Introduction

Significance of allocating minimum flows in a river for sustainable management of water and environmental resources has drawn worldwide attention in the recent past (Acreman and Dunbar 2004; Joshi et al. 2014; O’Keeffe et al. 2012; Richter et al. 1997; Soni et al. 2014; Tharme 2003; Yang et al. 2008). Emphasis on provision of minimum flows or widely known as “Environmental flows” (hereafter, Eflow) is highlighted in the Brisbane Declaration in 2007. The major factors, separately or combinedly, accounted by various studies estimating Eflow are the following: (i) prevention of algal bloom choking, (ii) sediment transportation, (iii) minimum flow for delta and sea interface, (iv) sustainability of dolphins and biodiversity, (v) recharge flow provision, and (vi) dilution water (Joshi et al. 2014; O’Keeffe et al. 2012; Soni et al. 2014; Yang et al. 2008). A review of literature on the estimation of Eflow and methods adopted indicates the frequent adoption of desktop-based methods for estimation of Eflow (Acreman and Dunbar 2004; Soni et al. 2014; Tharme 2003). These studies have estimated Eflow, either by considering various aspects such as sediment transportation, delta and sea interface, aquatic biodiversity, recharge flow, algal choking, dilution flow, river characteristics, seepage loss, and river evaporation or by considering a mere percentage of dependable flow in the river. However, most of the past Eflow studies (e.g. Acreman and Dunbar 2004; Joshi et al. 2014; O’Keeffe et al. 2012; Richter et al. 1997; Soni et al. 2014; Tharme 2003; Yang et al. 2008) have not given emphasis on water quality aspect while estimating Eflow. River water quality, which has been significantly affected by human-induced pollution activities, has raised serious concerns (Angelidis et al. 1995; Li et al. 2016). Since poor water quality undoubtedly affects the aquatic species and drinking water specifications adversely (Bunn and Arthington 2002; Malmqvist and Rundle 2002), which could alter the Eflow demands, it is necessary to incorporate the possible implications of water quality criterion during estimation of Eflow of any river stretch. The estimation of Eflow based on water quality had not been widely encouraged in order to exempt the polluters from the freedom of discharging polluted water into the river with minimum treatment. Depending upon the pollution level in a river stretch, either a minimum continuous flow or stringent treatment policies or a combination of the above needs to be implemented to handle the (i) frequent violation of the treatment policies, (ii) self-cleansing ability of rivers, or (iii) inconsiderate upstream restrictions and diversions of water. Hence, it is imperative to change the common perception about Eflow and to incorporate the water quality criterion, in addition to the above-mentioned aspects for estimation of Eflow, for making river water usable for different activities.

Another significant factor that needs to be considered in the Eflow estimation framework is the possible effect of future climate changes and future water availability, which may in turn affect the water quality. The effect of climate change is more distinct through increased mean global air temperature and modified precipitation patterns across the world (IPCC 2007). Air temperature is the dominant variable, which determines water temperature (Mohseni et al. 1998) and, hence, has a strong influence on reactions and water quality parameters, such as dissolved oxygen (DO) and biochemical oxygen demand (BOD) (Delpla et al. 2009; Rehana and Mujumdar 2011). According to Arrhenius relation between temperature and reaction rates, physico-chemical equilibriums and biological reactions are significantly altered by the variations in temperature. Hence, the increase in water temperature enhances various processes such as solubilization and degradation. While these processes increase the existing matter in the water, it reduces the dissolved gases and, hence, reduces the dissolved oxygen content in water. Dissolved oxygen saturation concentration decreases at almost 10% with a 3 °C increase in air temperature (Delpla et al. 2009). Hence, an expected average increase of 1.8–4.0 °C in global air temperature in the twenty-first century as reported by the Intergovernmental Panel on Climate Change (IPCC) (Bates et al. 2008) will have an adverse impact on the water quality (Rehana and Mujumdar 2011; Rehana and Mujumdar 2012; Whitehead et al. 2009). In addition to the changes in temperature, an increase in precipitation extremes is also observed across the globe, which is also attributed to climate change (IPCC 2013). The abrupt changes in the precipitation patterns has altered the streamflow, especially the extremes, i.e. floods and droughts (Mimikou et al. 2000; Tang et al. 2014), which may enhance the decomposition or even flushing of organic matter into the streams (Evans et al. 2005). Hence, the change in air temperature and precipitation will have a direct effect on surface water variables, i.e. water temperature and streamflow (Ozaki and Fukushima 2003; Tang et al. 2014). Although limited studies (e.g. Thompson et al. 2014) have addressed the impact of climate change on various factors of Eflow, however, the possible variations in water quality factor under impact of climate change have not been considered so far in Eflow estimation studies. Therefore, in addition to existing river water quality conditions, the possible variations in water quantity and quality under different plausible future scenarios of climate change also need to be considered during estimation of Eflow.

In short, the major knowledge gaps in the existing literature on Eflow estimation are (i) omission of water quality criterion in Eflow methodology and (ii) exclusion of possible future changes in water quantity and quality and a systematic methodology thereof. Hence, the present study is focussed to develop a framework for estimating Eflow for any river stretch considering water quality criterion and possible variations in future water quantity and quality. The developed framework is demonstrated by estimating the Eflow for the polluted stretch of the Yamuna River passing through Delhi, India. Additionally, the impact of climate change and human activities on river water quality is also investigated in this study.

Study area and methodology

Study area

The proposed framework is employed on the 21.9-km stretch of Yamuna River passing through Delhi, India. The stretch starts at Wazirabad barrage and ends in Okhla barrage as shown in Fig. 1 and is considered as the most polluted stretch in the country (CPCB 2006). The flow in Yamuna River varies significantly during monsoon and non-monsoon seasons. The river constitutes maximum flow, i.e. around 80% of the total annual flow during monsoon period, while the condition of the river during non-monsoon season is critical. The 20% of annual flow remaining in the river is diverted and abstracted recklessly for drinking, agricultural, and industrial purposes (CPCB 2006). The condition becomes critical during the dry season period when almost no flow is available in the river. Further, contribution of partly treated and untreated wastewater effluents from 15 different drains (indexed 1 to 15 in Fig. 1) increases the pollution severity of the stretch. Among the 15 drains, the Najafgarh drain (indexed 1 in Fig. 1) carries significantly high organic pollution load. The stretch also receives wastewater from Hindon Cut canal (indexed 16 in Fig. 1). Along the stretch, water is abstracted at two locations, viz., Wazirabad barrage for drinking purposes and Agra Canal (indexed 17 in Fig. 1) for irrigation purpose (CPCB 2013). Hence, it is necessary to ensure the water quality standards for drinking purposes and for other domestic uses at these locations.

Fig. 1
figure 1

Line diagram of drains entering into the Yamuna River in the Delhi (India) stretch

Full size image

A few studies have recommended Eflow values for this stretch, considering different aspects. For instance, the High Powered Committee appointed by the Supreme Court of India recommended a provisional minimum flow of 10 m3/s at the downstream of Wazirabad barrage. Soni et al. (2014) proposed Eflow equal to 268.13 m3/s without considering the water quality factor. However, the recommended Eflow values failed to meet the water quality requirements of different water usages recommended by the Central Pollution Control Board (CPCB), India (Table 1). Therefore, this stretch of Yamuna River is taken as the study region to implement the proposed framework on Eflow estimation.

Table 1 Primary water quality criteria for various uses of river water (source: http://www.cpcb.nic.in)
Full size table

Proposed framework to estimate Eflow

The step-by-step description of the proposed framework to estimate Eflow incorporating water quality constraints and future climatic changes is illustrated in Fig. 2 and described below. The generic methodology is described first, with necessary description to the study region, wherever required.

Fig. 2
figure 2

Flowchart of the methodology adopted to estimate the future water quality and Eflow under impact of climate change and human activities (numbers indicate order of steps)

Full size image

Step 1. Data collection

In order to model water quality in any river stretch, information about following parameters is needed: (i) flow and water temperature at headwater of the stretch; (ii) elevation, location, width, slope, etc. of the river stretch; (iii) air temperature, dew point temperature, wind speed, cloud cover, and shade over the river stretch; (iv) water quality parameters such as DO and BOD along the river and at headwater; and (v) point sources and non-point sources of pollution contributing the river stretch.

Step 2. Derivation of a relationship between air temperature and water temperature of the river stretch

The relationship between air temperature and water temperature is derived to estimate the future water temperature values associated with any variation in future air temperature. Various studies have developed different models to relate air temperature and water temperature (Mohseni et al. 1998; Rehana and Mujumdar 2011). Most of these studies have adopted different regression models (linear, exponential, logarithmic, polynomial, power, logistic, inverse, and cubic) to derive the relationship. However, since those relationships cannot be directly used over the present study region due to the differences in climatic conditions and catchment characteristics, a relationship between air temperature and water temperature needs to be developed, specific to the study region. Hence, in the present study, various regression models are experimented. The performance of these models is analysed using regression coefficient or coefficient of determination (R 2). The model with the highest regression coefficient is selected as the best model to describe the relationship between air temperature and water temperature for the study region.

For the Yamuna River stretch, monthly water temperature data (Source: CPCB 2013) and monthly minimum, maximum, and mean air temperature for a period of 6 years (from 1999 to 2005) for three different locations (Palla, Nizamuddin, and Agra Canal) (Source: Indian Meteorological Department, Pune) are used. Various regression models (as mentioned above) are employed to relate observed monthly water temperature values with monthly minimum, maximum, and mean air temperature values from 1999 to 2005 (6 years).

Step 3. Generation of hypothetical future air temperature and streamflow scenarios

Plausible future climate scenarios, i.e. possible variations in water temperature and streamflow, have been addressed in various studies (Arnell 1999; Xu 1999; Rehana and Mujumdar 2011). In order to determine any future variations in air temperature and precipitation and related variations in water temperature and streamflow, two approaches can be undertaken. The first approach is to obtain the future air temperature and precipitation projections from various general circulation models (GCMs), generally available for a coarser region and to downscale these variables for the study region (Ficklin et al. 2013). Any future variation in water temperature and streamflow can then be determined by using the relationship as developed in step 2 and performing hydrological modelling for the study region, respectively. The second approach is to use hypothetical future scenarios, i.e. approximate plausible variations for the future, knowing the general characteristics of the region and catchment. While the first approach is preferable (Xu 1999), it requires development of any downscaling technique for the region and hence is quite data intense. The second approach is simple, quick, and easy to implement.

For the Yamuna stretch, we used the second approach of generating hypothetical scenarios due to the unavailability of observed hydrological data. The flow and pollution conditions in the stretch on March–June 2002 (for which the data is available) are selected as the baseline condition. Scenarios are generated by varying the air temperature and streamflow with reference to the baseline input conditions. The change in precipitation and evaporation due to changing climate ultimately gets reflected in the streamflow of the catchment, which is a driving input to a water quality model. Therefore, instead of generating hypothetical scenarios of precipitation and evaporation, the hypothetical scenarios of streamflow are only used in the study. Information about the projected possible changes in air temperature and streamflow over the study region is obtained from the Intergovernmental Panel on Climate Change (IPCC) Climate Change report (IPCC 2007). IPCC (2007) has projected a maximum of +4 °C increase in air temperature at Delhi at the end of the twenty-first century. Streamflow is found to have exhibiting both increasing and decreasing tendency and is associated with high uncertainty (IPCC 2007). Hence, it is decided to study the possible effect of climate change impact on water quality by varying air temperature as follows: ΔT = 0, ±1, ±2, ±3, and ±4 °C and by varying streamflow as follows: ΔSF = 0, ±10, and ±20% with respect to the baseline condition. Hence, a total of 45 hypothetical scenarios are considered. The decreasing scenarios of streamflow, i.e. −10 and −20%, symbolize the further worsening of low-flow conditions during dry season, which may prevail if unscrupulous diversion and extraction of river flow upstream of Yamuna River are continued. Similarly, the increasing scenarios of streamflow, i.e. +10 and +20%, are used to simulate the water quality conditions under lesser diversions and controlled abstraction.

Step 4. Generation of future water temperature scenarios

Using the plausible future air temperature scenarios from step 3, future water temperature scenarios are determined using the developed relationship between air temperature and water temperature (from step 2).

Step 5. Generation of hypothetical scenarios of drain flow and pollution load

Hypothetical scenarios of human-induced pollution are generated by varying the drain flow (ΔDF = ±20, ±100%) and the organic load concentration (ΔOLC = ±20, ±100%) in the drains. Change in drain flow of ±100% signifies the extreme flow conditions coming from the domestic or industrial sources. Variation of organic load concentration within ±100% signifies both, the rampantly increasing pollution scenarios (+100%) and pollution treatment scenarios (−100%).

Step 6. Water quality modelling under impact of human-induced pollution activities

QUAL2K, a one-dimensional river and stream water quality model (Chapra et al. 2008), is employed to simulate water quality under influence of the Yamuna River stretch. QUAL2K represents river as a series of reaches with constant hydraulic characteristics and has been frequently used in different water quality modelling studies (Chapra et al. 2008; Chaudhary et al. 2017; Kalburgi et al. 2010; Rehana and Mujumdar 2011; Vasudevan et al. 2011). Calibration of the model is performed using a trial-and-error approach, in which the model parameters are varied within their expected range until satisfactory performance measures are obtained for DO and BOD simulations. Four performance metrics, namely root mean square error (RMSE), coefficient of determination (R 2), coefficient of correlation, and index of agreement (IOA), are used to derive the best combinations of model parameters. Range of different model parameters can be obtained from QUAL2K manual (Chapra et al. 2008), Brown and Barnwell (1987), Zhang et al. (2012), etc. Further, the model is validated and the performance of the model in simulating DO and BOD values is evaluated using the same performance measures.

For modelling water quality in the Yamuna River stretch, the whole stretch is divided into 16 reaches (as shown in Fig. 1). Information about different input hydraulic and water quality variables is collected from Parmar and Keshari (2012), CPCB (2005), and DJB (2005). The meteorological data, such as air temperature, dew point temperature, and wind speed, are assumed to be constant for all reaches. The cloud cover and shade data are taken as 25% (partly cloudy) for daylight hours and 100% shade is used from 1 h before sundown to 1 h after sunrise. As mentioned earlier, there are 16 identified drains or point sources between Wazirabad barrage and Agra Canal and the only withdrawal/abstraction point is at Agra Canal. The flow and water quality details of the point sources and abstractions are also taken from Parmar and Keshari (2012).

QUAL2K model is then calibrated to simulate water quality indicators, DO and BOD, in Yamuna River by adjusting various model parameters for the average condition data of March–June 2002. Sediment oxygen demand (SOD) is assumed to be zero for the entire course of the river (Paliwal et al. 2007). The contribution of the BOD removal by settling is neglected since it accounts for only about 2% of the total BOD (Vasudevan et al. 2011). Photosynthetic oxygenation is taken as zero since high turbidity in the stretch diminishes the penetration of light to deeper layers, preventing the growth of phytoplankton (Kazmi 2000). Hence, the parameters to be calibrated are only three, i.e. oxygen re-aeration rate, BOD hydrolysis rate, and BOD oxidation rate. Calibration is done by using a trial-and-error approach where the model parameters, oxygen re-aeration rate (K a), BOD hydrolysis rate (K h), and BOD oxidation rate (K o) are adjusted within the expected ranges of 0–100 (Brown and Barnwell 1987), 0.04–4.2 (Zhang et al. 2012), and 0.02–4.2 (Zhang et al. 2012), respectively. The calibrated QUAL2K model is validated for average conditions of February 2003.

Step 7. Simulation of water quality variables (DO and BOD) under impact of plausible scenarios of climate change and human activities

The calibrated and validated QUAL2K model is run to simulate water quality variables (DO and BOD) under the baseline conditions of March–June 2002. In order to assess the impact of climate change on water quality of Yamuna River, QUAL2K model is run under different scenarios of change in air temperature (ΔT = ±2 and ±4 °C) and streamflow (ΔSF = ±20%). The percentage changes of DO and BOD values are quantified with respect to the baseline scenario. Further, the variations in water quality resulting from the plausible changes in human activities (change in drain flow, ΔDF = ±20, ±100% and organic load concentration, ΔOLC = ±20, ±100%) are also considered.

Step 8. Estimation of Eflow required in the river stretch

Future Eflow requirements along the river stretch are estimated considering the water quality criteria for various streamflow values. The streamflow satisfying the water quality requirements in the stretch is selected as the desired Eflow for the stretch.

For the Yamuna stretch, Eflow values required from Wazirabad barrage are estimated by simulating the river water quality for 18 extreme hypothetical scenarios, which are generated by varying air temperature as follows: ΔT = 0, ±4 °C; ΔDF = 0, ±100%; and ΔOLC = 0, ±100%. QUAL2K model is run by considering various streamflow values at the head (downstream of Wazirabad barrage). Eflow values are finally determined based on the satisfaction of CPCB standards of water quality (Table 1) on three different locations along the stretch, viz., location at the entry (downstream of Wazirabad barrage), location at the midstream (11 km downstream of Wazirabad barrage/downstream of Delhi Gate drain), and location at the exit (21.9 km downstream of Wazirabad barrage/exit of Okhla barrage). Eflow estimation is conducted for two cases: (i) baseline conditions under air temperature change scenario and (ii) combined extreme air temperature and pollution load change (quality and quantity) scenarios.

Results and discussion

The main results obtained at various stages are discussed below.

Relationship between air temperature and water temperature for the Yamuna River stretch

River water temperature is found to be more significantly correlated with minimum air temperature rather than maximum and mean air temperatures. Among the various regression models (linear, exponential, logarithmic, polynomial, power, logistic, inverse, and cubic) employed, a cubic polynomial regression model (y = −0.002x 3 + 4.331x − 46.411 where y is the mean monthly water temperature in °C and x is the minimum monthly air temperature in °C) is found to give the best fit to the observed data (R 2 = 0.627). The developed cubical polynomial relationship is used to predict future water temperature for the corresponding hypothetical future air temperature scenario.

Calibration and validation of QUAL2K model for simulating water quality in Yamuna River

The combination of reaction rate coefficients for 16 divisions of the Yamuna River, which give the best performance measures during simulation of DO and BOD values, is obtained. O2 re-aeration rate is found to be 0.1/day for the first 11 stretches and 3.2/day for the last 5 stretches whereas BOD hydrolysis rate and BOD oxidation rate are found to be 0.9/day for all stretches. The simulated DO and BOD profiles for the calibration period are shown in Fig. 3. While DO concentration shows a large drop (5 mg/L approx.) at the entry of Najafgarh drain, it regains some DO (2 mg/L approx.) at the downstream locations of the stretch. Similarly, BOD concentration undergoes a sharp increase and falls down at the downstream locations. In order to examine the efficiency of the calibration approach adopted, the simulated values of DO and BOD are compared with those reported in previous studies (Fig. 4). It is observed that the simulated values of DO and BOD in the present study give a better fit to the observed data than the results highlighted in previously reported studies.

Fig. 3
figure 3

Simulated a DO and b BOD along the Yamuna stretch for the calibration period (March–June 2002)

Full size image
Fig. 4
figure 4

Comparison of simulated a DO and b BOD along the Yamuna stretch using re-aeration rate coefficients obtained from present and previous approaches, for the calibration period

Full size image

The DO and BOD simulations of QUAL2K model during the validation period also show a good match with measured data (Fig. 5). The performance measures for simulated DO and BOD values for validation period are listed in Table 2. Significantly high performance measures clearly indicate the model’s ability to successfully simulate the water quality of the river stretch.

Fig. 5
figure 5

a DO and b BOD along the Yamuna stretch for the validation period (February 2003)

Full size image
Table 2 Model performance for the calibration and the validation period
Full size table

Impact of plausible future scenarios on water quality

Impact of climate change scenarios on water quality

Table 3 shows the maximum percentage change in DO and BOD at the upstream, middle, and downstream locations of the river stretch due to variations in streamflow and air temperature. When streamflow is kept constant, DO is found to be inversely proportional to air temperature. Maximum improvement of DO is obtained for a decrease in air temperature by −4 °C. Maximum deterioration of BOD is observed for a temperature increase by +4 °C. Similarly, effect of variation of streamflow on water quality is also investigated, keeping air temperature constant. For all the air temperature change scenarios (ΔAT = 0, ±4 °C), maximum increase in DO and maximum reduction in BOD are observed when streamflow from the upstream is increased by 20%. Among the scenarios of streamflow and air temperature, streamflow fluctuations have significantly high impact on water quality.

Table 3 Maximum percentage changes in DO and BOD with respect to baseline condition (March–June 2002) simulated data for extreme scenarios
Full size table

A simultaneous 20% decrease in streamflow and 4 °C increase in air temperature (i.e. scenario with ΔSF = −20% and ΔAT = +4 °C) give maximum deterioration in DO values throughout the stretch, indicating the worst-case scenario. It indicates that the combination of low streamflow and high air temperature conditions might drastically decrease DO values. Similarly, a simultaneous 20% decrease in streamflow and 4 °C decrease in air temperature (i.e. scenario with ΔSF = −20% and ΔAT = −4 °C) give maximum deterioration (or increase) in BOD values.

Impact of extreme pollution and climate change scenarios on water quality

Figure 6 shows variations in DO along the river stretch under extreme air temperature for two scenarios: (i) change in drain flow magnitude only and (ii) change in organic load concentrations in drain flow only. Reduction in DO of the downstream stretches is observed for all the scenarios of perturbations in air temperature and drain flow (Fig. 6a). Increase of organic concentration load in drains reduces the DO content of the latter stretches for both scenarios of ΔAT = −4 and +4 °C (Fig. 6b). Moreover, the reduction of DO in +4 °C scenario is higher than in −4 °C scenario. In the case of decrease in OLC, increase in DO level is observed for both the temperature extreme scenarios.

Fig. 6
figure 6

Variation of DO along the stretch varying only a drain flow by ±20% and b organic load concentrations of drain flow by ±20% under air temperature change conditions

Full size image

Similar simulations are done using a one-way sensitivity approach to determine the effect of each variable on water quality parameters. This approach involves varying of one variable at a time, keeping other variables constant. Table 4 presents percentage changes in DO at the exit location of the river stretch with respect to the baseline condition (March–June 2002) obtained for different scenarios. It is apparent that the variation in DO is significantly higher for change in organic load concentration than that due to change in any other variable. Similar results are also obtained for BOD simulations under various scenarios.

Table 4 Percentage changes in DO with the baseline condition at exit location of the river stretch under change in factors affecting water quality for different fluctuations in air temperature
Full size table

Estimation of Eflow for Yamuna River stretch

Eflow under baseline condition and air temperature change scenarios

Different streamflow values are considered for Eflow estimation. As a guideline, the Eflow suggested by the Supreme Court of India, i.e. 10 m3/s, and the Eflow suggested by Soni et al. (2014), i.e. 268.13 m3/s, which is around 50–60% of total flow of Yamuna (taking into account various aspects such as transport sediments, delta and sea interface, dolphins and biodiversity, recharge flow, and prevention of algal bloom choking), are taken. The simulated values of DO at three locations for different streamflows and different extreme air temperature change with respect to baseline air temperature are shown in Table 5 (for constant drain flow and organic load concentration). It can be seen that the required streamflow from the upstream of Wazirabad barrage, as per the directive of the Supreme Court of India (flow rate = 10 m3/s), does not meet DO requirements of any of the CPCB-recommended water classes (Table 1) at any of the three locations. The Eflow suggested by the Soni et al. (2014) for Yamuna River (i.e. 268.13 m3/s) shows an improvement in the DO values along the river stretch considered. This flow satisfies the minimum DO requirement as per the CPCB specification for the entry and exit locations of the river stretch under consideration, irrespective of the extreme air temperature changes. However, it fails to satisfy the minimum DO requirement at the middle stretch of the river. Similar results are obtained for BOD simulations too.

Table 5 Simulated values of DO at three locations for different streamflows and different extreme air temperature changes with respect to baseline air temperature (March–June 2002) keeping drain flows and organic load concentrations constant with respect to baseline drain characteristics
Full size table

A tenfold increment in the streamflow value, suggested by Soni et al. (2014) (i.e. 2681.3 m3/s,) seems to satisfy the DO criteria for water classes B, C, and D usages, irrespective of extreme air temperature changes at all locations of the river. Water quality meets the class A usage requirement at exit of Okhla Barrage only. Hotspot along the stretch with respect to class A type water usage requirement could be the stretch of the river between downstream of Najafgarh drain and downstream of Delhi Gate drain. The reason for not satisfying the water quality requirements for class A type usage at the upstream and middle locations of the river could be insufficient re-aeration and oxidation rates to meet the water quality criteria under existing headwater condition. Re-aeration rates at the downstream locations are found to be higher compared to that at upstream and middle locations. Thus, the self- purifying capacity of the river helps to meet the required specifications of different water usage requirements. The streamflow requirements for satisfying class B and class C water type usages are the same. The streamflow requirement is very high at the upstream location compared to that at the middle and exit locations. However, the requirement of streamflows at the middle location for satisfying the class D water type requirement is higher in the middle location than the upstream and exit locations.

Eflow value for combined extreme air temperature and pollution load change (quality and quantity) scenarios

Future changes in pollution loadings are more detrimental than future air temperature scenarios. Hence, water quality simulations are conducted with different streamflow values for different possible pollution load scenarios and air temperature change scenarios. The simulated values of DO and BOD for these scenarios at three different locations are obtained. As an illustration, only the DO values for various scenarios at the three locations considered are plotted in Fig. 7. The minimum Eflow value satisfying the water quality criteria is determined by considering only the maximum deteriorated values of DO and BOD for different streamflows at upstream of Wazirabad barrage. This value is calculated without considering the effect of extreme air temperature fluctuations on different drain flow characteristics at location downstream of Najafgarh drain. It is observed that water quality at the upstream location of the river is not suitable for class A type water usage, irrespective of any change in scenario and streamflow value. For other water usage class types, both DO and BOD criteria are met as streamflow is increased.

Fig. 7
figure 7

Maximum deteriorated values of DO at different Eflow from upstream of Wazirabad barrage irrespective of extreme air temperature fluctuation for different drain flow characteristics at location a downstream of Najafgarh drain, b downstream of Delhi Gate drain, and c exit of Okhla barrage

Full size image

The best-case scenario for BOD is found when the drain flows are maintained the same way but organic load concentration reduced to nil. Almost similar results are obtained for middle of the river stretch (i.e. downstream of Delhi Gate drain) too. At the exit location, it can be observed from the results that water quality is suitable for class A water type usage for Eflow starting from 268.13 m3/s value for some scenarios. For other water usage classes, specifications are met, as streamflow from the upstream is increased. Same trends of DO values are found at other two upstream locations too.

Overall, the framework provides a generic guideline to estimate the Eflow based on water quality factor under different scenarios of climate change and pollution load. However, the study is based on the following assumptions and limitations:

  1. 1.

    The present study has considered only DO and BOD water quality parameters due to non-availability of observed data of other parameters.

  2. 2.

    Pollution sources considered are only point sources whereas diffuse sources can also be of prime importance.

  3. 3.

    As results are based on water quality data of the dry season period only, it is applicable only for such extreme periods. Scenario combinations considered for climate change impact studies are limited to the hypothetical values assumed (streamflow ±10 and ±20% and air temperature ±1, ±2, ±3, and ±4 °C) whereas in reality, the limits may vary.

  4. 4.

    Water quality data of the headwater flow and drains and meteorological data are assumed to be constant throughout the day.

  5. 5.

    The relationship developed between air temperature and water temperature is limited only to streamflow of Yamuna River and is not applicable for drain flow. In addition, it is also assumed that the increase/decrease in drain flow rates is uniform for different drains for different scenarios generated.

The proposed procedure can very well be replicated, given the information of all data limitations are addressed.

Conclusions

In the present study, a framework is proposed to estimate Eflow incorporating water quality variations under climate change. The framework is applied on Yamuna River stretch in Delhi using QUAL2K and the impact of climate change on water quality is analysed using 45 hypothetical climate change scenarios. A cubic regression equation, developed for relating air temperature and water temperature, is used to obtain projected water temperature from projected future air temperatures. DO and BOD values of river are significantly affected by change in streamflow values than any change in air temperature. Of the three factors (i.e. streamflow from the upstream, drain flow rates, and its organic load concentrations), organic load concentration is found to be the principal factor governing the changes in water quality under air temperature change conditions. Hence, reducing the organic load concentration can be an effective method for improving water quality. The guidelines provided will be useful to undertake necessary treatment policies, especially during dry flow seasons.

It is found that different values of Eflow for the Yamuna River as recommended by previous studies did not meet requirements of DO and BOD for any of the CPCB-recommended water usage classes. With the increasing streamflow values, specifications of class B, C, and D type usages are met. However, it is impossible to meet the specification of class A type usage for any flow value. Stretch between Najafgarh drain and Delhi Gate drain is found to be a hotspot, which failed to meet the class A type water specifications even with the increased Eflow value. An approximate value of streamflow between 700 and 800 m3/s is found to satisfy the water quality demands of the Yamuna River in Delhi for all conditions. Such huge discharge requirements indicate the level of pollution on one hand and the inconsiderate upstream discharge restrictions on the other hand. It would be appropriate to investigate the actual availability of water in the basin through hydrological modelling. Findings of the present study, along with the actual water availability in the basin, may help decision-makers to derive upstream release policies to ensure minimum Eflow or to derive treatment policies for the river stretch during periods of low flow.

For climate impact assessment studies, it would be desirable to downscale the future streamflow from precipitation and other climate variable outputs of various climate models. It is also necessary to obtain current water quality data and re-examine the changes in water quality parameters for different scenarios and conditions. The monitoring of water quality should be carried out more frequently by CPCB to obtain monthly time series of water quality indicators. More efforts are required for monitoring other water quality parameters like ammonium and phosphate, so that these can be included in determining the Eflow. The feasibility of the proposed framework should also be checked for river systems with diverse characteristics.

Overall, this study provides a framework for estimating Eflow considering water quality variations under climate change and illustrates the applicability of the framework for the Yamuna River. The proposed framework could also be applied to any river stretch to estimate the Eflow considering water quality under any climate change scenario. This framework could aid policymakers in (i) determining the minimum flow to be maintained in a river and (ii) deriving release policies of storage structures located upstream.