Complex dynamics of water quality mixing in a warm mono-mictic reservoir Science of the Total Environment

Vertical mixing of water quality constit- uents in a reservoir Incomplete mixing of constituents observed, in homo-thermal conditions. Pollutant loads reservoir operation regulated vertical mixing of constituents. on vertical mixing of constituents in new impoundedreservoirs,especially thoseconstructed tosupply domestic water. In thisstudy,samplingcampaigns were conducted in the Sabalan reservoir, Iran, to investigate vertical changes in constituent concentrations dur-ingtheyearinperiodswiththermalstrati ﬁ cationandhomo-thermalconditions.Theresultsrevealedincomplete mixing of constituents, even during cold months when the reservoir was homo-thermal. These conditions interacted to create a bottom-up regulated reservoir with sediment that released settled pollutants, impairing water quality in the Sabalan reservoir during both thermal strati ﬁ cation and homo-thermal conditions. Analysis of total nitrogen and total phosphorus concentrations revealed that the reservoir was eutrophic. External pollution loads, internal cycling of pollutants diffusing out from bottom sediments, reductions in in ﬂ ow to the reser- voir, and reservoir operations regulated vertical mixing and concentrations of constituents in the Sabalan reservoir throughout the year. © 2021 The Author(s). Published by B.V. This is an open access article under the CC BY license


Introduction
Lake thermal budgets strongly affect the depth profile of water quality constituents (hereafter 'constituents') (Noori et al., 2019a). The constituents are layered during thermal stratification and usually show decreasing/increasing trends with depth in the water body. Under strong thermal stratification, suspended substances and non-motile organisms can even be expected to accumulate in thin strata by rising or sinking to neutral buoyancy depth (Alldredge et al., 2002;Durham and Stocker, 2012). In the absence of thermal stratification, i.e., when the thermal gradient vanishes (homo-thermal conditions), constituents are usually mixed throughout the depth of lakes. Homo-thermal conditions can thereby improve water quality in deep layers through penetration of dissolved oxygen to the depth. On the contrary, homothermal conditions may further degrade water quality in upper layers by the transport of sulfide, nutrients and organic matter from sediments to the water surface. This may expose lake water to processes with direct consequences for ecosystem health (Lawson and Anderson, 2007;Yu et al., 2014a).
Variation of lake constituents with depth does not follow a simple pattern, e.g., homo-thermal conditions do not guarantee that lakes will be mixed chemically (Calamita et al., 2019). For example, there may be a lag in chemical mixing under homo-thermal conditions in lakes due to redox conditions in deep layers, causing the release of pollutants to the water column (De Jonge et al., 2012;Aradpour et al., 2020). Homo-thermal conditions can also occur in meromictic lakes, which are chemically layered. In addition, the development of turbulence in bottom layers may diffuse out settled pollutants in shallow lake sediments, e.g., suspended and dissolved solids such as nutrients, to the water column (Beutel, 2003;Markou et al., 2006;Noori et al., 2019b). Therefore, a depth concentartion gradient may be created for some constituents from bottom to surface layers when a lake is homothermal. Changes in other environmental conditions, such as oligotrophication, eutrophication, lake morphology, and brownification, may also influence the variation with depth in constituents in lake water (Fennel and Boss, 2003;Håkanson, 2005;Clegg et al., 2007;Johansson et al., 2007;Leach et al., 2018). Carpenter (1983) and Read et al. (2015) concluded that both depth and surface area contribute to the variation in constituents with depth in lakes. Wetzel (2001) reported that the variation in nutrients with depth in (unproductive) oligotrophic lakes is different from that in (productive) eutrophic lakes. Leach et al. (2018) found that the variation in chlorophyll concentrations with depth in lake water is influenced more by light attenuation than thermal stratification. In addition to these factors, hydrological variability induced by anthropogenic manipulation of inflow/outflow can affect the variation with depth in constituents in lakes and reservoirs (Ford, 1990;Cott et al., 2008;Baldwin et al., 2008;Gikas et al., 2009;Nowlin et al., 2009;Noori et al., 2018;Mi et al., 2019). Reservoirs differ from natural lakes in terms of hydraulic residence time (HRT), loads of total suspended solids, and productivity (Walker, 1985;Søballe et al., 1992). Reservoirs, especially those impounded to supply domestic water, require competent management of constituents since increases in nutrients can significantly increase eutrophication symptoms. Eutrophication can threaten the positive functions of lakes (Modabberi et al., 2020) and potentially producing blooms of odor-and taste-affecting aquatic species and cyanobacteria (Schindler, 1977).
Considering the large number of factors affecting water quality in reservoirs/lakes, more investigations are needed to determine the variation with depth in constituents in new impounded reservoirs supplying domestic water. Therefore, in this study, sampling campaigns were conducted in Sabalan reservoir, Iran, to: (i) investigate vertical changes in constituents during thermal stratification and homo-thermal periods; (ii) calculate the strength of thermal stratification and oxycline depth; (iii) explore the possibility of chemical stratification in the reservoir; and (iv) identify important factors influencing nutrient concentrations in reservoir water and associated quality deterioration. The aim was to improve understanding of the fate of physiochemical contaminants in the water system and assist those managing water supply reservoirs in less studied areas, including Iran.

Study area
Sabalan reservoir is located in Ardabil province in northwestern Iran, a temperate zone. The dam (clay-core, rock-fill) was constructed on the Qareh-su river in 2006 to annually supply irrigation and domestic water amounting to around 105 million cubic meters (MCM) at normal operating level (reservoir water surface at 1123 m above sea level) (Fig. 1).
The active and dead storage capacity of the Sabalan reservoir is 94 and 11 MCM, respectively. The maximum depth of the reservoir is around 70 m, and its surface area at the normal operating level is around 4.4 km 2 . During our sampling campaigns, which ran from May 27, 2017 to August 5, 2018, the HRT of the reservoir varied from 144 to 5850 days (Fig. 2a). Since inflows (Q inflow ) and outflows (Q outflow ) were found to be unbalanced, the HRT was calculated as: 2V/(Q inflow + Q outflow ), where V is the reservoir volume (Strasškraba and Hocking, 2002). Details about volume-area characteristics of the reservoir at different depths, obtained by mass balance computations using the CE-QUAL-W2 model (Noori, 2020), are shown in Fig. 2b.
The Qareh-su river basin occupies an area of around 5469 km 2 , and the mean slope of the river is about 10%. Annual potential evaporation in the dam location (1380 mm/yr) exceeds annual precipitation (440 mm/yr). According to 43-year meteorological data, January and July are the coldest and warmest months in the region, with a mean monthly temperature of around −0.7 and 20.8°C, respectively. During the study period, February 2018 (mean monthly air temperaturẽ 1.1°C) was the coldest month and July 2018 the warmest month (mean monthly air temperature~23.6°C). The coldest and warmest days during the study period were on January 10, 2018 (~−9.5°C) and July 21, 2018 (~28.7°C), respectively. The maximum monthly average wind speed during the study period was around 3.4 m/s (February-March), and the minimum was around 2.5 m/s (September-October) (Table S1). According to 43-year meteorological data, the winds predominantly blow from the west and southwest to the east in the dam valley ( Fig. S1). Around 70%, 6%, and 2% of winds blow at speeds of 0.5-3.6 m/s, 5.7-8.8 m/s, and >8.8 m/s, respectively (Fig. S2). According to 20-year historical data, the mean river inflow to the Sabalan reservoir is around 2.12 m 3 /s (Fig. 3).
In terms of geology, the Sabalan dam is located on igneous rock (e.g., basaltic-andesite, basalt, and agglomerate), and alluvial depth is about 8 m (Noori, 2020). The reservoir water is thermally stratified in most months of the year and only mixes during cold months (usually December-March) (Aradpour et al., 2020). Therefore, it can be classified as a warm mono-mictic lake (Lewis, 1983).
Frequent drought events and climate change impacts have exacerbated water problems in Ardabil province, by changing the spatiotemporal features of precipitation and temperature and by restricting the renewable water supply (Vousoughi et al., 2013;Modarres et al., 2018). Thus, the authorities in Ardabil province decided to construct the Sabalan dam to meet the growing demand for water in the agricultural, domestic, and industrial sectors. Against the background of reductions in rainfall in Ardabil province during recent decades, conserving water in the Sabalan reservoir by controlling the outflows has been one of the main dam operation strategies to mitigate drought events in the province.
Downstream irrigation systems and drinking water supplies are not yet completely equipped to withdraw water from the reservoir. Thus, the reservoir is mainly regulated by the manipulation of outflows to meet downstream environmental flow, which dramatically increases the HRT in the reservoir. However, water quality in the reservoir can be impaired by increasing the HRT. In addition, due to the unsustainable practices of industrial and especially agricultural areas (Maghrebi et al., 2020), the Qareh-su river, which is the main drainage outlet for the Ardabil plain, receives large pollutant loads originating from anthropogenic and natural sources and transports them to Sabalan reservoir. Some evidence of water pollution in the Sabalan reservoir was presented by Aradpour et al. (2020). However, there is a lack of information regarding the cycling of the contaminants in the reservoir and the contribution of different factors to water quality deterioration. Therefore, it is important to investigate the variation in concentrations of constituents with depth in the reservoir.

Sampling campaigns
According to UNEP/WHO (1996), sampling at the deepest point is sufficient for monitoring water quality in a lake/reservoir with good horizontal mixing. In reservoirs, the deepest point is usually located close to the dam structure. Areas close to the intake should also be considered for sampling if the intake is located far from the dam structure (ILEC, 1999). For large lakes/reservoirs, the number of sampling points can be computed as 'log 10 β', where β is a lake area in km 2 (UNEP/ WHO, 1996). Considering the area of the Sabalan reservoir at a normal operating level (about 4.4 km 2 ), one sampling point would be sufficient {log 10 (4.4) = 0.64~1} to monitor water quality in the reservoir. Therefore, sampling was performed on 12 occasions at the deepest point in Sabalan reservoir, adjacent to the dam structure ( Fig. 1). To assess the effects of thermal stratification and homo-thermal conditions on constituents in the Sabalan reservoir, sampling covered more than a year (May 27, 2017 to August 5, 2018).
On each sampling occasion, eight samples were taken at different depths i.e., 0.5, 3, 6, 10, 15, 20, 25, and 35 m since the active storage capacity of reservoir extends to about 40 m. Points deeper than 35 m were excluded since they were located close to the dead storage zone, where water is not usually withdrawn for usage. Also, two samples were taken from the reservoir inflow and outflow. In each water sample, 13 constituents were analyzed: pH, dissolved oxygen (DO), water temperature (Tem), conductivity, total dissolved solids (TDS), total suspended solids (TSS), turbidity (Tur), nitrate expressed as N (NO 3 − -N), nitrite expressed as N (NO 2 − -N), ammonia expressed as N (NH 3 -N), total nitrogen expressed an N (TN), orthophosphate expressed as P (PO 4 3− -P), and total phosphorus expressed as P (TP). Note that regular monthly sampling was not conducted during the study period. For example, due to unsuitable weather conditions and equipment constraints, sampling was not performed in June 2017, January 2018, and July 2018. In addition, only DO and water temperature were analyzed in samples taken in February 2018. In total, 13 constituents were analyzed through 1576 experiments during the study period to determine their variation with depth in the reservoir. Hydro-Bios Standard Water Sampler acc. to Ruttner 1L (Hydro-Bios Company, Germany) was used to take samples at different depths in the Sabalan reservoir. This instrument is equipped with a thermometer for in situ measurement of water temperature at each sampling depth (accuracy~0.2°C). Dissolved oxygen concentration and pH in water were also measured in situ using intelligent DO and pH meters, i.e., Lutron YK-2001DO, Taiwan, with an accuracy of around 0.5 mg/L and 0.02, respectively. After collection, samples were filtered through a fine membrane (0.45 μm) except those samples for analysis of TSS, TN, and TP. All samples were acidified by sulfuric acid to reduce the pH < 2 (except samples for PO 4 3− measurements), and stored in polyethylene vessels. All samples were kept in a refrigerator (temperaturẽ 4°C) during transport to the laboratory and analyzed within 48 h. In general, all analyses were done according to the guidelines suggested by standard methods for the examination of water and wastewater (Baird and Eaton, 2017). Conductivity and TDS were analyzed by conductivity meter 'WTW inoLab Cond 7110'. 'Hach DR 5000™ UV-Vis Spectrophotometer' was used to analyze NO 3 − -N (Hach method 10,020 and λ = 410 nm), NO 2 − -N (Hach method 8507 and λ = 507 nm), NH 3 -N (Hach method 8038 and λ = 425 nm), TN (Hach method 10,071 and λ = 410 nm), PO 4 3− -P (Hach method 8048 and λ = 880 nm) and TP (Hach method 8190 and λ = 880 nm). 'Hach DR 2000™ Spectrophotometer' has also been used to measure turbidity. TSS was trapped by a fine Whatman membrane filter (0.45 μm).
All constituents were analyzed in duplicate, and mean values were used in calculations. Deionized water blanks were used to assure quality control of the analyses and revealed a satisfactory recovery rate (95-107%) and relative standard deviation (<5%).
Since DO is an important water quality index, we also calculated the depth of the oxycline, essential information for the authorities managing the Sabalan reservoir. Oxycline depth is the depth at which a sudden decline in DO occurs (Zhang et al., 2015;Valerio et al., 2019). In this study, a critical DO value of about 0.5 mg/L/m was selected (Valerio et al., 2019). To better illustrate the stratification of DO in the Sabalan (1) (Yu et al., 2010).
where, DO s and DO d are DO concentrations in surface (depth = 0.5 m) and deep (depth = 35 m) layers, respectively. A SCS-DO value of zero indicates holo-mixing conditions (Yu et al., 2010).

Variation in water temperature and DO with depth
The variation in water temperature with depth in the Sabalan reservoir is shown in Fig. 4 (left panel), while raw data on measured water temperature at different depths in the reservoir during the sampling period are presented in Table S2. Thermal stratification gradually developed in early April and lasted until late November, when the difference in water temperature between the surface and bottom layers reached more than 8°C (Fig. 4). Combined evaluation of Table S1 and Fig. 4 revealed that environmental conditions such as warm air temperature and weak winds contribute to stratification onset. Strong thermal stratification formed a distinct thermocline in May, and this then progressed downwards until November.
Low inflow to the reservoir during summer (July, August, and September) and no inflow in August and early warm September were observed during the study period (Fig. 3). Considering the environmental conditions (e.g., warm air temperature, higher atmospheric radiation, and weak winds) and no inflow (inflow is usually cooler than the reservoir surface water) to the reservoir in warm months, the differences between water temperature in surface and deep layers increased in summer, as shown in Fig. 4. Inflow to the reservoir increased after September 27, as shown in Fig. 3 (with a maximum value of 6.5 m 3 /s on October 8), causing cold water to be pushed to the bottom. However,  during the cold months, beginning in late fall, and with intensifying external forces such as strong winds (Table S1), higher rainfall, and greater inflow to the reservoir (Fig. 3), the epilimnion gradually deepened so that thermocline started at a depth of more than 15 m and 20 m in October and November, respectively. Eventually, thermal stratification disappeared in early December, and this situation lasted until late March when the difference in water temperature between the surface and bottom layers was still small (<2°C), and wind power was at its maximum (Table S1). The water temperature did not reach below 3°C in cold months (winter). All these findings confirm that the Sabalan reservoir is a warm mono-mictic lake that only experiences annual mixing in winter (Lewis Jr., 1983). Compared with long-term hydro-meteorological averages, the sampling period was warmer and drier in terms of mean monthly and minimum daily air temperature and inflow in almost all months (Table S1 and Fig. 3). Such temporal changes would not alter the reservoir type from warm mono-mictic to dimictic since the reservoir surface area has never frozen since its impoundment in 2006, according to the dam operator's observations (2020 data). However, thermal stratification in the reservoir is long-lasting and severe. During the period of thermal stratification, the maximum difference in water temperature between the surface and bottom layers reached 18.6°C in August 2017. To better understand the strength of thermal stratification in the Sabalan reservoir, we calculated the difference in water temperature (ΔTem,°C) between the surface (depth = 0.5 m) and deep (depth = 35 m) layers at the time of sampling, as suggested by Calamita et al. (2019). The results revealed a gradual increase in ΔTem from February to August, followed by a decrease from August to February (Fig. 5a). These increasing and decreasing trends are characteristic of lakes located in the mid-latitude zone (Noori et al., 2019b). Note that since no data were available for January, ΔTem was not calculated for that month.
The variation in DO with depth in the reservoir is shown in Fig. 4 (right panel), while raw data on measured DO at different depths in the reservoir during the sampling campaigns are presented in Table S3. Also, the calculated oxycline depths in the reservoir are shown in Fig. 5b. Values in green in this radar diagram show the oxycline depth in the reservoir for each month in the study period except January, when no sampling was performed. However, the critical DO value did not occur within the reservoir depth profile in February, so the maximum depth of active storage capacity in the Sabalan reservoir, i.e., 40 m, was selected as the oxycline depth in that month.
Dissolved oxygen concentration in surface water (as a result of direct contact with the atmosphere and wind action) ranged between 7.2 and 11.2 mg/L, while a DO deficit was observed in deeper layers (range 1.1-4.5 mg/L). During periods of thermal stratification, oxygen intake from the atmosphere to deep layers is restricted by thermocline. Additionally, organic-rich sediments present in the reservoir bottom (Aradpour et al., 2020) may contribute to sediment oxygen demand and reduce DO in deep layers. Although no information was available on water transparency (Secchi disk readings), we observed that the sampling devices we employed were not visible well at depths greater than 3 m during all sampling campaigns. This indicates that there is no effective light penetration to layers deeper than 3 m to effectively support primary production, and consequently to produce oxygen. Chemical oxygen demand, total organic carbon, and biochemical oxygen demand can also contribute to DO variation, although no data existed to investigate their contribution to DO changes in the depth of Sabalan reservoir.
DO depletion in deep layers was also observed even when the Sabalan reservoir was in its homo-thermal state in the winter months (Fig. 4), indicating that the reservoir is chemically stratified in terms of DO throughout the year. Declining inflow to the reservoir as a result of low rainfall during the study period (Fig. 3) and long HRT (Fig. 2a) contributed further to DO depletion in deep layers. The highest DO concentration in deep layers (4.5 mg/L) was observed during February, as a result of more severe wind action and lower air temperature (Table S1), and oxygen transfer from the surface to deep layers due to homothermal conditions in the reservoir (Fig. 4).
The SCS-DO values found in this study (~0.78-1.58) indicate that the Sabalan reservoir is chemically stratified in terms of DO (Fig. 5a). Such incomplete chemical mixing, even during holo-mixing conditions, would change the chemical composition and redox condition in bottom layers of the Sabalan reservoir, as also reported for Lake Zürich (Naeher et al., 2013;North et al., 2014). However, our findings indicated hypoxia (DO concentrations <6.5 mg/L; Nürnberg, 2004) in layers deeper than 20 m during both thermal stratification and homo-thermal conditions in the reservoir (Fig. 4). Noted that, the threshold value for hypoxia determination can fall into the range of 4.8 and 10.5 mg/L based on the type of waterbody, agencies, and the future applications of water (Nürnberg, 2002). Here, the threshold value of 6.5 mg/L was selected to specify the hypoxia in the reservoir.
A closer look at Fig. 4 shows that deep layers in the reservoir experienced anoxic conditions (DO concentration <2 mg/L; Nürnberg, 2004) during October, November, and March. A similar situation has been observed previously in many deep lakes, with DO close to zero in deep layers (Satoh et al., 2000;Yacobi, 2006;Nishri et al., 2011;Marshall et al., 2013;Yu et al., 2014a;Fukushima et al., 2017;Noori et al., 2018;Valerio et al., 2019). In the present case, DO decreased continuously with depth in the Sabalan reservoir as a result of the thermal stratification from late March, and reached its minimum in October and November. In addition to the thermal stratification effect, methane production due to mineralization of organic carbon in sediment, oxidation of some minerals in sediment, and decomposition of the settled nutrient may contribute to DO depletion in deep layers of lakes (Di Toro et al., 1990;Livingstone and Imboden, 1996;Müller et al., 2012;Noori et al., 2015;Torabi Kachoosangi et al., 2020).

Variation in concentrations of nitrogen and phosphorus compounds with depth
The -P, TN, and TP in the Sabalan reservoir is shown in Fig. 6, while raw data on these constituents in different depths in the reservoir during the sampling campaigns are presented in Tables S4-S9. All these parameters, except NO 3 − -N, showed a general increasing trend with depth in the Sabalan reservoir during thermal stratification and even in the homo-thermal period. This has been reported previously for many meso-eutrophic thermally stable lakes characterized by a nutrient-poor epilimnion and nutrientrich hypolimnion (Gervais et al., 2003). In general, thermal stratification contributes to variation in nutrient concentrations with depth (Holm-Hamen et al., 1976;Varela et al., 1994;Mellard et al., 2011). This creates a nutricline depth, where severe gradients can be observed with depth in concentrations of nutrients (Leach et al., 2018). NO 3 − -N concentration in Sabalan reservoir increased from the water surface downwards during thermal stratification, to reach a peak in the metalimnion. Thereafter, the concentration decreased or became constant with depth (Fig. 6). Considering the anoxic conditions prevailing in deep layers (Fig. 4), the rate of nitrate consumption by denitrification is likely to exceed its production rate by nitrification, as nitrification and denitrification usually occur in water enriched and depleted of oxygen, respectively (Xiong et al., 2017). This can reduce nitrate concentrations to close to zero in the bottom of stratified anoxic lakes, according to Satoh et al. (2000). A similar trend has been reported for the nitrate profile in deep lakes/reservoirs by Miyajima (1994), Miyajima et al. (1997), Satoh et al. (2000), Kim et al. (2016), andNoori et al. (2018). As a byproduct of denitrification, NO 2 − -N peaked in the bottom layers of the Sabalan reservoir during thermal stratification (Fig. 6).
Ammonia is released under anoxic conditions following ammonia build-up in bottom sediments due to a loss of biological conversion of ammonia to nitrate and a drop in ammonia uptake by anaerobic microorganisms (Beutel, 2006). Therefore, NH 3 -N concentrations were low in surface layers and peaked in deep layers of the Sabalan reservoir (Fig. 6). In general, nitrogen and phosphorus compounds (TN and TP) decrease at the water surface because their mineral forms may be consumed by algae and plants in the water.
Similarly to the thermal stratification period, an increasing trend with depth in the Sabalan reservoir was observed for all nitrogen and phosphorus compounds except NO 3 − -N under homo-thermal conditions (Fig. 6). Vertical distributions of NO 3 − -N were similar to those observed in the stratification period. After the long stratification period, the transfer of nutrients to the surface layers could expose the Sabalan reservoir to eutrophication, directly affecting water quality.
TP concentration varied between 0.11 and 0.57 mg/L at the surface (depth = 0.5 m) and between 0.68 and 1.85 mg/L at 35 m depth in the reservoir. TN concentration varied between 2.2 and 4.7 mg/L in surface water (depth = 0.5 m) and between 6.1 and 8.2 mg/L at 35 m depth. Considering the threshold value for the occurrence of eutrophication in freshwater systems of greater than 1.39 mg/L and 0.084 mg/L, respectively for TN and TP (Vollenweider, 1982), our data revealed that the Sabalan reservoir is eutrophic. In general, oligotrophic lakes under natural succession require long times (up to thousand years) to become eutrophic (Shen et al., 2013). However, our results indicated that the Sabalan reservoir, impounded in 2006, is already an eutrophic aquatic ecosystem. This rapid conversion from oligotrophic to eutrophic condition in the reservoir can be considered as a consequence of the following factors: i. Discharge of large pollution loads originating from both natural (e.g., weathering of rocks) and anthropogenic (e.g., industrial and agricultural zones) sources. Unfortunately, populated areas upstream of the Sabalan reservoir usually use shallow pit latrines (<10 m) for wastewater disposal or directly discharge wastewater into the Qareh-su river. These effluents ultimately discharge into the reservoir, which is the main drainage outlet of the Qareh-su river basin. Excess mineral and organic fertilizers used in agriculture also discharge into the Qareh-su river and its tributaries, resulting in large external nutrient loads to the reservoir, especially during warm months (April to September) when agricultural activities in the watershed are at their peak (Fig. 7). According to the Director of Agriculture for Ardabil Province, this province has the highest proportion of agricultural land of all provinces in Iran, with the rapid expansion of the agricultural area in the province from about 1000 to 10,000 ha during the past decade. The expansion in agricultural land has been accompanied by increasing fertilizer use in the province. Given the unbalanced use of fertilizers in Iran (FAO, 2005) and lack of treatment systems for agricultural effluents, it can be concluded that nutrients from agricultural land have likely contributed to the conversion of Sabalan from oligotrophic to eutrophic status. In general, anthropogenic activities play a major role compared to natural sources, e.g., transport of phosphorus compounds from agricultural and residential areas can be fivefold and 10-fold higher, respectively, than transport from forests (Shen et al., 2013). ii. Internal cycling of pollution loads in the Sabalan reservoir. As shown in Fig. 7, it can be concluded that the maximum TN and TP loads leaving the reservoir occurred in the months October to March when the input loads were very small. This presumably indicates that the TN and TP loads leaving the reservoir may originate from internal nutrients settled in bottom sediments, leading to an internal fertilizer in lakes (Augustyniak et al., 2019). However, there is no data of TN and TP neither at the over-bottom water nor at the deeper part of the reservoir than 35 m to completely support this hypothesis. Therefore, more studies are required to investigate this hypothesis. iii. Reservoir management strategy. Although the Sabalan reservoir was impounded in 2006, downstream irrigation systems and drinking water supplies are not yet completely equipped to withdraw water from the reservoir. Given the reduction in rainfall and increase in air temperature in the study area in recent decades, which has resulted in higher evaporation rate and decline of water level in the reservoir, conserving water in Sabalan reservoir has been one of the main strategies to mitigate drought events in the Ardabil province. This has resulted in increasing the HRT over short periods in the reservoir, e.g., HRT varied from 144 to 5850 days during our study period (Fig. 2a). The longer the HRT of water in the reservoir, the more time there is for nutrients to accumulate in both sediments and water strata. iv. Reduced inflow. A decline in the inflow to the Sabalan reservoir was observed during the study period (Fig. 3). Due to the reduction in inflow, the reservoir experienced a decline in water level and volume during the period (Fig. S3). Declining inflow, an important driver of dilution of water in lakes/reservoirs, can negatively influence reservoir water quality.

Variation in pH, turbidity, TSS, TDS, and conductivity with depth
Measured pH was higher in surface layers (range 8.6 to 10.9), with a vertical decreasing trend with depth in the Sabalan reservoir during stratification and even during homo-thermal conditions ( Fig. 8 and Table S10). Eutrophic condition is likely to contribute to higher pH in the surface layers in lakes/reservoirs due to algal growth (Yu et al., 2014b). During the day, consumption of dissolved CO 2 by algae for growth increases pH in surface layers. Releases of CO 2 during organic matter decomposition in the sediments may decrease pH in deep layers (Miyajima et al., 1997;Leidonald et al., 2019). In addition, CO 2 concentration increases with depth in the Sabalan reservoir because photosynthesis is not possible in deep layers due to light attenuation (water depth > 3 m).
It was found that turbidity and TSS were not significantly correlated (correlation coefficient~0.20), although both showed similar decreasing trends with depth in the Sabalan reservoir ( Fig. 8 and Tables S11 and S12). The values mainly peaked close to the metalimnion, depending on density gradients during thermal stratification. Considering the eutrophic condition in the reservoir, deep chlorophyll maximum most likely contributes to the peak in turbidity and TSS close to the metalimnion (Milstein and Feldlite, 2015). Deep chlorophyll maximum, i.e., the subsurface maximum in chlorophyll in thermally stable lakes/ reservoirs, usually occurs close to the metalimnion (Williamson et al., 1996).
Experimental data show that large TSS loads in seasonal riverine inputs can also influence plankton dynamics and lake metabolism (Johengen et al., 2008). Deep layers in the Sabalan reservoir showed maximum turbidity and TSS during mixing (Tables S11 and S12), which probably resulted from external sediment loads discharged to the reservoir after heavy rainfall (usually in cold months, when the reservoir is homo-thermal). Similar vertical trends in TSS and turbidity in lakes/reservoirs have been reported by Elçi (2008), Bonalumi et al. (2012), and Kim et al. (2016).
Total dissolved solids and conductivity were significantly correlated (correlation coefficient~0.97) and showed an increasing trend with depth in the Sabalan reservoir ( Fig. 8 and Tables S13 and S14). Greater values of TDS and conductivity at depth compared with surface layers are probably due to lower pH in bottom layers, where CO 2 is released from bottom sediments. The lower the pH, the higher the ion concentration, resulting in higher conductivity and TDS values in deeper layers (Chimney et al., 2006;Elçi, 2008). As can be seen from Fig. 8, there was little variation in conductivity and TDS with depth when the Sabalan reservoir was homo-thermal, but they were affected in deep layers (35 m) due to the anoxic conditions in the bottom of the reservoir.

Strength of stratification of reservoir water constituents
In general, the results obtained in this study indicated an incomplete mixing of constituents in the Sabalan reservoir. The results for chemical stratification of reservoir water (Table 1) indicated no complete mixing of constituents (SCS values above zero) even under homo-thermal conditions. However, the strength of stratification was close to zero in months when the reservoir was homo-thermal (Dec 2017 to Mar 2018) for some constituents, such as pH, NO 3 − -N, NO 2 − -N, TDS, and conductivity, indicating weak gradients of these constituents with depth in the reservoir (weak mixing). It should be noted that, although the SCS threshold suggested by Yu et al. (2010) produces helpful information for quantifying the strength of stratification of constituents, it does not fully address the strength in lakes/reservoirs. This index seems to work well for those constituents that show regular increasing or decreasing trends with depth in lakes/reservoirs, such as TN, TP, and TSS in the Sabalan reservoir. Meanwhile, some constituents such as chlorophyll-a usually increase from surface water to deep chlorophyll maximum and then decrease to the bottom of lakes (Williamson et al., 1996;Milstein and Feldlite, 2015;Leach et al., 2018). Our results revealed that NO 3 − -N increased from surface water to reach a peak close to the metalimnion and then decreased or became constant with depth (Fig. 6). These trends may give inconsistent results when the strength of stratification is quantified using the SCS index suggested by Yu et al. (2010). For example, the index indicated almost complete mixing for NO 3 − -N in May 2017 and June 2018 (SCS value of around −0.06 and −0.03, respectively, i.e., very close to zero) ( Table 1), but vertical profiles showed a severe gradient for NO 3 − -N with depth in Sabalan reservoir during these months in the present case (Fig. 6).

Conclusions
Vertical mixing of water quality constituents in the Sabalan reservoir (impounded 2006) was investigated. The results showed that high external and internal pollution loads, long HRT over short periods in the reservoir, and severe thermal stratification resulted in incomplete chemical mixing, even during cold months when the reservoir was homo-thermal. Deep layers in the Sabalan reservoir showed high concentrations of nutrients, mainly diffused from bottom sediments to the water column, resulting in eutrophication. Based on these findings, the authorities responsible for managing the Sabalan reservoir water resources must urgently improve reservoir water quality because if incomplete mixing continues over time, it will alter the chemical composition and redox conditions in deep zones of the reservoir. Considering the very long HRT in the reservoir, changing the current reservoir operation strategy would provide faster water quality improvements than reducing external/internal pollution loads. Conserving water in the reservoir by controlling the outflow is one of the main strategies to mitigate drought events in downstream regions. This has dramatically increased the HRT in the reservoir, allowing more time for pollutants to accumulate in sediments and water.
The study period was warmer and drier than the long-term average, and the Sabalan reservoir seemed to be in critical condition in terms of water quality. Our observations cover only a short period but represent the most extreme conditions imposed on the reservoir over the year. Thus, our results can be used to refine future reservoir operation strategies, with a reasonable safety factor considering possible impacts of future climate change.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.