distance weighted interpolation of point rainfall data in GIS platform.Rank and weights were assigned to each layer.Further, these layers were integrated using weighted overlay analysis (WOA).The weight for the selected thematic layer was calculated based on Saaty's scale (1980).The particular layer and their class has assigned a weightage and a rank ranging from 1 (low GWPZ) to 5 (high GWPZ), respectively.
The GWPZ map was prepared and categorized into five classes namely very good, good, moderate poor and very poor.The Cumulative Score Index (CSI) method was used for this classification.The CSI expression is as follows (eqn.where W j is the normalized weight of the j th factor; X i is the rank of the i th classes/features of athematic layer; m (8) is the total number of thematic layers; and n is the total number of classes/features (5) in a thematic layer (Suja Rose & Krishnan, 2009).The schematic flow chart of the methodology is showed in Figure 2.

Result and discussion
The detail of different factors mentioned in following section given below:

Drainage density (D d )
Majority of drainage is found in upper and middle portion and low flow of water at upper to lower part of watershed due to low slope.In rainy season, all drainages have maximum of water and almost flood condition because the watershed is located in low lying area.In case of GWPZ, natural drainage systems (Figure 3(a)) have importance to govern the water directly on the surface because more number of drainage is a sign of more possibility of water to drain out ( Yadav et al., 2020Yadav et al., 2020;Kumar andSingh, 2021 Singh andSingh, 2022).While poor drainage network in any region will have low possibility of good GWPZ.The study area comes under the 6 th order drainage system.The D d map was developed (Figure 3(b)) using following formula (eq.2) proposed by Horton (1932); where D i = Total lengths of channels in km; A = Area of the basin in sq.km.

Geomorphology
Geomorphologic features gives a direct possibility about groundwater existence, flow and evolution (Machiwal et al., 2011) 5).Geomorphic and morpho-structural study area advises neotectonic activity during the past period (UNDP, 1987).Among these morphological features, Coastal Plain (CP), Water body-River (WR) and Waterbodies-Other (WO) are lies under very good condition hence assigned as very high rank of 5, PPC and QMD are good (rank of 4).However, LDDHV, FP and DP are fall under poor condition for water potential zones hence assigned as very poor rank of 1 (low priority).

Soils
Soil map show mean soil texture or soil type (Figure 6 (a)).Soil texture played a vital role in groundwater recharge and GWPZ.Runoff characteristics of any watershed are depend on water holding capacity of soil properties (porosity, thickness, etc.) and runoff characteristics have a good correlation with GWPZ.Zomlot et al. (2015) reported that groundwater recharge in loamy soils has a highest positive possibility which is an indication of more suitable area for groundwater potential.The area has clay loam (CL), loam (L) and sandy loamy (SL) soil and its distribution is as follows: 55%, 37% and 6% respectively (Figure 6 (b)).In study area majority of soil is CL and L types based on FAO soil data.

Rainfall
Rainfall is a vital component to assess potential zone of groundwater.Rainfall, infiltration and percolation are three factors which directly control groundwater table of an aquifer (Maity & Mandal, 2019).Das (2019) has found a good co-relationship between rainfall and groundwater table fluctuation.Considering this a score for rainfall was allotted as 1 to 5. The average annual rainfall is about 1213.3 mm and behavior is erratic.Based on average annual rainfall a distribution map (Figure.8), has been categorized into five category (score is also 1 to 5), they are, less than 1627.76mm or 1627.76 to 1708.13 mm, 1708.13 to 1761.01 mm, 1761.01 to 1803.31 mm, 1803.31 to 1841.38 mm, and 1841.38 to 1897.43 mm and score has been allotted as 1, 2, 3, 4 and 5, respectively (Table 1).

Slope information
The spatial information of elevation is shown in Figure 9 (a).Slope gradient directly impact over the infiltration and percolation process of rainfall (Figure 9(b)).Therefore, slope information can be used as one of the vital key for identification of GWPZs (Singh et al., 2018).
Therefore, slope help to identify the infiltration and percolation rate of rainfall into ground, however, higher degree of slopes means a rapid runoff and negatively correlated with infiltration and percolation rate.Whereas, at level and gently sloping area, where rain water always flows on the surface with lower velocities and positively correlated with infiltration and percolation rate.The slope varies between 0 and 28.1° in the study area (Figure 9(b)).Flat slope area give more possibility infiltration and percolation process in comparison to the steep slope.This fact was considered while assigning the weightage or score for various slope classes.Watershed slope has been categorized into five classes.In study area a large part come under the nearly level slope (0° to 0.5°, Figure 9(b)) class which indicate toward the presence of good GWPZs (Pande et al., 2020).Based on Figure 9(b), cover with nearly level (0 to 0.5°, with highest score of 5 because of nearly level, which give more possibility of more infiltration and percolation process of rainfall with late runoff), very gently sloping (0.6 to 3.3°, with score of 4 because of late runoff), gently sloping (3.4 to 8.3°, with score of 3), moderate sloping (8.4 to 15.2°, with score of 2), and steep sloping (15.2 to 28.1°, with score of 1 because of higher degree of slope results of quick runoff which give almost zero present of infiltration and percolation process) and slope covers about 60.53%, 37.55%, 1%, 0.6% and 0.29% of the total area of watershed, respectively.

Weighted overlay analysis (WOA)
The WOA is a pixel based analysis technique.Different layers have different measurement units like rainfall layer in mm, DEM in meter, slope layer in degree, D d in per kilometer (Table 1) and others.For WOA these

Validation of GWPZ
The GWPZ map was validated using the existing wells data using Receiver Operating Characteristics (ROC) or Area Under the Curve (AUC) method (Naghibi et al., 2018).The wells data within the watershed were overlaid on GWPZ.In the graphical representations, the false positive rate (FPR) and the true-positive rate (TPR) values are representing in the X and Y axis, and it is denoted by equation 3 and 4. Value range of the ROC curve from 0.5 (or 50%, moderate) to 1.0 (or 100%, best).If the model unable to forecast the existences of the groundwater zones the AUC will be 0.5.

Groundwater water potential zone (GPZ)
The GWPZ has been divided into five zones, a brief description in given in following sections:

Zone-I (Very Poor)
This zone has a steeper slope and high runoff hence a very low groundwater potential.Slope is a very steep 15.2 to 28.1° and this zone is located at northern and western part.It covers an area of 2.92% (or 55.81 km 2 , Figure 10

Zone -II (Poor, P)
In this zone slope ranges from 8.4-15.1°,such watershed area has D d range of 0.20 to 0.51 km/km 2 .This area also has clay loam soil and relief is also high.Therefore, zone-II also has few possibilities of groundwater occurrences.This area cover 21.72% (or 414.80 km 2 , Figure 10

Zone-V (Very Good)
In this class GWPZ has favorable condition but major contribution of slope as it range of 0° to 0.5° and it has good groundwater occurrence.But very low area comes under this class.It may be due to eroded pediplain and agricultural land with high infiltration and percolation capacity.Moreover, during both monsoons, the study area receives very good amount of rainfall which drains these watershed therefore zone-V have a high potentiality of retaining groundwater.This excellent zones or class contribute covers 1.23% (or 23.45 km 2 , Figure 10 We also found that wells total 9 (11%), 18 (22%), 20 (24.4%), and 16 (19.5%)wells come under the poor potential zone (Figure 11 and 2015 moderate potential zones converged into poor category (Table 2) it is due to high exploration of groundwater and less rainfall or raising of drought condition.Almost each type of water conservation practices (plantation, rainwater harvesting structure, artificial recharge bore well, and water conservation structure, etc.) can be implementing in this category for development of groundwater resource.Accuracy of the GWPZ map was validated using ROC or AUC analysis with the existing 82 wells.During the ROC analysis the value of AUC for each five years interval and these were 0.829 (or 82% at a significance value of less than 0.001, Figure 12

Conclusions
The GWPZ delineation was carried out using seven criteria in GIS platform.The GWPZ map was categorized into different classes and very poor-GWPZs have fall in clay loam with rocky soils and in very poor GWPZs the rooftop rainwater harvesting structures are recommended.The areas under poor-GWPZs are required a rigorous appropriate planning and management because a large area of watershed approx.(around) 24 km 2 area come under poor-GWPZs class.In such area small check dam and artificial recharge structure or bore well can be developed for improving groundwater resources.The medium-GWPZ category also covers a large part of watershed around 49% of total area of watershed.This category has a large area of watershed because some part of agriculture land and barren land are contributed in this category.The medium-GWPZ, has vast scope for implementation of groundwater resource improvements programs because barren land give the opportunity for it and for improving this category is required less efforts in comparison of poor-GWPZs category.Around 25% watershed area falls in good category.Moreover, present research work has good possibility to improve the irrigation facilities as well as agricultural productivity.

Figure 1 .
Figure 1.The map show a location of study area.
1): CSI = Σ (Drainage density rank x weight + Geology rank x weight + Geomorphology rank x weight +LULC rank x weight + Rainfall rank x weight + Soil rank x weight + Slope rank x weight)

Figure 2 .
Figure 2. The schematic diagram show the adopted methodologies in the study.

Figure 3 .
Figure 3. (a) A drainage network map of study area, and (b) A drainage density map of study area (c) The drainage density distribution graph of study area.

Figure 4 .Figure 5 .
Figure 4. (a) The geology map for study area, and (b) The geology class distribution graph of study area.

Figure 6 .
Figure 6.(a) The soil map of study area, and (b) The soil classes distribution graph of study area.

Figure 7 .Figure 8 .Figure 9 .
Figure 7. (a) The LULC map of study area of 2005 (b) The LULC distribution (in %) of year 2005 (c) The LULC map of study area of 2010 (d) The LULC distribution (in %) of year 2005 (e) The LULC map of study area of 2015 (f) The LULC distribution (in %) of 2015 (g) The LULC map of study area of 2020, and (h) The LULC distribution (%) of 2020.

Figure 11 .
Figure 11.(a) The GWPZ map of 2005 (b) The distribution GWPZ of 2005 (c) The GWPZ map of 2010 (d) The distribution GWPZ of 2010 (e) The GWPZ map of 2015 (f) The distribution of GWPZ in percentage 2015 (g) The GWPZ map of 2020, and (h) The distribution of GWPZ in percentage 2020.

Figure 12 .
Figure 12.(a) The ROC curve for validation of predicted map 2005 (b) The ROC curve for validation of predicted map 2010 (c) The ROC curve for validation of predicted map 2015, and (d) The ROC curve for validation of predicted map 2020.

Table 1 .
Influence and scale value of each feature.

Table 2 .
Number of wells under different classes during five years intervals.