4.1 Coverage and intensity of hydro-meteorological droughts:
Figure 2 evident that the study area has experienced three major and region-wide hydro-meteorological drought events (1985-86, 2002-03 and 2011-12) which persisted consecutively for two years. It is interesting to note that the succeeding year of these meteorological drought events registered considerably large spatial coverage (100% area) with moderate to extreme intensities (Fig. 2a). As the semi-arid hydrology of Maharashtra is rain-dependent, the ripple effect of these meteorological droughts can be observed in runoff and groundwater droughts (Fig. 2b and 2c, respectively). Particularly, coverage and intensities of the groundwater droughts show remarkable harmony with rainfall droughts. The rainfall droughts in 1986, 2003 and 2012 were the region-wide droughts, as they experienced in all the meteorological sub-divisions of Maharashtra (Chavadekar and Kashid 2018). In terms of coverage (100%) and intensity (SPI = -2.2), drought in 2003 was the worst experience of water scarcity for Maharashtra after the 1972 drought (Todmal 2019a). During this drought event, all the present study basins were observed with zero/nil runoff (discharge) and severe to extreme groundwater deficiency. On account of successive and intense drought conditions during 2002/03, the vegetation cover was also adversely affected (Messina 2013). As a result of this, first time in Maharashtra, government cattle camps were started to protect the livestock. Therefore, from the economic viewpoint, it was the most expensive disaster for Maharashtra during the past few decades (World Bank 2008). Similarly, the drought event in 2012 was observed over the entire study area. Inappropriate management of water resources for sugarcane cultivation during this drought event was one of the reasons to intensify the water scarcity (Purandare 2013).
Though the rainfall drought in 2012 resulted in more intense (SWI = 1.5 to 2.5) groundwater drought condition (as compared to 1985 drought event), the spatial coverage was condensed by ~ 22%. An increased number of surface impoundment structures during the past two decades may have caused this, as some domains in the study area with dense water bodies have registered a positive effect on groundwater resource (Todmal 2019b). It is also observed that the antecedent groundwater in the selected basins has a notable effect on the succeeding year’s groundwater levels (Udmale 2014a; Todmal 2019b). Therefore, it can be stated that the groundwater deficiencies in 1985, 2002 and 2011 were partially responsible for the intensification of region-wide groundwater droughts observed in 1986, 2003 and 2012 (Fig. 2c). Moreover, perhaps due to groundwater recharge (SWI = -1.7) and copious availability of surface water (SRI = 1.2) to harvest (in 2010), the rainfall deficit in 2011 could not result in groundwater drought (Fig. 2). The marginal decline in SRI value in 2011, which is calculated by using discharges at dam sites, also corroborates this fact (Fig. 2b). In a nutshell, these observations highlight the role of water harvesting structures constructed during the past two decades and the antecedent surface and groundwater to buffer the regional water scarcity. The study carried out by Udmale et al. (2014a) supports this finding. Another observation that can be made from Fig. 2 is that the runoff and groundwater droughts can occur even during the normal or mild rainfall deficit year. This is observed in 1984, 1992, 1994, 1997 and 2000. It should be noted that the rainfall droughts in 1992, 1994 and 1997 covered 20–70% of the study area (Fig. 2a). During these drought events, runoff and groundwater droughts were experienced over a remarkably larger area (85–100% and 70–90%, respectively) (Fig. 2b and 2c). It is very likely due to a number of surface impoundment structures constructed during the last two decades (Biggs et al. 2007; GoM 2014), which trap the surface water. On account of this, a very marginal amount of discharge reaches the downstream gauging sites, particularly in basins with small catchment areas. On the other hand, the rainfall deficits before 1990 could not result in runoff droughts with > 20% of spatial coverage (Fig. 2). Moreover, on account of rising temperature, the study area experiencing an increase in PET (Todmal et al. 2018). Therefore, the higher hydrological losses can also be one of the responsible factors for such a phenomenon (Todmal 2019b). These observations pointed out the adverse effects of anthropogenic warming and human interventions in the river basins during post-1990. The findings of the study conducted by Biggs et al. (2007) corroborate these observations.
Table 2
Hydro-meteorological drought return periods for the study region.
Index range
|
Category
|
Number of times in 100 years
|
Frequency of drought event
|
Years of drought events recorded
|
Rainfall droughts (SPI)
|
-1.0 to -1.49
|
Moderate drought
|
10
|
1 in 10 years
|
1985, 1986, 1997, 2011 and 2012
|
-1.5 to -1.99
|
Severe drought
|
4
|
1 in 28 years
|
1994, 2003 and 2011
|
= < -2.0
|
Extreme drought
|
2
|
1 in 50 years
|
2003
|
Runoff droughts (SRI)
|
-1.0 to -1.49
|
Moderate drought
|
10
|
1 in 10 years
|
1992 and 1997
|
-1.5 to -1.99
|
Severe drought
|
3
|
1 in 36 years
|
1994 and 2002
|
= < -2.0
|
Extreme drought
|
--
|
--
|
1992, 1994, 1997, 2002 and 2003
|
Groundwater droughts (SWI)
|
1.0 to 1.49
|
Moderate drought
|
9
|
1 in 11 years
|
1986, 1992, 1994, 1997 and 2002
|
1.50 to 1.99
|
Severe drought
|
2.5
|
1 in 41 years
|
1986, 2002, 2003 and 2012
|
= > 2.0
|
Extreme drought
|
1.2
|
1 in 85 years
|
2003 and 2012
|
The listed drought events were observed with > 40% of the study area (refer to Fig. 2). The mild drought events are not treated as drought events. The frequencies of basin-wise moderate, severe and extreme drought events are averaged for the study area. The intensity of some drought events varies basin-wise, therefore, listed in more than one category. As the groundwater level readings are taken from the ground surface, positive values of SWI imply drought conditions. # Drought event observed in the Karha and Agrani Basins (< 15% area).
4.2 Hydro-meteorological drought frequency:
The frequencies or return periods of meteorological and hydrological droughts calculated based on the non-exceedance probabilities of drought events are given in Table 2. In order to provide a generalized picture of the study area, the drought events with > 40% of coverage were considered. In the case of rainfall and groundwater droughts, it is observed that the lower intensity droughts are observed to be more frequent and vice versa. Previous studies carried out by Singh et al. (2019) and Amrit et al. (2018) were evident comparable frequencies of meteorological droughts over the study area. However, the runoff droughts with extreme intensity were experienced more often. It is pertinent to mention that about 80% of extreme runoff droughts are observed with zero runoff. Hence, the estimation of the average return period for the SRI < -2.0 appears to be unreliable (Table 2). The previous studies registered a significant increase in frequency and intensity of meteorological droughts over the Indian sub-continent (Mallya et al. 2016; Gore et al. 2010) and Maharashtra (Amrit et al. 2020) during the recent past. The present investigation confirmed this observation in the semi-arid region of Maharashtra, as the higher frequency of intense hydro-meteorological droughts (severe and extreme) was experienced during the post-1994 period (1994, 2002, 2003, 2011 and 2012) (Table 2). Another interesting fact emerges from Table 2 that the low-intensity rainfall droughts resulted in extreme hydrological deficits. Irrespective of coverage (by considering all drought events in Fig. 2), the rainfall, runoff and groundwater droughts were observed for about 32%, 56% and 29% of gauge time, respectively. From these observations, it can be inferred that apart from natural variability in monsoon rainfall, anthropogenic activities have played a crucial role in the increased frequency of surface water deficiency, particularly after 1994. A significant decline in discharge volume in the upper Krishna Basin (including study area) attributed to human interventions (Biggs et al. 2007) supports this finding. The groundwater in the study area can buffer the rainfall deficits for about two years (Udmale et al. 2014a). Very likely due to this reason, as compared to runoff droughts, the majority of groundwater droughts were observed in the moderate and severe category with 11 and 41 years return period, respectively. It suggests that even during the acute surface water deficits, the groundwater resource is available, at least for drinking and domestic use.
The region-wide meteorological and hydrological drought of 2003 was the rare catastrophic event, which occurs once in 30 to 50 years and > 40 years, respectively. All study rivers registered absolutely zero discharges during this drought event. It can also be noticed that the rainfall drought of 2012 resulted in extreme groundwater deficit of 40 to 85 years return period. Perhaps, due to the persistence of rainfall and surface water deficit consecutively for two years (2011 and 2012), the groundwater resource was over-exploited to irrigate high water requiring crops (Purandare 2013). Table 2 also shows that the groundwater deficiencies were intensified during the rainfall droughts in 1986, 2002, 2003 and 2012. This observation highlights the uncontrolled increase in groundwater withdrawal during meteorological droughts for agricultural and domestic purposes as well.
4.3 Semi-arid hydro-meteorology and agriculture:
Although the ENSO-rainfall relationship is weakening at the all-India level (Feba et al. 2019; Krishnamurthy and Kirtman 2003), it is observed to be significant over the present study area (Fig. 3a). About 63% of the negative SOI events found to be associated with below-average rainfall events (Fig. 3c). It signifies that the monsoon rainfall registered a significant decline during the El Niño events. The findings of previous regional studies (Takle and Pai 2020; Rishma and Katpatal 2016; Todmal and Kale 2016) are in good agreement with this observation. As the monsoon rainfall in the study area explained variations in the monsoon runoff and post-monsoon groundwater to a considerable extent (65% and 80%, respectively), these water resources exhibit a significant relationship with SOI (Fig. 3a and 3b). It is worth mention that Rishma and Katpatal (2016) and Kale et al. (2014) have observed the analogous linkage between surface water and SOI in central India and the semi-arid region of Maharashtra, respectively. An interesting feature of this relationship can be noticed that as compared to the monsoon rainfall, monsoon runoff and post-monsoon groundwater level in the study area have a stronger connection with SOI (r2 = 0.48 and 0.28, respectively). Almost a comparable connection between ENSO and groundwater was ascertained in different parts of Maharashtra including the present study area (Wable and Jha 2017; Janga Reddy and Ganguli 2012; Rishma and Katpatal 2019). The study also observed that about 79% and 58% of the negative SOI events were encountered with below-average surface runoff and groundwater level, respectively during the gauge period (Fig. 3c). On account of increased surface storage structures in the semi-arid region of Maharashtra during the last three decades (Todmal 2016; Biggs et al. 2007), the marginal amount of runoff generated during the below-average monsoon rainfall year could not reach the downstream discharge gauging sites (Kale et al. 2014). Therefore, the ENSO induced minor decrease in rainfall (SPI > -1.0) resulted in severe to extreme runoff droughts (SRI < -1.0). Perhaps due to this reason, SRI has a much stronger connection with SOI (Fig. 3a). In a nutshell, this study confirms that the El Niño events (weak to very strong) not only retard the monsoon rainfall but also negatively affect the surface and sub-surface water availability in the study area.
Irrespective of El Niño intensity, the study area has experienced severe hydrological droughts during the last two decades (Fig. 3c). The area under cash/high water-requiring crops is steadily increasing in the semi-arid region of Maharashtra (Todmal 2019a; Kalamkar 2011). Therefore, apart from the rainfall deficiency, the anthropogenic interventions in river basins (Biggs et al. 2007) and over-exploitation of groundwater resource (Purandare 2013) to fulfill the additional agricultural water demand may have attributed to amplification of surface and groundwater water scarcity in the post-1990 period. Another interesting observation emerged from Fig. 3c is that albeit rainfall has a considerable link with SOI, the region-wide droughts of 1985-86, 2003 and 2012 occurred in non-El Niño year. Apart from ENSO, the Indian Ocean sea surface temperature anomalies (IOD) significantly affect the monsoon rainfall over India (Vishnu et al. 2018; Ashok et al. 2004) including the semi-arid region of Maharashtra (Todmal 2019a). The unfavorable phase of Equatorial Indian Ocean Oscillation (EQUINOO), which is an atmospheric component of IOD, causes below-average rainfall over India. This type of unfavorable condition of EQUINOO existed in 1985, 1986, 2002-03 (Gadgil et al. 2003). Perhaps due to this reason the present study area experienced severe drought events during the neutral phase of ENSO (Fig. 3c). Moreover, it is found that the co-occurrence of positive IOD and El Niño reduces the intensity of drought in India (Ashok and Saji 2007). Therefore, it can be stated that the intensity of El Niño-induced meteorological droughts over the study area in 1994 and 1997 (SPI > -1.4) was normalized by the positive phase of IOD (~ 0.90) to some extent (Fig. 3c).
From Table 3 it is clear that the monsoon rainfall and groundwater exercise a prominent role in determining agricultural productivity, as all the selected crops (rainfed and irrigated) exhibit a statistically significant relationship with SPI and SRI. However, the results obtained from partial correlation analysis indicate the indirect effect of monsoon rainfall to strengthen the relationships between SCPI and SWI. Being a major source of irrigation in the semi-arid region of Maharashtra (Udmale et al. 2014a), groundwater availability determines the productivity of sorghum, gram and sugarcane to a considerable extent (Table 3). Among these crops, sorghum and gram are the post-monsoon crops whereas sugarcane is a long duration (12 to 15 months) and high water-requiring crop. It should be noted that less than 19% of the net sown area in the study area is under surface irrigation (GoM 2014). As a result, farmers have to rely on groundwater resource to irrigate long duration and high-water requiring crops during the post-monsoon season. During the last few decades, the area under cash/high water-requiring crops is increasing at the cost of rainfed crops (Kalamkar 2011; Todmal 2019a), which has increased the agricultural water demand in the semi-arid region of Maharashtra. Under such circumstances, the uncontrolled withdrawal of groundwater and mismanaged irrigation may have aggravated the recent hydrological and agricultural droughts. Among all the selected crops, pearl millet (drought-resistant crop) is observed to be heavily reliant on rainwater (Table 3). The productivity of pearl millet is neither dependent on the surface water nor the groundwater (Table 3), as it is cultivated during the monsoon season. On the contrary, sugarcane is cultivated in areas where surface/canal irrigation is well-developed (Todmal and Kale 2016). Even though the partial effect of SPI on SRI and SWI was removed, the obtained results suggest that the surface and sub-surface water resources significantly explained (18% and 26%, respectively) variations in regional sugarcane productivity.
Table 3
Coefficient of correlation (r2) between water resources and crop yield.
|
SCPI (Rainfed crops)
|
SCPI (Irrigated crops)
|
|
Sorghum
|
Pearl millet
|
Gram
|
Sugarcane
|
Wheat
|
SPI
|
0.49*
|
0.20*
|
0.38*
|
0.29*
|
0.26*
|
SRI
|
0.26*
|
0.03
|
0.21*
|
0.41*
|
0.06
|
SWI
|
0.59*
(0.21*)
|
0.22*
(0.03)
|
0.52*
(0.24*)
|
0.46*
(0.26*)
|
0.36*
(0.15)
|
* denotes significant relationships at 95% confidence levels. Values in bracket are coefficient of partial correlation (r2) by removing the effect of SPI (rainfall). |
4.4 Coverage and intensity of agricultural droughts:
Figure 4 exhibits year-wise coverage and intensities of agricultural droughts over the study area calculated based on agricultural productivity and cropped area. The agricultural productivity during the 1980s was observed notably below-average over 10–80% of the study region, irrespective of rainfall condition. By excluding the agricultural drought-affected area (SCPI) during major rain deficit events (1986, 1994, 1997, 2002 and 2003), the total area under all drought categories suggests a declining trend. It is an established fact that fertilizer consumption has significantly increased in western Maharashtra during the past four decades (Dahiwade et al. 2018). Therefore, the condensation of the spatial coverage of agricultural drought events during post-1990 is very likely associated with an increase in fertilizers consumption and irrigation facilities.
Another observation that can be made from Fig. 4 is that the severe rainfall deficits notably reduced the agricultural productivity of rainfed and irrigated crops as well. Particularly, the meteorological droughts in 1985-86, 1997, 2002-03 and 2011-12 were affected agricultural yield over the considerable area. It should be noted that despite increased consumption of fertilizers and irrigation facilities, the productivity of major crops in 2003 and 2012 was declined significantly with moderate to extreme intensity. These two dry events caused agricultural droughts of moderate to extreme intensity in 60–80% and 30–70% of the study area, respectively. The region-wide and intense water scarcity in 2003 hammered the yield of irrigated crops (sugarcane and wheat) in about 80% of the study area. Figure 4 also evident that the effect of water scarcity in 2003 on agriculture was continued in the succeeding year (2004). This can be noticed from the coverage and intensity of the sugarcane yield deficit in 2004, as it is a high water-requiring and long duration crop. It is very obvious when the entire area experiences severe to extreme rainfall, surface and sub-surface water scarcity. Although the drought of 2012 was observed with lower intensity (moderate to severe and SPI = -1.5), caused considerable damage to agriculture. During this drought event, about 70% and 50% of area/talukas registered a decline in productivity (moderate to extreme) of sorghum and wheat, respectively which are the post-monsoon crops. The study was undertaken by Udmale et al. (2014b) corroborates this observation. As mentioned earlier, the effect of El Niño on the monsoon rainfall in 1994 and 1997 was normalized by the positive phase of IOD. It can be considered as an indirect connection of IOD with agriculture in the study area, as the coverage (< 40%) and intensities (SCPI > -1.5) of agricultural droughts were observed to be controlled during these drought events.
The temporal variations in the agricultural cropped area under principal crops are also given in Fig. 4. The records show almost comparable departures in crop acreage during normal and below-average monsoon years, except during the major drought events. Therefore, the decline in the cropped area during low-intensity droughts could not be exhibited properly (Fig. 4). It is due to the reason that the government agencies collect crop acreage data at the time of sowing and not at the time of harvest. However, the sharp decline in area under all the principal crops during the region-wide drought of 2003 highlights the impact of acute water scarcity on agriculture. It can be observed that as compared to sorghum and pearl millet (declined by 20–25%), sugarcane and wheat registered a notably higher decline in acreage (50–60%). It indicates farmer's decision to cultivate high water-requiring crops in the semi-arid region is based on the availability of water resources. The reduction in area under sugarcane crop (by 60%) in 2004, which was a normal monsoon year, supports this explanation (Fig. 4).
The present investigation highlights that rainfall and groundwater are the chief determinants of agricultural productivity in the semi-arid region of Maharashtra (Table 3). Alike to crop productivity, water availability has a pronounced effect on the agricultural cropped area in western Maharashtra (Todmal 2019a). Unfortunately, due to limitations in data, the present study could not ascertain it properly (Fig. 4). The NDVI is widely used to examine the condition of biomass, vegetation cover and cultivated crops over a region (Revadekar, et al. 2012). In the upper Krishna Basin (including the present study area), a significant positive relationship between NDVI and SPI was observed (Dodamani et al. 2015). Due to sparse vegetation cover in the semi-arid region of Maharashtra, changes in NDVI considerably represent variations in agriculture cropped area. Therefore, in the present investigation, obtained basin-wise average NDVI values during wet, normal and deficit monsoon years (1998, 2013 and 2002, respectively) were correlated with basin-wise SPI and SWI (Fig. 5). It can be noticed that the winter NDVI has a statistically significant connection with rainfall and groundwater resources. The monsoon rainfall explained about 80% of variations in vegetation cover including agricultural cropped area. The availability of groundwater resource influences agricultural crops rather than other vegetation. Perhaps due to this reason, SWI has a comparatively weaker connection with NDVI (62% explained variance). Three clusters of scatter points in Fig. 5 indicate distinctive variations in vegetation cover and agricultural cropped area during wet, normal and deficit years. Among these selected years, the lowest value of NDVI (-0.169) for the entire study area was observed in 2002 which was a region-wide meteorological (SPI = -1.06) and hydrological (SWI = 1.47) drought year. On the other hand, remarkably higher positive NDVI (+ 0.142) was observed in 1998. As it was a wet year, the study region received a copious amount of rainfall (SPI = 1.67) and the groundwater level (SWI = -1.12) was significantly close to the ground level. It can also be noticed that during the normal monsoon year (2013), the study area registered NDVI, SPI and SWI values around zero (-0.003, -0.05 and − 0.69, respectively). In addition to this, the results of the Student's t test also suggested a significant difference in basin-wise average NDVI values during wet and drought years. From these observations, it can be inferred that the vegetation cover including the cultivated cropped area in the study region was considerably dropped during the hydro-meteorological drought events and vice versa.
4.5 Near-term future droughts and temperature:
The Intergovernmental Panel on Climate Change (2013) has reported that future climatic changes may adversely affect the semi-arid regions of South Asia. As the climatic variables considerably drive the semi-arid hydrology of Maharashtra, this study attempts to examine future variability in monsoon rainfall and temperature (Fig. 6). The results obtained from the regional climate projection models reveal that alike droughts in 2003 and 2012, the Madhya Maharashtra Sub-division (including the present study area) is very likely to experience severe rainfall droughts in 2029–2030, 2040 and 2050. It should be noted that these expected drought events with comparable intensity were identically reflected from the CORDEX RCA4 simulation and REMO 2009 model as well (Fig. 6a and 6b). The results of previous studies undertaken by Todmal et al. (2018) and Patil (2013) are in good agreement with this finding.
Under the climate change scenario, the majority of districts in Maharashtra are very likely to experience a significant rise in annual mean temperature (AMT) by < 2.5°C during the near-term future (IPCC 2013; Niu et al. 2015; Todmal 2021). In the present investigation, both of the considered climate projection models suggested a significant increase in AMT over Madhya Maharashtra Sub-division (and study area) with a decadal rate of 0.29°C (Fig. 6). Based on these estimates, it can be stated that the semi-arid region of Maharashtra may witness a significant rise in AMT by 1.0 to 1.1°C up to 2050. The results obtained from Student’s t-test indicate that there will be an abrupt increase in AMT during the post-2035 period. As the rate of potential evapotranspiration (PET) is directly associated with temperature (Purnadurga et al. 2017), the hydrological losses over the study area are very likely to increase in the near-term future, which may lead to augment the agricultural water demand. A rise in PET due to warmer conditions can increase aridity in the semi-arid regions (Ramarao et al. 2018). Therefore, the future warming scenario may exacerbate the challenge of water management during the estimated drought events (2029–2030, 2040, and 2050). From the agricultural viewpoint, AMT increase is a serious concern, as it can adversely affect agricultural crop productivity (Todmal 2021). With reference to the previous experimental studies (Boomiraj et al. 2011; Ong and Monteith 1985; World Bank 2003; TERI 2014), it can be stated that the future (up to 2050) rise in temperature may reduce the productivity of sorghum, pearl millet, wheat and sugarcane in the present study area. Secondly, the rapid expansion of areas under high water-requiring crops (Kalmkar 2011) is adding stress to available water resources (Todmal et al. 2018). Under such circumstances, the warming-induced increase and decrease in agricultural water demand and crop productivity, respectively may accountable for the intensification of agricultural droughts in the near-term future (up to 2050).