Assessing characteristic time and space scales of in-sewer processes by analysis of one year of continuous in-sewer monitoring data

Abstract

Long-term and high-frequency in-sewer monitoring opens up a broad range of possibilities to study (influences on) water quantity and quality variations. Using data from the Eindhoven wastewater system in The Netherlands both dry weather flow and wet weather flow situations have been studied. For approximately 160 dry weather days mean diurnal variations of flow and pollutant concentrations have been derived. For wet weather situations (≈ 40 storm events) peak load factors have been studied. Generally, peak load factors for all considered pollutant parameters are larger than one. Peak load factors for particulate matter are larger than for dissolved constituents. Also, the smallest catchment area consistently shows the largest mean peak factors and vice versa.

Full text

Assessing characteristic time and space scales of
in-sewer processes by analysis of one year of continuous
in-sewer monitoring data
R. P. S. Schilperoort, J. Dirksen, J. G. Langeveld and F. H. L. R. Clemens
ABSTRACT
Long-term and high-frequency in-sewer monitoring opens up a broad range of possibilities to study
(influences on) water quantity and quality variations. Using data from the Eindhoven wastewater
system in The Netherlands both dry weather flow and wet weather flow situations have been
studied. For approximately 160 dry weather days mean diurnal variations of flow and pollutant
concentrations have been derived. For wet weather situations (≈40 storm events) peak load factors
have been studied. Generally, peak load factors for all considered pollutant parameters are larger
than one. Peak load factors for particulate matter are larger than for dissolved constituents. Also, the
smallest catchment area consistently shows the largest mean peak factors and vice versa.
R. P. S. Schilperoort (corresponding author)
J. G. Langeveld
Royal Haskoning, PO Box 151, 6500 AD Nijmegen,
The Netherlands
E-mail: r.schilperoort@royalhaskoning.com
R. P. S. Schilperoort
J. Dirksen
J. G. Langeveld
F. H. L. R. Clemens
Delft University of Technology,
PO Box 5048, 2600 GA Delft,
The Netherlands
J. Dirksen
Waternet, PO Box 94370, 1090 GJ Amsterdam,
The Netherlands
F. H. L. R. Clemens
Witteveen þBos, PO Box 233, 7400 AE Deventer,
The Netherlands
Key words |dry weather flow, in-sewer processes, monitoring data, peak factors, wastewater
composition, wet weather flow
INTRODUCTION
In the last decade wastewater monitoring equipment has seen
a drastic development, which enables continuous and high-
frequency monitoring over long time periods. This shifts the
focus in monitoring projects from collecting samples and
laboratory labour to data validation and subsequent data
analysis. For instance, Mourad & Bertrand-Krajewski
()and Liefting & Langeveld ()describe in detail
the required data validation steps, resulting in a set of vali-
dated data. This paper focuses on the subsequent data
analysis and interpretation to obtain information on charac-
teristic time and space scales of in-sewer processes in
urban wastewater systems. For this, a 19-month dataset on
wastewater quantity and quality parameters was available
from the Eindhoven area in The Netherlands. Distinct
analyses have been made for dry weather flow (DWF) and
wet weather flow (WWF) situations, focusing on differences
due to catchment characteristics and inter-event variations.
METHOD
The Eindhoven wastewater treatment plant (WWTP,
750,000 population equivalents) treats the wastewater
from ten municipalities divided over three catchment
areas that are very different in size and character, and
each has a separate inflow to the WWTP; see Figure 1.
Wastewater from Eindhoven Stad (ES, municipality of
Eindhoven) accounts for approximately 50% (or
17,000 m
3
/h) of the hydraulic capacity and discharges
directly to the WWTP. The other nine (much smaller)
municipalities are each connected to one of the two waste-
water transport mains, one to the north (Nuenen/Son or
NS, 7 km in length) and one to the south (Riool-Zuid or
RZ, 32 km in length), accounting for respectively 7%
(3,000 m
3
/h) and 43% (15,000 m
3
/h) of the hydraulic
capacity. An elaborate description of the studied waste-
water system can be found in Schilperoort ().
At each of the three inflows into the WWTP (locations ‘A’
in Figure 1) on-line sensors have been installed that measure
concentration values of wastewater quality parameters total
suspended solids (TSS), chemical oxygen demand (COD)
and filtered COD (CODf) (dissolved fraction) at an interval
of 2 min. The used sensors are Ultraviolet–visible (UV/VIS)
spectrometers (type spectro::lyser by manufacturer s::can,
Austria) that relate ultraviolet and visible light absorp-
tion to the aforementioned quality parameters. A suffix -eq
1614 © IWA Publishing 2012 Water Science & Technology |66.8 |2012
doi: 10.2166/wst.2012.115
Downloaded from https://iwaponline.com/wst/article-pdf/66/8/1614/441935/1614.pdf
by guest
on 14 July 2020

indicates that measurement results constitute an equivalent
value based on optical measurements instead of ‘standard’
laboratory analyses. For a detailed description of the sensor,
refer to e.g. Langergraber et al.().
All three UV/VIS sensors have been installed in their
own by-pass installation. Each installation is fed by a 6 L/s
shredding pump that is positioned in a pump suction
chamber of the WWTP influent pumping station; see Figure 2.
The pumps are installed after the 25 mm bar screens to
prevent too frequent clogging of the by-pass pumps. Their suc-
tion mouths are located at a height of approximately one
third of the normal DWF target water level, which is in
accordance with ISO ()to obtain a value representative
of the solids present in the wastewater flow.
The UV/VIS sensors have been calibrated per sensor
and per parameter to the local wastewater matrix by
means of two calibration campaigns: one for DWF con-
ditions (24 hourly samples on a dry weather day) and one
for WWF conditions (10 hourly samples during one storm
event). Using the reference values (grab samples, subsequent
standard laboratory analysis), a linear regression model has
been derived that corrects the sensor measurement results
(on manufacturer’s settings or ‘global calibration’) for the
local wastewater matrix (‘local calibration’); see Figure 3.
Despite known matrix variations during dry weather
conditions (e.g. Gruber et al.;Maribas et al.)as
well as (large) inter-event variations for wet weather
situations (e.g. Stumwöhrer et al.), budget constraints
have limited the calibration campaigns to a single dry
weather day and a single storm event. This should be
appreciated when studying the measurement results.
Two-minute interval flow rates at all three WWTP inflows
have been derived using monitoring data from full-pipe elec-
tromagnetic flow sensors installed at the discharge lines of
the WWTP influent pumps. The combination of the UV/VIS
and flow data yields pollutant loads per 2 min for the afore-
mentioned pollutant parameters. In total, a 19-month
(1 April 2007 –1 November 2008) dataset is available for
all parameters. After extensive data validation the data sets
have been reduced to a net time-span of approximately 1 year.
RESULTS AND DISCUSSION
Dry weather conditions
DWF days have been selected using precipitation data. The
applied definition of a DWF day is: ‘if during a 2-days time-
span in total less than 0.5 mm of precipitation has been
recorded the last day can be considered a DWF day’. This
way, around 200 DWF days have been identified. Per
2 min time-step mean values are calculated, then normalised
with respect to the overall DWF mean flow value and
plotted in Figure 4. The results constitute the mean DWF
patterns of wastewater flow from the three catchment
areas to the treatment plant.
For all three catchments a clear diurnal variation can be
observed. Compared with typical DWF patterns (e.g. Met-
calf & Eddy ) the patterns are shifted in time due to
the relatively long wastewater travel times that differ per
catchment area. As expected, peak factors reduce with
increasing catchment size. Peak values are of the same
order as for catchment areas of similar sizes (e.g. CIRIA
;Krebs et al.). Analysis of data distribution at the
six peak moments (maximum and minimum values for the
three DWF patterns) has yielded normal distributions in
five cases; only data at the minimum peak of Nuenen/Son
flow data are lognormally distributed, which might be due
to infrequent and relatively large industrial discharges
during the night.
For the same dry weather days mean DWF patterns for
catchment Eindhoven Stad have been plotted for pollutant
parameters TSS
eq
, COD
eq
and CODf
eq
(and flow Q for com-
parison) in Figure 5. DWF pollutant patterns are based on
Figure 1 |Schematic lay-out of the wastewater system in the Eindhoven area.
Figure 2 |Longitudinal profile of the pump suction chambers of the WWTP influent
pumping station (not to scale) showing the location of the by-pass installations
and pumps.
1615 R. P. S. Schilperoort et al.|WWTP influent assessment based on high-frequency monitoring data Water Science & Technology |66.8 |2012
Downloaded from https://iwaponline.com/wst/article-pdf/66/8/1614/441935/1614.pdf
by guest
on 14 July 2020

fewer days than for flow (79 for COD
eq
, 85 for TSS
eq
and
113 for CODf
eq
versus 158 for Q) as more data were dis-
carded during data validation. Pollutant concentrations
show a similar diurnal variation to flow. However, peaks
are less pronounced (closer to one, especially for night-
flow) and occur slightly later. Suspended compounds show
the largest DWF variation over a day associated with the
variation in flow values. Pollutant peak values may be less
pronounced, but –as indicated by results in the box plot –
the variation around mean values is larger than for flow.
Diurnal patterns for pollutants and flow that are similar in
shape are observed for Eindhoven Stad and Nuenen/Son,
but not for Riool-Zuid. Although for Riool-Zuid a regular
diurnal DWF pattern was observed (see Figure 4), Figure 6
shows an odd COD
eq
DWF pattern. A clear night minimum
is lacking and around noon a sharp increase in particulate
matter is observed. After ample consideration, this atypical
pollutant pattern has been attributed to the arrival of cen-
trate from the WWTP sludge processing installation. This
installation (situated 7 km further upstream in the system)
is operated 24 h per day, but centrate is only discharged to
the sewer system during working hours. At 08h00 every
morning the complete night stocks are discharged, arriving
at the WWTP around noon.
Wet weather conditions
During the 19-month monitoring campaign a large number
of storm events have occurred. To illustrate the short-term
fluctuations in concentration levels of pollutants in WWTP
Figure 4 |Mean DWF patterns for the three catchment areas. On the right, statistics at the minimum and maximum values of the three DWF patterns. The box has lines at the lower
quartile, median and upper quartile values. Whiskers extend to the most extreme values that are not considered outliers. Outliers are indicated by þ.
Figure 3 |For inflow Riool-Zuid under dry weather conditions: COD
eq
(UV/VIS sensor, global calibration) versus COD
lab
(laboratory reference values) with linear regression model and
confidence bounds. One of the 24 dry weather samples has been rejected due to improper laboratory analysis.
1616 R. P. S. Schilperoort et al.|WWTP influent assessment based on high-frequency monitoring data Water Science & Technology |66.8 |2012
Downloaded from https://iwaponline.com/wst/article-pdf/66/8/1614/441935/1614.pdf
by guest
on 14 July 2020

influent as a result of storm events, Figure 7 gives an
example for catchment area Eindhoven Stad. In the morn-
ing of 12 June 2008 a 16 mm storm event causes flow to
increase by a factor ∼6. The moment flow rises above
DWF levels large peaks in mainly particulate matter concen-
trations (TSS
eq
and COD
eq
) can be observed. Approximately
7 h into the storm event all parameter concentrations have
reduced by a factor ∼2 with respect to pre-storm DWF
values. After precipitation ceases, flow values return to
DWF values within a time-span of roughly 6 h. Pollutant
concentrations, however, recover more slowly to pre-storm
DWF levels. At the end of the event TSS
eq
and COD
eq
con-
centrations are still below normal DWF levels and continue
to recover.
Even though concentration levels fall during the storm
event, the relatively larger increase of flow rates leads to an
overall increase in total loads discharged to the WWTP from
the catchment area; see Figure 8.Thefigure shows flow,
COD
eq
loads and the 12 and 24 h moving average of loads
(all normalised to the mean DWF value). It can be observed
that directly after the onset of the event COD
eq
loads briefly
increase to roughly 10 times the mean DWF load. Con-
sidered at larger time scales, the variation of pollutant
loads during the storm event is more gradual and peak
load factors (PLF, i.e. the maximum attained normalised
load during a storm event) are hence smaller. This is illus-
trated in the figure with the application of symmetrical
moving average filters with spans of 12 and 24 h. For the
parameter COD
eq
this yields a PLF
12
of 4.8 and a PLF
24
of 3.5. The PLF
24
is an interesting parameter for WWTP per-
formance and future WWTP design criteria (Langeveld
). Essentially, it expresses the arriving load as a multiple
Figure 6 |Mean DWF patterns for COD
eq
concentrations for all three studied catchment areas. On the right, statistics (box plots) of the minimum and maximum values of the three DWF
patterns.
Figure 5 |Mean DWF patterns for flow and pollutant concentrations for catchment area Eindhoven Stad. On the right, statistics (box plots) of the minimum and maximum values of the
four DWF patterns. Please note that the vertical axis of the box plot is twice the length of the vertical axis of the line plot. Both plots are vertically centred at <1>.
1617 R. P. S. Schilperoort et al.|WWTP influent assessment based on high-frequency monitoring data Water Science & Technology |66.8 |2012
Downloaded from https://iwaponline.com/wst/article-pdf/66/8/1614/441935/1614.pdf
by guest
on 14 July 2020

of normal DWF load (which –normalised to DWF mean
and averaged over 24 h –consistently equals one). The
PLF enables easy comparison of peak loads from different
catchments without the need to define a storm event.
In total, per catchment area and per pollutant par-
ameter approximately 40 PLF
24
have been calculated.
Results are given in Figure 9. For all three catchment
areas and for all three parameters mean PLF
24
values are
larger than one. This means that for a ‘mean storm event’
the arriving pollutant load over 24 h from all inflows is sys-
tematically larger than during dry weather. The magnitude
of this ‘mean storm peak load’varies with parameter and
catchment area. The largest mean PLF
24
values are found
for TSS
eq
(2.0–4.3), followed by parameters COD
eq
(1.7–
3.3) and CODf
eq
(1.6–2.3). This suggests that during wet
weather conditions the additional discharge of suspended
solids is consistently larger than for dissolved compounds.
The smallest catchment area (Nuenen/Son) shows the lar-
gest mean peak factors whereas values are smaller for
Riool-Zuid and Eindhoven Stad, the largest catchments.
This contradicts earlier research (Kafi et al.) where
no significant variability between catchments of different
sizes could be observed. For parameter CODf
eq
the catch-
ment size effect is less pronounced than for parameters
TSS
eq
and COD
eq
.
The smallest PLF
24
values in Figure 9 are equal or
close to one. In other words, for ‘small storm events’arriv-
ing pollutant loads over 24 h are on the same order of
magnitude as mean dry weather loadings. For relatively
large storm events (i.e. the largest values in Figure 9)24-h
Figure 8 |Concentrations, loads and 12 and 24 h moving average of loads of COD
eq
for the same storm event as presented in Figure 7.
Figure 7 |Variation of flow and pollutant concentrations for catchment area Eindhoven Stad during a storm event on 12–13 June 2008.
1618 R. P. S. Schilperoort et al.|WWTP influent assessment based on high-frequency monitoring data Water Science & Technology |66.8 |2012
Downloaded from https://iwaponline.com/wst/article-pdf/66/8/1614/441935/1614.pdf
by guest
on 14 July 2020

loadings can become much larger than mean DWF values:
for parameter TSS
eq
a factor 3–6 for large areas such as
Riool-Zuid and Eindhoven Stad and up to a factor 10 for
area Nuenen/Son. For parameters COD
eq
and CODf
eq
these values are smaller, but remain much larger than
any dry weather variation.
CONCLUSIONS
Long-term and high-frequency in-sewer monitoring opens
up a broad range of possibilities to study (influences on)
water quantity and quality variations. Based on data from
the Eindhoven area, both DWF and WWF situations have
been studied. For dry weather a clear diurnal variation of
flow can be observed for all three studied catchment areas.
At peak moments, flows are often normally distributed.
For two catchment areas associated pollutant concen-
trations show a similar diurnal variation to flow. For one
area, however, a specific catchment characteristic causes
an odd diurnal variation of pollutant concentrations. For
wet weather situations the peak load factor (i.e. the largest
value of the 24 h moving average of normalised loads
during a storm event) is an interesting parameter for
WWTP performance and WWTP design criteria. For all
three catchment areas, TSS
eq
mean and individual peak
load factors are the largest, followed by COD
eq
and
CODf
eq
. The smallest catchment area consistently shows
the largest mean peak factors, and the largest and most com-
pact catchment the smallest.
ACKNOWLEDGEMENTS
The authors would like to acknowledge Waterschap De
Dommel for the use of their data in this project. We thank
E. Liefting of Royal Haskoning for his assistance in data
processing.
REFERENCES
CIRIA  Dry Weather Flows in Sewers. CIRIA Report 177,
London, UK.
Gruber, G., Winkler, S. & Pressl, A.  Continuous monitoring
in sewer networks an approach for quantifications of
pollution loads from CSOs into surface water bodies. Water
Sci. Technol. 52 (12), 215–223.
ISO  Standard 5667–10:1992. Water quality –Sampling –Part
10: Guidance on sampling of waste waters.
Kafi, M., Gasperi, J., Moilleron, R., Gromaire, M. C. & Chebbo, G.
 Spatial variability of the characteristics of combined
wet weather pollutant loads in Paris.Water Res. 42 (3), 539–
549.
Krebs, P., Merkel, K. & Kühn, V.  Dynamic changes in
wastewater composition during rain runoff. In: Proceedings
of the 8th International Conference on Urban Storm
Drainage, 30 August –3 September 1999, Sydney, Australia,
pp. 920–927.
Langergraber, G., Fleischmann, N. & Hofstädter, F.  A
multivariate calibration procedure for UV/VIS spectrometric
quantification of organic matter and nitrate in wastewater.
Water Sci. Technol. 47 (2), 63–71.
Langeveld, J. G.  Interactions within Wastewater Systems.Ph.D.
thesis, Delft University of Technology, The Netherlands.
Liefting, H. J. & Langeveld, J. G.  Sewer monitoring projects:
data collection, data handling and data quality in Arnhem.
Figure 9 |Twenty-four-hour peak load factors for TSS
eq
, COD
eq
and CODf
eq
for all observed storm events (Riool-Zuid n¼47, Eindhoven Stad n¼39, Nuenen/Son n¼42) over a 19-month
time-span.
1619 R. P. S. Schilperoort et al.|WWTP influent assessment based on high-frequency monitoring data Water Science & Technology |66.8 |2012
Downloaded from https://iwaponline.com/wst/article-pdf/66/8/1614/441935/1614.pdf
by guest
on 14 July 2020

In: Proceedings of the 11th International Conference on
Urban Drainage, 31 August –5 September 2008, Edinburgh,
Scotland, UK, CD-ROM.
Maribas, A., do Carmo Lourenço da Silva, M., Laurent, N.,
Loison, B., Battaglia, P. & Pons, M.-N.  Monitoring of
rain events with a submersible UV/VIS spectrophotometer.
Water Sci. Technol. 57 (10), 1587–1593.
Metcalf & Eddy  Wastewater Engineering. Treatment and
Reuse. Fourth edition, revised by Tchobanoglous, Burton and
Stensel, McGraw-Hill, New York, USA.
Mourad, M. & Bertrand-Krajewski, J.-L.  A method for
automatic validation of long time series of data in urban
hydrology. Water Sci. Technol. 45 (4–5), 263–270.
Schilperoort, R. P. S.  Monitoring as a Tool for the Assessment
of Wastewater Quality Dynamics. Ph.D. thesis, Delft
University of Technology, The Netherlands.
Stumwöhrer, K., Matsché, N. & Winkler, S.  Influence of
changes of the wastewater composition on the applicability
of UV-absorption measurements at combined sewer
overflows. Water Sci. Technol. 47 (2), 73–78.
First received 9 December 2011; accepted in revised form 26 January 2012
1620 R. P. S. Schilperoort et al.|WWTP influent assessment based on high-frequency monitoring data Water Science & Technology |66.8 |2012
Downloaded from https://iwaponline.com/wst/article-pdf/66/8/1614/441935/1614.pdf
by guest
on 14 July 2020