Reviving the “ Ganges Water Machine ” : where and how much ? 2547

Runoff generated in the monsoon months in the upstream parts of the Ganges River basin (GRB) contributes substantially to downstream floods, while water shortages in the dry months affect agricultural production in the basin. This paper examines the potential for subsurface storage (SSS) in the Ganges basin to mitigate floods in the downstream areas and increase the availability of water during drier months. The Soil and Water Assessment Tool (SWAT) is used to estimate “sub-basin” water availability. The water availability estimated is then compared with the sub-basinwise unmet water demand for agriculture. Hydrological analysis reveals that some of the unmet water demand in the subbasin can be met provided it is possible to capture the runoff in sub-surface storage during the monsoon season (June to September). Some of the groundwater recharge is returned to the stream as baseflow and has the potential to increase dry season river flows. To examine the impacts of groundwater recharge on flood inundation and flows in the dry season (October to May), two groundwater recharge scenarios are tested in the Ramganga sub-basin. Increasing groundwater recharge by 35 and 65 % of the current level would increase the baseflow during the dry season by 1.46 billion m3 (34.5 % of the baseline) and 3.01 billion m3 (71.3 % of the baseline), respectively. Analysis of pumping scenarios indicates that 80 000 to 112 000 ha of additional wheat area can be irrigated in the Ramganga sub-basin by additional SSS without reducing the current baseflow volumes. Augmenting SSS reduces the peak flow and flood inundated areas in Ramganga (by up to 13.0 % for the 65 % scenario compared to the baseline), indicating the effectiveness of SSS in reducing areas inundated under floods in the sub-basin. However, this may not be sufficient to effectively control the flood in the downstream areas of the GRB, such as in the state of Bihar (prone to floods), which receives a total flow of 277 billion m3 from upstream sub-basins.

Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | crops and infrastructure. In these basins, upstream storage is generally the preferred solution to buffer the variability of flow and reduce floods downstream (Khan et al., 2014). Traditionally, dams are the major surface water storage structures. However, the construction of large dams requires huge investments, displaces people, submerges forests, and some of the water released is lost to non-beneficial evaporation (Pavelic 5 et al., 2012). In contrast, underground aquifers are efficient water reservoirs with minimum evaporative losses, no displacement of people or submergence of land (Bouwer, 2000;Dillon, 2005;Ghayoumian et al., 2007). For centuries, the utilization of water resources in the Ganges River Basin has been severely hampered by substantial seasonal variation in river flows. In the basin, the 10 main source of water is the (southwest) monsoon rainfall, and also the snowmelt and ice melt in the Himalaya during the summer season (Sharma and de Condappa, 2013) which is about 1170 billion cubic meters (B m 3 ). Of this, around 500 Bm 3 becomes stream flow with the rest directly recharging groundwater or returned to the atmosphere through evapotranspiration (Jeuland et al., 2013). The monsoon (between June and Introduction ural level is the best way to control floods downstream. Subsurface storage (SSS) also allows meeting water requirements during the dry months. Popular belief is that having large dams is the only option to meet the basin's water storage needs (Onta, 2001). However, contrary to that, the Ganges strategic basin assessment conducted by the World Bank (2012) found that the sustainable use of the basin's vast groundwater 5 aquifers can store far greater volumes of water compared to the potential of man-made storage in the basin, which is about 130-145 B m 3 (Sadoff et al., 2013). For instance, the estimated storage available in the shallow alluvial aquifers of eastern Uttar Pradesh and Bihar, which could be utilized in the dry season and naturally recharged in the wet season, is 30-50 B m 3 (SMEC, 2009) 10 From a purely biophysical perspective, four conditions are necessary to develop sustainable SSS solutions (that involve groundwater recharge beyond the natural levels) to tackle water scarcity and flood damage in the basin: -Existence of adequate un-met demand (e.g., for agriculture and other uses) to deplete the water pumped from the aquifers in a basin/sub-basin.

15
-Existence of adequate flows for capture during the monsoon season.
-Existence of extra space underground which can be created by pumping and depleting groundwater before the onset of the monsoon.
-Ability to actually capture the excess monsoon surface runoff to recharge that additional space created -naturally (through surface water and groundwater in-20 teractions) or artificially (through managed aquifer recharge (MAR)). Amarasinghe et al. (2015) examined the first condition above and estimated un-met demand throughout the basin under two scenarios of irrigation expansion. The main objective of this paper is to examine the second condition above, i.e., assess the potential availability of runoff, by conducting a hydrological analysis of the sub-basins of 25 the Ganges River Basin.

The model
Many models have been developed (e.g., Eastham et al., 2010;Gosain et al., 2011;World Bank, 2012) to study water issues in the Ganges River Basin (Johnston and Smakhtin, 2014). However, they are not available to the public. To overcome this re-5 striction and provide the research community with a working hydrological model for the Ganges River Basin, the International Water Management Institute (IWMI) has developed a publicly available hydrological model for the basin (Muthuwatta et al., 2014) using the Soil and Water Assessment Tool (SWAT) (Arnold et al., 1998). The model set up files can be downloaded from the website http://waterdata.iwmi.org/model_inventory. php, and used in further applications and scenario analyses in a variety of projects. SWAT is a widely used, semi-distributed conceptual hydrological model developed by the Agricultural Research Service of the United States Department of Agriculture (USDA) over the last 30 years, and is available free of charge as a public domain model (Arnold et al., 1998;Gassman et al., 2007;Sood et al., 2013). Broadly, the SWAT input data can be grouped into five categories: topography or terrain, land use, soil, land use management and climate (Neitsch et al., 2002). SWAT possesses adequate representation of processes governing hydrology and is particularly suitable for application in large river basins. In SWAT, a river basin is subdivided into a number of catchments, so that each catchment has at least one representative stream. Based on unique combi-20 nations of soil, land use and slope, the catchments were further divided into hydrological response units (HRUs), which are the fundamental units of calculation. Subdividing the watershed into areas having unique land use, soil and slope combinations enables the model to reflect differences in evapotranspiration and other hydrologic conditions. Runoff is predicted separately for each HRU and routed to obtain the total runoff for the 25

catchment.
SWAT simulates the local water balance of the catchment through four storage volumes -snow, soil profile, shallow aquifer and deep aquifer -based on the soil water 9745 Where: SW t is the soil water content minus the wilting-point water content at time t, and R t , SR t , ET t , P t , and G t are the daily amounts (in mm) of rainfall, runoff, evapotranspiration, percolation and groundwater flow, respectively, at time t. SW 0 is the initial soil 5 water content. The simulated processes include surface runoff, infiltration, evaporation, transpiration, lateral flow, and percolation to shallow and deep aquifers.

The data and model setup
The model used in this study was set up using the datasets shown in Table 1. The Ganges River Basin was delineated using 3000 ha as the minimum area threshold and has resulted in 1684 catchments (Fig. 1). The model was initially developed to study streamflow entering Bangladesh. Therefore, the spatial domain of the SWAT model developed for the Ganges does not entirely cover the areas that belong to West Bengal and Bangladesh. However, this does not affect the current study, as its focus is to assess water availability in the upstream sub-basins of the Ganges River Basin. Figure 1 shows the catchments delineated for SWAT, 22 major sub-basins (Table 2) in the Ganges River Basin and the area covering Nepal. The 19 main sub-basins in the Indian part selected in this paper are those considered by the Central Water Commission (CWC) of India, which is the main government agency responsible for water resources development and management in the Ganges River Basin. Since the focus 20 of this study is to estimate water availability in the sub-basins within India, Nepal is considered as one region. Hereafter, in this paper, "sub-basins" are referred to as the 22 major areas shown in Fig. 1, while the smaller spatial units inside those 22 subbasins and Nepal are termed "catchments". For details of the model setup, including calibration and validation, please refer to Muthuwatta et al. (2014)

Simulating sub-basin runoff
Annual time series of catchment-scale surface runoff from 1991 to 2010 were constructed by aggregating daily surface runoff simulated by SWAT. Next, using geographic information system (GIS) techniques, annual runoff time series were estimated for all sub-basins within the modeled area of the Ganges River Basin. The study uses the hy-5 drographs of the simulated runoff (SR) to estimate the 75 % dependable runoff (SR 75 ). SR 75 is an estimate of the runoff that can be expected in the basin, on average, every three out of 4 years, and is considered to be a reliable estimate of water availability for augmenting groundwater storage.

Surface runoff of the sub-basins
The spatial and temporal distribution of the annual surface runoff is analyzed to determine the water availability in different sub-basins. Streamflow includes surface runoff and baseflow from groundwater, which can be captured by diversion or from dams. Surface runoff is part of the precipitation that is left after evapotranspiration and infil- 15 tration, which can be captured for MAR before it reaches the stream. Therefore, only the surface runoff portion was considered for augmenting SSS. Figure 2 shows the simulated catchment-scale mean annual surface runoff. The surface runoff of catchments ranges from less than 0.1 B m 3 to more than 2.0 B m 3 . The statistics of the estimated surface runoff for the sub-basins is given in 20 than 80 % of the annual surface runoff in most sub-basins (Table 3, last two columns). This intra-and inter-annual variability of the flows clearly indicates the need for storages to capture the excess surface runoff during the monsoon season, which could be a SSS. For this analysis, SR 75 was used to identify the sub-basins that are consistently producing higher volumes of surface runoff. Figure 3 shows the spatial distribution of 5 SR 75 of sub-basins. Ghaghara (10) sub-basin and Nepal have, by far, the largest SR 75 . The Kali Sindh (13), Ramganga (16), Son (17) and Yamuna Lower (20) sub-basins have more than 10 B m 3 of SR 75 . The Gandak (9) also produces higher surface runoff, but the subbasin is located in the downstream area of the Ganges River Basin. Because of the 10 high monsoon runoff, the upstream sub-basins contribute substantially to flooding in the downstream areas of the Ganges River Basin. Amarasinghe et al. (2015) estimated the un-met agricultural water demand. Two scenarios were considered in the analysis (Table 4).

15
-Scenario 1 assumed that all irrigable land will be irrigated in the Rabi (November to March) and summer (April-May) seasons.
-Scenario 2 considered all cropland to be irrigated in the Rabi and summer seasons.
As of now, all the sub-basins in the Ganges River Basin have substantial un-met water 20 demand for agriculture in the dry season. Therefore, capturing a substantial portion of the surface runoff during the monsoon months can help close the gap between current supply of water and demand in the dry months. Therefore, there is potential for increasing agricultural productivity in these sub-basins with more irrigation in the dry months. In the sub-basins, the total un-met demands are 55.03 and 108.4 B m 3 under scenarios 1 and 2, respectively. The values presented in Table 4 show that, for some sub-basins, annual un-met demand exceeds the annual water availability. In these subbasins, only a part of the un-met demand can be satisfied by additional underground storage. In some other sub-basins, the un-met demand is less than 30 % of the SR 75 5 of surface runoff. These sub-basins have the potential to meet all the un-met demand with SSS. For instance, in the Ramganga sub-basin, the SR 75 of surface runoff is about 10.1 B m 3 , and approximately 83 % of this runoff is occurring during the wet season. To meet the maximum un-met agricultural water demand in the Ramganga sub-basin only requires capturing 33 % of the monsoon surface runoff.

Impact of sub-surface storage on flood control
Floods are a recurrent phenomenon in some parts of the Ganges, such as the State of Bihar, which is located in the middle part of the basin (Fig. 4) Spectroradiometer (MODIS) satellite data, showed that the flood inundated area can be more than 14 000 km 2 during the peak flood period (Amarnath et al., 2012). Flow coming from upstream of the Ganges plays a substantial role in floods in the state. The mean annual streamflow from the upstream sub-basins from 2001 to 2010 was estimated and is presented in Fig. 4. indicates that the water volumes received from upstream flows are more than twofold the amount of rainfall in Bihar. As presented in Fig. 3, five sub-basins located in the upstream part of the Ganges River Basin consistently produce higher surface runoff compared to other sub-basins. These sub-basins are Ghaghara (10), Son (17), Yamuna Lower (20), Ramganga (16) 5 and Kali Sindh (13). Figure 5 shows the relationship between the outflows from four upstream sub-basins with inflow to the State of Bihar. Coefficient of determinant (R 2 ) in each graph indicates that annual outflows from the upstream sub-basins are strongly correlated with the inflows to the State of Bihar. R 2 for Kali Sindh is 0.71 and not presented in Fig. 5. 10 Flow from Ghaghara and Yamuna Lower sub-basins is approximately 30 % of the total inflow from the upstream Ganges River Basin to Bihar. The contributions from Son, Kali Sindh and Ramganga are 17, 10 and 7 %, respectively. This implies that, by capturing a portion of the upstream flows during peak runoff periods would reduce the flow to Bihar, which creates floods during the wet season.

Conclusions
Creating additional SSS beyond the current levels in the Ganges River Basin can simultaneously enhance water supply for beneficial depletion and control downstream floods. Water availability analysis conducted and based on time series of simulated surface runoff using SWAT showed that annual total surface runoff generated in the 20 Ganges River Basin is about 298 ± 99 B m 3 , and runoff in the monsoon months contributes to 80 % of this total runoff. The sub-basin-wise mean annual surface runoff ranges from 2. 24  runoff with the estimated un-met water demand indicated that capturing only a portion of the wet-season runoff would be sufficient to provide water to irrigate all the irrigable land in the dry months. Further analysis revealed that the annual surface runoff from the upstream of the Ganges River Basin to the State of Bihar, a flood-prone area located downstream, is 5 twice the amount of rainfall in the same area. Sub-basin-wise streamflow analysis in the Ganges River Basin showed that approximately 60 % of the upstream flow to Bihar comes through the Ghaghara and Yamuna Lower sub-basins. This runoff contributes to the recurrent floods in Bihar. As shown in Fig. 5, there are strong linear correlations between annual outflows from the upstream sub-basins and the inflow to the State of 10 Bihar. This suggests that SSS upstream has the potential to control floods downstream, by capturing a portion of the surface runoff during the wet season in the upstream subbasins.
This study only discusses the surface water availability for SSS, and further analysis is needed to ascertain the storage capacity of the aquifer and how much additional stor- 15 age capacity can be created by pumping groundwater during the dry months. Further, a detailed analysis of the soil, topographic and geological characteristics is required to determine the suitable areas for groundwater recharge.
Finally, it is pertinent to understand the interactions between groundwater and surface water in the sub-basin. This requires coupling a groundwater-surface water model 20 to run some scenarios to investigate the effect of pumping and recharging of groundwater on the hydrology of the Ramganga sub-basin. Introduction