A parsimonious approach to delineating groundwater potential zones using geospatial modeling and multicriteria decision analysis techniques under limited data availability condition
Abstract
This study delineates groundwater potential zones by following an “equifinality” approach and adopting a standard methodology using remote sensing, geospatial modeling, geographic information system (GIS) and multicriteria decision analysis (MCDA) techniques. A total of 11 thematic layers (ie, rainfall, topographic elevation, slope, slope length, slope steepness, soil, geomorphology, geology, drainage density, and pre- and post-monsoon groundwater levels) which have an influence on the occurrence of the groundwater are developed. The suitable weights to themes and their features are assigned and then normalized by using an analytic hierarchy process (AHP)—MCDA technique. All themes are integrated in GIS for generating a groundwater potential index (GPI) map, which classifies the study area into three zones of “good” (588.5 km2, 34.1%), “moderate” (933.4 km2, 54.1%), and “poor” (203.3 km2, 11.8%) groundwater potential. Furthermore, the accuracy of the developed GPI map is verified from the coherent estimates of rainfall-recharge. Unlike earlier studies, this study further evaluates relative sensitivity of the themes, and develops a novel “parsimonious” groundwater potential index (PGPI). The PGPI is a cost-effective and time-efficient method that excludes redundant parameters from the analysis and employs only the most sensitive themes in assessing groundwater potential. The results of both GPI and PGPI are found in good harmony over 86.7% area, which confirms the efficacy of the developed PGPI. The results of this study may be of interest to planners and policymakers as a guideline for locating appropriate groundwater development sites and managing sustainable water supplies, especially under data scarcity conditions and/or in developing countries.
1 INTRODUCTION
Two-thirds of the current world's population live in areas that experience water scarcity for at least 1 month a year, and it is remarkable that about half of this population live in China and India.1 Water consumption over large areas including parts of India, China, the Mediterranean region and the Middle East, Central Asia, arid parts of Sub-Saharan Africa, Australia, Central and Western South America, and Central and Western North America is exceeding the locally renewable water resources by a factor of two, and about 500 million people are reported to be affected with such water scarcity.2 In addition, the predicted increase in global water demands for agriculture, industry and energy sectors can exacerbate the problem of water shortage in the coming decades.3 Water crisis has been determined as the global risk of highest concern for people and economies for the next decade by the World Economic Forum.4 Thus, it is emphasized to effectively address the water scarcity through water stewardship by efficiently managing water resources worldwide to ensure social equity and sustainability of our freshwater ecosystems.5
Groundwater, the major source of freshwater, is now the significant resource for human consumption, which supplies nearly half of all drinking water in the world6 and around 43% of all water effectively consumed in irrigation.7 Consequently, the majority of the world's groundwater systems are no longer in dynamic equilibrium but do show significant declining trends.8 About 700 to 800 km3 of groundwater has been depleted from the aquifers in the USA during the 20th century.9 A World Bank Report10 states that India is the largest consumer of groundwater in the world, with an estimated annual groundwater use of 230 km3. Satellite data used in the Gravity Recovery and Climate Experiment (GRACE) indicated that the groundwater in northwest Indian states of Rajasthan, Punjab and Haryana is being depleted at a rate of 17.7 ± 4.5 km3.y−1.11 It is revealed that these Indian states lost about 109 km3 of groundwater between August 2002 and December 2008, which is double the capacity of India's largest reservoir “Wainganga” and almost triple the capacity of the USA's largest man-made reservoir “Lake Mead.” Therefore, it is very much needed to pay adequate attention for widespread assessment of this vital but invisible resource based on scientific knowledge in order to manage it in a sustainable manner.
Accurate assessment of groundwater potential is a challenging task as it requires all recharge and discharge parameters to be precisely evaluated.12 Knowledge of such parameters is critical in arid and semi-arid regions where natural recharge and discharge rates are often low,13 and groundwater pumping can rapidly dominate the dynamic equilibrium of the system. Also, proper quantification of groundwater generally requires long-term data, which are not available in many parts of the world, especially in the developing world.14 This limitation has been curtailed to some extent with emergence of the modern techniques such as geospatial modeling, which include geographical information system (GIS) and remote sensing (RS) technologies. The RS provides a powerful and cost-effective tool to obtain spatial and temporal coverage of information over large areas within a short timeframe whereas the GIS technique offers a framework for capturing, storing, analyzing, manipulating, retrieving and displaying spatial data.15 In literature, integration of RS and GIS techniques is emerging as a useful tool for assessment of groundwater potential.14-31 In past studies, researchers considered as many different themes/factors as possible for groundwater assessment mainly depending upon the extent of data availability in their area.25, 32-39 This practice presumes that all the selected themes have significant influence on the occurrence of the groundwater, which may not always be true.25, 40 Sometimes, a problem of “equifinality” may also be faced.41 Also, such a large number of thematic layers may not be found for data scarce regions especially in the developing countries.
In this study, a novel “parsimonious” approach involving minimum number of influential parameters is adopted to assess groundwater potential where only the most influential themes are selected for the analysis. The parsimonious approach also avoids the problem of “equifinality.” The concept of “equifinality,”41 in simple words, implies that there may be a possibility of different sets of model parameters yielding similar output.42 Hence, the adopted approach identifies and discards redundant parameters to make the analysis time-efficient and economic, which may also be preferable under data scarcity conditions. Such a cost-effective approach can be practical, usable and largely applicable in the developing nations. This study is carried out with two major objectives: (i) to delineate groundwater potential zones using geospatial modeling and multicriteria decision analysis (MCDA) techniques adopting a standard methodology, and (ii) to develop a novel “parsimonious groundwater potential index” (PGPI) for data scarce regions using the selected most-sensitive parameters. To date, such kind of “parsimonious” approach has never been employed in the studies dealing with assessment of groundwater potential. Furthermore, the developed methodology is demonstrated through a case study of Saraswati River basin in Gujarat, India. The study area is situated very near to the driest “white desert” of Kachchh, India where water shortages pose a severe water crisis to completely meet domestic and agricultural water requirements during dry periods.
2 MATERIALS AND METHODS
2.1 Description of study area
Saraswati River basin (study area), located in the northern part of Gujarat, India, extends over 1725.3 km2 area, and is confined between 71°45′26.84″ to 72°50′9.28″ E longitudes and 23°44′27.32″ to 24°18′39.04″ N latitudes (Figure 1). The area is surrounded by the plains of Gujarat in eastern and southern directions, Kachchh district in the western and the Aravallis towards northern directions. The Saraswati River originates from the Banaskantha district of Gujarat and spreads up to the Little Rann of Kachchh. The boundaries of the basin extend partly over three districts, that is, Banaskantha, Patan and Mehsana of Gujarat, India. The catchment of Saraswati River is the subbasin of the west-flowing rivers of Kachchh and Saurashtra regions including Luni basin flowing towards the Arabian Sea. The Saraswati River flows from the northeast towards the southwest directions with total length of 182.8 km.
Climate of the Saraswati River basin is characterized by four seasons of winter, summer, monsoon and post-monsoon. In the area, January is the coldest month of the year. Most of the winds generally blow from northwest to southeast directions. The hot weather sets in March month and continues till June. The mean annual rainfall of the study area is 640 mm (1981-2012), and more than 90% occurs during June through September by the southwest monsoon.43 The rainfall has erratic pattern over the years, its distribution is uneven over the space and the annual rainfall (1971-2004) indicates a decreasing temporal trend.43 The maximum (1488 mm) and minimum (155 mm) average rainfall occurred in the years 2006 and 1987, respectively over the period from 1981 to 2012. The major portion of the land is comparatively flat with slope ranging from 0% to 1% except some area in the extreme eastern side where the land has considerable slope. Topographic elevation of the land varies from 32 to 783 m from the mean sea level (m MSL). Texturally, the soil consists of mixed fine, coarse rocky and majority in clay. The distribution of the temperature within the basin area is uneven and erratic in nature.
2.2 Geology and hydrogeology
Geologically and mineralogically, the basin is formed due to prolonged alluvial action in Quaternary period and is dominated by pegmatites, amphibolites, quartz, feldspar, china clay and fire clay. The study area is geotectonically a graben of early Tertiary period, which offers a space for large-scale alluvial deposition through major river systems from the Aravallis. Presently, whole study area has become a blanket of thick alluvium.44
Hydrogeologically, quaternary formations of the recent period occupy large area in the basin, and it consists of fine to coarse sand, gravel, silt and clay. The surface area of the Saraswati basin is mostly consisted by wind-blown or Aeolian deposits. The most of the area form prolific, multilayer aquifer system and groundwater occurs in unconfined to semi-confined/confined conditions. Majority of the area has unconfined aquifers with very high yield ranging from 1440 to 4320 m3 day−1.44 The quality of groundwater is generally potable, and it is being exploited to a great extent. Depth to groundwater level remains from 7 to 22 m from mean sea level (m MSL) in a large portion of the area during pre-monsoon (1371.9 km2, 79.6% of total area) and post-monsoon (1325.4 km2, 76.8% of total area) seasons.
2.3 Data collection
In this study, pre- and post-monsoon groundwater level data of 74 sites located in an unconfined shallow aquifer, shown in Figure 1, are collected for 8 years (1995, 1998, 2000, 2003, 2006, 2009, 2012 and 2013) from the Central Ground Water Board, Ahmedabad, Gujarat. Toposheets of the area at 1:50 000 scale are collected from the office of Survey of India, Gandhinagar, Gujarat and digital elevation model (DEM) at 90-m resolution is downloaded from the Shuttle Radar Topographic Mission (SRTM) website. Both toposheets and DEM are used to prepare topographic elevation and slope maps. The SRTM-DEM is also used to develop slope length and slope steepness maps of the study area. In addition, annual rainfall data of 13 stations for 32-year (1981-2012) period are acquired from the India Meteorological Department (IMD), Pune. The soil and geomorphology maps are procured from the Bhaskaracharya Institute for Space Applications and Geo-Informatics (BISAG), Gandhinagar, Gujarat. Drainage density map is prepared from SRTM-DEM and toposheets using option of “hybrid drainage” in ArcGIS software, version 10.2.2. The district resource maps including geology are collected from the office of Geological Survey of India, Government of India. The succinct methodology used in preparing all these themes is described ahead.
2.4 Developing thematic layers in GIS
A total of 11 themes namely rainfall, topographic elevation, percent slope, slope length, slope steepness, soil, geomorphology, geology, drainage density, and pre- and post-monsoon groundwater levels are prepared using GIS. All GIS analyses including preparation of thematic layers and their geospatial integration are performed using ArcGIS software (ArcMap version 10.2.2). This study used proximity analysis to derive Thiessen polygons for 13 rainfall stations of the study area, which are assigned values of the mean annual rainfall to develop spatial map of the rainfall.
For developing topographic elevation map, SRTM-DEM is processed to remove sinks and peaks using tool available in ArcGIS. Then the processed DEM without having depressions/holes and sudden peaks, is sliced into a number of classes to obtain a classified elevation map. The slope of the area is directly calculated using “slope tool” in ArcGIS using the processed DEM, and the slope map is classified.
Maps of soil and geomorphology are registered in the GIS platform with the help of geo-referencing tool, and then digitized for the desired extent of the study area. Boundaries of all the classes for soil and geomorphology themes are digitized and few of the similar kind of classes are merged to avoid any recurrence resulting in unreasonably large number of classes. The geology map is geo-referenced using GIS software and is classified into seven classes.
Drainage density, defined by proximity of spacing of drainage channels,46 is calculated by dividing total length of all drainage channels with the catchment area of these channels. In this study, SRTM-DEM and the toposheets are used to compute drainage density map that divides the study area into 37 microcatchments. Then, length of drainage channels encountered in individual microcatchments is calculated through GIS tools. The drainage density for a particular catchment is estimated as the ratio of drainage-channel length to area of the catchment.
All 74 groundwater monitoring sites are registered in GIS using their location coordinates, that is, latitude and longitude. Both the pre- and post-monsoon groundwater levels are spatially interpolated for individual 8 years using geostatistical modeling, that is, kriging technique. Of the several existing spatial interpolation techniques, kriging is considered as the best linear unbiased estimation (BLUE), which has been most frequently applied in mining, geology, and hydrology.47 One of the major advantages of kriging is more flexibility than other interpolation methods.47 In kriging, weights are not selected by an arbitrary rule that may not be applicable in some cases; rather weights depend on variability of the function in space. Another advantage of kriging is that data can be analyzed in a systematic and objective way by deriving a variogram using prior experience to determine the appropriate weights.47 In other words, kriging is an “exact interpolator.” Three widely-used geostatistical models, that is, circular, spherical and exponential are employed for spatial interpolation. It is worth mentioning that the sites located beyond 10 km distance from boundaries of the study area are also considered in spatial interpolation of groundwater levels. The best-fit model is selected based on four goodness-of-fit criteria, that is, mean error (ME), root mean square error (RMSE), mean standardized error (MSE) and root mean square standardized error (RMSSE). The acceptable values of the goodness-of-fit criteria are: (i) ME and MSE should be near to zero, (ii) RMSE should be as small as possible, (iii) RMSSE should be nearest to 1, and (iv) MSE should be close to RMSE.48, 49 Finally, the best-fit model is used for generating spatial maps of groundwater levels. Thereafter, the mean of the pre- and post-monsoon groundwater level raster is generated through “cell statistics tool” of ArcGIS software.
2.5 Assigning and normalising weights by MCDA technique
In this step, the relative weights are assigned to 11 themes and their features on Saaty50 scale ranging from 1 to 9 depending upon their relative influence on the occurrence of the groundwater availability. First, opinion of the international and national experts in the field of hydrogeology is sought, and then the final weights are chosen after consulting with the local hydrogeologists and experts. This type of approach is generally followed in applying the Saaty's analytic hierarchy process (AHP) technique,50 which is one of the widely-used MCDA techniques in groundwater potential zoning studies. The final assigned weights consider the relative importance of one theme over another in influencing the occurrence of groundwater potential. An assigned weight closer to 9 indicates excellent potential for the groundwater occurrence for a theme or its class while the assigned weight closer to 1 suggests relatively poor potential for the groundwater availability.
- The principal Eigenvalue (λmax) is computed by eigenvector technique for different number of themes and their subclasses.
- Then, consistency index (C) is calculated using following formula50:
(4)where n, number of themes included in weight assignment.
- Finally, consistency ratio (CR) is determined using the following formula50:
(5)where Crandom, random consistency index, whose values are taken from the standard table provided by Saaty50 depending upon the number of considered themes.
2.6 Computing groundwater potential index
The GPI values on the developed map are classified into three classes following the natural break method. The classes of natural breaks are based on internal natural grouping, which is inherent in the particular datasets. The identification of the class breaks is made through the best group of similar values, and that maximize the variation between the distinct classes. Thus, based on the relative differences in the data values, the corresponding features are divided into three classes according to the set boundaries.48
2.7 Validating developed GPI map
The mean annual recharge map is crossed in GIS with the developed GPI map through overlay analysis using the “Union tool” of ArcGIS software.
2.8 Identifying most-influential parameters through sensitivity analysis
2.9 Preparing parsimonious GPI map
2.10 Verifying parsimonious-GPI approach
The PGPI map developed by using the most-influential themes is verified by comparing it with the original GPI map. Both the GPI and PGPI maps are subjected to overlay analysis in geospatial domain using “Union tool” of the ArcGIS software. Prior to applying “Union tool” in ArcGIS, both the raster GPI maps are converted into vector format. A confusion matrix is prepared to depict the similar and dissimilar results of two approaches.
3 RESULTS AND DISCUSSION
3.1 Features of themes for groundwater potential mapping
Features of the themes used for groundwater potential mapping are briefly described in the following subsections.
3.1.1 Mean annual rainfall map
The mean annual rainfall map of the study area for 13 Theissen polygons classified into five classes is shown in Figure 3A. Occurrence of relatively high rainfall quantities at a place has scope for the better groundwater recharge, and thus, the better may be the prospect of groundwater potential. It is seen from Figure 3A and Table 1 that the mean annual rainfall is the lowest (less than 550 mm) in 369.39 km2 (21.4%) area located in the southwest and middle portions while the highest mean annual rainfall of more than 700 mm occurred in 408.62 km2 (23.7%) area existing in the northeast portion. The most of the portion with highest rainfall overlapped with hilly area of the basin. The mean annual rainfall in the study area over a period of 32 years (1981-2012) varied from 510 mm at Sami raingauge station located in western side to 765 mm at Danta raingauge station located in eastern side. A major portion of the study area, that is, 628.3 km2 (36.4%) lying in the middle reaches of the basin received the mean annual rainfall of 600-650 mm (Table 1). Thus, rainfall amount is found decreasing from upper to lower elevation zones in the basin. High groundwater potential is favored by high amount of rainfall and the relative weights are assigned accordingly.56
| Theme | Subclass/feature | Groundwater potential | Area (km2) | Area (%) | Assigned weight | Normalized weight | Consistency ratio (%) |
|---|---|---|---|---|---|---|---|
| Rainfall | <550 mm | Poor | 369.4 | 21.4 | 4 | 0.13 | 3.3 |
| 550-600 mm | Moderate | 162.7 | 9.4 | 5 | 0.17 | ||
| 600-650 mm | Good | 156.3 | 9.1 | 6 | 0.20 | ||
| 650-700 mm | Very good | 628.3 | 36.4 | 7 | 0.23 | ||
| >700 mm | Very good | 408.6 | 23.7 | 8 | 0.27 | ||
| Topographic elevation | <50 m | Very good | 121.0 | 7.0 | 8 | 0.25 | 1.0 |
| 50-75 m | Good | 124.0 | 7.2 | 7 | 0.22 | ||
| 75-100 m | Good | 77.8 | 4.5 | 6 | 0.19 | ||
| 100-150 m | Moderate | 267.9 | 15.5 | 4 | 0.13 | ||
| 150-200 m | Poor | 399.0 | 23.1 | 3 | 0.09 | ||
| 200–300 m | Poor | 597.1 | 34.6 | 2 | 0.08 | ||
| >300 | Very poor | 138.5 | 8.0 | 1 | 0.04 | ||
| Percentage slope | <1% | Good | 738.5 | 42.8 | 7 | 0.37 | 7.3 |
| 1%-3% | Good | 711.0 | 41.2 | 5 | 0.26 | ||
| 3%-10% | Moderate | 149.4 | 8.7 | 4 | 0.21 | ||
| >10% | Poor | 126.4 | 7.3 | 3 | 0.16 | ||
| Slope length | 2.017 | Poor | 171.1 | 9.9 | 4 | 0.27 | 3.5 |
| 1.753 | Moderate | 104.6 | 6.1 | 5 | 0.33 | ||
| 1.523 | Good | 1449.6 | 84.0 | 6 | 0.40 | ||
| Slope steepness | 0.03-0.15 | Good | 809.6 | 46.9 | 7 | 0.39 | 2.4 |
| 0.15-0.50 | Moderate | 728.7 | 42.2 | 6 | 0.33 | ||
| 0.50-12.55 | Moderate | 187.0 | 10.8 | 5 | 0.28 | ||
| Soil | Coarse sand (river bed) | Very good | 66.2 | 3.8 | 8 | 0.25 | 9.1 |
| Coarse loam | Good | 265.2 | 15.4 | 7 | 0.22 | ||
| Mixed coarse to fine loam | Moderate | 409.3 | 23.7 | 6 | 0.16 | ||
| Fine loam | Moderate | 778.4 | 45.1 | 5 | 0.13 | ||
| Loamy skeletal | Poor | 26.1 | 1.5 | 4 | 0.19 | ||
| Rocky outcrop | Very poor | 180.1 | 10.4 | 2 | 0.06 | ||
| Geomorphology | Waterbody/river | Excellent | 50.6 | 2.9 | 9 | 0.21 | 6.3 |
| Pediplain | Very good | 207.6 | 12.0 | 8 | 0.05 | ||
| Alluvial plain | Very good | 878.0 | 50.9 | 8 | 0.04 | ||
| Flood plain | Moderate | 47.4 | 2.7 | 5 | 0.10 | ||
| Aeolian plain | Moderate | 347.6 | 20.1 | 4 | 0.13 | ||
| Built-up land | Very poor | 1.6 | 0.1 | 2 | 0.21 | ||
| Denudation hill | Very poor | 4.3 | 0.2 | 1.5 | 0.03 | ||
| Structural hill | Very poor | 188.2 | 10.9 | 1 | 0.23 | ||
| Geology | Channel fill deposit | Excellent | 100.7 | 5.8 | 9 | 0.18 | 1.2 |
| Flood plain and levee deposit | Excellent | 592.0 | 34.3 | 9 | 0.17 | ||
| Flood plain | Very good | 90.8 | 5.3 | 8 | 0.15 | ||
| Flood plain and channel fill deposit | Very good | 512.8 | 29.7 | 8 | 0.16 | ||
| Sand sheet and dune deposit | Good | 279.0 | 16.2 | 7 | 0.14 | ||
| CalcSchist/gneiss, marble and biotic gneiss | Moderate | 7.9 | 0.5 | 6 | 0.12 | ||
| Granite, leucogranite, quartz porphyry | Poor | 142.1 | 8.2 | 5 | 0.10 | ||
| Drainage density | <0.60 | Good | 29.2 | 1.7 | 7 | 0.26 | 8.4 |
| 0.60-0.70 | Good | 330.6 | 19.2 | 6 | 0.22 | ||
| 0.70-0.80 | Moderate | 655.8 | 38.0 | 5 | 0.19 | ||
| 0.80-0.90 | Moderate | 398.2 | 23.1 | 4 | 0.15 | ||
| 0.90-1.00 | Very poor | 254.6 | 14.8 | 3 | 0.11 | ||
| >1.00 | Very poor | 56.9 | 3.3 | 2 | 0.07 | ||
| Premonsoon groundwater level (m bgl*) | <7 | Very good | 54.3 | 3.1 | 8 | 0.23 | 8.8 |
| 7-10 | Very good | 79.7 | 4.6 | 7 | 0.20 | ||
| 10-13 | Good | 201.0 | 11.7 | 6 | 0.17 | ||
| 13-16 | Good | 431.7 | 25.0 | 5 | 0.14 | ||
| 16-19 | Moderate | 349.4 | 20.3 | 4 | 0.11 | ||
| 19-22 | Poor | 310.1 | 18.0 | 3 | 0.09 | ||
| >22 | Very poor | 299.1 | 17.3 | 2 | 0.06 | ||
| Postmonsoon groundwater level (m bgl*) | <7 | Very good | 89.0 | 5.2 | 8 | 0.23 | 8.8 |
| 7-10 | Very good | 197.9 | 11.5 | 7 | 0.20 | ||
| 10-13 | Good | 234.3 | 13.6 | 6 | 0.17 | ||
| 13-16 | Good | 380.4 | 22.0 | 5 | 0.14 | ||
| 16-19 | Moderate | 250.4 | 14.5 | 4 | 0.11 | ||
| 19-22 | Poor | 262.4 | 15.2 | 3 | 0.09 | ||
| >22 | Very poor | 310.9 | 18.0 | 2 | 0.06 |
- Note: *m bgl, meter below ground level.
3.1.2 Topographic elevation map
Topographic elevation map of the study area is classified into seven classes as shown in Figure 3B where a linear gradient of the topography from northeast to southwest direction along the river course can be clearly seen. A general perception of the researchers is that the higher the topographic elevations, the less are the chances to have good potential for the groundwater availability, and vice versa.15 Land elevation is found as the highest, that is, more than 300 m MSL, in 138.5 km2 (8%) towards the extreme northeast portion while the elevation is the lowest, that is, less than 50 m MSL, in 121 km2 (7%) southwest part of the area (Table 1). A large portion of the land (597.07 km2, 34.6% area) is situated at an elevation range of 200-300 m MSL (Figure 3B). The portion comprised of highest elevation is distributed over very small area of hilly tracks and slope of the land is drastically reduced in adjoining range of lower elevation.
3.1.3 Percentage slope map
Percentage slope map sliced into four groups, that is, <1, 1-3, 3-10 and >10% is depicted in Figure 3C. On the lands having higher values of the percent slope, there is less opportunity for the flowing water to get infiltrated into the land surface, and hence, the lesser is the likelihood of groundwater availability.57 It is seen from Figure 3C that the land in northeast portion has relatively high percentage slopes (>3%) in 275.8 km2 (16%). The mean land slope of the area is towards the southwest direction and more than 84% (1449.6 km2) of the area is having more or less flatlands with less than 3% slope (Table 1).
3.1.4 Slope-length map
Classified into three classes, slope-length map of the area is shown in Figure 3D. It is believed that the groundwater potential should be relatively good at a place where slope length is lesser.57 Of the total, about 84% of the area (149.6 km2) excluding the hilly portion in the upper reaches has the lowest slope length value of 1.523. However, the highest slope length value (2.017) is distributed over 171.1 km2 (9.9%) land (Table 1) in the extreme northeast portion of the area. Very less spread of the land (104.6 km2, 6.1%) is under the moderate slope length of 1.753 that is distributed in the pockets mostly in the middle portion.
3.1.5 Slope steepness map
Spatial distribution of three classes of slope steepness can be seen in Figure 3E. Slope steepness of a land should be as much less as possible from prospective of the groundwater availability under the land surface.57 Among the three classes, the slope steepness is the lowest (<0.15) over a major portion of the area, that is, 809.6 km2 (46.9%) that is mostly extended in the lower reach towards the southwest portion with moderate and little stretches towards the middle and upper reaches of the basin. The highest slope steepness up to a value of 12.55 exists in the hilly portion of the upper reach covering 187 km2 (10.8%) area (Table 1). The moderate slope steepness values ranging from 0.15 to 0.50 are mostly seen over the middle portion of area encompassing an area of 728.6 km2 (42.2%).
3.1.6 Soil texture map
The study area has six classes of the soil texture: (i) mixed coarse to fine loam, (ii) coarse sand, (iii) loamy skeletal, (iv) rock outcrop, (v) coarse loam, and (vi) fine loam (Figure 3F). Presence of fine loam soil is predominant in the area with more than 45% (778.4 km2) coverage of land area (Table 1). Soils with coarse sand and loam content provide adequate space for the rainwater to get recharged through the land surface, and hence, there is high possibility of good groundwater potential.15 On contrary, the skeletal type of soils or rock outcrop area with negligible soil depth has less influence on the recharge and groundwater occurrence.57 Fine loam soil is available over entire study area with dominance in the middle and lower reaches of the Saraswati River and is nearly absent in the upper hilly portion. The second major type of the soil texture available in area is mixed coarse to fine loam that is present in 409.3 km2 (23.7%) area. In the area, loamy skeletal soil is distributed over a small extent of 26.1 km2 (1.5%). Whereas, coarse loam and coarse sand soils in 265.2 km2 (15.4%) and 66.2 km2 (3.8%) areas, respectively are mainly available along the river course between the middle and the lower reaches.
3.1.7 Geomorphology map
Entire study area is divided into eight classes according to type of geomorphology (Figure 3G). With negligible potential of the groundwater availability, the structural hills are present over 188.2 km2 (10.9%) towards the northeast portion of the area. On the other hand, waterbody mainly present in the form of river courses have a good prospective of groundwater availability,57 and is spread over 50.6 km2 (2.9%) area. It is seen that the alluvial plain located in the middle and lower reaches of the basin is the prominent type of geomorphology that exists over more than 50% area (878 km2) followed by the Aeolian plains that cover 20.1% (347.6 km2) area in the middle-northern portion (Table 1). Extent of built-up land and denudation hill in the area is limited up to 1.6 (0.1%) and 4.3 (0.2%) km2 lands, respectively.
3.1.8 Geology map
There exist seven types of geology classes in the study area (Figure 3H). It is seen from Table 1 that a major portion of the area, that is, 592 km2 (34.3%) is covered by flood plain and levee deposits followed by flood plain and channel fill deposits, that is, 512.7 km2 (29.7%). The flood plains with both levee and channel fill deposits have an excellent potential for the occurrence of groundwater resources. The levee deposits have more clay content in comparison to channel deposits, and thus, the former type of geology as compared to later is less favorable for the occurrence of groundwater. Among all the geology classes, calcareous schist/gneiss, marble and biotic gneiss has negligible area (7.9 km2, <0.5%) within the upper reach of the basin. Besides, hard-rocks with granite, gneiss, leucogranite, and quartz porphyry type of geology has only 8.2% (142.07 km2) area entirely in the extreme northeast portion (Figure 3H). The middle-north portion of the area contains sand sheet and dune deposits over 279.04 km2 (16.2%) of the area, which may have sufficient groundwater resources as reported by Agarwal and Garg.16
3.1.9 Drainage density map
The calculated values of drainage density for different subcatchments are classified into six classes. The Drainage density of the basin ranges from less than 0.6 to more than 1.0 (Figure 3I). There are two major streams of the Saraswati River flowing from northwest to the southeast direction (Figure 1). Both the streams are met at a point near the middle constricted portion where the width of catchment is drastically reduced, and afterwards, the river continues to flow towards the lower reach. Drainage density in a large portion, that is, 655.8 km2 (38% area), ranges between 0.7 and 0.8 that is mainly located in the upper reach of the river (Figure 3I). Relatively moderate drainage density in the upper reach may be attributed to the presence of hilly area comprised of hard-rocks with highly undulating topography. However, in the lower reach of the river, land is comparatively flat with geological feature of flood plain levee deposits and the drainage density is found maximum (>1) within this extent. The mean drainage density is found comparatively higher in the lower reach of the river, which is generally true for all the river basins. Relatively large drainage density in the lower reach of the river basins is mainly responsible for accumulation of vast quantities of surface water flows there.57 The drainage density is the lowest (<0.6) in very small area (29.2 km2, 1.7% area), located in the lower-middle portion after the convergence of the two major streams (Figure 3I,G). The land area is mostly comprised with flat topography in lower reaches, and increasing small branching channels or more drainage density can lead to less groundwater recharge and more evaporation losses. The more drainage density can be responsible for generating more runoff and less infiltration and vice versa, the same criteria are adopted in the previous studies.16, 34
3.1.10 Pre- and Post-monsoon groundwater level maps
The values of prediction errors for three semi-variogram models (circular, spherical and exponential) fitted for interpolating groundwater level data are presented in Table 2. The circular model is fitted for 50% datasets of both pre- and post-monsoon groundwater levels.
| Year | Season | Model | Mean error (m) | Root mean square error (m) | Mean standardized error (m) | Root mean square standardized error (m) | Average SE (m) |
|---|---|---|---|---|---|---|---|
| 1995 | Premonsoon | Circular* | −0.29 | 3.99 | −0.07 | 0.96 | 5.64 |
| Spherical | −0.38 | 4.08 | −0.09 | 0.97 | 5.64 | ||
| Exponential | −0.47 | 4.15 | −0.12 | 0.97 | 5.60 | ||
| Postmonsoon | Circular* | 0.03 | 4.42 | −0.04 | 0.84 | 7.76 | |
| Spherical | −0.02 | 4.43 | −0.05 | 0.86 | 7.76 | ||
| Exponential | −0.28 | 4.32 | −0.09 | 0.87 | 7.35 | ||
| 1998 | Premonsoon | Circular | 0.11 | 4.79 | −0.06 | 0.97 | 4.07 |
| Spherical* | 0.09 | 4.77 | −0.06 | 0.97 | 4.04 | ||
| Exponential | −0.12 | 4.76 | −0.11 | 0.99 | 3.81 | ||
| Postmonsoon | Circular | −0.03 | 5.20 | −0.04 | 0.60 | 9.16 | |
| Spherical | −0.18 | 5.12 | −0.06 | 0.64 | 8.83 | ||
| Exponential* | 0.05 | 5.03 | −0.01 | 0.57 | 8.80 | ||
| 2000 | Premonsoon | Circular* | −0.21 | 4.58 | −0.23 | 1.83 | 4.92 |
| Spherical | −0.43 | 4.32 | −0.24 | 1.57 | 5.07 | ||
| Exponential | −0.41 | 4.21 | −0.24 | 1.58 | 4.96 | ||
| Postmonsoon | Circular | −0.30 | 4.57 | −0.20 | 1.40 | 5.46 | |
| Spherical* | −0.28 | 4.49 | −0.20 | 1.40 | 5.37 | ||
| Exponential | −0.30 | 4.51 | −0.21 | 1.40 | 5.34 | ||
| 2003 | Premonsoon | Circular | −0.22 | 5.49 | −0.09 | 1.16 | 5.61 |
| Spherical | −0.24 | 5.51 | −0.09 | 1.15 | 5.65 | ||
| Exponential* | −0.02 | 5.51 | −0.04 | 1.10 | 5.91 | ||
| Postmonsoon | Circular | 1.03 | 6.56 | 0.03 | 0.68 | 11.53 | |
| Spherical | 1.01 | 6.57 | 0.04 | 0.68 | 11.49 | ||
| Exponential* | 0.62 | 6.47 | −0.02 | 0.78 | 10.17 | ||
| 2006 | Premonsoon | Circular* | 0.03 | 5.18 | −0.08 | 1.17 | 5.77 |
| Spherical | −0.01 | 5.19 | −0.09 | 1.18 | 5.77 | ||
| Exponential | 0.04 | 5.36 | −0.05 | 1.07 | 6.09 | ||
| Postmonsoon | Circular | −0.32 | 4.39 | −0.23 | 1.48 | 5.75 | |
| Spherical | −0.36 | 4.44 | −0.24 | 1.47 | 5.73 | ||
| Exponential* | −0.33 | 4.62 | −0.23 | 1.49 | 5.66 | ||
| 2009 | Premonsoon | Circular* | −0.06 | 4.62 | −0.04 | 0.99 | 5.39 |
| Spherical | −0.05 | 4.67 | −0.04 | 0.98 | 5.42 | ||
| Exponential | −0.13 | 4.85 | −0.05 | 1.03 | 5.16 | ||
| Postmonsoon | Circular | 0.06 | 4.74 | −0.04 | 1.03 | 5.40 | |
| Spherical* | 0.06 | 4.76 | −0.04 | 1.02 | 5.40 | ||
| Exponential | 0.15 | 4.89 | 0.02 | 0.88 | 5.90 | ||
| 2012 | Premonsoon | Circular* | 0.09 | 8.18 | −0.04 | 1.02 | 7.74 |
| Spherical | 0.12 | 8.19 | −0.05 | 1.05 | 7.66 | ||
| Exponential | 0.50 | 8.06 | −0.01 | 0.99 | 9.22 | ||
| Postmonsoon | Circular* | 0.59 | 8.55 | 0.03 | 0.80 | 11.53 | |
| Spherical | 0.75 | 8.57 | 0.03 | 0.77 | 12.17 | ||
| Exponential | 0.98 | 8.53 | 0.07 | 0.67 | 13.57 | ||
| 2013 | Premonsoon | Circular* | 0.02 | 7.36 | 0.04 | 0.90 | 9.68 |
| Spherical | 0.01 | 7.39 | 0.04 | 0.90 | 9.71 | ||
| Exponential | −0.34 | 8.13 | 0.01 | 0.82 | 9.93 | ||
| Postmonsoon | Circular | 0.69 | 7.96 | 0.04 | 0.74 | 13.73 | |
| Spherical* | 0.65 | 7.98 | 0.04 | 0.75 | 13.63 | ||
| Exponential | 1.73 | 8.86 | 0.13 | 0.50 | 18.79 |
- Note: *Model Selected for Interpolation.
Spatially-distributed groundwater level maps for pre- and post-monsoon seasons are depicted in Figure 3J,K. It is apparent from both the maps (Figure 3J,K) that the spatial patterns of the groundwater levels remain similar during both the pre- and post-monsoon seasons in the area. Both the pre- and post-monsoon groundwater level maps are classified in to same classes based on the range of the groundwater variation over the area. The groundwater levels in both the seasons are found to be deepest as more than 22 m below ground level (bgl) in the central-north portion. Whereas, the groundwater levels in northeast and southwest portions stay at relatively shallow depths (<7-10 m bgl) during both the seasons. It is seen that the groundwater levels at a depth of 13-16 m bgl are dominant over 431.7 km2 (25% area) and 380.4 km2 (22%) during the pre- and post-monsoon seasons, respectively (Table 1). Area under the shallow groundwater levels (<7 to 10 m bgl) is found as 134 km2 (7.8%) during the pre-monsoon season, which increased up to 286.9 km2 (16.6%) during the post-monsoon season. In contrast, the area under the deep groundwater levels (more than 16 m bgl) get decreased by 14% from 958.6 km2 (55.6% area) during the pre-monsoon season to 823.7 km2 (47.7% area) during the post-monsoon season. These findings clearly indicate the impact of the groundwater recharge occurring during the monsoon season in the area; similar findings are reported in the literature.58 Overall on the mean groundwater levels, the largest change in the groundwater levels is observed in the upper reach of the basin during post-monsoon season in comparison to that in the middle and lower reaches.
3.2 Spatial map of groundwater potential zones
Weights assigned to 11 themes are normalized by computing pair-wise comparison matrix as shown in Table 3. Similar procedure is adopted for normalizing the assigned weights to features of all the themes, and the assigned and normalized weights are presented in Table 3. The consistency ratio for all themes and their individual features is found to be less than 10%, which indicated that the weights are consistent and can be used for integration. The GPI is computed by integrating normalized weights of all themes and their features. Spatial map of the GPI for the study area is classified into three categories based on natural break classification scheme: (i) “good” (GPI = 2.7-3.6), (ii) “moderate” (GPI = 1.9-2.7), and (iii) “poor” (GPI = 0.4-1.9) (Figure 4). The same kind of classification is also adopted in previous studies.34, 59 The central-north portion of the area (588.5 km2, 34.1%) has “good” potential for the groundwater. The zone having “moderate” groundwater potential covers 933.4 km2 (54.1%) land, and is mainly located in the upper hilly reach, central-south and central-lower reach of the basin. The groundwater potential is found “poor” in 203.3 km2 (11.8%) area situated in the lower reach and a small pocket in the central-south portion of the basin. Overall, a major portion of the area, that is, more than 50% is having “moderate” to “good” potential for the groundwater occurrence.
| Theme | Assigned Weight | R | TE | PS | SL | SS | SO | GM | GE | DD | GWpre | GWpost | Normalized weight |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| R | 8.0 | 8/8 | 8/4 | 8/5 | 8/4.5 | 8/4.5 | 8/4.5 | 8/9 | 8/7 | 8/8 | 8/7 | 8/8 | 0.12 |
| TE | 4.0 | 4/8 | 4/4 | 4/5 | 4/4.5 | 4/4.5 | 4/4.5 | 4/9 | 4/7 | 4/8 | 4/7 | 4/8 | 0.06 |
| PS | 5.0 | 5/8 | 5/4 | 5/5 | 5/4.5 | 5/4.5 | 5/4.5 | 5/9 | 5/7 | 5/8 | 5/7 | 5/8 | 0.07 |
| SL | 4.5 | 4.5/8 | 4.5/4 | 4.5/5 | 4.5/4.5 | 4.5/4.5 | 4.5/4.5 | 4.5/9 | 4.5/7 | 4.5/8 | 4.5/7 | 4.5/8 | 0.06 |
| SS | 4.5 | 4.5/8 | 4.5/4 | 4.5/5 | 4.5/4.5 | 4.5/4.5 | 4.5/4.5 | 4.5/9 | 4.5/7 | 4.5/8 | 4.5/7 | 4.5/8 | 0.06 |
| SO | 4.5 | 4.5/8 | 4.5/4 | 4.8/5 | 4.5/4.5 | 4.5/4.5 | 4.5/4.5 | 4.5/9 | 4.5/7 | 4.5/8 | 4.5/7 | 4.5/8 | 0.06 |
| GM | 9.0 | 9/8 | 9/4 | 9/5 | 9/4.5 | 9/4.5 | 9/4.5 | 9/9 | 9/7 | 9/8 | 9/7 | 9/8 | 0.13 |
| GE | 7.0 | 7/8 | 7/4 | 7/5 | 7/4.5 | 7/4.5 | 7/4.5 | 7/9 | 7/7 | 7/8 | 7/7 | 7/8 | 0.10 |
| DD | 8.0 | 8/8 | 8/4 | 8/5 | 8/4.5 | 8/4.5 | 8/4.5 | 8/9 | 8/7 | 8/8 | 8/7 | 8/8 | 0.12 |
| GWpre | 7.0 | 7/8 | 7/4 | 7/5 | 7/4.5 | 7/4.5 | 7/4.5 | 7/9 | 7/7 | 7/8 | 7/7 | 7/8 | 0.10 |
| GWpost | 8.0 | 8/8 | 8/4 | 8/5 | 8/4.5 | 8/4.5 | 8/4.5 | 8/9 | 8/7 | 8/8 | 8/7 | 8/8 | 0.12 |
Presence of the “moderate” groundwater potential zones in the central-south portion may be likely due to soils comprising of fine loam and geology of flood plain and levee deposits. On the other hand, the “good” groundwater potential in the central-north portion may be attributed to the existence of sand sheet and dune deposit type of geology and soil types varying from sole fine loam or mixed coarse to fine loam. Findings obtained in this study are also supported by previous studies carried out under similar kind of geological settings.16 Furthermore, the Aeolian and alluvium plains type of geomorphology along with topographic land slopes of less than 3% in the central-north portion may be responsible for occurrence of the “good” groundwater potential there.56 The downstream part of the basin is mainly comprised of shallow water table with less variation in pre- and post-monsoon water levels and showed availability of “poor” groundwater potential. The area consisting steep slopes, subsurface strata of granite, leucogranite, quartz porthyry and rocky outcrop soils is categorized as “moderate” potential zone of groundwater availability. The fine loam soil with flood plain channel deposit responded well to groundwater recharging. This area can used for developing recharge strategies in upper reaches of the basin. The area classified under the “good” potential of groundwater is found having pressure of excessive groundwater withdrawal, and the area is turning towards highly water-scarce region. Hence, there is an urgent need for making the balance between recharge and withdrawal of groundwater for sustainable management of water resources in the region.
3.3 Efficacy of delineated groundwater potential zones
Efficacy of the delineated GPI zones is evaluated from the groundwater recharge estimated by rainfall infiltration factor (RIF). The standard values of RIF taken from the authentic sources for the lithology classes available in the area are used in this study (Table 4). The maximum (0.225) and minimum (0.075) values of RIF are assigned to sand sheet/dune deposits (alluvium formation) and granite/leucogranite/quartz porphyry (hard-rocks) formations, respectively. The RIF values are multiplied with the mean annual rainfall map drawn for Thiessen polygons of the raingauge stations to obtain geospatial distribution of the rainfall-recharge for the study area over the rainy season. Rainfall-recharge is classified into three classes as (i) “low” (recharge <70 mm), “moderate” (recharge = 70-120 mm) and “high” (recharge >120 mm) (Figure 5). By comparing spatial distribution of the GPI (Figure 4) and rainfall-recharge (Figure 5) over the study area, it is obvious that the zones characterized by the “good” and “poor” groundwater potential are geographically associated with zones having “high” and “low” rainfall-recharge classes, respectively. This finding reveals adequacy of the delineated GPI zones. In order to further evaluate the efficacy of the GPI zones, both the maps (Figures 4 and 5) are crossed to each other in GIS platform, and the results of cross-validation are presented in Figure 6. It is seen that more than 85% (1307.6 km2) area of the “good” to “moderate” GPI exactly intersects with “high” and “moderate” rainfall-recharge zones. Besides, 88% (272 km2) area of “high” rainfall-recharge category is perfectly overlapped by the “good” groundwater potential zone. Similarly, 75 and 82% of the area under “moderate” and “low” groundwater potential zones is overlapping with “moderate” and “low” classes of the rainfall-recharge. These findings very clearly revealed that the delineated groundwater potential zones are effectively captured by following the methodology adopted in this study.
| S. no. | Lithology | Rainfall infiltration factor* | Area (km2) | Area (%) |
|---|---|---|---|---|
| 1 | Flood Plain and Channel Deposit | 0.150 | 512.8 | 29.7 |
| 2 | Flood Plain and Levee Deposit | 0.125 | 592.0 | 34.3 |
| 3 | Granite, Leucogranite, Quartz Porphyry | 0.075 | 142.1 | 8.2 |
| 4 | Sand Sheet and Dune Deposit | 0.225 | 279.0 | 16.2 |
| 5 | Calc Schist/Gneiss, Marble and Biotic Gneiss | 0.065 | 7.9 | 0.5 |
| 6 | Flood Plain | 0.165 | 90.8 | 5.3 |
| 7 | Channel Fill Deposit | 0.175 | 100.7 | 5.8 |
- Source: *modified after GEC.53
It is observed that the behavior of groundwater levels is not uniform in response to the rainfall. The area showed the declining groundwater level after the rainfall due to poor response of immediate recharge. This is a common phenomenon of such areas with much pressure of groundwater withdrawal in comparison to recharge. It is very difficult to collect information of groundwater withdrawals every year in developing countries like India. Therefore, the RIF method can be an easier and better option for estimation of groundwater recharge in data scarce regions, and the same is demonstrated adequately in this study. The portion of study area responsible for “high” recharge potential are required to emphasize on groundwater recharge programs including surface and subsurface recharge techniques for rejuvenating the depleted regions. The areas with “low” groundwater potential are required to promote the use of harvested rainwater prior to more evaporation losses in such areas.
3.4 Results of sensitivity analysis
The sensitivity analysis is accomplished through removal of each theme one-by-one from 11 themes, and then integrating the normalized weights of the themes and their features to compute the GPI. Thus, the process of map removal is repeated 10 times for 11 themes. Values of the spatial variation index (VI) indicating relative variation in two GPIs is presented in Table 5 for removal of every theme. It is seen from Table 5 that the minimum values of the VI are more than 4% for three themes, that is, geomorphology, geology and soil. Likewise, the maximum values of the VI are more than 40% for three themes of topographic elevation, slope length and drainage density. Range of the VI is obtained as the maximum (48.16%) for topographic elevation and the minimum (2.56%) for soil. The large variability of the VI over the area is reflected for the themes of drainage density, post-monsoon groundwater level, and mean annual rainfall with more than 4% of the SD values. However, based on the mean values of the VI, all themes can be categorized into (i) highly sensitive (VI > 10%), (ii) moderately sensitive (VI = 5%-10%, and (iii) less sensitive (VI < 5%) classes from the groundwater potential aspect. Thus, it is revealed that three themes, that is, geomorphology, post-monsoon groundwater level and geology, are highly sensitive for occurrence of the groundwater potential. Similarly, three themes having moderate sensitivity towards the groundwater potential are drainage density, soil and topographic elevation. Geology and soil are found most sensitive to good groundwater potential in earlier study.56 Whereas, the remaining five themes are less sensitive for the groundwater potential mapping.
| Layer that was excluded | Minimum (%) | Maximum (%) | Range (%) | Mean (%) | SD (%) |
|---|---|---|---|---|---|
| Geomorphology | 10.46 | 15.27 | 4.82 | 13.74 | 0.57 |
| Post-monsoon groundwater level | 0.50 | 28.14 | 27.64 | 10.68 | 5.95 |
| Geology | 8.44 | 11.89 | 3.46 | 10.39 | 0.22 |
| Drainage density | 0 | 41.11 | 41.11 | 6.76 | 6.93 |
| Soil | 4.84 | 7.42 | 2.58 | 6.12 | 0.20 |
| Topographic elevation | 0 | 48.16 | 48.16 | 6.05 | 3.75 |
| Mean annual rainfall | 0 | 16.62 | 16.62 | 4.94 | 4.15 |
| Percentage slope | 0.01 | 26.89 | 26.88 | 2.90 | 1.66 |
| Pre-monsoon groundwater level | 0.01 | 26.89 | 26.88 | 2.90 | 1.66 |
| Slope steepness | 0 | 19.47 | 19.47 | 2.48 | 1.40 |
| Slope length | 0 | 46.42 | 46.42 | 2.46 | 2.69 |
The themes such as slope parameters (percent slope, slope steepness and slope length) are not found much influenced with availability of groundwater, as the variation in slope parameters is not much more over most of basin area and only a small portion in the upper reaches contains varying values of slope parameters. Another important finding of the sensitive analysis is that pre-monsoon groundwater level is found very less significant in comparison to post-monsoon groundwater levels while pattern of weight assignment is taken same for both the themes. Similar findings are reported in the earlier studies.15, 25, 55 Knowledge about the most-sensitive themes having major influence on the occurrence of the groundwater potential is imperative for managing this vital and scarce resource in sustainable manner over drylands of arid and semi-arid regions of the world. Due to availability of sandy and loamy soil type, the adequate amount of water for irrigation purpose could not be stored in surface water reservoirs, even when the good amount of rainfall is experienced in the area. The rainwater flowing on gentle slope with highly permeable soil increases percolation rate of the soils, which is found responsible for recharging groundwater.60 This finding is also supported by the results of present study as geomorphology is found most sensitive parameter for delineating groundwater potential. Thus, the approach for delineating the groundwater potential through the recharge capacity of the aquifer is also found acceptable.
3.5 Spatially-distributed parsimonious-GPI
Six themes identified as moderately to highly sensitive (VI > 5%) through sensitivity analysis are used for geospatial modeling to generate the PGPI map. Spatial distribution of the PGPI classified into classes of (i) “good” potential (PGPI = 2.3-3.1), (ii) “moderate” potential (PGPI = 1.5-2.3), and (iii) “poor” potential (PGPI = 0.1-1.5) according to the occurrence of groundwater is shown in Figure 7. Zones of “good” potential for the groundwater occur in 508.7 km2 (29.5%) area located in the central-north portion of the area. A major portion of the area, that is, 63.3% (1091.3 km2), is occupied by the “moderate” potential zone of the groundwater. Whereas, the “poor” potential for the groundwater is seen in very small area (125.3 km2, 7.3%) towards the lower reach of the basin.
3.6 Accuracy of parsimonious-GPI map
In order to assess the accuracy of the “parsimonious”-GPI map (Figure 7) generated from six themes, it is compared with the original GPI map (Figure 4) generated from 11 themes. It is evident from comparison that the areas containing “good,” “moderate,” and “poor” potential zones of the groundwater have more or less similar geographical extent in both the GPI maps. Hence, the accuracy of the “parsimonious”-GPI map is qualitatively validated. The minor deviation in both the GPIs is due to slight difference in extent of the “good” groundwater potential zone (original GPI = 589.2 km2, 34.1% and “parsimonious”-GPI = 508.7 km2, 29.5%). Furthermore, the “parsimonious”-GPI or PGPI map is quantitatively verified through overlay analysis by crossing it with the original GPI in GIS platform. Result of the crossing of two GPI maps using union tool showing four combinations of similar and dissimilar results is shown in Figure 8. It is seen from Figure 8 that the results of both the original GPI and “parsimonious”-GPI are in strong harmony to each other in majority of the area, that is, 86.7% (1496.4 km2). This finding further verified the accuracy of the generated “parsimonious”-GPI map. Thus, it is evident that the PGPI is the cost-effective and time-efficient method over the original GPI as it involves relatively less number of observed data that saves both financial resources and time. Also, the PGPI may be a suitable method to assess groundwater potential under the condition of data scarcity that is often experienced in the developing nations.
4 CONCLUSIONS
- Integration of geospatial modeling and multicriteria decision analysis delineated “good,” “moderate” and “poor” zones of groundwater potential in 34.1%, 54.1% and 11.8% area, respectively.
- The study area is divided into three zones of rainfall-recharge. The “moderate” rainfall-recharge zone shares the maximum percentage (60%) of the total area. The “good” and “poor” rainfall-recharge zones extend over 22.1% and 17.9% of the area, respectively.
- Areas under the “high,” “moderate,” and “low” rainfall-recharge zones are found in good agreement with “good,” “moderate” and “poor” groundwater potential zones in 88%, 75%, and 82% area, respectively. This finding validated the delineated groundwater potential zones.
- Relatively high values of variation index (>10%) indicated highly-significant influence of geomorphology, post-monsoon groundwater level and geology on the occurrence of groundwater potential. Likewise, three themes, that is, drainage density, soil and topographic elevation, had moderately-significant influence (variation index = 5 to 10%) on the groundwater potential.
- From “parsimonious” approach, the maximum area (63.2%) is under “moderate” groundwater potential zone. The “good” and “poor” potential zones cover 29.5% and 7.3% area, respectively.
- Groundwater potential zones delineated from 11 themes and “parsimonious” approach are found in good harmony to each other in 86.7% area.
This study demonstrated that the multicriteria decision analysis technique can be successfully integrated with remote sensing and geographical information system techniques following a “parsimonious” approach for an accurate and cost-effective assessment of groundwater potential under the limited data-availability conditions. The methodology and results obtained in this study can be easily adopted for sustainable planning, development and management of the precious groundwater resources in other data scarce regions of the developing nations. The validation of groundwater potential is somewhat a challenging task in the areas of continuously decreasing groundwater table, and the same is the major limitation of this study. This study emphasizes the use of well-yield data for validation of the GPI in future research. This study demonstrates application of the developed PGPI in a small basin in this study, which may be further extended at large scale up to regional level studies in future. In future, the work may be extended for developing region-specific technologies for rejuvenation of groundwater resources based on most influential factors responsible for availability of groundwater. The extension of this study may also be in the direction of developing policy guidelines and future strategies for monitoring the groundwater potential at frequent intervals for balancing the extraction and recharge capacity of aquifers in water scarce regions of the world.
ACKNOWLEDGEMENTS
Authors gratefully acknowledge the Central Ground Water Board, Ahmedabad and State Ground Water Department, Gandhinagar, Gujarat for providing groundwater level data. They are also grateful to Editor-in-Chief and four anonymous reviewers for their meticulous comments that helped to improve the earlier versions of this article.
CONFLICT OF INTEREST
The authors declare no conflict of interest.
