Climate variability, rainwater-harvesting structures and groundwater levels in Odisha, India: an empirical analysis

ABSTRACT

To examine the driver of groundwater depletion, this paper estimates a dynamic panel regression model using district-level secondary data from 1995 to 2017. Results show that the number of rainwater-harvesting structures, annual average rainfall, forest cover and net sown area improve groundwater levels. Tube well irrigation and rice cultivation can adversely affect groundwater levels. The positive relationship between surface water and improved groundwater level is consistent in all the estimated models. Hence, efficient use and management of surface and groundwater are crucial for the long-term sustainability of water resources.

KEYWORDS:

• Climate variability
• groundwater
• rainwater-harvesting structures
• SDG 6: Clean water and sanitation
• Odisha
• India

Introduction

India’s groundwater resources have been under tremendous stress for years because of overuse and over-extraction due to increasing requirements from the growing population for water, urbanization and economic growth (Tortajada et al., 2018). Rapidly changing climatic conditions have further depleted groundwater (Shiferaw et al., 2023; Swain et al., 2022). More specifically, about 239 billion cubic metres (bcm) of groundwater are extracted annually to meet various water requirements, including irrigation and domestic water supplies, making the country one of the biggest users of this resource in the world (CGWB, 2022a). According to the CGWB (2022a), more than 62% of farmland in India is irrigated through groundwater, which is critical in ensuring food security. Similarly, Mukherji (2022) observed that, globally, groundwater supports around 70% of the water supply in arid and semi-arid areas. Hence, the overexploitation of groundwater has led to the irreversible depletion and degradation of groundwater (Prasad & Rao, 2018; Sekhri, 2022). More than 60% of districts in India have exhibited groundwater depletion, which endangers India’s food security (Chowdhury & Behera, 2018). Jain et al. (2021) predicted that by 2025 up to 68% of areas in India would lose groundwater resources, which could reduce food crop production by up to 20% across the country.

Many empirical studies identified multiple factors responsible for the rapid decline of groundwater in India (Dangar et al., 2021; Roy et al., 2022). The main factors include the introduction of high-yielding crop varieties that are highly water-intensive such as rice (Jain et al., 2021), the availability of electric and diesel pumps for irrigation (Mishra et al., 2023; Mukherji, 2022), the expansion of rural electrification (Sarkar & Das, 2014), and the supply of subsidized and/or free electricity to farmers (Tortajada et al., 2018). In addition, the use of private wells in tank areas (surface irrigation water source) also significantly contributes to this groundwater extraction process (Amarnath & Raja, 2006). The installation and operation cost of pumps and private wells has considerably declined in recent decades, further incentivizing farmers to establish their own private wells (Kajisa et al., 2007; Palanisami et al., 2008). Furthermore, the degradation of rainwater harvesting systems, such as traditional water bodies and rivers, has also compelled farmers to instal private wells to meet the growing water demand in their fields (Balasubramanian & Selvaraj, 2003).

Groundwater depletion occurs when the recharge rate is lower than the groundwater extraction rate. Interestingly, it is seen that the groundwater recharge rate is dependent on several composite elements of the hydrological systems regulated by key natural and anthropogenic factors (Bhatti et al., 2020; Li et al., 2021; Shah et al., 2022). For instance, groundwater recharge is directly influenced by the hydraulic properties of soils, climatic parameters, irrigation practices, and cropping intensity and patterns (Freeze & Cherry, 1979). In addition, the volume of precipitation and the rate of evaporation through surface temperatures are also among the key natural factors that influence groundwater recharge (Atoui & Agoubi, 2022; Roy et al., 2022). For instance, Srivastava et al. (2017) observed that precipitation positively correlates with groundwater recharge. In this context, even more noteworthy results are reported by Jan et al. (2007), which indicate that a lower intensity rainfall for a longer time could result in more groundwater recharge than a shorter duration with high-intensity rainfall.

Forests and vegetation cover are other key variables that enormously influence groundwater recharge (Bhattacharjee & Behera, 2018). For instance, Tobella (2016) noted that the size of trees and their spatial distribution influence the magnitude of groundwater recharge significantly. It is found that higher groundwater recharge is detected in the areas with dense forest in the dry tropical region compared with the areas with less or without forest cover. Studies also mention the adverse impact of rapid urbanization on groundwater, variations in the groundwater recharge cycle and its depth, and groundwater contamination (Foster et al., 1993; Vázquez-Suñé et al., 2005).

Several studies explain the interactions between surface water and groundwater (Crosbie et al., 2023; Geris et al., 2022; Ledford et al., 2022; Wang et al., 2023). It is observed that increasing the productivity of well water through groundwater recharge can benefit farmers whose agricultural lands are located close to tanks (manufactured structures with horse-bow shapes having about 1–2 to 50–100 acres of land size; Shah & Raju, 2002). It is also observed that because of the prevalence of prolonged drought conditions in tank-dominated regions, people are mostly dependent on wells for water; hence, the water-holding capacity of tanks is reduced, which causes degradation of the tanks and may lead to loss of groundwater recharge (Raju & Reddy, 1998).

The eastern state of Odisha has experienced an acute problem of groundwater shortage, especially between 2009 and 2017, as reported by the State Government of Odisha in its several reports on water resources (Government of Odisha, 2017, 2019, 2021). In addition, the Central Ground Water Board (CGWB) has further indicated that there have been instances of groundwater depletion in 24 out of 30 districts of Odisha (CGWB, 2021). According to a report by the CGWB (2021), seven districts in Odisha have reached a critical level of groundwater depletion. According to Badiani-Magnusson and Jessoe (2019), agricultural electricity subsidies lead to groundwater exploitation. These subsidies encourage the production of water-intensive crops such as rice, the staple diet in eastern India, including Odisha. In addition, Odisha has rich natural resources such as mineral and forest resources (Hota & Behera, 2015). The mineral industries and their competing use for water have resulted in groundwater depletion and deforestation in Odisha (Hota & Behera, 2015; Mishra et al., 2023). Odisha has also witnessed considerable population growth (having a decadal growth rate of 13.97%) and rapid urbanization (about 26.8% of urban growth), which has adversely impacted water resources (Government of Odisha, 2018, 2021).

Studies show that decreased rainfall due to climate change has also largely contributed to the depletion of groundwater in India (Roy et al., 2022; Shiferaw et al., 2023; Swain et al., 2022). Rice has a quadratic relationship with the amount of rainfall (Bhargava, 2018); hence, climate variability is widely associated with rice production, a staple food for the people of Odisha. As cited in the Groundwater Yearbook 2020–21, the average rainfall in Odisha has decreased over the years, right from 1980 to 2020. For instance, in 2022, Odisha faced more depletion in groundwater level due to a 47% rainfall deficit in the last two quarters of the year (CGWB, 2022b). This situation has appeared mainly due to the lack of moisture-based wind flow from the Bay of Bengal (CGWB, 2022b).

Hence, against the backdrop of rapidly changing climatic conditions, groundwater depletion, and a decline in crop productivity in Odisha, this study attempts to establish the links between tanks and groundwater in Odisha by classifying and/or factoring in vulnerable regions such as tribal, coastal and drought prone. Several empirical studies explore the link between traditional water bodies and groundwater depletion using macro-level data in India in general, and in Odisha in particular (Swain et al., 2022; Watto et al., 2018; Yan et al., 2022). The present paper classifies regions based on various socio-economic and ecological vulnerabilities and examines the linkage between groundwater and traditional water bodies to understand the variability of groundwater sustainability across the regions. The findings of the study contribute significantly to the current literature on factors influencing groundwater depletion and sustainability across socio-economically and ecologically vulnerable areas to ensure food security in Odisha.

Study area description

Odisha is located in eastern India from 17°31ʹN to 22°31ʹN latitude and from 81°31ʹE to 87°.29ʹE longitude (Singandhupe & Sethi, 2016). The state has 30 administrative districts spread over an area of 155,707 km². In the present study, 29 districts have been included (Kalahandi district was dropped from the study due to lack of data). The state has a moist and subhumid climate with an average annual rainfall between 1000 and 1800 mm (Guhathakurta et al., 2020). The state receives most (80%) of its rainfall (1160 mm) from the south-west monsoon, which continues from June to September. A spatial disparity in the amount of rainfall is observed across the districts in the state. For example, more rainfall (1600–1800 mm) is recorded in the northern plateau than in the southern coastal plains (1100–1400 mm) (Guhathakurta et al., 2020). The average annual temperature in the state also varies widely, ranging from 15 to 38°C.

Odisha has about 16.69 bcm of groundwater resources available (Government of Odisha, 2017). However, according to the CGWB (2021), the groundwater level in the state varies widely across the districts: 0.11 m below ground level in the Anugul district to 15.60 m below ground level in the Koraput district. It is also observed that out of 30 districts, groundwater extraction in 21 districts increased from 2017 and 2020, resulting in a decline in groundwater levels in the state (Government of Odisha, 2021). Rice, a water-intensive crop, is the main crop grown in the state (Panigrahi & Pathak, 2018) with around 77.7% of the total cropped area under rice cultivation. The coverage under rice cultivation is about 4.12 million ha and 0.3 million ha during the kharif (crops grown between April and October) and rabi (crops grown between November to March) seasons, respectively (Singandhupe & Sethi, 2016); therefore, to improve the livelihood of smallholder rice farmers and to ensure food security in the long run, efforts are being made to scale sustainable agriculture practices that include efficient use of water resources (Mohapatra et al., 2022; Spiertz, 2009). As Odisha’s agriculture sector contributes to 20.61% of the state’s gross domestic product (GDP) and employs about 48.3% of the state’s total workforce (Government of Odisha, 2022), groundwater level depletion and water stress would have a devastating effect on Odisha’s livelihood and food security.

Table 1 reports the district-wise rainfall, forest cover, rice cultivated area, net area sown, areas sown more than once, groundwater depth, district GDP, number of tanks and tube well-irrigated area. Overall, it shows a significant variation in groundwater depletion across the state, which provides a strong motivation to explore the factors that influence the groundwater levels in the state and whether rainwater-harvesting infrastructure is likely to play any role in improving the groundwater levels.

Table 1. District-wise information relating to agriculture land-use practices, natural resource endowments and economic status in Odisha (average 1995–2017).

Districts	Area sown more than once (thousands ha)	Net area sown (thousands ha)	Forest cover (thousands ha)	Rice cultivated area (thousands ha)	Groundwater level (mbgl)	Tanks (thousands)	Tube well-irrigated area (thousands ha)	Rainfall (mm)	District GDP (Rs. thousands lakh)
Anugul	69.12	153.50	258.00	107.25	254.36	6.93	0.87	1277.19	540.03
Balangir	73.17	298.09	138.77	238.60	3.93	3.34	2.72	1232.67	254.93
Baleshwar	67.61	219.29	29.17	247.22	3.66	9.68	1.02	1604.97	349.04
Bargarh	97.31	300.34	114.43	333.54	3.10	15.88	13.63	1195.66	227.50
Bhadrak	35.58	167.75	7.76	186.74	2.99	1.32	0.20	1419.97	191.95
Boudh	25.67	76.38	111.97	74.17	3.73	0.93	0.57	1171.35	66.36
Cuttack	94.75	153.79	70.34	157.01	2.72	2.43	1.01	1486.03	547.59
Deogarh	20.89	56.81	138.86	52.80	3.90	2.34	0.36	1254.94	51.40
Dhenkanal	56.05	137.92	157.38	116.87	4.24	3.07	0.93	1333.41	209.49
Gajapati	32.97	67.21	216.82	40.56	4.30	31.09	11.14	1181.16	96.57
Ganjam	178.30	331.15	274.50	285.84	3.09	59.53	38.65	1177.70	565.37
Jagatsinghapur	62.23	93.02	11.44	99.34	2.01	0.81	0.16	1478.48	255.26
Jajapur	75.22	143.01	61.26	140.96	3.09	2.58	0.80	1478.15	360.78
Jharsuguda	15.56	63.50	23.47	58.91	3.88	6.31	1.15	1314.11	220.86
Kalahandi	123.97	304.39	263.64	274.28	3.75	3.03	1.26	1364.10	222.47
Kandhamal	30.99	89.60	498.49	52.39	6.40	0.42	0.02	1319.36	170.41
Kendrapara	71.27	138.79	21.64	149.07	2.34	1.95	0.47	1516.00	178.81
Kendujhar	74.04	243.49	278.82	209.76	4.22	1.66	1.10	1400.99	578.65
Khordha	55.51	116.88	62.06	120.33	4.05	2.48	0.48	1308.11	550.05
Koraput	65.06	242.90	173.59	139.43	4.43	1.75	0.95	1389.12	267.45
Malkangiri	37.11	126.60	294.63	102.55	3.44	0.26	0.09	1301.53	80.20
Mayurbhanj	57.53	355.86	387.21	346.05	4.85	19.67	1.98	1502.59	355.94
Nabarangpur	49.48	203.29	213.68	169.58	3.82	1.93	1.05	1538.23	138.17
Nayagarh	61.88	115.95	182.63	107.17	3.62	1.23	0.39	1387.02	117.12
Nuapada	50.21	142.13	137.69	112.17	4.12	0.89	0.24	1141.24	89.00
Puri	91.80	134.98	14.01	163.43	2.32	5.66	1.74	1301.45	238.80
Rayagada	45.17	144.51	265.42	68.03	4.18	3.48	2.93	1230.36	166.14
Sambalpur	52.78	157.08	322.75	164.91	3.79	10.90	2.55	1319.58	259.36
Sonepur	53.53	105.19	37.77	145.11	2.96	1.60	0.83	1222.54	83.97
Sundargarh	43.87	263.41	439.62	239.75	4.33	3.66	1.00	1305.63	672.71

Note: GDP, gross domestic product; mbgl, Metres below ground level; 1 lakh = 0.1 million.

Sources: Minor Irrigation Census, Odisha (2017); Census of India (2011); Economic Census Odisha (2016); and INDIA–WRIS (2017).

Analytical framework, econometric model specification and description of the variables used

Figure 1 shows the analytical framework explaining various factors influencing groundwater levels in Odisha. Climate change and agricultural systems significantly influence the groundwater level in several ways, determining the region’s overall water and food security. Sufficient and regular rainfall, dense forest cover, and topography can influence groundwater levels positively (Devineni et al., 2022). The growing concern of climate change in terms of erratic precipitation, rising temperature and drought can adversely impact groundwater levels, which have severe implications for agriculture production (Dangar et al., 2021) and, hence, the food security of the people (Devineni et al., 2022). On the other hand, manufactured rainwater harvesting structures such as tanks under agricultural systems can positively influence groundwater levels if groundwater extraction is less than groundwater recharge (Madhnure et al., 2023). Under the agricultural system in the framework, both tube well irrigation and the number of tanks are intricately connected. Both surface water and groundwater resources contribute to the increase in the area irrigated more than once, rice cultivation, and the net cultivated area. Economic variables (GDP and mining) and demographic drivers (population density, literacy and tribal population) influence groundwater levels by influencing natural and agricultural systems.

Figure 1. Analytical framework explaining the factors influencing groundwater levels.

Note: GDP, gross domestic product.Source: Authors.

The groundwater table in Odisha tends to vary widely across districts and changes over the years. A dynamic panel regression model, based on the generalized methods of moments (GMM) (Arellano & Bond, 1991), using district-level secondary data from 1995 to 2017, has been estimated to identify the determinants of groundwater depletion. The GMM obtains estimators by applying the moment condition generated by the lower lagged groundwater level. The dynamic panel data models are estimated here based on the assumption that the groundwater levels of the present year would depend on those of the last year. In order to understand the role of different parameters such as climate change, the concentration of tribal population, types of ecosystems, and level of economic development and their vulnerabilities to groundwater level, the districts are broadly classified into drought and non-drought prone, tribal and non-tribal, coastal and non-coastal, and low GDP districts (below US$300 million), medium GDP districts (US$300–600 million), and high (above US$600 million). Based on the frequency of drought events across Odisha, districts are classified as drought and non-drought districts (Gulati et al., 2009). If a district has faced 10 droughts since 1901, it is classified as drought prone; otherwise, it is termed a non-drought-prone district. Since droughts were not so frequent earlier, 10 droughts are assumed to be a reasonable threshold for such classification.

Similarly, based on the 2011 census, if a district has more than 50% tribal population, it is categorized as a tribal district; otherwise, it is a non-tribal district. The following equation is applied for the estimation of the dynamic panel data models on the determinants of the groundwater levels:

Groundwater levels = f (forest cover, net area sown, area irrigated more than once, rice cultivated area, literacy rate, population density, number of tanks, tube well-irrigated area, annual rainfall, district GDP).

The rainfall trend is crucial to understand groundwater depletion (Kotchoni et al., 2018). Timely and sufficient rainfall reduces the requirement for irrigation water extraction from the ground by depleting its water level. Therefore, it is argued that increased annual rainfall may improve the groundwater level. In addition, a lagged level of annual rainfall is also used in the model because the present groundwater levels depend on the amount of the previous year’s rainfall. However, the monsoon rainfall is confined mainly to four months a year (June–September). Therefore, the presence of traditional rainwater-harvesting bodies that store rainwater and help recharge groundwater throughout the year is another important determining factor influencing the groundwater depth.

Cropping patterns and profitability of cultivation can influence the choice of farmers’ irrigation system, which may support the need for uninterrupted water supply during cropping time. Therefore, tube well irrigation has turned out to be a more reliable source of irrigation for farmers. It is hypothesized that more tube wells in the region may lead to more groundwater extraction and rapid depletion of groundwater (Watto et al., 2018). The net cultivated area also indicates the water requirement for farming and the water extraction largely from groundwater-based wells; hence, an increase in net sown area under groundwater irrigation may lead to more groundwater extraction and lower groundwater levels. Since tube well irrigation is considered more reliable, farmers tend to go for multiple cropping in a year, adversely affecting the groundwater table. Rice is grown in the state both intensively and extensively (Swain et al., 2022). During the rabi season, rice is cultivated using tube well irrigation that relies on groundwater resources. Therefore, it is hypothesized that the groundwater table is inversely related to the rice cultivated area (Shome & Upadhyay, 2022); hence, the regions with more land under rice cultivation will have a lower groundwater table.

Forest cover helps to reduce runoff during rain, and it retains water through its leafy surface, which helps to recharge groundwater through infiltration (Hitinayake et al., 2008). Therefore, extensive forest areas can increase the infiltration rate, improving groundwater levels. In recent decades, the rapid growth of population and associated economic development has greatly increased the consumption of water, which has frequently resulted in the overuse and over-extraction of water resources (Ojeda Olivares et al., 2019). The literacy rate is another important determinant of groundwater depletion. Literacy among water users can significantly improve knowledge and awareness about water-related problems and hence can positively influence the perception and action of users through an active and integrative understanding of water sources, management and security challenges (Maniam et al., 2021). Hence, it is hypothesized that a higher literacy rate of the population may help improve the groundwater tables.

Table 2 presents a description of the variables used in the models and their expected effect. The summary statistics of variables used in the different regression models are reported in Table 3. The mean variance inflation factor values of all the models range between 3 and 5, indicating that multicollinearity among the models’ determinant variables is within the acceptable level (Hair et al., 1995).

Table 2. Description of the variables used in the models and their expected effect.

	Description	Expected effect
Dependent variable
Groundwater level	Depth of availability of groundwater (mbgl)
Independent variables
Lag groundwater level	1-year lag of the depth of groundwater level (mbgl)	+
Log of forest cover	Area under forest cover (ha)	–
Log of net area sown	Net area sown (ha)	–
Log of area irrigated more than once	Area irrigated more than once (ha)	+
Log of rice cultivated area	Rice cultivated area (ha)	+
Literacy rate	Literacy rate	–
Log of population density	Population density	+
Log of number of tanks	Number of tanks	–
Log of tube well-irrigated area	Tube well irrigated area (ha)	+
Log of annual rainfall	Average annual rainfall (mm)	–
1-year lag of log of annual rainfall	One-year lag of annual rainfall (mm)	–
Log of DGDP	District gross domestic product (Rs. lakh)	+

Note: mbgl, Metres below ground level.

Table 3. Summary statistics of the variables.

Variables		Groundwater level	Log of forest cover	Log of net area irrigated	Log of area irrigated more than once	Log of rice cultivated area	Literacy rate	Log of population density	Log of number of tanks	Log of tube well-irrigated area	Log of annual rainfall	1-year lag of log of annual rainfall	Log of DGDP
Odisha Overall	Observations	667	580	580	579	580	667	667	666	665	667	638	580
	Mean	3.67	11.47	11.94	10.57	11.81	0.61	5.53	7.97	5.96	7.18	7.18	12.08
	SD	0.84	1.23	0.53	1.26	0.6	0.18	0.61	1.34	2.05	0.19	0.19	0.93
Drought-prone district	Observations	437	380	380	379	380	437	437	436	435	437	418	437
	Mean	3.71	11.61	11.87	10.6	11.69	0.59	5.47	7.88	5.69	7.17	7.17	12.25
	SD	0.83	1.13	0.47	1.3	0.53	0.19	0.58	1.45	1.82	0.19	0.19	0.99
Non-drought-prone district	Observations	230	200	200	200	200	230	230	230	230	230	220	230
	Mean	3.6	11.21	12.07	10.51	12.03	0.66	5.65	8.14	6.48	7.21	7.21	12.28
	SD	0.85	1.36	0.59	1.2	0.65	0.15	0.63	1.1	2.35	0.18	0.18	1.08
Tribal district	Observations	184	160	160	160	160	184	184	183	183	184	176	160
	Mean	4.27	12.45	12.12	10.19	11.79	0.47	5.08	7.87	5.5	7.19	7.19	12.09
	SD	0.64	0.5	0.52	1.35	0.7	0.17	0.31	1.7	2.05	0.21	0.21	0.96
Non-tribal district	Observations	483	420	420	419	420	483	483	483	482	483	462	420
	Mean	3.45	11.1	11.87	10.71	11.82	0.67	5.71	8	6.14	7.18	7.18	12.07
	SD	0.79	1.22	0.51	1.2	0.55	0.15	0.6	1.18	2.03	0.18	0.18	0.92
Coastal district	Observations	161	140	140	140	140	161	161	161	161	161	154	140
	Mean	2.96	10.13	11.96	10.97	11.99	0.73	6.3	8.15	7.6	7.22	7.22	12.39
	SD	0.83	1.27	0.4	0.85	0.42	0.13	0.2	1.49	1.52	0.17	0.18	0.78
Non-coastal district	Observations	506	440	440	439	440	506	506	505	504	506	484	506
	Mean	3.9	11.9	11.92	10.44	11.74	0.57	5.28	7.9	5.43	7.16	7.16	12.15
	SD	0.7	0.84	0.55	1.34	0.62	0.17	0.47	1.28	1.91	0.18	0.18	1.04
Low DGDP	Observations	253	220	220	219	220	253	253	252	251	253	242	253
	Mean	3.77	11.42	11.57	9.93	11.37	0.55	5.17	7.41	4.73	7.14	7.14	11.57
	SD	0.68	1.25	0.44	1.45	0.54	0.19	0.51	1.43	1.9	0.2	0.2	0.82
Medium DGDP	Observations	207	180	180	180	180	207	207	207	207	207	198	207
	Mean	3.36	11.21	12.11	11	12.02	0.64	5.61	8.05	6.36	7.19	7.19	12.33
	SD	0.93	1.31	0.45	0.81	0.46	0.18	0.53	1.07	1.59	0.17	0.17	0.76
High DGDP	Observations	207	180	180	180	180	207	207	207	207	207	198	207
	Mean	3.87	11.8	12.21	10.92	12.13	0.67	5.89	8.56	7.06	7.22	7.23	13.02
	SD	0.84	1.04	0.42	1.07	0.45	0.15	0.54	1.22	1.84	0.17	0.17	0.9

Note: DGDP, district gross domestic product.

The data relating to the number of rainwater-harvesting structures (tanks), area sown more than once, tube wells, net area sown, forest cover areas and rice-cropped area are taken from the Department of Agriculture and Farmers Welfare’s web portal (https://agcensus.dacnet.nic.in). Rainfall data were obtained from the India water portal (www.indiawaterportal.org) and Customized Rainfall Information System (https://hydro.imd.gov.in) for 1995–2017. Groundwater level data are taken from the India Water Resources Information System, from 1995 to 2017. Literacy and population density data are collected from the census data for 1991, 2001 and 2011. These decadal data have been interpolated with respect to population growth.

Results

Table 4 reports the regression results of the estimated econometric models on the factors influencing the variations in the depth of groundwater availability across the districts of Odisha. While the regression results differ across the models, the coefficients of several important independent variables are statistically significant.

Table 4. Category wise econometric results on the determinants of the depth of groundwater availability.

			Overall Odisha	Drought prone	Non-drought prone	Tribal district	Non-tribal district	Coastal district	Non-coastal district	Low DGDP	Medium DGDP	High DGDP
			Coefficient (RSE)	Coefficient (RSE)	Coefficient (RSE)	Coefficient (RSE)	Coefficient (RSE)	Coefficient (RSE)	Coefficient (RSE)	Coefficient (RSE)	Coefficient (RSE)	Coefficient (RSE)
Lag groundwater level			0.154** (0.063)	0.088 (0.079)	0.305*** (0.067)	0.127 (0.105)	0.144** (0.058)	0.226*** (0.046)	0.152** (0.066)	0.175* (0.099)	0.197*** (0.072)	0.342*** (0.089)
Log of forest cover			0.003 (0.267)	−0.099** (0.040)	0.048 (0.056)	−0.072 (0.208)	−0.081 (0.209)	0.001 (0.138)	−0.064 (0.064)	−0.078 (0.057)	−0.079** (0.035)	0.127 (0.088)
Log of net area sown			−0.427* (0.258)	−0.839*** (0.220)	−0.494 (0.490)	−1.130*** (0.382)	−0.235 (0.362)	0.472 (0.835)	−0.814*** (0.203)	−0.329 (0.338)	−0.819*** (0.309)	−0.250 (0.309)
Log of area irrigated more than once			0.049 (0.033)	0.069** (0.033)	0.020 (0.042)	0.043 (0.033)	0.039 (0.066)	−0.118 (0.081)	0.042* (0.023)	0.072*** (0.022)	0.090 (0.085)	0.005 (0.036)
Log of rice cultivated area			0.339*** (0.073)	0.290*** (0.062)	0.321* (0.187)	0.441*** (0.140)	0.305*** (0.095)	0.691*** (0.257)	0.286*** (0.067)	0.475*** (0.151)	0.122* (0.071)	0.326*** (0118)
Literacy rate			−3.478** (1.541)	−2.551 (2.175)	−0.159 (1.002)	−10.292* (5.850)	−3.140** (1.246)	−0.137 (1.111)	0.158 (1.381)	−2.961 (3.270)	−2.707 (1.803)	2.421 (1.662)
Log of population density			4.776** (2.102)	4.217*** (1.601)	0.829 (1.024)	12.112*** (4.304)	2.493 (1.186)	2.903 (2.414)	0.193 (1.402)	5.922* (3.351)	−0.311 (0.728)	−3.175* (1.859)
Log of number of tanks			−0.212* (0.012)	−0.233** (0.114)	−0.185* (0.100)	−0.095 (0.124)	−0.319*** (0.104)	−0.309*** (0.112)	0.023 (0.111)	−0.225** (0.090)	−0.142* (0.074)	−0.078* (0.045)
Log of tube well-irrigated area			0.073** (0.036)	0.050* (0.030)	0.126*** (0.047)	0.083*** (0.028)	0.105** (0.051)	0.050 (0.093)	0.036 (0.041)	0.056* (0.029)	0.093*** (0.019)	0.017 (0.041)
Log of annual rainfall			−0.701*** (0.145)	−0.480*** (0.148)	−0.633*** (0.179)	−0.434* (0.230)	−0.699*** (0.185)	−0.101 (0.189)	−0.521*** (0.141)	−0.739*** (0.202)	−0.249 (0.169)	−0.799*** (0.210)
1-year lag of log of annual rainfall			−0.439** (0.185)	−0.364*** (0.121)	−0.315 (0.296)	−0.680** (0.267)	−0.073 (0.222)	−0.003 (0.273)	−0.394** (0.159)	−0.354* (0.199)	0.032 (0.150)	−0.684*** (0.258)
Log of DGDP			−0.008 (0.190)	−0.194 (0.290)	−0.082 (0.211)	−0.077 (0.432)	0.225 (0.249)	−0.350 (0.343)	−0.252 (0.185)	−0.170 (0.329)	0.287 (0.297)	0.028 (0.375)
Constant			−11.085 (9.867)	−1.166 (10.216)	7.835 (7.995)	−34.711 (24.070)	−4.783 (10.714)	−21.258** (10.941)	18.178 (6.568)	−16.132 (18.784)	13.148** (5.772)	28.133*** (6.218)
Observations			462	359	190	128	334	112	416	207	171	171
Wald chi2 (11)			81.06	188.51	98.29	33.73	69.6	50.92	306.71	3,740,000	35.88	127.44
Prob > chi2			0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000
Number of instruments			180	237	174	125	180	113	240	183	162	162
Arellano–Bond test	1st	z	−4.441	−3.669	−2.793	−2.574	−3.521	−2.231	−4.165	−3.039	−2.593	−2.667
		p	0.000	0.0002	0.005	0.010	0.000	0.0257	0.000	0.002	0.009	0.007
	2nd	Z	−0.777	−0.474	1.718	−0.801	1.254	0.515	1.038	1.239	0.538	0.514
		p	0.437	0.635	0.086	0.423	0.208	0.606	0.299	0.215	0.590	0.609
Mean VIF			3.51	3.14	3.84	3.52	3.83	5.55	3.19	4.57	6.89	6.44

Note: Significance level: * = 10%, ** = 5%, *** = 1%.

DGDP, District gross domestic product, RSE, robust standard error; VIF, variance inflation factor.

It is found that the coefficient of the lagged depth of groundwater availability is significant and positive in all the estimated models (except in the models for the drought-prone and tribal districts), whereas that of the area under forest cover is significant and negative in the case of the models for the drought-prone and medium GDP districts. This implies that as the area under forest cover increases, the depth of groundwater availability decreases and hence the groundwater level improves significantly. Similarly, the coefficient of net area sown is significant and negative in the models for the drought-prone, tribal, non-coastal and medium GDP districts. As expected, the coefficient of area irrigated more than once is significant and positive in the estimated models for the drought-prone, non-coastal and low GDP districts, implying that in such districts, when the areas are irrigated more than once, the groundwater level tends to decline. This is possibly because of the withdrawal of groundwater in excess of its replenishment through recharge.

Interestingly, the coefficient of the rice cultivated area is significant and positive in all the estimated models, which suggests that as the area under rice cultivation increases, the groundwater level declines in general. The coefficient of literacy level is found to be significant and negative in the estimated models for the tribal and non-tribal districts as well as for the combined (pooled) model. This implies that the groundwater level improves with a higher literacy rate, possibly through rational and efficient water use due to awareness and knowledge. It is also found that the coefficient of population density is significant and positive in the estimated models for the tribal and low GDP districts, implying that higher population density results in greater water demand and hence more groundwater extraction in these districts. However, the coefficient of the district’s GDP is not significant, indicating that the size of a district’s economy does not significantly impact the depth of groundwater availability. As hypothesized above, economic development is expected to be associated with a greater groundwater depth because of over-exploitation and over-use of groundwater. This relationship seems to hold in the case of both medium and high GDP districts, as the signs of their coefficients are positive, although not statistically significant.

The coefficient of the number of rainwater-harvesting structures (tanks) is significant and negative in all the estimated models except for the tribal and non-coastal districts model. This indicates that as the number of tanks rises in the district, the groundwater level would improve due to the percolation of surface water through such harvesting structures. Similarly, the coefficient of annual mean rainfall is also significant and negative in all the estimated models except for the coastal and medium GDP districts. This implies that the groundwater table would improve with an increase in average annual rainfall. On the contrary, the coefficient of the area under tube well irrigation is significant and positive in all the estimated models except for the coastal and non-coastal districts model, indicating the inverse relationships between groundwater recharge and tube well irrigation.

Discussion

The regression results show that the previous year’s depth of groundwater availability has a significant and direct impact on its present level. This is possibly because the current groundwater level depends on the extraction and use of this resource in previous years (Chowdhury & Behera, 2018). Hence, in the event of limited rainwater harvesting and low recharge, it would be expected that more groundwater extraction would reduce its availability next year, posing a serious threat to water and food security. In drought-prone and medium GDP districts, forest cover helps to increase the groundwater level significantly. This could be because forest cover acts as a sponge, which helps to hold the rainwater for a longer time and releases water slowly, increasing the infiltration rate in the region (Ilstedt et al. 2016). Similar empirical results on the role of forest cover in conserving water in the Indian context are reported by Bhattacharjee and Behera (2018). Broadly, it can be argued that forest cover plays a crucial role in protecting water resources in general and groundwater in particular, leading to a decrease in groundwater depth. However, several studies show that decreasing forest cover generally increases water recharge, and afforestation reduces it (Farley et al., 2005; Jackson et al., 2005).

Similarly, the negative coefficient of the net sown area in the estimated models is contradictory to what is expected. This is because as the net area sown increases, the gross sown area is also likely to increase, raising groundwater extraction and, hence, the depth of its availability. However, how an increase in the net sown area would affect the depth of groundwater availability also depends on other factors, including the nature of crops cultivated (crops are not all equally water intensive), cropping intensity, the extent of crop diversification, efforts towards rainwater harvesting and use of surface water for irrigation. Hence, further exploration using micro-level evidence would provide better insights. Besides, it should be noted that with financial support from the National Bank for Agriculture and Rural Development and through the Mahatma Gandhi National Rural Employment Guarantee Act scheme, the government of Odisha has been investing in constructing rainwater harvesting structures in rural areas to enhance the groundwater table and improve the drought-resilient capacity of the state. In addition, the role of pani panchayats (farmer-led bodies engaged in water management) in the sustainable use and management of minor irrigation in Odisha has been frequently cited in the empirical literature (Behera & Mishra, 2018). This could be one of the main reasons behind the negative relationship between net area sown and groundwater depth in the state, in addition to soil type, cropping patterns, better irrigation water management, and others.

On the contrary, the coefficient of areas irrigated more than once is found to be positive, possibly because the increase in such irrigated areas leads to the over-extraction of groundwater and hence a decline in its level, particularly when the withdrawal is in excess of replenishment through recharge. Similarly, the significant positive coefficient of rice cultivated areas indicates that greater emphasis on the cultivation of this crop would lead to the decline of the groundwater table. This is because rice, particularly the high-yielding varieties, is a water-intensive crop, and its cultivation during the pre-summer period requires flood irrigation using groundwater for a considerable period. Besides, cultivation of such rice also requires sufficient use of chemical fertilizers in combination with the necessary water, raising further pressure on groundwater. This is a critical issue, given that surface water sources in Odisha are insufficient to cater to the water requirement for rice cultivation. Consequently, the major portion of irrigation water is supplied from groundwater sources. As a result, the groundwater table in the state is declining in the rice-cultivated areas.

The regression results further suggest that the groundwater level is likely to improve with the rise in the literacy rate. This may be because literate people tend to be more aware of the problem of groundwater depletion and hence adopt rational and efficient use of groundwater resources, limiting its excess extraction. Further, literate people will likely have more information about modern farming technologies and practices and greater capabilities to use them. Similarly, the positive coefficient of population density suggests that a greater concentration of population in a district would increase groundwater use for various economic as well as domestic activities, raising the possibility of a decline in groundwater levels. Broadly, higher population density results in greater economic and household activities and, hence, greater water consumption, leading to the over-extraction of groundwater and a decrease in its level, particularly when surface water sources are not adequately available or used.

Notably, the regression results show that the number of tanks improves the groundwater level significantly by lowering the depth of its availability. This is because rainwater harvesting structures such as tanks store rainwater and help it percolate into the ground throughout the year, which in turn helps to recharge groundwater. Chowdhury and Behera (2018) also reported a similar empirical finding in West Bengal, and Madhnure et al. (2023) found that the tank restoration programme increased groundwater storage in Telangana state. On the contrary, the positive coefficient of the area under tube well irrigation indicates that the groundwater level tends to decline as the area under such an irrigation system increases. It is often observed that when farmers use tube wells for irrigation, pressure on groundwater increases, whereas surface irrigation water sources, such as tanks, ponds, etc., are not adequately explored. As a consequence, the depth of groundwater availability increases, causing a hydrological imbalance between surface and groundwater, which has serious consequences in coastal districts where saline water from seas tends to seep into the ground. Similar findings are reported in many of the earlier studies in the Indian context (e.g., Dangar et al., 2021; Prasad & Rao, 2018).

The negative coefficient of annual average rainfall implies that the depth of groundwater availability is likely to decrease with an increase in average annual rainfall. This is consistent with the findings of Kotchoni et al. (2018). Heavy rainfall often results in more water infiltration, enhancing the groundwater table. However, wide fluctuations in annual rainfall in recent years, primarily due to climate change, may not necessarily help in groundwater recharge, particularly in the absence of proper rainwater harvesting structures. When so, groundwater depletion is likely to accelerate even though monsoon and non-monsoon rainfall increases. This has a severe cascading effect regarding groundwater extraction for irrigating crops, resulting in a further decline in the groundwater table (Jain et al., 2021; Swain et al., 2022).

Conclusions and policy implications

The current study makes an in-depth assessment of the groundwater situation and examines the causes of groundwater depletion, with a particular focus on understanding the relationship between surface water and the groundwater table using a district-level panel dataset from 1995 to 2017 in the Indian state of Odisha. Results clearly highlight that the annual average rainfall, number of rainwater-harvesting structures (tanks), forest cover, and net area sown positively influence groundwater levels. In contrast, the area under tube well irrigation and the rice-cultivated area has a negative effect on the groundwater level. In particular, the influence of rainwater-harvesting structures on the groundwater table in all the estimated models indicates a strong positive relationship between rainwater-harvesting structures (surface water) and groundwater levels. Rainwater-harvesting structures could contribute to reducing the depletion of the groundwater table through two channels: (1) by directly recharging the groundwater table and (2) less extraction of groundwater for cultivation and feeding livestock as the water stored in the rainwater-harvesting structures could be used for some time of the year. This has important policy implications in resolving the problem of groundwater depletion and thereby ensuring water and food security in the medium to long terms. The finding suggests that investment in renovating and rejuvenating rainwater-harvesting structures, such as ponds and tanks, can significantly improve the groundwater table. The water policies in other states of India as well as other regions of the world, which aim to ensure water table management, water security, food security, and sustainable water management, should invest in restoring and building rainwater-harvesting structures.

The positive influence of average annual rainfall on groundwater tables also has important implications, particularly in changing climatic conditions. As a result of climate change, rainfall patterns have become erratic, resulting in increased pressure on groundwater. Initiatives for rainwater-harvesting are necessary to facilitate groundwater recharge, improve its level, and provide water for farming and other domestic purposes directly from the rainwater-harvesting structures, thereby reducing pressure on groundwater. This is crucial as irregular and untimely rainfall can adversely affect the hydrological balance. Further, while rice is one of the major crops cultivated in Odisha, it is highly water-intensive and relies largely on groundwater. Over the years, this has resulted in the over-extraction of groundwater in the state. In addition, large-scale adoption of groundwater extraction technology such as tube wells has also contributed to groundwater depletion and the disappearance of surface water irrigation structures due to a lack of continuous investment and maintenance. Hence, to reverse the trend of groundwater depletion, changes in cropping patterns from water-intensive rice to less water-intensive crops such as pulses and crop diversification can be an important policy option for sustainable use and management of water resources. Regulating the installation of tube wells in critical zones can also help improve the groundwater situation in the state. Groundwater pricing, removal of subsidies on fuel for farming, and awareness about the importance of maintaining water structures could be important measures to maintain groundwater table and hence water and food security for the long term.

The study findings further indicate that there has been an overemphasis on the use of scarce groundwater resources for irrigation and other economic activities in the state of Odisha, neglecting the importance of surface water sources. Hence, the efficient use of groundwater and surface water is critical to support the hydrological balance for long-term water and agricultural sustainability. Additionally, groundwater should be treated as a complement to surface water instead of considering it as a substitute, as found in the present study. Hence, the government should frame sustainable water policies to reverse the declining trend of traditional water bodies such as tanks and ponds. Such policies will help conserve rainwater and groundwater recharge and reduce excess groundwater extraction, as stored rainwater can be deployed for irrigation and farming. In this regard, creating awareness about the strong relations between surface and groundwater among farmers and other water users will go a long way in conserving these critical resources in rural areas of Odisha.

Thus, the strength of the paper lies in establishing the relationship between groundwater and surface water empirically, which would encourage policy makers to adopt a governance system towards the rational use and efficient management of both surface as well as groundwater resources in a sustainable manner. In this context, a sustainability index can also be constructed to monitor and conserve groundwater. Nonetheless, like every empirical study, this research also suffers from various data and methodological limitations. In particular, the present study is based on secondary data at the district level, and systematic data on many relevant variables were unavailable. In particular, examining the hydrological link between surface water, groundwater, and forest cover requires routine data on a wide range of variables. Besides, deeper scrutiny of the socio-economic and institutional dynamics of the groundwater situation at the local level should be carried out using primary data for robust policies and institutional arrangements. These aspects collectively indicate areas that require more in-depth research.

Acknowledgement

The authors express their gratitude for the invaluable feedback and suggestions provided by the journal’s Editor-in-Chief and the anonymous reviewers, as well as the organisers, discussants and participants of the conference entitled Water Resource Management for Achieving Food Security in Asia Under Climate Change, October 26–27, 2022. Furthermore, the authors deeply appreciate the Asian Development Bank Institute (ADBI) and the guest editors for initiating this special issue. The authors also acknowledge the assistance of Dr Panharoth Chhay (Research Associate, ADBI) during the manuscript submission process. In addition, they are thankful to Adam Majoe, ADBI, for help with the editing.

Disclosure statement

No potential conflict of interest was reported by the authors.