Causes of continuous and short-term hypoxia in rivers entering the sea: a case of Minjiang River in Fujian Province

Dissolved oxygen (DO) plays a significant role in maintaining the health of an aquatic ecosystem, and significantly affects the ecosystem and biogeochemical cycle of rivers and oceans (Herrmann & Keister 2020; Li et al. 2020b; LaBone et al. 2021). The occurrence and long-term persistence of hypoxia (low DO) negatively affects aquatic and benthic organisms, resulting in the deterioration of water quality and the loss of water ecosystem (Rabalais et al. 2002; Turner et al. 2005; Yang et al. 2020). Hypoxic and anoxic aquatic environments are normally observed in naturally formed oxygen minimum zones, fjords, and upwelling zones (Diaz 2001; Rabouille et al. 2008; Rabalais et al. 2009, 2018). However, anthropogenically induced hypoxia and low DO development and persistence in rivers, estuaries, and adjacent coastal waters around the world are commonly associated with an increase in nutrient loading as population growth and resource intensification rises (Wei et al. 2007; Rabouille et al. 2008; Rabalais et al. 2009, 2010; Xu et al. 2020; Zhang et al. 2020; Testa et al. 2021).

It is now widely accepted that low DO in water column is associated with eutrophication and increasing allochthonous inputs of organic matter and nutrients, such as dissolved organic carbon (DOC) and ammonia (NH₄), that require oxygen for oxidation (Rabouille et al. 2008; Buranapratheprat et al. 2021; Rigaud et al. 2021). Additionally, the phenomenon of low DO develops when organic matter-rich silt at the bottom of deep-water lakes or reservoirs accelerates the absorption of DO with minimal reaeration under the situation of thermal stratification (Noori et al. 2021, 2022). Moreover, oxygen depletion in estuaries is related to the residence time of waters, which is determined by freshwater inflow, depth, water volume, wind mixing, and tidal exchange (Wang 2009; Quinones-Rivera et al. 2010; Wang et al. 2012; Hong & Shen 2013; Rabalais et al. 2018). Therefore, the investigation of these factors that affect the kinetic process of DO will facilitate the prediction of the spatial and temporal distributions of DO, for determining the extent of hypoxic water and suggesting mitigation measures (Turner et al. 2006; Zhang et al. 2016; Wenjing et al. 2020).

The hydrodynamic changes and pollutant migration and transformation processes involved in estuarine hypoxic water environment are very complex (Cai et al. 2020; Sun et al. 2020a). Numerous researchers have qualitatively analysed the response state of the physical, biological, and chemical factors in hypoxic water environment using observation data (Chi et al. 2020; Guo et al. 2020; Sun et al. 2020b). For example, Wehmeyer & Wagner (2011) evaluated the relationship between water flows and DO in the Roanoke River between Roanoke Rapids Dam and Jamesville, North Carolina using the hydrological, water quality, and meteorological data from 2005 to 2009. However, observational analyses cannot meet the requirements of an extensive, detailed, and quantitative research for evaluating the causes of the long-term evolution of watershed water environment. In the past decades, numerous numerical modelling studies that simulate the underlying dynamic mechanisms of the development and variability of hypoxia in numerous rivers and estuaries have been developed (Scavia et al. 2004; Wei et al. 2016; Yau et al. 2020; Jarvis et al. 2021; Yu et al. 2021). For example, Xia et al. (2010) investigated the effects of river discharge, atmospheric winds, and tidal forcing on the spatial and temporal distributions of DO in the Caloosahatchee River Estuary using a numerical estuarine and coastal ocean circulation hydrodynamic and eutrophication model based on an Environmental Fluid Dynamics Code (EFDC).

However, there is an increase of peer-reviewed articles reporting the different types of models to detect DO variation trends in many coastal regions mainly ascribed to individual hydrodynamic factors or water quality factors (Xia et al. 2011; Zhang et al. 2016; Lajaunie-Salla et al. 2017). Low DO occurs when the amount of DO in the water column is decreased by the process of respiration at a faster rate than resupply through air–water exchange or advection (Kumarasamy 2015; Swanson et al. 2016; Harano et al. 2018). How to quantitatively describe the reaeration rate and DO consumption involved in the kinetic process of DO is the key to the study of low DO. In this study, the water age concept and average dissolved oxygen consumption (ADOC) concept were used to investigate the development of continuous and short-term low DO in the Minjiang River Estuary by using a three-dimensional EFDC model. The EFDC is a multi-purpose, open-source, and free-of-charge 3D fluid dynamics and water quality model that has been extensively used and documented to simulate circulation, thermal stratification, water quality and eutrophication, and sediment transport in a number of lakes, reservoirs, rivers, and estuaries (Hamrick 1992; Li et al. 2011). Therefore, EFDC could more accurately simulate the spatial and temporal variation of DO in the Minjiang River than CE-QUAL-W2 and other vertical two-dimensional models. As water age reflects the time elapsed for water body or dissolved substances to be transported from one point to another, making it a useful timescale for describing the reaeration time or contaminants residence time in rivers and estuaries (Deleersnijder et al. 2001; Shen & Haas 2004). In addition, four sensitivity analysis model scenarios were developed to examine the impacts of normal power generation and rapid increase in the water flow from the Shuikou dam on the spatial and temporal variability of the water age, ADOC, and DO in the lower reaches of the Minjiang River.

The Minjiang River model was developed by Zhang et al. (2016, 2021) based on the EFDC. This model used quadrilateral grids consisting of 2,743 active cells with a cell size varying from 50 to 1,000 m resulting from grid optimisation and orthogonality. Three uniform vertical sigma layers were applied to better fit the bottom topography. The model was driven by the tidal level, discharge of Shuikou Hydropower Station, inflow tributaries, atmospheric forcing, and surface wind field. The 15 state variables in the water column were simulated by the water quality model, which contained green algae, three types of carbon, five types of nitrogen, four types of phosphorus, DO, and chemical oxygen demand (Zhang et al. 2021). The model was initially run for several days for each flow condition to obtain a dynamic equilibrium condition. For this model, a 60-second time step was used in the simulations with no signs of numerical instability.

The Year Book of Hydrology P.R. China and the China Meteorological Network provided the daily atmospheric data for 2013. Since 2013, the Hydropower Station, hydrological stations (at YBHC), and frequent monitoring locations along the Minjiang River have provided data on the Shuikou daily discharge, water quality, and inflows of tributaries. The daily tidal level, salinity, and temperature data were provided by the hydrodynamic model of Shuikou Hydropower Station–Minjiang open (Zhang et al. 2021). In addition, the refined data of 2013 used three automatic water quality monitoring stations (Zhuqi, Wenshanli, and Baiyantan), including turbidity, DO, conductivity, pH, permanganate index (COD_Mn), total phosphorus (TP), and ammonia nitrogen (NH₃-N), with a frequency of daily or 3 h.

The data used to fuel the model came from the China Meteorological Network and include the wind field, air pressure, temperature, relative humidity, evaporation, rainfall, solar radiation, and cloud cover. The nearby hydrological stations were where the mainstream and tributary flow information was collected. Each tributary's water border in the basin where the pollutants were estimated and assigned. The Oregon State University-developed TPXO 6.2 global tidal model calculated the tide level. The detailed boundary settings for this model had been described in our previous research (Zhang et al. 2021).

In the initial test settings of the model, the DO, temperature, and other water quality parameters were set in accordance with the measured values, and the water surface height was set at 2.5 m. For each flow condition, the flow field, salinity, and water quality concentration were set as the equilibrium conditions derived from the model simulation over a number of days.

Model parameter calibration was performed using data from January through December 2012, and verification was performed using data from January through December 2013. Hydrological calibration mainly verified the tide level, flow, and temperature. The model validation showed that the average absolute error of the tidal level, water temperature, and the average relative error of flow were 0.23 m, 1.04 °C, and 19.9%, respectively. Water quality calibration mainly verified the DO, TN, TP, NH₃-N, and BOD₅. Each monitoring station's relative error range for DO concentration was 3.86–36.83%. The detailed model calibrations and verification could be found in Zhang et al. (2021). The Minjiang River model could better describe the hydrodynamic and water quality in the lower reaches of the Minjiang River, and could fully reflect the real-time (frequency per hour) variation of DO in Minjiang River.

Estimate the effect of river discharge on DO dynamics by the validated EFDC model. Three model scenarios (Table 1) were developed at the most unfavourable water temperature (30 °C), the minimum ecological flow, average annual flow, and maximum generation flow of the water discharge from the Shuikou dam were 308, 1,360, and 2,100 m³· s⁻¹, respectively. The influence of river flow on the kinetic process of DO was investigated from three major aspects (Wehmeyer & Wagner 2011): (1) the discharge of water from the bottom of the Shuikou dam for electricity generation or the spillway for discharging flood. The DO of the water from the bottom of the dam was very low with a minimum monitoring value of approximately 1.5 mg · L⁻¹. In contrast, the DO of the spillway water was very high; thus, leading to the over saturation of DO. (2) The decrease in the concentrations of Bx, NH₄, DOC, and COD in the lower reaches of the Minjiang River with an increase in the water discharge from Shuikou, which resulted in a decrease in DO consumption. (3) Reaeration time (∂t) (Shen & Wang 2007; Alosairi et al. 2011) and reaeration coefficient (KR) (Bansal 1973; Palumbo & Brown 2014; Kumarasamy 2015) were determined by the discharge of Shuikou (Equation (1)), which affected the kinetic process of DO in the lower reaches of the Minjiang River. The flow diversion ratio of the North Channel (FDRNC) had a significant impact on the DO of the North and South Channels (Zhang et al. 2015). To analyse the impact of the flow diversion ratio on the DO of the North Channel under the L0 model condition, a retaining dam was set at the bifurcation of Huaiantou to increase the FDRNC to approximately 50%.

Compared with that of the L0 model, the ADOC of the L2 model decreased by 0.14 gO_{2 ·}m⁻² · day⁻¹. Furthermore, the ADOC in the South Channel and the North Channel of the L2 models decreased by 0.22 and 0.14 g · O₂· m⁻² · day⁻¹, respectively. In addition, the concentration of DO consumption factors in the downstream decreased, which could be attributed to the rapid washing away of the pollutants with an increase in the river runoff. Compared with the L0 model (FDRNC = 26%), there was a change in the ADOC of the South and North Channels of the L3 model (the FDRNC was approximately 50%). Particularly, the ADOC of the North Channel of the L3 model decreased by 0.24 g·O₂ · m⁻²· day⁻¹, and the proportion of the DO consumption of NH₄ in the North Channel decreased by 2.92% compared with that of the L0 model, which could be attributed to the increase in the river flow (Figure 6).

The K_R of each reach in the L0–L3 models was calculated using Equation (2) according to the water depth and the flow velocity. Compared with the L0 model, the flow velocity in each reach of the L1 model decreased. In addition, the flow velocity in the downstream of the Shuikou dam decreased by 0.5 m · s⁻¹; however, there was no significant change in the flow velocity of the open sea owing to the influence of the tide. In addition, compared with the L0 model, the water level of the L1 model in the reaches from Shuikou–Geyangkou decreased by 1.47 m; however, that in the reaches from Min'an–Open sea increased by 0.45 m, which could be attributed to the enhancement in the tidal jacking. Furthermore, K_R decreased with a decrease in the river runoff. Compared with those of the L0 model, the K_R of each reach in the L1 model decreased slightly from 0.15 day⁻¹ in the upstream to 0.02 day⁻¹ in the downstream. However, compared with that of the L0 model, the DO from Shuikou to Huaiantou reaches in the L1 model increased by 1.40–1.80 mg · L⁻¹. In addition, compared with that of the L0 model, the DO in each reach of the L2 model decreased by 0.10–0.50 mg · L⁻¹ at a constant K_R. This indicated that K_R was not the main factor affecting the decrease in the DO in the lower reaches of the Minjiang River with an increase in the water discharge from the Shuikou dam.

At the minimum ecological flow (L1, Figure 7(b) and 7(c)), the time taken for the water to reach the intersection of the South Channel and the North Channel at a HWS and LWS was approximately 14 and 16 days, respectively. In addition, the water age of Min'an at HWS and LWS was 0.23 and 12.28 days, respectively, which was smaller than that at the North Channel, and this could be attributed to the tidal advection and dilution of the seawater at high tide. Compared with that of the L0 model, the water age of the North Channel export and the South Channel export of the L1 model increased by 12.12 and 11.16 days, respectively, at a LWS, and 7.49 and 9.15 days, respectively, at high tide. This indicates that the time taken for hypoxic water to reach the downstream in the L1 model was several days longer than that of the L0 model, which resulted in an increase in the ∂t.

The ∂t of water is affected by river runoff. With an increase in the discharge runoff, the flow velocity increased, and the ∂t decreased. The K_R of the hypoxic water from Shuikou was significantly larger than the DO consumption, which could be attributed to the low water quality concentration of the discharged water from the Shuikou reservoir. In addition, the decrease in the DO caused by a decrease in the time was more notable than the change in K_R caused by the increase in the river runoff.

At the maximum generation flow condition (L2, Figure 8(a) and 8(b)), the time taken for water to reach Min'an at a LWS was 4.26 days. In addition, compared with the L0 model, the water ages of the North Channel export and South Channel export in the L2 model decreased by 1–2 days. Consequently, the ∂t increased, which resulted in a decrease in the DO of the L2 model compared with that of L0 in the lower reaches of the Minjiang River. Compared with the L0 model, there was no significant change in the water age of Min'an in the L3 model (Figure 8(c) and 8(d)); however, the water age of the North Channel increased by 0.83 days and that of the South Channel decreased by 1.07 days.

The aforementioned results confirmed that under the same meteorological conditions, the increase in the runoff and the decrease in the water age, ∂t, and DO in the lower reaches of the Minjiang River resulted in a low DO and low water quality concentration of the water discharged from the Shuikou reservoir (refer to L0–L3). In addition, it should be noted that the DO of the downstream might decrease with a decrease of runoff when a large amount of pollutants are discharged (refer to L1). For example, the residence time of L1 increased by 9 days in the North Channel, which received 80% of the domestic and industrial pollutants of the municipal district of Fuzhou. Consequently, the concentration of DO-consuming pollutants gradually accumulated and could not be discharged into the ocean quickly owing to tidal jacking. Thus, the amount of DO in the water column of the North Channel decreased by the process of respiration was higher than the quantity of DO re-supplied via air–water exchange or advection (Rabalais et al. 2009; Kumarasamy 2015).

The impact of river runoff on the DO and water age in the lower reaches of the Minjiang River at low flow was more significant than that at high flow. In addition, the water age in the upper reaches of the Minjiang River was mainly controlled by river runoff, and the water age in the lower reaches near the open sea was mainly determined by tidal advection (Zhang et al. 2015, 2016, 2021).

There was a negative correlation between the DO in the downstream reach of the Shuikou dam and river flow when the water discharge from the Shuikou dam was less than the maximum generation flow (2,100 m³ · s⁻¹). During the period of continuous low flow, the Shuikou reservoir capacity was released via drainage during the initial stage of a rainstorm, resulting in a sharp increase in the discharge from Shuikou. First, this affected the original balance process of DO along the lower reaches of the Minjiang River, which resulted in the rapid flow of the hypoxic water downstream; thus, leading to a sharp increase in the DO until the lowest point owing to the reduction of the ∂t. As the water discharge from the Shuikou dam decreased the previous magnitude of discharge, the DO in the reach from Geyangkou to Zhuqi returned to the previous concentration. Consequently, the concentration of the DO in the downstream exceeded the standard for a short time at high temperatures (Lanoux et al. 2013; Li et al. 2020a).

The above results were supported by the correlation analysis of daily average DO and discharged from the Shuikou dam in 2012. The full load power generation flow of the Shuikou power station was about 2,100 m³·s⁻¹. DO and runoff presented a significant negative correlation in Zhuqi, Wenshanli when Q < 2,100 m³· s⁻¹. The correlation coefficients were −0.270 and −0.526, respectively. The runoff was not correlated with DO at the Baiyantan section near the estuary. Zhuqi, Wenshanli, Baiyantan DO, and runoff showed a significant positive correlation when Q > 2,100 m³· s⁻¹, the correlation coefficients were 0.436, 0.287, and 0.229, respectively. It could be seen that when the discharge of Shuikou Reservoir was all used for power generation (Q < 2,100 m³· s⁻¹), the larger the discharge, the lower the DO in the downstream. When the discharge further increased beyond full load power generation flow (2,100 m³· s⁻¹), as the flow rate increased, the downstream DO increase (Table 3).

In this study, the spatial and temporal distributions of the water age and DO were simulated by using the Minjiang River model. The water age concept was introduced to quantitatively investigate the causes of DO evolution in a typical estuary area, which was mainly affected by runoff and tide. The results revealed that the spatial distribution of DO was affected by temperature, runoff, pollution emission, tidal advection, and hypoxic water discharge from the reservoir bottom. In addition, the results revealed that continuous low DO in water could be attributed to was caused by low flow, high temperature, pollution emission, and tidal jacking, in which, the low flow and tidal jacking led to an increase in the water age and accumulation of pollutants. In addition, a high temperature resulted in a decrease in the saturated DO and an increase in the DO consumption coefficient. Consequently, there was a continuous low DO in the water of the North Channel.

With an increase in the pollution emission along the Minjiang River, the ADOC increased from upstream to downstream. The proportion of the four DO consumption factors was 45–54% (DOC) > 37–40% (SOD) > 7–10% (NH₄) > 2–5% (Bx). The largest DO consumption proportion of NH₄ was observed in the North Channel, and the largest DO consumption proportion of DOC was observed in the estuary.

There was a negative correlation between the DO in the downstream reach of the Shuikou dam and the river flow when the water discharge from the Shuikou dam was less than the maximum generation flow (2,100 m³· s⁻¹). The short-term low DO was attributed to the rapid increase in the water flow from the Shuikou dam, the decrease in water age (∂t), and the delayed reaeration of the hypoxic water along the Minjiang River when the Shuikou reservoir capacity was released via drainage at the initial stage of a rainstorm. With a decrease in the water flow to the previous level, the water age and DO in the downstream water recovered. In addition, the results revealed that ∂t rather than K_R was the main factor affecting the decrease in DO in the lower reaches of the Minjiang River when the water discharge from the Shuikou dam was large.

Part of this work was supported by High-level Personnel Research Startup Project of North China University of Water Resources and Electric Power (NO.40768). We are grateful to the journal experts for their valuable comments on this paper.

P.Z.: Investigation, EFDC model building, Formal analysis, Writing – original draft, Writing – review & editing. B.W.: Data curation, Mapping, Writing – review & editing. Y.S.: Mapping, Writing – review & editing. Y.P.: Project administration, Writing – review & editing. C.S.: Project administration, Supervision, Writing – review & editing. R.X.: Investigation, Formal analysis, Writing – review & editing.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.