Evaluation of predictive accuracy of a (micro)pollutant influent generator Laura Snip, X. Flores-Alsina, I. Aymerich, B.G. Plósz, S. Rodríguez-Mozaz, D.

Evaluation of predictive accuracy of a (micro)pollutant influent generator

Laura Snip, X. Flores-Alsina, I. Aymerich, B.G. Plósz, S. Rodríguez-Mozaz, D. Barceló, I. Rodriguez-Roda, Ll. Corominas, U. Jeppsson and K.V. Gernaey

[email protected] webinar 27th of February

27/02/2015Influent [email protected]

2 DTU Chemical Engineering, Technical University of Denmark

Outline

• Introduction• Materials and Methods:

– Catchment– Compounds studied– Influent generator– Quantitative evaluation methods

• Results:– Calibration WWTP Puigcerdà

• Conclusion



Introduction

• WWTP modelling studies need good quality influent data (Rieger et al., IWA STR no.22)

– Sampling campaigns: high work load and costs• Influent generators receive increased interest

(Martin & Vanrolleghem, 2014 Environ. Modell. Softw. 60, 188-120)

• ‘Traditional’ variables– Flow rate, ammonium etc.

• Micropollutants– Dynamics in influent and effluent– Concentrations affect reaction rate



Outline

• Introduction• Materials:



• Conclusion• Further work



Catchment - Puigcerdà

• Wastewater from Spain and France• Widespread catchment (area of 100 km2)• No industry present• Population equivalent 16.000 PE

– Fluctuating due to touristic activities– Average flow rate of 4100-8300 m3/day– Organic load of 595-1785 kg BOD/day – Nitrogen load of 123-349 kg N/day

• 60% from Puigcerdà



• Pharmaceuticals– Ibuprofen (IBU) and metabolite 2-Hydroxyibuprofen (IBU-2OH)

• Non-steroidal anti-inflammatory compound• 6 hours body residence time• Excretion in urine, 15% IBU, 9% IBU-2OH

– Sulfamethoxazole (SMX) and metabolite N-Acetyl Sulfamethazine-d4 (SMX-N4)

• Antibiotic• 10 hours body residence time• Excretion in urine, 14% SMX, 44% SMX-N4

– Carbamazepine (CMZ) and metabolite 2-Hydroxy Carbamazepine (CMZ-2OH)

• Mood stabilising drug• 8-72 hours body residence time• Excretion in urine 1% and faeces 28 % as CMZ,

4% urine CMZ-2OH

Compounds studied

Laura Snip

wat is de dose per person en hoeveel mensen zouden er dan op de 16,000 het medicijn nemen.



Influent generator

HOUSEHOLDS (HH)

INDUSTRIES (IndS)

SEASONAL CORRECTION

FACTOR

RAINFALL

HOUSEHOLDS (HH)

INDUSTRIES (IndS)

SOIL MODEL

FIRST FLUSH EFFECT

MODELASM FRACTIONATION

TEMPERATURE

FLOW RATE MODEL BLOCK

POLLUTANTS MODEL BLOCK

TEMPERATURE MODEL BLOCK

TRANSPORT MODEL BLOCK

100-aH

aH

SEWER SYSTEM

MODEL

infiltration

FIRST FLUSH EFFECT

MODEL

Gernaey et al., 2011

Environ. Modell. Softw. 26(11)



MP model - Influent generator

HOUSEHOLDS (HH)

INDUSTRIES (IndS)

SEASONAL CORRECTION

FACTOR

RAINFALL

HOUSEHOLDS (HH)

INDUSTRIES (IndS)

SOIL MODEL

ASM-X FRACTIONATION

TEMPERATURE


POLLUTANTS MODEL BLOCK

TEMPERATURE MODEL BLOCK


100-aH

aH

infiltration

PHARMACEUTICALS

Snip et al., 2014, Environ. Model. Softw., 62, 112-127

SEWER SYSTEM

MODEL

FIRST FLUSH EFFECT

MODEL



MP DAILY PROFILE

t (hours)

0 5 10 15 200,0

0,5

1,0

1,5

2,0

2,5

3,0

3,5

Phenomenological approach of occurrence MPs

MP_mg_d1000PE

MP load =

78 mg/(day*1000PE)

X PE

PE = 80.000 inhabitants

Average of the daily profile is 1

Snip et al., 2014, Environ. Model. Softw., 62, 112-127



• Peak evaluation– Magnitude of peak (PDIFF & PEP)– Timing of peak (MSDE)

• Absolute criteria– Bias of prediction (ME)– No cancelling out of errors (MAE)– Emphasis on large errors (RMSE)

• Relative criteria– Bias of prediction (MPE)– No cancelling out of errors (MARE)– Emphasis on large errors (MSRE)

• Other criteria– Index of Agreement (IoAd)– Correlation data with simulation (Corr.)

Quantitative evaluation methods

Dawson et al., 2007, Environ. Model. Softw., 22, 1034-1052

Hauduc et al., 2011, Watermatex, San Sebastian, Spain

Laura Snip



Outline

• Introduction• Materials:



• Conclusion• Further work



• Flow rate– Dry and wet weather– Blocks HH, Rainfall, Soil and Sewer system

• Soluble pollutant, ammonium – Block HH

• Particulate pollutant, COD particulate – Block HH and First flush effect

• Temperature– Block Temperature

• Automatic calibration procedure with Bootstrap (optimization of error)

Results of influent generator – Traditional compounds



HOUSEHOLDS (HH)

SEASONAL CORRECTION

FACTOR

RAINFALL

SOIL MODEL



100-aH

aH

infiltration

SEWER SYSTEM

MODEL

FIRST FLUSH EFFECT

MODEL


• Calibrated parameters:– HH: flow per PE = 110 m3/d; PE = 16,000– Soil: area connected to sewer pipes = 27,916 m2

– Sewer system: HRT = 3 h; area of sewer pipe per tank = 853.17 m2

– Rainfall: flow per mm rainfall = 823 m3/mm



• Calibrated parameters:– HH: flow per PE = 110 m3/d; PE = 16,000– Soil: area connected to sewer pipes = 27,916 m2

– Sewer system: HRT = 3 h; area of sewer pipe per tank = 853.17 m2

– Rainfall: flow per mm rainfall = 823 m3/mm

October5 10 15 20 25 30F

low

rat

e (

m3/d

ay)

0

5000

10000

15000

20000




Results of influent generator – Traditional compounds• Ammonium and COD particulates• Calibrated parameters:

– HH: • Ammonium per day per PE = 5.95 mgN/(day.PE)• COD particulate per day per PE = 55 mg/(day.PE)

– First flush effect:• Trigger flow rate = 10.000 m3/day• Maximum accumulated mass = 700 kg/SS• Fraction of settling particles = 0.40

October

5 10 15 20 25 30NH

4+ l

oad

(kg

/d)

0

100

200

300

400

500

October

5 10 15 20 25 30CO

D l

oa

d (

kg/d

)

0100020003000400050006000



Results of influent generator – Pharmaceuticals

• Correlation ‘traditional’ pollutants– IBU and IBU-2OH with ammonium 0.78 & 0.77 – SMX and SMX-N4 with ammonium 0.63 & 0.58– CMZ with TSS 0.82 and CMZ-2OH with ammonium 0.63

• Calibrated parameters:– IBU and IBU-2OH = 20.79 and 5.45 g/(day.PE)– SMX and SMX-N4 = 0.1227 and 0.08 g/(day.PE) – CMZ and CMZ-2OH = 8.86*10-2 and 0.1538 g/(day.PE)

MP DAILY PROFILE

t (hours)

0 5 10 15 200,0

0,5

1,0

1,5

2,0

2,5

3,0

3,5



Date

11/10/2012 12/10/2012 13/10/2012

Po

lluta

nt

load

(g

/d)

0

200

400

600

800

1000

Results of influent generator – Pharmaceuticals - Ibuprofen

IBU

Date

11/10/2012 12/10/2012 13/10/2012 P

ollu

tan

t lo

ad (

g/d

)0

200

400

600

800

1000

IBU-2OH



Results of influent generator – Pharmaceuticals - Sulfamethoxazole

• Considerably lower load than IBU and IBU-2OH

• Pattern less distinctive, decrease of HRT needed

• Missing of toilet flush/bad mixing

Date

11/10/2012 12/10/2012 13/10/2012

Po

lluta

nt

load

(g

/d)

0

2

4

6

8

10

Date

11/10/2012 12/10/2012 13/10/2012 P

ollu

tan

t lo

ad (

g/d

)

0

2

4

6

8

10

SMX-N4SMX



Results of influent generator – Pharmaceuticals - Carbamazepine

CMZ CMZ-2OH

• CMZ lower load than CMZ-2OH

• Different pattern due to different excretion paths

Date

11/10/2012 12/10/2012 13/10/2012

Po

lluta

nt

load

(g

/d)

0

2

4

6

8

10

Date

11/10/2012 12/10/2012 13/10/2012 P

ollu

tan

t lo

ad (

g/d

)

0

2

4

6

8

10



Results of influent generator – Quantitative evaluation

Quantitative method

Peak evaluation

Compound evaluated

PDIFF PEP MSDE

Flow rate 623.34 3.42 1.46*106

Ibuprofen -0.0045 -11.75 0.0013




Quantitative method

Absolute criteria Relative criteria

Compound evaluated

ME MAE RMSE MPE MARE MSRE

Flow rate 181.5 1.1*103 1.7*103 0.12 0.13 0.03Ibuprofen

0.0078 0.0245 0.031 4.93 0.42 0.39

Laura Snip

MAE indoen ook.




Quantitative method

Other criteria

Compound evaluatedIoAd Corr.

Flow rate 0.82 0.70

Ibuprofen 0.71 0.49



• Influent generator is capable of generating the dynamic profile of both ‘traditional’ variables and pharmaceuticals

• According to the excretion patterns of the micropollutants, different user defined profiles should be used

• Quantitative evaluation methods can help with identifying points of concern in the calibration

Conclusions



The research leading to these results has received funding from the People Program (Marie Curie Actions) of the European Union’s Seventh Framework Programme FP7/2007-2013 under REA agreement 289193.This presentation reflects only the author’s views and the European Union is not liable for any use that may be made of the information contained therein.

Thank you for your attention!

Questions?

www.sanitas-itn.eu

http://www.sanitas-itn.eu/

Evaluation of predictive accuracy of a (micro)pollutant influent generator Laura Snip, X. Flores-Alsina, I. Aymerich, B.G. Plósz, S. Rodríguez-Mozaz, D.

Documents

dtu chemical engineering

puigcerd slide

costs influent generators

work slide

reaction rate slide

flush effect model gernaey

urine cmz

catchment compounds