INTRODUCTION

Urban stormwater runoff has a sediment load that is retained by different structures that allow their sedimentation and accumulation processes (^{Nawrot et al., 2020}). Dredging and subsequent disposal of these sediments are taken into account in the management strategies of urban stormwater structures (^{Schwartz et al., 2017}). However, urban stormwater sediments adsorb different pollutants, such as heavy metals (HM) that have received increasing attention in recent years because they are not biodegradable and generate risk for the environment and human health (^{Joshi et al., 2009}; ^{Zhang et al., 2017}). Especially in water bodies, they have an impact due to their prevalence, persistence, bioaccumulation, and toxicity caused mainly by anthropogenic activities (^{Xiao et al., 2019})

Recently, different sediment quality indexes have been proposed to assess HM contamination and their toxicity evolution in bed sediments, road deposited sediments, and sediments in wetland areas. Some of them are the geo-accumulation index (Igeo) to analyse natural fluctuations in the content of a given substance in the environment (^{Barbieri, 2016}), the pollution load index (PI) which is the ratio of its concentration to their background concentration (Faiz et al., 2009), the potential ecological risk index (RI) which represents the sensitivity of the biological community to toxic substances (^{Zhao & Li, 2013}), and the enrichment factor (EF) which helps to establish the principal sources of the metals (natural or anthropogenic) (Barbieri, 2016; ^{Joshi et al., 2009}; R. ^{Kumar et al., 2019}). Besides, human health risk assessment has become a widely applied methodology to evaluate the potential risks arising from exposure to environmental contaminants (^{Ferré-Huguet et al., 2009}).

Physical and hydrological analyses have been carried out on the sediments collected in runoff waters, finding variations in the HM concentrations and the associated risk results according to their particle size distribution (PSD) and the antecedent dry weather period (ADP) (^{Deletic & Orr, 2005}; ^{McKenzie et al., 2008}; ^{Pimiento et al., 2018}).

Despite the progress reported above, there is a need for field studies that focus on understanding how human activities combined with hydrological conditions variables could generate diffuse HM pollution, especially for those that harmful to flora and fauna in low levels (^{Lynch et al., 2018}). Also, the relationship between risks associated with sediment deposition and granulometric and hydrometeorological characteristics need to be studied (^{Jang et al., 2010}). On the other hand, it is crucial to assess the influence of environmental factors such as the ADP, the rainfall characteristics (i.e. intensity and duration), anthropogenic factors such as land use, runoff contact surface, the PSD and the surface loads of HM in urban watersheds on the toxicity of the sediments (^{Guo et al., 2020}; ^{Zhan et al., 2020}; ^{Zhang et al., 2017}) in order to relate these variables with a possible risk alert.

Studies on the pollution process and efficient measures of its concentrations in urban stormwater runoff are lacking, especially in tropical regions such as Colombia (^{Liu et al., 2018}; ^{Ma et al., 2018}). Additionally, Colombia does not have a regulation that allows establishing the permissible ranges of pollutants in urban stormwater runoff.

This study aims to evaluate the risk of contamination from runoff sediments, collected in a small experimental catchment at Pontificia Universidad Javeriana (PUJ), associated with hydrological, PSD, geo-accumulative and pollutive characteristics that will be able to contribute to the decision making for the adequate management of dredged sediments in the different urban stormwater bodies.

MATERIALS AND METHODS

The field experiments were carried out on the constructed-wetland/storage-tank structure from Pontificia Universidad Javeriana (PUJ) which receives runoff from a soccer field and a parking lot building (^{Galarza-Molina et al., 2015}). The sediments were collected from the sand traps for five months (May 2016 to September 2016). The heavy metals (HM) concentrations were measured using ICP for Cooper (Cu), Chromium (Cr), Lead (Pb), and Zinc (Zn). The rainfall data were obtained from El Paraíso rain gauge station, located approximately 500 meters from the experimental site in latitude 4.62802611849305 and longitude -74.05863996311022 (Figure 1). This station is part of the early warning system of the district risk management institute of Bogotá (IDIGER, for its name in Spanish). The variables analysed were the following ones: total precipitation height (TPH), average intensity (AI), maximum intensity (MI), number of days with dry weather (DW), number of days with rainfall weather (RW), net average intensity (NAI) and antecedent dry weather period (ADP) (Table 1).

Source: "Lluvias en tiempo real" htttp://app.sab.gov.co:8080/sab/lluvias.htm

Figure 1 location of hydrometeorological station and sampling site 

Table 1 Precipitation data 

Source: authors own creation.

RISK ASSESSMENT

Igeo was calculated using equation (1) initially proposed by ^{Müller, (1979)} to analyse the pollution in aquatic botttom sediments, where Cn is the measured HM concentration in sediments and Bn is the geochemical background value of the pollutant: Cu = 40 mg/kg, Cr = 74 mg/kg, Pb = 17 mg/kg and Zn = 65 mg/kg (^{Rauch & Pacyna, 2009}). Igeo is classified as G0 unpolluted environment - Igeo≤0, G1 unpolluted to moderately polluted - 0<Igeo≤1, G2 moderately polluted - 1<Igeo≤2, G3 moderately to strongly polluted - 2<Igeo≤3, G4 strongly polluted - 3<Igeo≤4, G5 strongly to extremely polluted - 4<Igeo≤5, and G6 extremely polluted - Igeo > 5 (Faiz et al., 2009; ^{Loganathan et al., 2013}):

PI (see equation 2, where Cn and Bn are the same as for equation (1)) is classified in low level of pollution -PI<1, medium level of pollution PI≤1, medium level of pollution - 1<PI≤3, high level of pollution - PI > 3 (Faiz et al., 2009).

RI expresses the ecological risk caused by sediments. We calculated it using equation (3) proposed by ^{Hakanson, (1980)}, where Cn and Bn are the same as for equation (1), and Trn is the metal toxic response factor taken as 5 for Cu, 2 for Cr, 5 for Pb, and 1 for Zn (^{Zhao & Li, 2013}). RI is categorized as low ecological risk - RI≤150 (G0), moderate ecological risk - 150<RI≤300 (G1), considerable ecological risk - 300<RI≤600 (G2), and very high ecological risk - RI≥600 (G3) (Hakanson, 1980).

EF (equation 4) is the normalization of an HM concentration respect to a reference element. In this study, we used Aluminum (Al) which is a stable element and without vertical mobility and degradation. In equation 4, Cn and Bn are the same as for equation (1), and C_Al (7000 mg/kg) (^{Feria et al., 2010}), B_Al (15 mg/ kg) are the measured and the background concentrations of Al respectively. EF is classified as deficiency to minimal enrichment - EF<2, moderate enrichment - 2<EF<5, significant enrichment 5<EF<20, very high enrichment - 20<EF<40, and extremely high enrichment - EF>40 (^{Barbieri, 2016}):

The previously defined indexes were calculated to determine the risks of sediments in a small experimental catchment at PUJ, and the relationships between their size distributions and rainfall characteristics.

STATISTICAL ANALYSIS

The Data analysis was done using R software (^{R Core, 2020}). The ade4 library was implemented (^{Dray et al., 2017}) for Principal Component Analysis (PCA) (^{Lebart, L., Morineau, A., & Piron, 1995}) to determinate which component explains the variance and which variables are the most influential on risks outcomes. Only numerical variables of precipitation and particle diameters were taken, as well as Cd, Cu, Zn and Pb concentrations, each one with their risk indexes. Additionally, corrplot library (^{Wei et al., 2017}) was used for Spearman correlation instead of Pearson Correlation, due to the non-normal distribution of the data (Minitab 18, 2019; ^{Solutions, 2020}; ^{Statistics, 2018}) searching the variables which are most related to geo-accumulation and pollution risks by HM on sediments. The Spearman correlation coefficient is classified into five statements according to their absolute magnitude: 1) 0.00-0.1 as a negligible correlation, 2) 0.1-0.39 as a weak correlation, 3) 0.40 0.69 as a moderate correlation, 4) 0.70-0.89 as a strong correlation and 0.9-1 as the strongest correlation factor (^{Schober et al., 2018}).

RESULTS

Risk Assessment

The results obtained for risk assessment demonstrate that according to the Igeo analysis, the sediments are unpolluted (G0) by Cr, moderately polluted (G3) by Cu, and heavily polluted (G4) by Pb and Zn. PI values obtained show that the sediments have a low level of pollution by Cr, a middle level of pollution by Cu, and a high level of pollution by Pb and Zn. From the weighted result RI, the sediments are classified as moderate ecological risk. Some specific samples were classified as low ecological risk with few samples categorized as moderate ecological risk, as shown in Figure 2.

Source: authors own creation.

Figure 2 Risk assessment by geo-accumulation index - Igeo (left) (classified from G0: unpolluted environment to G6: extremely polluted), pollution load index - PI (center) (L: low level of pollution, M: medium level of pollution, H: high level of pollution), and potential ecological risk index - RI (right) (classified from G0: low ecological risk to G3: very high ecological risk) 

EF values show that the HM are generated as a result of anthropogenic activities because the values obtained for all the HM measured are higher than 40, which means that the sediments have an extremely high enrichment: pollutants come from different sources of non-cortical materials.

Statistical Analysis

The analysis showed that PICu, PIPb, IgeoCu, and IgeoPb are more associated with higher diameters (D50-D80), and also with IgeoZn and PIZn. The sediments associated with diameters between D50-D80 have lower Igeo and PI values for Cu and Pb (Figure 2 - Upper).

As with the sediment sizes, NAI is inversely related to the IgeoZn (r=-0.422) and PIZn(r=-0.416). If the magnitude of this variable is higher, then Zn indexes would be lower. It could be confirmed with the reported results by ^{Wicke et al., (2012)}. They establish that as the number of dry days in the background increases, there will be a decrease in the accumulation of pollutants, such as the accumulation rates of Zn as shown in Figure 3 - Lower

Source: authors own creation.

Figure 3 Upper: First segmentation of the Spearman matrix, correlation between granulometric diameters (D10 to D100) and risk indexes (geo-accumulation index - Igeo and potential ecological risk index - PI). Lower: Second segmentation of the Spearman matrix, correlation between hydrological variables (total precipitation height -TPH, average intensity -AI, maximum intensity -MI, number of days with dry weather - DW, number of days with rainfall weather - RW, net average intensity - NAI and antecedent dry weather period - ADP) and risk indexes (geo-accumulation index - Igeo and potential ecological risk index - PI). 

On the other hand, it is confirmed that the PCA methodology is viable to determine the influence of the PSD and hydrological variables on the values of the indexes of each element, since the first two components represent 75% of the total variance of the data as shown in Figure 4. Also, it is possible to determine that the second component shows the ADP as a variable inversely related to the IgeoCr and PI, while the DW, presents a small directly proportional relationship. Additionally, it is possible to affirm that the IgeoCu, IgeoPb, PICu, PIPb, and PIZn present high statistic variability regarding the particle diameters D50 and NAI.

Source: authors own creation.

Figure 4 Principal Components Analysis (PCA). Medium diameter - D50, hydrological variables (number of days with dry weather - DW, net average intensity - NAI and antecedent dry weather period - ADP) and risk indexes (geo-accumulation index - Igeo and potential ecological risk index - PI). 

DISCUSSION

The results obtained by EF show that the HM are supplied by various sources of pollution, such as human activities typical of urban areas, "the emissions of contained Cu, Pb and Zn particles are associated with nearby vehicular traffic, including both the combustion process and tire wear" (^{Machado et al., 2008}).

The results related to ADP and DW were obtained and could be contrasted with literature findings. The Igeo and PI values for Cr have a little slightly directly related to DW, and this relation can be observed in ^{Romero-Barreiro et al., (2015)}. Different authors have agreed with this behavior, stating that the more DW, the more pollutants charge will be found (^{Wicke et al., 2012}). Romero-Barreiro et al., (2015) found a strong correlation of ADP with the Igeo and PI risk indices, this results agrres with those obtained in this research, suggesting. that ADP is an important explanatory factor of the inverse variation of geo accumulation and pollution risks.

These relations observed between PSD and Cu and Pb risks agree with those reported in the literature concerning the importance of the PSD over the control of HM concentrations (^{Morelli et al., 2012}; ^{Yao et al., 2016}): urban runoff sediments with a coarser grain size tend to have significantly lower HM concentrations and higher heterogeneity, which could explain their high variability in HM concentrations relative to their low HM concentration (^{Kang et al., 2017}; ^{Zhao & Li, 2013}). Finer grain fractions could be reached in HM and be related to the large surface area of finer sediments with high adsorption capacity (Yao et al., 2016).

A possible accumulation derived from DW variables could affect different environment matrixes such as air, soil and surrounding vegetation since this phenomenon can generate toxic effects on the health of users and residents of areas near to road corridors (^{Kang et al., 2017}; ^{Yao et al., 2016}). In this case, higher risk indexes will be presented in the area, if the variables of NAI and ADP have lower values.

CONCLUSIONS

From the risk assessment, we can conclude that the elements that present the main geo-accumulation indexes are mainly given by Pb and Zn, and by Cu pollution index, presenting certain high risks in the samples.

Some RI values were classified as low and moderate on most of the campaigns; however, a few samples are evaluated as considerable risks which could be derived from high concentrations of Pb induced by vehicular traffic from the surrounding main roads.

Events can be classified, as representing a high or low risk according to hydrological and particle size variables. In a specific way, the D50 manages to represent the variability of the geo-accumulation and contamination indexes, opening the possibility of assessing the sediment quality indexes based on its median particle size.

Hydrological variables are important to determine the risks of urban runoff sediments. In this study, we found that the variable of dry weather is related to the values of geo-accumulation indexes and contamination of Cr and Cu. Based on this, it is possible to propose risk management tools of these sediments based on the climatological characteristics of the sector.

The inverse relationship between diameters and geo-accumulation indexes and contamination for the rain event allows determining that, if the rain accumulates sediments that present more fine particles, the risk will be higher; in the same way, for the relationship between Zn indexes and the net average intensity, since at low net average rainfall intensities, higher values of Zn contamination and geo-accumulation in the sediment can be presented.

The findings of this work reinforce the possibility of developing early warning systems for sediment risks using key hydrological and sedimentological variables. To achieve this, we recommend, as future work, to analyze more extensive databases to reach more generalizable relationships, as well as it is possible to establish relationships regarding stormwater quality and sediment transport during runoff (^{Ji, 2017}; V. ^{Kumar et al., 2019}). Additionally, this could only be achieved through hydrological measurements with better spatial-temporal resolutions, so that they can be processed as part of an online decision system.