of the water purification ecosystem service regarding in ... scarce international... · waste water...
TRANSCRIPT
L. Boithias, R. Marcé, V. Acuña, J. Aldekoa, V. Osorio, M. Petrović, A. Ginebreda, F. Francés, S. Pérez, S. Sabater
4th SCARCE International Conference25‐26 November 2013, Cádiz, Spain
Assessment of the water purification ecosystem service
regarding in‐stream pharmaceutical residues:
Exploring the GREAT‐ER model parameters
based on data uncertainty
Barcelona
Igualada
Manresa
Anoia river
Llobregat river
Cardener river
N
Drinking water treatment facilityWaste water treatment plants
Pharmaceuticals in the Llobregat basin
Boithias et al. SCARCE 20132
• 5000 km2
• High industrial, agricultural and urban activity: 60 WWTP
• PCP = 700 mm . At outlet, discharge is 19 m3 s‐1
• The basin supplies drinking water for the 3 million inhabitants Barcelona area
• Throughout the basin, discharge may be provided by the WWTP
• High concentrations of pollutants, including pharmaceuticals
• Water purification service from WWTP and ecosystems is a major issue
Modelling the fate of pharmaceuticals
• Pharmaceuticals are ubiquitous in densely urbanized areas
• Attenuation of pharmaceutical contamination in the aquatic environment depends on:
• Few studies about the in‐stream attenuation of pharmaceuticals
• Objective: assess the ability of the spatially explicit GREAT‐ER model to simulate the concentration of 13 pharmaceuticals in two contrasted hydrological conditions
• Uncertainty analysis approach
Boithias et al. SCARCE 20133
‐ Physico‐chemical properties‐ Hydrology (sediments)
The GREAT‐ER model
• Steady‐state spatially explicit model (Boeije and Koorman, 2003) for personal care and pharmaceutical compounds
• Previously applied over several catchments
– UK: triclosan (Sabaliunas et al., 2003), LAS (Price et al., 2009), diclofenac and propranolol (Johnson et al., 2007)
– Austria: LAS, EDTA, triclosan (Wind et al., 2004)
– Germany: carbamazepine and diclofenac (Heberer et al., 2005)
– Swiss: Estrogens (Vermeirssen et al., 2006)
– Spain: Diclofenac in the Llobregat (Aldekoa et al., 2013)
Boithias et al. SCARCE 20134
The GREAT‐ER model
• 3levels of complexity ‐> simplest one
• Inputs:
Boithias et al. SCARCE 20135
Attenuation = Degradation / Sorption
Catchment descriptionPharmaceuticals
properties
‐ River stretches location (confluences, dams, WWTP, gauging stations)‐ River stretches annual discharges, velocity, depth
‐WWTP location ‐WWTP annual discharge
‐ Annual pharmaceutical emissions to sewage (kg cap‐1 yr‐1)
‐WWTP removal rates (percentage ‐ %)
‐ River removal rate (decay ‐ h‐1)
Outputs: annual average concentration in each river stretch
LLO1
LLO4
LLO3
LLO2
LLO5LLO6
LLO7
CAR3
CAR2
CAR1
CAR4
ANO1ANO2
ANO3
Data ‐ Pharmaceuticals of interest
Boithias et al. SCARCE 20136
• Discharge : ACA• Pharmaceuticals in‐stream
concentrations: 14 sampling points of SCARCE– 2 campaigns : high flow (2010) and low
flow (2011) conditions• Selected pharmaceuticals based on:
– Medical use (point source through WWTP)– At least 8 out of 14 samples with
concentration > LOQ for both campaigns– Availability of WWTP removal efficiencies
values (Gros et al., 2010; Jelic et al., 2011)– Availability of river removal efficiencies
values (literature review – 30 references)– 13 selected pharmaceuticals
• Calculated percentiles to provide a 95% confidence interval
Input data uncertainty
Boithias et al. SCARCE 20137
• Uncertainty is high, depends on:– Molecules– Number of available data
• Model minimal, median and maximal scenarios
• Simulation of the statistical distribution of observed data ‐> avoiding the calibration step
• Is the simplest model of GREAT‐ER able to model pharmaceutical fate in the Llobregat?
1<n<27
6<n<38
7<n<40
Results
Boithias et al. SCARCE 20138
Nonsteroidal anti‐inflammatory‐ 9 < RMSEmedian < 65‐Min & max encompass the 1:1 line‐Well simulated for both low flow and high flow
Results
Boithias et al. SCARCE 20139
Antiepileptic‐ 7 < RMSEmedian < 29‐Min & max encompass the 1:1 line‐Well simulated for both low flow and high flow
Antibiotic‐ 2 < RMSEmedian < 3‐Min & max encompass the 1:1 line‐Well simulated for both low flow and high flow
Sim. C
onc. (n
g L‐1)
Sim. C
onc. (n
g L‐1)
Results
Boithias et al. SCARCE 201310
Lipid regulator‐ 126 < RMSEmedian < 507‐Min & max encompass the 1:1 line‐Well simulated for both low flow and high flow
Sim. C
onc. (n
g L‐1)
Results
Boithias et al. SCARCE 201311
Sim. C
onc. (n
g L‐1)
Sim. C
onc. (n
g L‐1)
Antihistaminic‐ 4 < RMSEmedian < 5‐Min & max encompass the 1:1 line‐ Badly simulated for low flow
Beta‐blocker‐ 49 < RMSEmedian < 84‐Min & max encompass the 1:1 line‐ Badly simulated for low flow
Results
Boithias et al. SCARCE 201312
Beta‐blocker‐ 1 < RMSEmedian < 217‐ Overestimated during high flow
Lipid regulator‐ 5 < RMSEmedian < 18‐ Overestimated during low flow
Results
Boithias et al. SCARCE 201313
Sim. C
onc. (n
g L‐1)
Analgesic‐ 2 < RMSEmedian < 6‐ Underestimated during both low flow and high flow
Analgesic‐ 43 < RMSEmedian < 62‐ Underestimated during high flow and ovestimated during low flow
Discussion
• For some molecules, the simplest model is enough to describe the fate, when using the median of the available data– Non‐steroidal anti‐inflammatory drugs were well simulated– A simple statistical analysis showed that the RMSE was lower for
lower Henry low constants ‐> more volatilizable molecules
• Some molecules are badly simulated during low flow: REWWTPand RERivers may increase during low flow depending on the pharmaceutical concentration ‐> this was not simulated:
– future work: simulate 27 Med/Min/max combinations to check thisassumption
• Other uncertainty source : emissions may change spatially, but not depending on the hydrological conditions
Boithias et al. SCARCE 201314
Conclusions and future work
• Well simulated molecules could already be used to assess the water purification service, i.e. the contaminants removal, at basin scale
• Next step: run the 27 scenarios of Min/Max/Median combinations, to get more details about the effect of the hydrological conditions on the removal efficiencies
• Ongoing work: automatic sensitivity analysis and an automatic calibration (GREAT‐ER more complex tiers)
Boithias et al. SCARCE 201315