Evaluating the impact of hydrometeorological conditions on E. coli concentration in farmed mussels and clams: experience in Central Italy Uncorrected

Highly populated coastal environments receive large quantities of treated and untreated wastewater from human and industrial sources. Bivalve molluscs accumulate and retain contaminants, and their analysis provides evidence of past contamination. Rivers and precipitation are major routes of bacteriological pollution from surface or sub-surface runoff ﬂ owing into coastal areas. However, relationships between runoff, precipitation, and bacterial contamination are site-speci ﬁ c and dependent on the physiographical characteristics of each catchment. In this work, we evaluated the in ﬂ uence of precipitation and river discharge on molluscs ’ Escherichia coli concentrations at three sites in Central Italy, aiming at quantifying how hydrometeorological conditions affect bacteriological contamination of selected bivalve production areas. Rank-order correlation analysis indicated a stronger association between E. coli concentrations and the modelled Pescara River discharge maxima ( r ¼ 0.69) than between E. coli concentration and rainfall maxima ( r ¼ 0.35). Discharge peaks from the Pescara River caused an increase in E. coli concentration in bivalves in 87% of cases, provided that the runoff peak occurred 1 – 6 days prior to the sampling date. Precipitation in coastal area was linked to almost 60% of cases of E. coli high concentrations and may enhance bacterial transportation offshore, when associated with a larger-scale weather system, which causes over ﬂ ow occurrence.


INTRODUCTION
Foodborne disease caused by enteric bacteria in bivalves, mainly due to sewage discharges into the sea, has been reported since the 19th century (Foote ) However, the results obtained from these models can vary significantly depending on the geographic region being investigated. In some cases, rainfall over coastal areas is the main environmental factor driving increased Hydrological conditions prior to peak river discharges are also important for determining the presence of E. coli in seawater, as weak precipitation events may cause significant increases of concentrations when they follow a dry period, whereas higher amounts of precipitation are needed to produce the same increase of faecal coliforms in wet conditions (Kashefipour 2002 In this study, we investigated the influence of the Pescara River on the E. coli concentration in molluscs harvested from areas that are located at various distances from the river mouth. The links between river discharge, precipitation in the catchment area, and mollusc E. coli concentrations were explored using a correlation analysis. Due to the lack of directly measured discharge data, we predicted river discharge rates using a hydrological simulation performed with the Cetemps Hydrological Model For our analysis, the CHyM was forced using observed precipitation data from a network of rain gauges. Using precipitation measurements from rain gauges (or other instruments) as an input for hydrological models avoids the issue of not having time series of hydrological data available (e.g., Berenguer et al. ; Nerini et al. ). This advantage is particularly important because hydrological data are not always available from hydrographic services or may not have recently calibrated through updated rating curves. Moreover, hydrological models can then be used to investigate areas that are not commonly observed with stream gauges (e.g., seasonal streams). Because the CHyM is also used for hydrological forecasts when coupled with a meteorological model (Colaiuda et al. ), our study provides the basis for implementing a deterministic early warning system for sanitary risks associated with E. coli pollution areas, existing marine uses (Ippoliti et al. ) and dispersion models (Ippoliti et al. ).
In the following sections, an overview of the E. coli concentration and hydrometeorological data sources is given, as

Geographical framework
We analysed the Aterno-Pescara catchment, which represents a natural transect of the Italian region of Abruzzo.
The catchment encompasses the inner part of the region, In the 'middle flow' section, the river turns to the northeast and receives further water inputs from other springs. In the 'lower flow' section, the Aterno and Pescara Rivers join. The Pescara River is short but contributes significantly to the total discharge of the Aterno-Pescara basin, making the lower flow of the catchment perennial. Table 1 shows hydrological data for the Santa Teresa hydrometric station (42.424 N and 14.163 E), which is located about 9 km upstream from the mouth of the river (Russo ).
The catchment is exposed to weather fronts from both the west and the east. Western fronts, which originate from the Atlantic Ocean and primarily contain humid air masses, mainly cause precipitation in the inner part of the region, whereas unstable weather associated with an concentration of E. coli in bivalve molluscan shellfish is In the framework of the CAPS2 project (www.caps2.eu), the data of microbial monitoring were uploaded in the project database.
A web-based geographical information system    replaced with their ranks, the Spearman's coefficient r is given by: Diff 2 i n(n 2 À 1) where Diff is the rank difference between pairs of data (i), and n is the number of observations. P-values were calculated for both correlations to assess whether the data reject the null hypothesis.  We further defined rainfall and discharge peaks for each sampling data. Discharge peaks were defined as the local

Overview from November 2015 to April 2016
We established two preliminary conclusions based on an overview of the sampling period we analysed: (i) E. coli  Land-use classifications are from the Corine Landcover project (Büttner et al. 2017). 'N' indicates the negligible area (i.e., less than 5 ha).
concentrations appear to be more closely linked to the Pescara River discharge peaks than to precipitation around the river mouth and (ii) most bacterial peaks exceeding the sanitary threshold occur 3-4 days following a discharge peak.
The Aterno-Pescara basin is a natural transect of the Abruzzo region because it includes inland areas as well as the whole hill slope from the Apennines Mountains to the sea. This territory is heterogeneous and characterized by different land uses, including urban areas. Precipitation measured at the river outlet is therefore only weakly representative of precipitation over the entire river basin, especially when watershed physiography is complex.
Tables 5 and 6 indicate whether the E. coli sanitary threshold was exceeded at P1/P2 (Table 5) and PM (Table 6)    preceded by a discharge maximum, and 10 were also preceded by precipitation at the Pescara River mouth. Only one bacterial maximum (24 December 2015) was not preceded by either a precipitation or a discharge peak.
In some cases, precipitation over coastal areas and a river discharge peak occurred simultaneously, and the contribution of the two effects therefore could not be discriminated. The rainfall effect may also include combined sewer overflows, direct land-runoff into the estuary, and resuspension of contaminated sediments within the estuary reported that microbial contamination can persist for up to 1 week after a rainfall event in a smaller watershed.
We must stress that the waters and molluscs in our study were monitored for E. coli contamination at fixed 15-day intervals, as dictated by local regulations. These regulations do not consider weather conditions, river flow rates, or other abiotic factors that may affect the concentration of FIOs, and our sampling intervals for detecting potential microbial contamination may therefore not be representative of changes in these conditions because they were planned independently from them. For this reason, we suspect that our data underestimate the strength of the correlations among bacterial concentration, rainfall, and river discharge.
Based on our correlation analysis, we found a significant association between E. coli concentrations and the magnitude of the antecedent discharge peak: Spearman's correlation coefficient r D calculated for the two variables was 0.69, and the associated p-value was low (∼4.5 × 10 À5 ), indicating that our hypothesis was not rejected. The correlation between rainfall maxima and E. coli concentrations resulted in a lower correlation coefficient (r R ¼ 0.35). The associated p-value was high (∼0.065), indicating that our hypothesis was rejected.

CS1: 11-14 January 2016
Most of the precipitation that fell in the 5-day period from 11 January to 14 January 2016 was concentrated inland, with rainfall maxima (yellow shades in Figure 6)    The bacterial concentration in Posidonia/Mitilmare on 21 January was higher than the previously measured concentration on 14 January, potentially due to the combined contribution of both the first discharge peak on 13 January and the second discharge peak on 16 January. Moreover, the precipitation pattern during this second case study was different from the first case study, as precipitation mainly fell in coastal areas around the Pescara River mouth  maximum for most of our simulated case studies. Moreover, increased flow through the Pescara River outlet was linked to 87% of cases of high bacterial concentrations, while increased precipitation was linked to almost 60% of cases.
These results suggest that systematically comparing discharge rates and E. coli concentrations would be a straightforward way to validate our hypothesis that the Pescara River discharge is the main hydrometeorological driver of E. coli concentrations in harvested bivalves. We therefore recommend more bivalve sampling and bacterial measurements, especially after discharge peaks, to enable a more robust statistical analysis. The case-study analysis may be useful to evaluate these effects, as the source of bacterial contaminants may often be human activities occurring in the inland areas of the simulated region. Anthropogenic microbiological contamination of coastal marine waters from freshwater streams and runoff presents an important public health risk, as humans may be exposed to these bacterial contaminants while bathing or if they consume bivalve molluscs.
This work demonstrates that, especially at the mouths of rivers, a holistic approach based on correlations between rainfall data (intensity and location) and variation in river flow rates can be used to predict the times and places of exposure to microbiological contamination. The combined evaluation of abiotic factors (physical and chemical), hydrometeorological components, and biotic factors (bivalve organisms considered in this study) also provides holistic information on the overall state of the ecosystem. In this study, the relationship between biotic and abiotic factors highlights the role of bivalve molluscs as biological indicators, able to provide responses with a high synthesis capacity, but with poor analytical ability.
Supplementing our experimental data with additional observations of E. coli concentrations will enable the development of a model that can predict areas of potential microbiological contamination at river mouths based on forecasts of rainfall and river flow rates. This information will allow competent authorities to intervene promptly before and during potential contamination events. This integrated information may additionally contribute to a deeper understanding of the estuarine system and could even be used to develop a decision support system for aquaculture activities if the hydrometeorological model was combined

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