Water quality index and spatio-temporal perspective of a large Brazilian water reservoir

The water spatio-temporal variability of the Irapé Hydroelectric Power Plant reservoir and its main tributaries was evaluated by analysing the temporal trend of the main parameters and applying the Canadian Council of Ministers of the Environment Water Quality Index (CCME WQI), considering data from 2008 to 2018. This reservoir is in Minas Gerais, Brazil, covering an area of approximately 142 km 2 , across seven municipalities. The dissolved iron (DFe) presented the highest percentage of standard violations (31.7% to 80.5%), with most frequencies being veri ﬁ ed in the reservoir tributaries. The Mann – Kendall test indicated that the monitoring stations showed an increasing trend of 78.5% N – NH 4 þ and 64.1% DFe. During the evaluated period, the reservoir waters were classi ﬁ ed as excellent (1.2%), good (61.3%), acceptable (29.5%), and poor (8.0%) according to the WQI for the proposed use. The poorest quality classes were more frequent in the tributaries, especially in the year 2009. The WQI seasonal assessment indicated a worsening during the rainy period in 57% of the stations, as a result of external material transport to the water bodies. The CCME WQI, in conjunction with temporal statistical analysis, contributed to the monitoring data interpretation, generating important information for reservoir water quality management.


INTRODUCTION
Quality assessment is essential for the proper use of water in different human activities and is affected by natural and anthropogenic factors as well as by hydrological dynamics.
Lentic ecosystems, such as reservoirs, have characteristics that affect the spatio-temporal scale, presenting a different dynamic in relation to lotic ecosystems. Thus, programs for monitoring physical, chemical, and biological parameters of water quality in these environments are indispensable for a better understanding and evaluation of water body conditions. However, these programs generate a complex interpretation data set, requiring specific tools for proper interpretation and evaluation of the water quality parameters temporal and spatial variability.
Several statistical techniques can be used to understand the temporal dynamics in water reservoirs, such as tests of seasonality and temporal trends of quality parameters (Penev et al. ). These analyses allow us to determine, statistically, if the values of a random variable are decreasing or increasing over a certain period as well as the seasonality of these variables (Helsel & Hirsch ; Yenilmez et al. ).
The complexity of interpreting many quality parameters can be reduced by applying quality indices that enable simultaneous evaluation of several parameters as well as natural and anthropic influences on the aquatic ecosystem's environmental dynamics. The water quality index (WQI), developed by the Canadian Council of Ministers of the Environment (CCME), aims to assess the distance between the current water quality and the goal established by the water resource framework (CCME ). Some studies have demonstrated the applicability and contribution of WQI to water quality diagnosis (Rosemond et al. ; Tyagi et al. ). These benefits can be extended with their use in conjunction with other statistical analyses.
The Irapé Hydroelectric Power Plant (HPP), located in the state of Minas Gerais, Brazil, is in a semi-arid region at risk of desertification (Tomasella et al. ). It has a reservoir of approximately 142 km 2 , whose waters are used by the population of seven municipalities. Population growth and climate change have increased concerns about the water quality in this reservoir. Thus, understanding the water quality spatio-temporal dynamics of the Irapé HPP reservoir and its tributaries is highly relevant, and may support the actions of decision makers (Helsel & Hirsch ; Penev et al. ) while contributing to a more efficient use of its waters by the local population.
In this study, the spatio-temporal variability of surface water in the Irapé HPP Reservoir and its main tributaries was evaluated using statistical techniques, comparison with environmental standards, and application of the CCME WQI. The association between statistical analysis and quality indices allows the identification of the points under greatest anthropogenic pressure in the reservoir and its surroundings, evaluates natural influences, and helps understand the temporal dynamics of pollutants in the reservoir and its tributaries. Notably, the results of the study can be applied to other reservoirs and can aid in multiple water use analysis. The Alto Jequitinhonha region predominantly produces forest products, specifically with Eucalyptus sp., agriculture, and livestock (Silva & Miranda ).

Water quality monitoring data
The secondary data used in the present study were obtained from the water quality monitoring carried out by the Companhia Energética de Minas Gerais (Cemig) at 14 sampling stations in the Irapé HPP reservoir and its tributaries, between 2008 and 2018. The geographical location and description of the monitoring stations are shown in Figure 1 and    Table 2.

Trend analysis
The temporal trend analysis of the parameters was performed by station, using the Mann-Kendall test (MK) or Mann-Kendall Seasonal test (MKS), commonly used in temporal analysis of environmental data owing to its simplicity and robustness (Yenilmez et al. ). One assumption for the tests that produced reliable results was the lack of autocorrelation in the analysed data. In this study, the autocorrelation was verified through the autocorrelation function (ACF), which measures the degree of variable correlation, at a given moment, with itself and at a later point in time.
The choice between MK and MKS was based on the presence or absence of seasonality within the data measured at different periods of the year, as this factor is a potential source of variation in the water quality data series (Helsel & Hirsch ). Seasonality was analysed using the Kruskal- the CCME WQI calculation. The index is a combination of three factors that represent non-compliance with the proposed quality criteria to produce a single value (between 0 and 100) that describes water quality (CCME ).

F1 (Scope) represents the percentage of variables that
did not meet the objectives at least once during the time period under consideration ('failed variables'), relative to the total number of variables measured (Equation (1)).
F2 (Frequency) represents the percentage of individual tests that did not meet the objectives ('failed tests') (Equation (2)). (a) The number of times by which an individual concentration is greater than (or less than, when the objective Water Temperature Temperature C * is a minimum) that of the objective is termed an 'excursion' and is expressed as follows.
When the test value did not exceed the objective (Equation (3)).
For cases in which the test value did not fall below the objective (Equation (4)).
(c) F3 is then calculated using an asymptotic function that scales the normalised sum of the excursions from objectives (nse) to yield a range between 0 and 100 (Equation (6)).
Finally, the WQI can be calculated using Equation (7).
The index ranges from 0 to 100, were divided into five categories by the CCME: (i) excellent (95-100), (ii) good (80-94), (iii) reasonable (65-79), (iv) marginal (45-64), and (v) terrible (0-44). The WQI methodology proposed by CCME does not define the parameters but recommends a minimum of eight and a maximum of twenty parameters to be used in the calculation (CCME ). In this study, all monitored parameters that defined legal standards were considered ( Table 2). The WQI was applied to each monitoring station and to each year. In the second step, the index was calculated by season, dry (April to September) and rainy (October to March), to analyse the influence of seasonality on the water quality.

Legislation standards violation
Box-plot graphs with the standard violations of each parameter for all seasons are shown in Figure 2. Trend analysis of the reservoir water quality parameters and main tributaries The autocorrelation coefficient did not reach a significant value for most parameters, except for isolated occurrences of autocorrelation for SO 4 À2 T ( Figure S1 to S14). Thus, the

Mann-Kendall tests (MK or MKS) were used for all par-
ameters at all stations.
The temperature showed a substantial seasonal variation, with values significantly higher (p < 0.05) in the rainy period in eleven stations. For the other parameters, seasonal differences, when detected, indicated higher levels during the rainy period (Table 3).
The N-NH 4 þ and DFe were the parameters that showed the highest time trend occurrences, detected in 78.6% and 64.3% of the monitoring stations, respectively. The tendency to increase N-NH 4 þ may be associated with a greater contribution from agricultural sources in the basin, which may result in greater toxicity to aquatic organisms as well as a reduction in DO concentration in the water. However, the concentrations of N-NH 4 þ did not exceed the legal standards. On the other hand, the DFe has a significant impact and a tendency to increase the reservoir water quality and its tributaries, which can be associated with an increase in soil exposure in the watershed over the study period.
Turbidity showed an upward trend at eight stations Only total solids dissolved (TDS) and thermotolerant coliforms showed significant reduction trends. All stations located in the reservoir presented a temporal tendency to reduce thermotolerant coliforms, two other stations are on   These results indicate that the point source effects of sanitary sewage on reservoir waters decreased over the study period.  There is a deterioration of the index values in the rainy season (October to March), indicating a greater supply of nutrients by rain during this period, suggesting a predominance of diffuse pollution sources (Barbosa et al. ). In addition, the amplitude of reservoir seasonal variation was less than that observed for the tributaries.

CCME WQI application
In general, it is perceived that the existing impacts on the water quality of the Irapé Reservoir have natural and anthropic origins and that they can be intensified during rainy seasons. However, even though some problems have a natural origin, it is noted that they can be aggravated by human actions, such as inadequate soil management in agricultural practices in the region. One strategy that can be adopted to minimise this problem is investment in reforesta-  The reservoir water quality and its tributaries were considered adequate during the study period, with approximately 61% of the 88 WQI qualified as good, 29.5% as acceptable, and 8.0% as bad. The reservoir points are of better quality than that are the tributaries, and the WQI assessment showed a deterioration in the rainy season, reinforcing the influence of diffuse sources on water quality.
The application of the CCME WQI together with temporal statistical analysis showed a potential contribution to the interpretation of environmental monitoring data, generating important information for reservoir water quality management. Therefore, the results can support control and management measures in specific stretches of the watershed, with a focus on the most relevant polluting sources.

ACKONOWLEDGEMENTS
The authors would like to thank the Energy Company of Minas Gerais State (CEMIG), Foundation for Research

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