Preliminary Analysis of the Water Quality Status in an Urban Mediterranean River

: Recreational use of urban rivers is becoming popular since rivers may act as amenity corridors with the potential for exhibition, recreation, leisure, relaxation and retreat. However, several point and non-point pollution sources contribute to the degradation of urban rivers’ water quality, thereby impeding their beneﬁcial uses and amenities. The physicochemical and microbiological quality of a Greek urban river (Kiﬁsos–Athens) was analyzed over a period of 12 months. A sampling campaign was implemented, collecting monthly data from ﬁve sites. Spearman’s analysis showed signiﬁcant correlation of the Hellenic Water Quality Index with speciﬁc nutrients. The total physicochemical status of all sampling stations was characterized as poor or bad. The annual average concentration of Escherichia coli ( E. coli ) was extremely high in four sites out of ﬁve, ranging from 16,822 to 26,780 cfu/100 mL. Bacteriological quality was unacceptable, as the study demonstrated the widespread occurrence of E. coli and low-quality physiochemical conditions. The spatiotemporal distribution of pollution levels revealed hotspots to be monitored further via automatic monitoring stations. A series of management and restoration measures, including tracing the exact pollution routes, should be initiated to minimize pollution pressures and establish the good ecological status of an important Mediterranean river.


Introduction
Urban rivers are important ecosystems that promote health, prosperity and wellbeing in cities, regarding both public welfare and community livelihood. Rivers offer places to walk next to river banks, boat or paddle, fish, or just sit and relax, refresh, and renew. It is therefore crucial that people are able to safely access and interact with these spaces. Nevertheless, several pollution sources driven by human activity contribute to urban river pollution, including direct discharge from industrial facilities and sewage treatment systems, agricultural runoff, recreational activities themselves and illegal actions [1]. Urban diffuse pollution from cities and transport carries pollutants from roads and urban surfaces into storm drains and out into watercourses at many locations along the river. Moreover, the atmospheric deposition, storm water and urban runoff from urban surfaces may carry various contaminants [2] including toxic metals, organic pollutants, pathogenic organisms and pharmaceuticals such as antibiotics, which create extensive environmental and human health concerns [3]. Additionally, many urban rivers have been channeled, buried or otherwise confined with the impacts of hydro-morphological alterations, enhancing the risk of further degradation in both water quality and quantity [4,5].
Water quality conditions are commonly assessed by indicators including physicochemical parameters [6], faecal indicator bacteria [7,8] and polycyclic aromatic hydrocarbons (PAHs) [9]. The latter pose a serious risk to humans, and are linked with infection or illness from waterborne pathogens during contact-recreation [10]. Guideline values for fresh water recreational water were first presented at the World Health Organization (WHO, 2003), The data presented here can be used to provide valuable insights and contribute to the development of comprehensive and inclusive management and restoration plans to tackle the degradation of urban water bodies and their in-stream habitats.

Study Area and Sample Collection
Kifisos river, including various spellings of the name (Kifissos, Cephissus, Kephisos), flows through Athens city and reaches the Saronic Gulf in the western part of Athens metropolitan area (~3.8 million inhabitants). It operates as major drainage channel for a large part of Athens. Its catchment area extends~420 km 2 with the river main route length being approximately 29 km. Most of the year, water discharge is low (~3.2 m 3 s −1 ), but during flood events, the water fluxes may reach up to 1400 m 3 s −1 [20]. According to a previous study [21], certain sections of the Kifisos drainage network, particularly in the central and north-western parts, are vulnerable to flood phenomena. Surface elevation differs between upland and low-land areas, ranging from 1350 m to sea level elevation in its estuary, while the mean elevation of the catchment area is 284.9 m. According to previous studies, approximately 65% of the geological formations are alpine, and the rest are meta-alpine formations [22].
Concerning land use, Kifisos catchment is characterized by multiple anthropogenic pressures, mainly industry, agricultural and residential use. The built-up areas (houses, roads, industries) cover most of the downstream parts of the catchment, and are estimated to exceed 70% while they are still increasing [23]. At two distinct parts along the Kifisos route, industries and manufacturing businesses have been operating on both sides of the river, functioning as unplanned and informal industrial zones. In the upstream areas, forests and shrubland are dominant, but have been significantly impacted by 2007 and 2019 wildfires. This is also where land used for agricultural and livestock use is located. Kifisos has undergone extensive river engineering in the last 13 km of its route, addressing issues of flood control, drainage system management and the road network; Kifisos river is canalized, alternating from fully boxed to open cross-section embankment. As a result, its riparian zone for this stretch was practically eliminated. Moreover, Kifisos Avenue a high-traffic-density road, and the central axis of Athens highway network was constructed over the canalized river. Five sampling locations were selected that are distributed along the drainage catchment from its upstream part, where limited pollution sources exist, to the river mouth ( Figure 1).
Monthly water surface collection was conducted in between 10:00 a.m. and 1:00 p.m. from May 2021 to April 2022 at five sampling sites (K Ekv, K 5, K 12, K 17, K MD). Physicochemical parameters such as dissolved oxygen (DO) (mg/L), electrical conductivity (EC) (µS/cm), temperature (T) ( • C), pH and salinity (ppt) were measured on-site by a portable water quality multiparameter instrument (YSI ProDSS). Surface water samples were collected in sterile bottles, preserved by the addition of HgCl 2 , stored in a cooler box and subsequently transferred to the Hellenic Centre for Marine Research (HCMR) laboratories to further analyze concentrations of NO 3 − -N (mg/L), NO 2 − -N (mg/L), DIN (mg/L), TN (mg/L) and TP (mg/L). Additionally, the output of the analyses was used to estimate the Hellenic Water Quality Index (HWQI) according to a Nutrient Classification System (NCS) initially developed by [18] and modified to include dissolved oxygen concentrations. Based on the values of five nutrient species along with dissolved oxygen levels, HWQI classifies water samples into five water quality categories: bad, poor, moderate, good and high, according to the requirements of the WFD (2000/60/EC).
For microbiological analysis, water samples were collected separately in 1 L sterile containers for the quantification of E. coli according to the International Organization for Standardization (ISO) method/protocol (ISO 9308-1:2014), which specifies the method for the enumeration of Escherichia coli (E. coli). Two additional water samples were collected in May 2021 and in September 2022 with precleaned glass bottles for TPAHs and THs analysis.  Monthly water surface collection was conducted in between 10:00 a.m. and 1:00 p.m. from May 2021 to April 2022 at five sampling sites (K Ekv, K 5, K 12, K 17, K MD). Physicochemical parameters such as dissolved oxygen (DO) (mg/L), electrical conductivity (EC) (µS/cm), temperature (T) (°C), pH and salinity (ppt) were measured on-site by a portable water quality multiparameter instrument (YSI ProDSS). Surface water samples were

Statistical Analysis
First, the Shapiro-Wilk's test was performed to assess for normality of distribution of the river water quality parameters due to the small sample size. Then, the Spearman's non-parametric correlation test was employed to reveal the relationships among the water quality parameters. We analyze the following groups of pollutants: nutrients (nitratenitrogen as NO 3 − -N (mg/L), nitrite-nitrogen as NO 2 − -N (mg/L), ammonium−nitrogen as NH4 + -N (mg/L), dissolved inorganic nitrogen as DIN (mg/L), total nitrogen as TN (mg/L), orthophosphate as PO 4 -P (mg/L), total phosphorus as TP (mg/L), pathogens (such as E. coli), and physicochemical parameters (dissolved oxygen as DO (mg/L), electrical conductivity as EC (µS/cm), pH, salinity (ppt) and temperature as T ( • C) using monthly water surface samples collected from May 2021 to April 2022 from five sampling sites. The data obtained were subjected to statistical analysis using Statistical Package for Social Sciences (SPSS), version 28 software. Box and whisker plots per sampling site were developed for nutrients, electrical conductivity (EC) and temperature (T) using an open-source application, (BoxPlotR, http: //boxplot.tyerslab.com/, accessed on 08 March 2023), to examine the distribution of the measured concentrations. Information about sample sizes was represented by the width of each box, where the widths are proportional to the square roots of the number of observations. Box and whiskers plots present the min-max and median in addition to the first and third quartiles of each variable and site for the entire sampling period.
Due to the large variability in the E. coli concentrations, the visualization of this pathogen was made via beanplots to provide a clear view of the central tendency and variability of the data. In the beanplots, the individual observations are presented as small white lines. Results were compared with EU recommendations (Directive 2006/7/EC; Annex 1 for inland waters). Comparisons were made based on the Hazen method according to the addendum of the World Health Organization (2009).
Quality classes were assigned to the sampling sites according to the Hellenic Water Quality Index (HWQI). Finally, Spearman's correlation was applied between the study variables to check for covariances. Correlations were used to identify monotonic relationships between two quantitative parameters simultaneously using the 1-tailed Spearman's rank correlation. The criterion used to check the reliability was the correlation coefficient (ρ).

Correlations between the Examined Water Quality Parameters along Kifisos River
Due to of the small size of the sample data, we used the Shapiro-Wilk test to test for normality of distributions of the following water quality parameters ( Table 1). The p values were considered statistically significant at a 95% confidence interval. According to the results of the Shapiro-Wilk test for normal distribution, most of the examined water quality parameters are not normally distributed. * p values in bold were considered statistically significant at a 95% confidence interval.
As the majority of the data did not fit a normal distribution, Spearman's correlation analysis was then applied to check for covariances. Spearman's correlation analysis showed that temperature has significant (p < 0.05) correlations with nutrients, particularly with phosphorus ( Table 2). Regression analysis for orthophosphates and total phosphorous as function of temperature yielded values (R 2 ) 0.347 and 0.346, respectively. Figure 2 also indicates a strong seasonal trend among temperature and high nutrient values, especially in the summer season (upward-sloping power curve; June-July-August).  HWQI and temperature are also strongly correlated, confirming that during warm periods, the physicochemical quality gets worse. The DO concentration is also strongly correlated with all nutrients, HWQI and E. coli, which implies that when these substances' levels increase, dissolved oxygen is decreased. This is expected since organic pollution, which also consumes dissolved oxygen due to intense microbial activity, is the main factor affecting these parameters. HWQI and E. coli are more strongly correlated with NO2 − -N and NH4 + -N than with the rest of nutrients, likely implying the frequent presence of human sewage in the river. Nevertheless, there are other sources of nutrients as well since the correlation between E. coli, NO3 − -N and HWQI is not very strong. The very high correlations between dissolved inorganic nitrogen (DIN) and total nitrogen (TN) as well as between total phosphorus (TP) and PO4-P indicates anthropogenic pollution as the dom- HWQI and temperature are also strongly correlated, confirming that during warm periods, the physicochemical quality gets worse. The DO concentration is also strongly correlated with all nutrients, HWQI and E. coli, which implies that when these substances' levels increase, dissolved oxygen is decreased. This is expected since organic pollution, which also consumes dissolved oxygen due to intense microbial activity, is the main factor affecting these parameters. HWQI and E. coli are more strongly correlated with NO 2 − -N and NH 4 + -N than with the rest of nutrients, likely implying the frequent presence of human sewage in the river. Nevertheless, there are other sources of nutrients as well since the correlation between E. coli, NO 3 − -N and HWQI is not very strong. The very high correlations between dissolved inorganic nitrogen (DIN) and total nitrogen (TN) as well as between total phosphorus (TP) and PO 4 -P indicates anthropogenic pollution as the dominant source of nutrients in the river ( Table 2). In Table 2 green-yellow-red color scale was applied to visually represent data values based on their relative magnitude compared to the rest of the values. The lowest values were colored with red shades, the middle values with yellow shades, and the highest values with green shades.

Nutrients' Statistics per Site
Nitrogen-nitrate in almost all samples was very high, with values ranging from close to 0 up to 20 mg/L ( Figure 3). More than 90% of the total samples fell in the category of "bad quality" for NO 3 − -N, and the sites having more than 50% of their samples with very high concentrations (more than 7 mg/L) are K 12 (approx. in the middle of the catchment) and K Ekv (close to the river mouth). K MD, which is in the upstream area, had the lowest NO 3 − -N values, but those values were still high enough to be classified as "bad quality." Total nitrogen (TN) values were very similar to NO 3 − -N values in most cases, which means that most of the nitrogen-related pollution is inorganic and in the form of nitrates, while a significant difference is observed at the K 5 site since TN values are generally much higher than NO 3 − -N concentrations ( Figure 3). This is due to the fact that NH4 + -N values are also very high at the specific site, which could be related to domestic or industrial sewage in close vicinity to the sampling area.  Similarly, for phosphorus, most of the samples indicated high concentrations and therefore poor and bad quality. Only the river mouth area (K Ekv) presented a slightly better picture, with more than 50% of the samples to fluctuate between the moderate and poor classifications (Figure 4). This is likely due to the effect of the sea, which generally has significantly lower P concentrations, and therefore, the mixing processes that occur at K Ekv improve slightly the quality of the site. The sites that present the highest PO 4 -P and TP concentrations are K 17 (close to the industrial area of Metamorphosis) and K 5 (in another industrial area close to Athens city center) that also had very high NH 4 + -N levels. Nevertheless, even in the upstream part of the catchment (K MD), more than 75% of the samples were classified as of bad quality. According to the HWQI, the sampling sites K Ekv, K 12 and K 17 could be characterized as of bad or poor physicochemical status in the majority of the measurements. Approximately 50% of the samples in K MD (upstream) and 25% of the samples in K Ekv and K 12 indicated moderate physicochemical quality ( Figure 5). The worst results in terms of WFD quality class were recorded in K 5, with more than 50% of the samples classified as of bad quality and all the rest as of poor quality. K 17 follows, with almost 80% of the samples classified as of poor physicochemical quality. According to the HWQI, the sampling sites K Ekv, K 12 and K 17 could be characterized as of bad or poor physicochemical status in the majority of the measurements. Approximately 50% of the samples in K MD (upstream) and 25% of the samples in K Ekv and K 12 indicated moderate physicochemical quality ( Figure 5). The worst results in terms of WFD quality class were recorded in K 5, with more than 50% of the samples classified as of bad quality and all the rest as of poor quality. K 17 follows, with almost 80% of the samples classified as of poor physicochemical quality.

Variations of E. coli Concentrations per Site
Sampling sites of the Kifisos river had a large variation in E. coli concentrations during the examined period (Figure 7). A comparative assessment was made using the Hazen method, as recommended by WHO 2009, and results are presented in Table 3. The maximum admissible number as set by the EU Directive 2006/7/EC indicative of good quality conditions was exceeded on all occasions, with the highest value found being 117,200 (cfu/100 mL) at the K5 site.

Variations of E. coli Concentrations per Site
Sampling sites of the Kifisos river had a large variation in E. coli concentrations during the examined period (Figure 7). A comparative assessment was made using the Hazen method, as recommended by WHO 2009, and results are presented in Table 3. The maximum admissible number as set by the EU Directive 2006/7/EC indicative of good quality conditions was exceeded on all occasions, with the highest value found being 117,200 (cfu/100 mL) at the K5 site. Extremely high values were also found at the K 12 and K 17 sites (100,000 cfu/100 mL), while the lowest median value of E. coli was observed at K MD (Figure 7). These results are in accordance with the nutrient findings mentioned above since there is a tendency for pollution levels to increase from the upstream to the downstream part of the river, with the exception of K 5, which is a small branch of the river that is inside an industrial area and is heavily polluted.
Moreover, the Pearson correlation coefficient (r) per site was calculated (Table 4) to examine the patterns between E. coli and the physicochemical parameters (T, EC, pH and DO). A strong negative correlation was estimated between EC and E. coli for two (K 12 and K MD) out of five sites. For these two sites, EC and E. coli concentration linear regression models were developed. These models resulted in a coefficient of determination R-squared (R 2 ) of 0.63 for the K MD site and an R 2 of 0.15 for K 12.
Extremely high values were also found at the K 12 and K 17 sites (100,000 cfu/100 mL), while the lowest median value of E. coli was observed at K MD (Figure 7). These results are in accordance with the nutrient findings mentioned above since there is a tendency for pollution levels to increase from the upstream to the downstream part of the river, with the exception of K 5, which is a small branch of the river that is inside an industrial area and is heavily polluted.  (Table 4) to examine the pa erns between E. coli and the physicochemical parameters (T, EC, pH and DO). A strong negative correlation was estimated between EC and E. coli for two (K 12 and K MD) out of five sites. For these two sites, EC and E. coli concentration linear regression models were developed. These models resulted in a coefficient of determination Rsquared (R 2 ) of 0.63 for the K MD site and an R 2 of 0.15 for K 12. Table 4. Pearson correlation coefficient (r) per site between E. coli and the physicochemical parameters (bold values are statistically significant with p < 0.01).

Discussion
As demonstrated by other researchers [3,8], the quality of recreational water must be of prescribed standards for both physicochemical and microbial quality. According to similar studies for urban rivers, Kifisos is not bathing water. However, it is used for recreational and commercial activities, where users can get in direct contact with the water via several activities (rowing, boating, kayaking, fishing and activities on the banks), making the Water Framework (2000/60/EC) as well as the bathing water (2006/7/EC) directives appropriate reference for risk assessment [7,24]. Additionally, the physicochemical water quality status of all five study locations during the sampling period was always below good threshold according to the WFD standards. Similar water quality conditions were estimated in previous studies for the Kifisos river [15].
Analysis with Spearman's correlation demonstrated that the physicochemical water quality index (HWQI) was significantly correlated with NO 2 − -N and NH4 + -N and to a lesser degree PO4-P, which implies the substantial presence of human sewage in the river system. This is attributed to the very high levels of E. coli, for which the maximum admissible E. coli concentration, as set by the EU directives, was exceeded on all occasions. EC can indirectly influence the presence and behavior of E. coli since it is often associated with the concentration of dissolved ions in water, including nutrients like phosphates and nitrates, which may support the growth and survival of E. coli bacteria. These contaminants could originate from various sources, such as industrial discharges or agricultural runoff, and can affect the viability of E. coli populations. However, EC is not a direct measure of E. coli presence or concentration. A coefficient of determination R-squared (R 2 ) of 0.63 for the K MD site indicated that the two variables are likely related to each other. However, more data are necessary for decisive validation of the relationships illuminated in this study.
Pollution of the Kifisos river with human sewage likely occurs when the sewer system gets overloaded during heavy rainfall events, during illegal disposal of industrial and domestic waste, livestock farming effluents and other illegal point pollution sources. Other studies [25,26] found similar causes that led to water quality deterioration in urban rivers. Citizen engagement is important to monitor and improve river water quality effectively and achieve the goal of a progressive reduction in discharges from the point and non-point pollution sources, coupled with knowledge on factors that are related to urban riverscapes and recreational uses [27].
We conducted this preliminary study examining the Kifisos river to analyze its water quality status, including physicochemical parameters, nutrients, E. coli and polycyclic aromatic hydrocarbons concentrations. The results indicated the significant water quality degradation of the river, particularly at downstream stations, between the stations K 17 and K Ekv, probably triggered by urban runoff, illegal sewage and industrial waste disposal. K 17 and K 5 are the stations with the worse quality status, which was expected due to their close proximity to industrial areas. Even the upstream station (K MD) illustrated a poor or bad physicochemical water quality status in more than 50% of the sampling efforts. An existing study [28] for the Kifisos river revealed similar influence by anthropogenic activities regarding major elements (nitrate and phosphate concentrations). Another study [29] revealed that the Kifisos river is highly polluted and ranked the river's quality as very poor regarding species diversity. In the current study, the average E. coli concentration was significantly high at all stations, characterizing the water as bad quality, while the total polycyclic aromatic hydrocarbons (TPAHs) results indicated the existence of significant pollution. More specifically, the TPAHs and TH concentrations detected along the river (particularly in K 5 site) indicated the presence of pollutants originating from petroleum sources. Our results regarding the levels of the TPAHs and THs are similar to other studies in urban areas [16,30]. High TPAHs and THs values should most likely be attributed to road runoff from heavy traffic highways adjacent to the river. Point source pollution from industries located along the route of the river should also be examined. Moreover, other manmade chemicals, such as heavy metals and pesticides, can also impact water quality and aquatic ecosystems. Therefore, monitoring should be done in conjunction with other water quality assessments to provide a comprehensive evaluation of river health. All of the above indicate the necessity of continuous environmental monitoring at certain points of the river and legal investigations to identify the specific polluters in order to undertake the necessary restoration actions.
In the future, a multi-pollutant modelling approach would help to better understand and manage water quality issues, since monthly sampling is unlikely to offer the necessary information for classifying the water quality status accurately [31]. These issues can be addressed by field-deployable water quality analyzers designed to continuously measure concentrations of pollutants in freshwater [32]. Switching to sustainable uses of freshwater resources is critical for minimizing land degradation and biodiversity loss [33].
From a water quality management perspective, it is evident that continuous environmental monitoring at certain points of the rivers is essential to reduce uncertainty in the results. High frequency samplings with the adoption of modern sensors can provide high frequency data [34] and substantially larger datasets to reduce uncertainty of pollution load estimates, particularly in urban rivers with multiple stressors, such as the Kifisos river [35].

Conclusions
This study is a scientific attempt to understand the dominant pollution pressures and their spatiotemporal patterns in the Kifisos river in the context of the EU Water Framework Directive. Bacteriological quality of the water in the Kifisos river was unacceptable, as the study demonstrated the widespread occurrence of E. coli and low-quality physiochemical conditions. There is no clear cut-off value at which health effects are excluded. However, the increased microbial numbers can cause recreational water quality deterioration, and it is a major problem for urban rivers such as the Kifisos river. In order to avoid further degradation and continuous decline of the ecological status of the Kifisos river, several protection and restoration measures should be undertaken within the framework of the existing EU and national policies. Additionally, we recommend legal investigations to identify polluters and undertake the necessary legal and environmental restoration actions. Surface runoff should be also included in future studies to investigate the accuracy and description of pollutant transfer processes during rainfall events and their effect on the total pollutant load of urban rivers.