Correlation between phytoplankton dynamics and hydrometeorological factors in the Three Gorges reservoir area tributary

In recent years, the rapid development of eutrophication in lakes and rivers has emerged as a significant global environmental issue, garnering widespread attention (Jin et al. 2005; Le et al. 2010). Research indicates that eutrophication in lakes and rivers can lead to the prolific growth of phytoplankton, resulting in decreased dissolved oxygen levels, deteriorating water quality, and potential threats to drinking water sources, posing risks to human survival (Anderson et al. 2002; Carvalho et al. 2013). Chlorophyll a (Chla) concentration in water is a crucial indicator representing phytoplankton biomass and indicating the degree of eutrophication in lakes. In eutrophic waters, Chla concentration is directly correlated with phytoplankton biomass (Zang et al. 2011). Therefore, studying Chla concentration and its relationship with various environmental factors contributes to the analysis and assessment of the current state and causes of eutrophication in lakes and rivers.

Phytoplankton growth is not only dependent on sufficient nutrient availability but also closely linked to hydrometeorological conditions such as river flows, temperature, sunlight, wind, and precipitation (Trombetta et al. 2019; Liu et al. 2024). Long-term studies on Lake Taihu have demonstrated that meteorological factors contribute significantly to phytoplankton, on par with nutrient inputs (Zhang et al. 2018). Global climate warming is expected to accelerate phytoplankton growth rates and alter the seasonal climate patterns, including periods of droughts, floods, and storms, leading to more frequent phytoplankton blooms (Paerl & Huisman 2009). Abundant precipitation can increase inflow and outriver flows, raise water levels, and alter the depth of the mixed layer and the sub-lake layer, consequently changing the availability of light and nutrients and impacting the functional groups of phytoplankton (Figueredo & Giani 2001; Zeng et al. 2006). Simultaneously, heavy precipitation may elevate nutrient loads in the water, further promoting phytoplankton blooms (Swinton & Boylen 2014). Under conditions of ample nutrient availability, slow water currents, favorable light exposure, and suitable temperatures, phytoplankton proliferate rapidly. Reduced wind speed and increased sunlight have been linked to the prolonged duration of cyanobacterial blooms in lakes (Neukermans et al. 2018; Oziel et al. 2020). In the case of the Three Gorges Reservoir, a decrease in rainy days has been associated with the summer cyanobacterial outbreak in the Shennongxi River in 2008 (Zhu et al. 2012). Wind also plays a significant role in shaping phytoplankton distribution, with strong winds causing phytoplankton to distribute uniformly throughout the water column. When wind speeds drop below 3.1 m/s, phytoplankton tend to accumulate on the water's surface, forming blooms. Wind speeds of 1–2 m/s are particularly conducive to the formation of thin, extensive surface blooms of Microcystis in Lake Taihu (Cao et al. 2006). Wind primarily affects the spatial distribution of phytoplankton and impacts the exchange of CO₂ and O₂ at the water–air interface, influencing phytoplankton photosynthesis and respiration processes. These findings highlight the complex interplay of hydrometeorological factors in governing phytoplankton growth and bloom dynamics (Mesman et al. 2022; Zhao et al. 2022).

The Xiangxi River is the largest primary tributary in the Hubei region of the Three Gorges Reservoir (Rao et al. 2022). There are significant seasonal and annual fluctuations in the water quality of the Xiangxi River, and studies have shown that such fluctuations are significantly related to climate change (Burgan et al. 2013; Hu et al. 2018). It is known for the frequent occurrences of phytoplankton blooms. Various scholars have discussed the potential causes of phytoplankton blooms in the Xiangxi River, examining factors such as the relationship between phytoplankton community structure and environmental factors (Cai & Hu 2006; Yang et al. 2014), meteorological conditions (Ji et al. 2013), and hydrodynamics (Ji et al. 2010). However, most of these studies have been conducted at monthly or interannual scales, focusing primarily on the relationship between water quality and algal growth. There has been limited research into the short-term responses of algal communities to hydrometeorological factors at the daily scale in different regions of the reservoir (Liu et al. 2023b). Furthermore, previous studies have typically focused on linear correlations, overlooking the complex non-linear relationships between Chla concentration and various environmental factors. The generalized additive model (GAM) is a non-parametric extension of generalized linear models, capable of directly addressing non-linear relationships between response variables and multiple explanatory variables (Liu 2008; Chouldechova & Hastie 2015; Wood 2017). It allows for a more in-depth exploration of the relationships between response and explanatory variables and the determination of the importance of each explanatory variable. GAM models have been applied in various research fields, including agriculture, disaster, and ecology (Yee & Mitchell 1991; Fewster et al. 2000; Liu et al. 2022; Liu et al. 2023a). In recent years, some researchers have also applied GAM models to studies of lake environments, analyzing the relationships between Chla concentration and various environmental factors. Therefore, this study employs the GAM model to analyze the relationship between Chla concentration and environmental factors (Chen et al. 2012; Zhang et al. 2021), providing a robust method for exploring the dynamic impact of environmental factors on the phytoplankton community in the Xiangxi River, a tributary within the Three Gorges Reservoir area.

This study focuses on the period of May to June, characterized by dramatic hydrometeorological fluctuations during the summer season. The long-term average concentrations of total nitrogen (TN) and total phosphorus (TP) in May are 1.76 ± 0.35 and 0.11 ± 0.04 mg/L, respectively, while in June, they are 1.75 ± 0.28 and 0.10 ± 0.04 mg/L, respectively. Throughout the entire study period, there is no significant variation in TN and TP concentrations, and they have already exceeded the water's nutrient threshold (TN > 0.2 mg/L, TP > 0.02 mg/L). This indicates an ample supply of nutrients required for phytoplankton blooms. Therefore, it is highly likely that the outbreak of phytoplankton blooms in the Xiangxi River is closely related to hydrometeorological factors. Through high-frequency observation in both the perennial backwater area and the dynamic backwater area, this study investigates the short-term hydrometeorological conditions’ impact on phytoplankton biomass. This research aims to provide valuable insights into the prediction and control of phytoplankton blooms in the Three Gorges Reservoir, offering a basis for early warning and prevention measures.

The Xiangxi River is situated in the northwest of Hubei Province, with a longitude range of 110.29°E to 111.13°E and a latitude range of 30.96°N to 31.67°N. It spans a length of 94 km and lies approximately 33 k downstream from the Three Gorges Dam, covering a total drainage area of 3,099 km². Following the impoundment of the Three Gorges Reservoir, the stretch of the Xiangxi River from Xiangxi Town in Zigui County to Zhaojun Town in Xingshan County has transformed into a reservoir bay. Due to the reduced flow velocity in this reservoir bay, instances of phytoplankton blooms in the Xiangxi River have been recurrent.

The collected data were preprocessed using MATLAB 2021 and correlation analysis was performed using Spearman with a confidence interval of 0.95.

Parameters of GAM such as edf, p-value, and deviance explained were used to characterize the statistical results of the model, where edf represents the degree of freedom of the estimated values, which is used to determine whether the dependent variable is linearly related to the respective variables (edf = 1 indicates that the environmental factor has a linear relationship with the Chla concentration, and larger values imply a greater ability to influence non-linearly); p-value represents the level of significance of the statistical results, which is used to assess the correlation between the effects of each factor on the dependent variable; deviance explained is the rate of explanation of the overall change of the dependent variable by the model. The value represents the level of significance of the statistical results, which is used to assess the correlation between the effects of each factor on the dependent variable; deviance explained is the rate of explanation of the overall changes in the dependent variable. This methodology was implemented using the mgcv package in the R programming language.

In order to quantify the importance of different hydrometeorological factors in influencing phytoplankton blooms, the importance of key hydrometeorological factors will be ranked in this study using a random forest model. The random forest algorithm, initially proposed by Breiman in 2001, is a powerful non-linear modeling tool that encompasses both classification and regression techniques (Breiman 2001). This approach involves utilizing the bootstrap sampling method, where in K samples are repeatedly drawn with replacement from the original training dataset N to create a new ensemble of training datasets. Subsequently, K decision trees are generated from these bootstrap samples, forming the random forest. In the context of regression analysis, the final prediction is derived by averaging the predictions of all decision trees. To assess the importance of each variable, we applied the prediction accuracy method, which measures how much the accuracy of random forest predictions decreases when the values of a variable are replaced with random values. This methodology was implemented using the randomForest package in the R (version 4.2.2) programming language (Liaw & Wiener 2002).

A random forest model was employed to rank the importance of different impact factors by extracting the factor, the prediction result is expressed as the mean square error (MSE), the larger the MSE is, the greater the influence weight of the factor. From the results listed in Table 2, the importance of the impact factors at the XK station were river flow, temperature, solar radiation, and wind speed in descending order of importance and this order is similar in the PYK station and the ZJ station, with river flow, solar radiation, and precipitation in that order.

The result indicates that hydrometeorological factors account for more than 30% of the variance in phytoplankton dynamics. In the perennial backwater area, the main influencing factors are river flow, temperature, solar radiation, and wind speed. In contrast, in the dynamic backwater area, the primary factors are river rates, solar radiation, and precipitation. This highlights that the hydrometeorological factors affecting phytoplankton dynamics vary in different regions of the reservoir river. The dynamic backwater area is particularly influenced by reservoir operations, experiencing changes over time, and is more susceptible to the scouring effects of upstream inflows. On the other hand, the perennial backwater area resembles a lake-like environment and is consistently influenced by the river flow in the main river channel.

The river flow of the Xiangxi River is the most direct influencing factor on phytoplankton dynamics. In May, there are two peak flow events, while in June, the river flow remains within a lower range, leading to higher phytoplankton in June compared to May (see Figure 3). The increase in river flow is somewhat correlated with precipitation (r = 0.45, P < 0.01). When the river flow exceeds 100⁠, the Chla concentration at the Xiakou, Pingyikou, and Zhaojun stations remains at lower levels (see Figure 4(a)). Some studies suggest that when the river flow of the Xiangxi River exceeds 100⁠, the flow velocity can reach 0.05 m/s, which has a certain inhibitory effect on phytoplankton growth (Yin et al. 2014). This indicates that an increase in the Xiangxi River's river flow, accompanied by a significant increase in flow velocity, is favorable for suppressing phytoplankton blooms (Li et al. 2022). This scouring effect is more pronounced in the dynamic backwater area than in the perennial backwater area. Previous research has shown that in the downstream of the Han River, a reduction in river flow and slower flow velocities can easily trigger phytoplankton blooms. Increasing river flows can effectively suppress phytoplankton blooms (Li et al. 2021). It indicates that tributary inflow can also inhibit phytoplankton growth by reducing water temperature and increasing turbidity. Therefore, increasing tributary flow directly from upstream is one of the means to control phytoplankton blooms.

Precipitation has differential effects on phytoplankton in different regions. In the perennial backwater area, the impact of precipitation on phytoplankton is not significant, while it directly affects phytoplankton in the fluctuating backwater area, as shown in Table 1. Precipitation exerts a certain inhibitory effect on phytoplankton growth, with some studies finding that the specific growth rates of phytoplankton are lower during rainy periods compared to non-rainy periods. During precipitation, increased vertical turbulence and shear forces disrupt vertical stratification, which is an important reason for inhibiting phytoplankton blooms (Weng et al. 2019). Additionally, precipitation impacts phytoplankton community structure through scouring, nutrient input, and enhanced water mixing (Reichwaldt & Ghadouani 2012). In this study, the different responses of phytoplankton to precipitation in the two locations may be due to variations in the extent of scouring caused by different precipitation events. Precipitation increases river flows, and in the fluctuating backwater area located closer to the upstream, the scouring effect of precipitation is more pronounced. Precipitation, by increasing river flows and reducing water temperature, helps disrupt thermal stratification, primarily by increasing the mixing layer. This disruption of the phytoplankton growth environment, coupled with favorable light and temperature conditions, leads to rapid phytoplankton growth and reproduction in the bay, resulting in a subsequent increase in the surface Chla concentration. Precipitation has a phase-specific inhibitory effect on phytoplankton blooms (Yang et al. 2017). The data from this study indicate that phytoplankton gradually increases after precipitation (see Figure 8), possibly due to the release of nutrients from particles introduced by precipitation. Therefore, it is essential to be mindful of the risk of phytoplankton blooms following precipitation events.

During the summer season, sunlight intensity and temperature are important influencing factors for phytoplankton growth in the perennial backwater area on the Xiakou station, while their impact on the Pingyikou station and the Zhaojun station in the fluctuating backwater area is weaker (see Table 1). Statistical analysis reveals that Chla concentration in the Xiakou station is significantly positively correlated with solar radiation and temperature, while phytoplankton at the Pingyikou station and the Zhaojun station show no correlation with temperature. Since the Xiakou station is located in the perennial backwater area, the water body is well-mixed and exhibits characteristics similar to a lake. Under conditions of lower inflow and stable water levels, increased sunlight intensity and temperature favor phytoplankton growth and reproduction. High sunlight intensity is conducive to the growth of phytoplankton populations, promoting the occurrence of phytoplankton blooms (Xu et al. 2016). During the observing period, from May to June, the temperature in the Xiangxi River fluctuates between 20 and 32°C. There is little difference in temperature between the Xiakou area, the Pingyikou station, and the Zhaojun station. However, due to the influence of upstream inflows, the water stability at the Zhaojun station and the Pingyikou station is weaker than at the Xiakou station.

The wind speed in the Xiangxi River has a certain impact on phytoplankton in the Xiakou station, and there is a certain synergistic effect between changes in wind speed and phytoplankton (see Figure 7(b)). However, the correlation coefficient is relatively low, indicating that wind speed is one of the explanatory factors for the variation in phytoplankton in the perennial backwater area. Phytoplankton variation is influenced to a significant extent by other environmental conditions such as water temperature, transparency, and unknown factors. However, wind speed does not significantly affect phytoplankton at the Pingyikou station and the Zhaojun station. Wind speed primarily influences water body stability. Its effects on phytoplankton blooms include direct disturbance, the suspension of sediments and release of nutrients due to wind-induced turbulence, and the transport of water caused by wind. Research in Lake Taihu has found that wind has the strongest effect on phytoplankton transport, followed by direct disturbance, with indirect disturbance having the least impact (Wang et al. 2016). Lake Taihu has an average depth of 1.9–2 m and wind-induced wave disturbance re-suspends sediments, becoming a significant source of nutrients in the lake (Pang et al. 2008). In the study area of this research, the water depth is approximately 10 m, reaching up to 40 m during high water periods. The study suggests that wind-induced wave disturbance is not the primary factor affecting nutrient release from sediments (Fengqing et al. 2008). In this study, wind speeds mainly range from 0 to 2.5 m/s (see Figure 7(a)), and rapid changes in wind speed are significantly correlated with Chla concentration (see Figure 7(b)). This indicates that, under the existing wind speed conditions, as wind speed increases, there is a trend of increasing phytoplankton, possibly due to weak wind conditions promoting phytoplankton aggregation on the water surface (Yang et al. 2020).

The Xiangxi River Reservoir Bay is abundant in nutrients which is no longer the limiting factor for the phytoplankton blooms. Therefore, changes in hydrometeorological conditions may have a more significant impact on the phytoplankton blooms process. This study demonstrates that inflow from tributaries into the reservoir significantly influences phytoplankton in the reservoir bay. An increase in inflow will enhance turbidity in the reservoir bay, lower water temperatures, increase river flows, and reduce the risk of phytoplankton blooms. In the prediction of phytoplankton blooms in Lake Taihu, some researchers have employed probabilistic empirical formulas that incorporate factors such as phytoplankton concentration, dissolved oxygen concentration, wind speed, and precipitation to predict the probability of phytoplankton blooms occurrence. Under conditions of extreme precipitation and strong winds, the probability of phytoplankton blooms is low, whereas under conditions of low oxygen, weak winds, and clear, rain-free skies, the probability of phytoplankton blooms is higher (Li et al. 2016; Li & Qin 2019). There are some distinctions between the Three Gorges Reservoir and shallow water lakes such as Lake Taihu. While in the reservoir, precipitation suppresses phytoplankton, and the probability of phytoplankton blooms decreases on rainy days, the impact of precipitation factors on phytoplankton is relatively small compared to river flows. Precipitation's role in inhibiting phytoplankton blooms is limited. Following precipitation, persistent sunny and warm weather increases the risk of phytoplankton blooms in the perennial backwater areas. In phytoplankton blooms prediction and warning, certain wind speeds favor the induction of phytoplankton blooms, although their overall impact is limited. Conducting phytoplankton blooms prediction and warning requires a comprehensive analysis based on tributary inflow, light exposure, temperature, and precipitation conditions.

These relationships between phytoplankton blooms and hydrometeorological factors offer the potential to establish a relationship between long-term trends in climate change and the risk of phytoplankton blooms. Linking these results to climate change, the IPCC reports indicate that changing precipitation patterns, increased frequency of extreme weather events, and rising temperatures due to climate change could significantly alter hydrometeorological conditions, thereby affecting phytoplankton bloom dynamics in reservoirs and lakes. These projected future climate conditions suggest that the risk and patterns of phytoplankton blooms could shift, necessitating adaptive management strategies to mitigate potential ecological impacts (Masson-Delmotte et al. 2021).

The above discussion provides a thorough and innovative approach to understanding phytoplankton blooms in large reservoir systems. However, there are limitations to this study. The geographic scope is limited to the Xiangxi River tributary, which may not fully represent the dynamics in other parts of the Three Gorges Reservoir or different types of water bodies. The temporal scope focuses on specific periods of heightened phytoplankton activity, potentially limiting the generalizability of the findings across different timescales. Additionally, while GAM and random forest models are powerful, they have limitations, including the potential for overfitting and the need for extensive computational resources. Finally, the study may not account for all environmental variables that could influence phytoplankton dynamics, such as nutrient loading from agricultural runoff or other anthropogenic factors.

This study provides a comprehensive examination of the hydrometeorological factors influencing phytoplankton dynamics in the Xiangxi River, a major tributary of the Three Gorges Reservoir, by uniquely combining GAMs and random forest analysis. Through the analysis of high-frequency observing data of phytoplankton and hydrometeorological conditions in the Xiangxi River during the summer season, this study demonstrates that inflow from tributaries into the reservoir is the most significant influencing factor on phytoplankton in the reservoir bay. An increase in river flow from tributaries significantly inhibits phytoplankton, while higher light exposure and temperatures promote phytoplankton growth in the backwater area. However, their impact on the fluctuation of the backwater area is not significant. Additionally, moderate wind speeds are conducive to phytoplankton aggregation in the backwater area. This research will provide a scientific basis for predicting the summer phytoplankton blooms period in the reservoir bay and implementing scheduling measures to manage phytoplankton development trends effectively.

This integrated approach allows for a robust analysis of the complex relationships between multiple hydrometeorological factors and phytoplankton dynamics, enhancing predictive capabilities and informing better management and mitigation strategies. Focusing on a major tributary rather than the main body of the reservoir offers valuable insights into localized hydrological changes, while detailed temporal analysis aids in understanding immediate and lagged effects of factors such as precipitation and river flow. Innovations include enhanced predictive models and comprehensive data collection from automatic observing buoy stations, which improve accuracy and reliability. The research highlights seasonal dynamics, offering insights for seasonal management practices and emphasizing the study's potential applications for early warning systems and management strategies to control phytoplankton blooms. However, limitations include the geographic focus on the Xiangxi River, potential temporal scope constraints, model limitations such as overfitting, and the exclusion of some environmental variables. In conclusion, this study provides a thorough and innovative approach to understanding and managing phytoplankton blooms in large reservoir systems, with significant implications for developing effective management strategies, particularly in the context of future climate change scenarios.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.