Statistical evaluation of trends of water quality monitoring parameters relevant to cyanobacterial blooms in the urban tropical reservoir

Good quality water supply is essential for maintaining the health of the population. However, the quality of water in urban supply reservoirs is susceptible to contamination and diseases that can be transmitted through the consumption of contaminated water and the agricultural irrigation using this resource (Dal'Olio Gomes et al. 2021). Due to the proximity of urban centers, reservoirs are subject to a myriad of contaminants, from domestic sewage, industrial effluents, and waste associated with agricultural activities that are rich in phosphorus, nitrogen, and potassium (Esteves 2012).

The elevated concentration of organic matter resulting from the contamination of reservoirs, combined with the increased concentrations of nitrogen, phosphorus and potassium caused by the improper disposal of domestic effluents and agricultural activities, associated with high average temperatures (15–30 °C) and a neutral to alkaline pH (range from 6 to 9), create conditions favorable to the eutrophication process of the water bodies (Paerl & Paul 2012). Once eutrophicated, a water body is subject to cyanobacterial blooms (Machado 2016). The microorganisms present in these blooms can, in turn, produce metabolites (cyanotoxins) that are highly harmful to human health, having cytotoxic, hepatotoxic, and neurotoxic effects (Uriza et al. 2017). Cyanotoxins can remain in the body of water for a long time. There are records of microcystins (main metabolites produced by cyanobacteria) that remained in bodies of water for up to six months (Uriza et al. 2017). In addition to cyanotoxins, some organisms produce organic compounds, such as geosmin and 2-methilisoborneol (MIB), which are difficult to remove by conventional water supply treatment. They are responsible for the earthy taste and musty odor of water, even in nanogram concentrations (Visentin et al. 2019). These compounds are not toxic; however, they may cause discomfort to consumers (Mustapha et al. 2021).

The regular water treatment processes carried out in the water treatment stations are not capable of efficiently removing cyanotoxins, nor the organic compounds that cause the unpleasant taste and odor, present in a eutrophic water body with cyanobacterial bloom, because of the high complexity of these molecules (Henrique et al. 2021). Hence, to remove these substances and the cyanobacterial cells present in the water body, complex techniques would be necessary, such as ultrafiltration, oxidative processes, and adsorption by activated carbon (Serrà et al. 2021). These alternative methods would make the water treatment processes more expensive and would require permanent or intermittent changes in the operation or configuration of the unit operations in the water treatment process. Therefore, the development of an active monitoring of water resources that assists in decision-making by managers is essential for maintaining the quality of water in reservoirs.

The Guarapiranga Reservoir is one of the main water suppliers in the São Paulo Metropolitan Area (SPMA), supplying about 4.9 million inhabitants (SABESP 2022) from a total of 22 million people in the SPMA (Instituto Brasileiro de Geografia e Estatística 2023). Some studies showed that the concentrations of cyanobacteria in the Guarapiranga Reservoir were higher than 50,000 cells/mL (Lizarralde 2014; Martins dos Santos 2014; Pires 2014). That is a threshold value or limit established (CONAMA 2005) for the water body classification. These results expose that the reservoir has constant and recurrent problems with cyanobacterial blooms (Sonobe et al. 2019).

Public reports of water quality parameters from reservoirs at São Paulo State in Brazil are provided by São Paulo Sanitation Company (SABESP 2022). Monthly monitoring data from the reports of the Guarapiranga Reservoir from 2010 to 2022, containing physical, chemical, microbiological, volumetric, and precipitation parameters, were used in the present work.

Most of the researchers using environmental and water quality parameters in reservoirs, including the Guarapiranga one, were carried out with in situ collections and with time intervals that rarely exceeded one year. This research project aims to understand how the dynamics of cyanobacterial blooms work on a broad time scale (over 10 years), seeking the conditions that can support the understanding of the water quality parameters that can contribute to cyanobacterial blooms. Knowing the more relevant parameters for the increasing cyano cells would assist the development of mathematical models that can predict the phenomena. This is important to avoid new blooms and mitigate them, or even if these blooms were already established, it could help in taking the appropriate decisions to preserve the quality of water.

The Guarapiranga Reservoir was impacted by irregular urban expansion, without any planning, specially related to domestic wastewater collection and disposal. Because of that, sewage is discharged into the reservoir (Nishimura et al. 2014). Besides, the reservoir receives waters from the Billings reservoir, which is impacted by the Pinheiros river, a very polluted water course, which has its waters pumped into it. The main objective of that pumping is maintaining the reservoir level and the diverse use of the water, specially providing water to other supplying systems and for energy production.

To determine the correlations of water quality parameters with cyanobacterial blooms, two statistical analyses, which are commonly used in research works in the environmental area, were applied, namely, Pearson's correlation matrix and principal component analysis (PCA) (Zeinalzadeh & Rezaei 2017; Pastro et al. 2020; Kothari et al. 2021).

The available data come from three monitoring stations: A, B, and C, at the Guarapiranga Reservoir. Details of the reservoir and the monitoring stations are shown in Figure 1. The stations were located near the dam, including water intake for supplying (station A), in the middle of the water body (station B), and near the transposition of the Billings reservoir (station C). That transposition from the Billlings to Guarapiranga Reservoir uses the Parelheiros river (Figure 1) (Secretary of the Environment of the State of São Paulo 2006).

Data obtained from reports (SABESP 2022) available were organized in spreadsheets on a time scale from 2010 to 2022. The parameters considered in this work were conductivity (μS/cm), reversal flow from the Billings reservoir (Taquacetuba branch) (m³/s), Capivari reversal flow (m³/s), turbidity (NTU), free cyanide (mg/L), chloride (mg/L), biochemical oxygen demand (BOD) (mgO₂/L), antimony (mg/L), arsenic (mg/L), barium (mg/L), cadmium (mg/L), chromium (mg/L), copper (mg/L), iron (mg/L), lead (mg/L), manganese (mg/L), mercury (mg/L), nickel (mg/L), selenium (mg/L), silver (mg/L), zinc (mg/L), phosphorus (mg/L), nitrate (mg/L), nitrite (mg/L), dissolved oxygen (mg/L), pH, total solids (mg/L), cyanobacteria (cells/mL), microcystins (μg/L), and saxitoxins (μg/L).

The parameters of precipitation (mm), Guarapiranga natural flow rate to the reservoir (m³/s), and volume (hm³) were obtained from the National Water and Sanitation Agency (ANA 2021). The meteorological data of total monthly insolation time (h) and air temperature (°C) were downloaded from the Brazilian Meteorological Institute (INMET 2021). Similarly, all data were organized in the same time scale from 2010 to 2022.

The regulatory data (Capivari Flow, Taquacetuba Flow, and Natural Flow) were calculated based on the monthly average, and the precipitation was accumulated over each month.

The correlation matrix analysis was chosen because it is widely used in the scientific literature for understanding water quality parameters and identifying correlations among these parameters (Khatoon 2013; Wang et al. 2017). The determination of the correlations between cyanobacteria and the water quality, and the meteorological and volumetric parameters were studied using the Pearson correlation coefficient with a significance of p < 0.05. This coefficient ranges from −1 to 1. Positive correlations from 0 to 1 indicate that the parameters are in consonance, that is, when one parameter increases, the others also increase, while values from 0 to −1 show that the increase of one parameter implies the decrease of another (Kothari et al. 2021). The correlation matrix analyses were performed as a guide for PCA.

With the results of the correlation matrices in hand, the PCA was performed, which allows correlating parameters of different dimensions by normalizing them and graphically and concisely displays the correlations among the analyzed data. It transforms the data sets into linear functions, so that they can be displayed bidimensionally on a diagram of two main factors (Mostafaei 2014). This analysis, in addition to allowing the visualization of correlations with cyanobacteria, also makes it possible to verify the correlations among the other water quality parameters, bringing a better understanding of how this aquatic ecosystem behaves.

The correlation matrix analysis for cyanobacteria at monitoring stations A, B and C yielded the results presented in Table 1. The values highlighted in bold were those with significant correlations (p < 0.05).

The results obtained from the correlation matrices reveal the presence of some chemical compounds not found in a reservoir in its natural preservation state. A significant finding is that these elements (antimony, arsenic, selenium, and mercury) exhibit low average concentrations. However, there is variation in the readings taken by the monitoring stations, with minimum and maximum values thus showing standard deviation, mean, and variance, which enabled the statistical analyses to be conducted. Hence, these compounds, such as antimony, arsenic, and selenium, showed positive correlations with cyanobacteria and are strong indicators of contamination of the reservoir waters by agricultural activities near the banks of the dam, as these compounds are very common in pesticides and herbicides (Souza et al. 2022). However, negative correlations were found with metallic ions (with the exception of mercury), as these compounds have a cytotoxic effect and would hinder the blooms of cyanobacteria (Esteves 2012). Mercury, in addition to being extremely toxic, exhibits bioaccumulative behavior (Li et al. 2020). When it is in environments rich in organic matter, it forms complexes through methylations of this metal with the organic matter present in the aquatic environment (Li et al. 2020). In environments with cyanobacterial blooms (eutrophicated and rich in organic matter), the complexation of this metal and bioaccumulation at trophic levels is expected (Drude de Lacerda Olaf Malm 2008).

The positive correlation for the chloride anion (stations A and B) may indicate the use of algicides in conjunction with peroxides, which are compositions widely used to combat cyanobacterial blooms and are rich in chloride ions (Ferreira 2018). As mentioned previously, station A is near the intake for supplying and one of the measures to control blooming is the addition of algicides in the reservoir (Ferreira 2018). The strong correlation with chloride may be related to the formation of trihalomethanes, which occurs when there is a high amount of organic matter in the water. This phenomenon usually happens when the reservoir is being treated with algicide mixtures containing compounds with a high presence of chlorine. This supports the elevated presence of chlorides in the water (Santos Franco et al. 2020). Notably, at station B, there is a weak positive correlation with the daily insolation rate, reinforcing what is described in the literature about these microorganisms being photosynthetic (Beltran-Perez & Waniek 2021).

Positive correlations for phosphorus and nitrites (an intermediate compound formed in the metabolization of the nitrogen matrix to nitrate) are related to the eutrophication process of the water body, which makes conditions favorable for cyanobacterial blooms that feed on these nutrients and multiply excessively, causing an imbalance in the epilimnion of the reservoir and contaminating it with cyanotoxins (Jankowiak et al. 2019). The positive correlation with turbidity matches what happens in eutrophicated reservoirs with cyanobacterial blooms; the high concentration of these microorganisms clouds the water and prevents light from reaching the lower levels of the reservoir. Total solids have a positive correlation because of the enormous amount of organic matter suspended in the reservoir and dead cyanobacterial cells.

At station C, the positive and significative correlation with saxitoxins is a strong indication of the presence of cyanobacterial blooms, as these metabolites are produced by cyanobacterial blooms (De Julio et al. 2010) and have neurotoxic action (Uriza et al. 2017). This is an alarming situation, showing the importance of controlling cyanobacterial growing. The positive correlation with phosphates and nitrites supports the indication of an eutrophicated environment with significant concentrations of cyanobacteria.

The strong positive correlation between BOD and cyanobacteria indicates that the aquatic environment in this section of the reservoir is in a state of high pollution, with the occurrence of hypereutrophication at this locale (Aprol et al. 2021). Additionally, the occurrence of blooms can be demonstrated because of the correlations with saxitoxins.

The results from the correlations of the regulation parameters of the Guarapiranga Reservoir (natural flowrate, Capivari flowrate, Taquacetuba flowrate, and volume of the reservoir) with the concentration of cyanobacteria were not significant, with the exception of station C, where a positive correlation with volume was observed. The analysis (p < 0.05) also revealed that, despite the lack of a direct correlation with the growth of cyanobacteria, the Taquacetuba flowrate branch shows significant and positive correlations with contaminants present in the reservoir such as chloride (0.2444), lead (0.2337), conductivity (0.6231), and manganese (0.2136). This may indicate that this flow is a strong contributor to the pollution and worsening of the water quality in the reservoir.

The implementation of the PCA method allowed the assessment of covariance and the behavior of relevant physical and chemical water quality parameters in relation to the average monthly count of cyanobacteria. The correlations for the two factors are presented in Table 2 for the collection data from stations A, B, and C. When analyzing the PCA results for the three monitoring stations, it can be understood that Factor 1 is highly associated with the water quality parameters, and Factor 2 is associated with the regulatory and climatic parameters of the reservoir. Thus, this observation is consistent with the results of the correlation matrices and what is described in the literature, which is that water quality parameters affect the concentration of cyanobacteria, and regulatory and climatic parameters do not show a direct correlation with the blooms of these microorganisms.

The results of the PCA are consistent with the correlation matrix results (Table 1). The covariances of the presence of cyanobacteria with antimony, arsenic, and selenium are in the same direction with angles less than 90°, which reinforces the possibility of using algicides and chlorine-based treatments (covariance with chloride). The other covariances with phosphates (total phosphorus) align with the literature for eutrophic environments rich in cyanobacteria.

The presence of mercury near the cyanobacteria with an extensive radial line shows that this compound was present when the amount of cyanobacteria was high, in line with the literature. Despite its high toxicity, this metal binds with the organic matter present in high amounts in the eutrophic aquatic environment through the phenomenon of bioaccumulation.

The diagram indicates that turbidity and total solids are in harmony with the increase of cyanobacteria. This fact is related to the low light incidence in the lower layers of the reservoir caused by the increase in cyanobacteria and the accumulation of organic matter in the lacustrine environment.

Figure 4 presents the principal components’ diagram for monitoring station B.

The results obtained from the monitoring station B data are similar to the results from station A, highlighting the relationships of the cyanobacteria with phosphates, compounds closely linked to favorable conditions for blooms of these microorganisms. As in the previously mentioned monitoring station, high concentrations of cyanobacteria in station B are also associated with high concentrations of arsenic, antimony, and chloride. This reinforces the possibility of using algicides with organochlorine compounds to control its population.

Figure 5 depicts the results of the analyses obtained by monitoring station C.

The results from station C are related to the other stations, showing the covariance of cyanobacteria in line with arsenic, antimony, and selenium, linked to the presence of chloride. Once again, this may indicate the use of algicides and compounds used in combating cyanobacterial blooms, which may be associated with the phenomenon of methylation. This occurs when a highly polluted water body (rich in organic matter) is treated with chlorinated compounds or highly oxidizing compounds (Santos Franco et al. 2020).

The pioneering nature of this work using a long-term database provided a useful understanding of the behavior of Guarapiranga, a tropical reservoir. It allowed us to identify significant correlations among water quality parameters, especially those related to algal blooms. These results can be used in developments in scientific technology to better understand the phenomena, contributing to improving the management of water resources and, consequently, preserving the health of the populations that are supplied by these reservoirs.

The results show that phosphates, nitrites, and turbidity related to water quality are crucial for the blooming of cyanobacteria, in consonance with the literature. This also provides potential evidence of measures (use of algicides and chlorine-based compounds) that are used to control these blooms.

The results found for metals and pollutants in the reservoirs are alarming, as these compounds are highly toxic to human life and are strong indicators of poor water quality. Since this reservoir is used for water supply, it is necessary to study the behavior of these pollutants in the water body and their possible origins. Some propositions have been indicated in the present study based on the literature; however, statistical results showed a positive correlation of the concentrations of some of these pollutants with the flow of the Taquacetuba branch, suggesting that this water input may be polluting the Guarapiranga Reservoir.

As a logical continuity of this research, a time-series analysis will be carried out to understand, over a long period of time, how the relevant parameters of cyanobacterial blooms behave in time and develop a predictive model. These could be useful to prevent and avoid these blooms. Moreover, diminishing the cyano blooms can bring savings in water treatment and consequently the improvement of water quality delivered to the population.

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

The authors declare there is no conflict.