2.1 Groundwater-level monitoring

Seasonal (i.e., quarterly or four times per year) ground-based monitoring data (n>15 000) for 20 years (from 1996 to 2015) were a part of the Central Ground Water Board's (CGWB, Government of India) groundwater monitoring mission. Most (∼85 %) of the studied wells represent shallow, unconfined aquifers (Central Ground Water Board, 2014); however, these aquifers have not been characterized in CGWB reports and products. Initially, groundwater-level data were filtered for attaining a temporally continuous dataset in the study area for each year. The data were collected once during the months of January, May, August and November in individual years (Groundwater Resource Estimation Methodology, 1997). Outliers were removed following Tukey's fence approach (Tukey, 1977), reducing the usable number of monitoring data points to 3468. The annual change in groundwater levels was determined using the highest and lowest groundwater head elevation (highest and lowest groundwater level generally occur in post-monsoon and pre-monsoon times, respectively) for each year. Finally, gridded $(\mathrm{0.1}{}^{\circ }\times \mathrm{0.1}{}^{\circ })\mathrm{\Delta }h$ data were obtained following the ordinary kriging technique.

2.2 Groundwater recharge estimation using the water-table fluctuation (WTF) method

One of the major challenges is the absence of high-resolution, observed data of climatic parameters and other controlling factors that could be used in groundwater recharge measurement. High spatiotemporal variability of groundwater recharge is another concern for the regional-scale recharge estimates (Healy, 2010). In the absence of field-scale data to estimate recharge, simpler methods, such as the water-table fluctuation (WTF) technique, are more widely used to estimate recharge (R_g; Healy and Cook, 2002). R_g may be defined as

where Δh represents seasonal changes in groundwater levels.

Specific yield (S_y) of the aquifer material is one of the important components for the recharge estimation. S_y information was retrieved following the Central Ground Water Board (2012) and Bhanja et al. (2016). Mean S_y values vary between 0.02 and 0.13, varying with identified aquifers (Figs. 3 and 4a). The IGB basins are heavily irrigated, with intensive groundwater withdrawals (Fig. 4b; Supplement Fig. S1) relative to other parts of the country.

Figure 4Maps of (a) specific yield data (modified from Bhanja et al., 2016). Blank areas represent areas of no data availability; (b) state-wide groundwater abstraction for irrigation (mm; estimates are for the year 2009; Central Ground Water Board, 2012).

Download

2.3 Field-scale recharge estimation using tritium injection approach

A tracer in the form of ions, isotope or gases move with water and can be tracked to estimate water infiltration through unsaturated zone (Healy, 2010). Here, the tritium injection method has been used to estimate recharge rates in 52 locations across India (Fig. 1). The equivalent depth of water of the peak movement can be represented as the total infiltrated water flux (Healy, 2010). The piston-flow method, which has been a major assumption in tracer movement technique, is linked with the vertically downward movement of water and dissolved solutes without mixing and change in velocity (Healy, 2010). Tritium injection is performed during the pre-monsoon seasons, and the soil core is collected after the monsoon in order to estimate the monsoon recharge rate (Rangarajan and Athavale, 2000). More details on the tritium injection method can be found in the Supplement and in Rangarajan et al. (2000).

2.4 Non-linear trend analysis

Non-linearity in the recharge and precipitation data can be represented through Hodrick–Prescott (HP) trend analysis (Hodrick and Prescott, 1997), which separates cyclical components present (c_t) in the data (y_t) from the trend (T_t) after solving the following equations:

where λ is a constant (Hodrick and Prescott, 1997; Ravn and Uhlig, 2002).

Figure 5Maps showing gridded ($\mathrm{0.1}{}^{\circ }\times \mathrm{0.1}{}^{\circ }$) groundwater recharge (mm yr⁻¹) calculated using WTF method. Blank areas represent areas of no data availability.

Download

Figure 6Map of mean recharge in the study period (1996–2015). Blank areas represent areas of no data availability.

Download

3.1 Recharge estimates from the water-table fluctuation method

Calculation of the R_g, based on water-level data from 1996 to 2015, exhibits both spatial and temporal variability over the years (Fig. 5). In general, the years of lowest groundwater recharge, i.e., 2002, 2004, 2009 and 2014, correspond to years of lowest precipitation (Figs. 2 and 5). An R_g exceeding 300 mm yr⁻¹ is found over most regions of the IGB basin (Figs. 5 and 6). Central and southern India have been subjected to comparatively lower recharge rates (< 200 mm yr⁻¹). Basin-wide data show highest recharge rates in basins 1, 2a, 2b, 2c, 10, 11, 14 and 20 (Fig. 6; Table 1); parts of the basins are composed of alluvial sediments that are also intensively irrigated and are linked to groundwater withdrawal. The highest annual recharge rate was calculated for 2010 in most basins. High recharge rates in 2010 are attributed to increased space for recharge related to lower precipitation in 2009 (Fig. 2), with 22 out of 26 meteorological subdivisions declared as rainfall deficit in India (National Climate Centre, 2013). Linear regression analysis of recharge rates between 1996 and 2015 show simultaneous occurrence of increased and decreased trends in recharge rates in different parts of India (Fig. 7). Parts of the Ganges basin (basin 2a) have been experiencing rapid declines in recharge rates, while parts of western and northwestern India are experiencing increases in recharge in the study period (Fig. 7). Non-linear trend analysis shows high temporal variation in recharge in all of the studied basins (Fig. 8); maximum recharge amplitudes are found in basins 2b, 2c, 11 and 14 (Fig. 8). The basins are located in comparatively higher precipitation zones of northeastern and western–coastal India, respectively.

Figure 7Maps of positive and negative trends of groundwater recharge in 1996–2015. Basin boundaries are overlaid. Blank areas represent areas of no data availability.

Download

Figure 8Basin-wide estimates of Hodrick–Prescott trend analysis of R_g (mm yr⁻¹) and precipitation (mm yr⁻¹). For basin locations, see Fig. 1.

Download

Table 2Field-scale recharge (R_n) estimates across 52 locations (Fig. 1), river basin ID number, hydrogeology and prevailing precipitation rates (mm yr⁻¹).

Download Print Version | Download XLSX

Figure 9Basin-wide box-whisker plot of precipitation (blue) and recharge rates (black). Symbols representing the mean, median, maximum value (max), minimum value (min), 10th percentile, 25th percentile, 75th percentile and 90th percentile are shown in inset of the bottom figure.

Download

3.2 Groundwater recharge estimates as a function of climate, hydrogeology and irrigation

Comparatively higher rates of precipitation partly explain the high recharge rates in the IGB basin. Precipitation data show high annual variability in all of the basins (Fig. 9). Highly fertile sedimentary formations of IGB basin facilitate both direct and indirect recharge. Higher agricultural groundwater withdrawal in the IGB basin (Fig. 4b) leads to decreases in water storage, which can result in increased recharge by generating more recharge space. Subsequently, recharge rates are not homogenous throughout the IGB basins (Figs. 5 and 6). The unconsolidated formations of the IGB basin are highly transmissive for the water flow (transmissivity values varies from 250 to 4000 m² day⁻¹; Fig. 3). Horizontal hydraulic conductivity values are found to be higher in the IGB basin (hydraulic conductivity varies from 10 to 800 m day⁻¹; Fig. 3). As a result, intense groundwater withdrawal at a region within the IGB basin would have a profound impact on the groundwater storage on the surrounding regions (particularly the areas within the periphery of the pumping influence). This would facilitate the creation of the additional recharge space within the entire region. In contrast, comparatively lower transmissivity and hydraulic conductivity values in aquifers of western, central and southern India restrict water to flow in the horizontal direction (Fig. 3). This inhibits the horizontal flow of water after a pumping event in those regions. Basin-wide mean recharge rates are found to be variable over the years (Fig. 9). Highest interannual variability has been obtained in basins 2b, 2c and 14 (Fig. 9); the basins are also experiencing highest precipitation rates (Table 1; Fig. 9). In contrast, basins located in Indian craton, i.e., basins 3, 4, 5 and 19, exhibit lowest interannual variability (Fig. 9). Lower recharge rates are found in central and southern parts of India (Figs. 5 and 6). The crystalline aquifers in these regions (Mukherjee et al., 2015) are not conducive enough for precipitation-based infiltration through the subsurface. The observation is consistent with Sukhija et al. (1996), who also found a lower recharge rate in fractured regions. Deeply weathered, lateritic soils that have developed on cratons often have low-matrix permeabilities due to concentration of kaolinite and development of ferricretes (laterites; Taylor and Howard, 1999). These low-matrix permeabilities promote infiltration via discontinuities. Soil permeabilities are substantially less than those of the sorted alluvial soils in the IGB basin.

3.3 Comparison with field-scale recharge estimates

We present field-scale recharge rates (R_n; Fig. 1) derived from the tritium injection approach (Rangarajan et al., 2000) in Table 2. The recharge rates were estimated in four different land use types, i.e., granite, basalt, sediment and alluvium. R_n varies within a range of 22–179 mm in locations within granite setting and the values reach 46–171 mm in basalt regions (Table 2 and Fig. 10). Recharge rates are found to be comparatively higher in the sediment and alluvial regions, with rates of 29–181 and 20–198 mm, respectively (Table 2 and Fig. 10). Similarly, R_g values also exhibit higher recharge in sediment and alluvium regions (Fig. 10).

Figure 10Comparison of groundwater recharge rates obtained from the present study (annual range) and some of the earlier studies conducted over the point locations shown in Fig. 1. Recharge estimates exceeding 700 mm yr⁻¹ data are indicated on the top of the column.

Download

Values of R_g from this study compare favorably with site-scale determinations in 42 out of the total 52 locations (Fig. 10). However, the R_g was found to be larger than the recharge rates in location identification (ID) numbers 44, 45, 49 and 50 in Figs. 1 and 10). Three out of four locations are located in IGB basin, which experiences intensive groundwater abstraction for irrigation (Fig. 4b). This discrepancy may be coming from the assumption that direct recharge is only through the unsaturated zone traced by radioisotopes (Healy, 2010). Also, the R_n includes only monsoon recharge, while the R_g includes both monsoon and pre-monsoon recharge that is dominated by the irrigation return flow. Hence, contributions from indirect recharge via irrigation return flows and recharge amplified by dry-season abstraction for irrigation are ignored in R_n estimates. R_g was found to be lower than R_n in some of the locations (location numbers 10, 11, 25, 28, 40 and 46 in Figs. 1 and 10). Each of the six locations is subjected to comparatively lower rates of irrigation-linked groundwater withdrawal (Fig. 4b). Therefore, the irrigation return flow and creation of additional recharge space are almost negligible in these locations. Furthermore, the locations are also experiencing comparatively higher rates of precipitation (Table 2). As the R_g will not consider a fraction of recharge particularly during aquifer full condition (Healy and Cook, 2002), it is likely the major reason for the observation of lower R_g values than R_n values in these six locations.

3.4 Groundwater recharge as a function of precipitation

Recharge rates are significantly (p value <0.01) correlated with precipitation in 10 out of the 22 basins. Non-linear trend analysis between basin-wide recharge rates and precipitation show good match in basins 2a, 3, 4, 12 and 20 (Fig. 8). In contrast, recharge and precipitation trends are negatively correlated in basins 2b, 2c, 7, 16 and 18 (Fig. 8). In order to study the relationship in more detail, we used the Granger causality analysis (Granger, 1988). Results show precipitation significantly (p value <0.01) causes R_g in six basins, i.e., 8, 11, 12, 13, 15 and 16. Most of these basins are not intensively irrigated; groundwater withdrawal rates are found to be lower in these regions, and irrigation groundwater withdrawal was found to be less than 75 mm in all of the basins (Fig. 4b). Therefore, natural processes, i.e., precipitation, still influence recharge rates in those basins. Alternatively, the relationship between recharge and precipitation is statistically insignificant in other basins, which are more intensively irrigated (50 to more than 300 mm; Fig. 4b). As a result, precipitation influence is found to be less dominant in recharge in these regions.

3.5 Assumptions and limitations

Minimal reliance on assumptions in measurement is an advantages of using the WTF method (Healy and Cook, 2002). Furthermore, the method can be followed to estimate recharge for a large region, simultaneously (Healy and Cook, 2002). Uncertainty in groundwater storage coefficients (e.g., S_y) influences the magnitude of computed recharge. In the WTF method, recharge rates exhibit minimum values in groundwater discharge regions. The antecedent recession during the peak groundwater level (Healy and Cook, 2002), which is a component of base flow and discharge, is difficult to measure accurately and hence results in underestimation of recharge during that time.